Context
Positron emission tomography (PET) of 68Ga-labelled prostate-specific membrane antigen (68Ga-PSMA) is an emerging imaging modality introduced to assess the burden of prostate cancer, typically in biochemically recurrent or advanced disease. 68Ga-PSMA PET provides the ability to selectively identify and localize metastatic prostate cancer cells and subsequently change patient management. Owing to its limited history, robust sensitivity and specificity data are not available for 68Ga-PSMA PET–positive scans.
Objective
A systematic review and meta-analysis of reported predictors of positive 68Ga-PSMA PET and corresponding sensitivity and specificity profiles.
Evidence acquisition
We performed critical reviews of MEDLINE, EMBASE, ScienceDirect, Cochrane Library, and Web of Science databases in April 2016 according to the Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) statement. Quality was assessed using the Quality Assessment if Diagnostic Accuracy Studies-2 tool. Meta-analysis and meta-regression of proportions were performed using a random-effects model with pre-PET prostate-specific antigen (PSA) levels as the dependent variable. Summary sensitivity and specificity values were obtained by fitting bivariate hierarchical regression models.
Evidence synthesis
Sixteen articles involving 1309 patients were analysed. The overall percentage of positive 68Ga-PSMA PET among patients was 40% (95% confidence interval [CI] 19–64%) for primary staging and 76% (95% CI 66–85%) for biochemical recurrence (BCR). Positive 68Ga-PSMA PET scans for BCR patients increased with pre-PET PSA. For the PSA categories 0–0.2, 0.2–1, 1–2, and >2 ng/ml, 42%, 58%, 76%, and 95% scans, respectively, were positive. Shorter PSA doubling time increased 68Ga-PSMA PET positivity. On per-patient analysis, the summary sensitivity and specificity were both 86%. On per-lesion analysis, the summary sensitivity and specificity were 80% and 97%, respectively.
Conclusions
In the setting of BCR prostate cancer, pre-PET PSA predicts the risk of positive 68Ga-PSMA PET. Pooled data indicate favourable sensitivity and specificity profiles compared to choline-based PET imaging techniques.
Patient summary
Positron emission tomography using 68Ga-labelled prostate-specific membrane antigen is an emerging radiological technique developed to improve the characterisation of metastatic prostate cancer. We summarised the data available to date and found that this new test provides excellent rates of detection of cancer spread in late-stage prostate cancer.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Prostate cancer is among the most prevalent cancers worldwide and is the third most common cause of cancer-associated mortality among men [1]. While the introduction of PSA-screening has led to earlier diagnosis of prostate cancer [2], a subset of patients develops high-risk or metastatic disease. Radiographic diagnostic studies are critical in the evaluation of these patients, particularly in assessing the presence of metastatic disease before definitive treatment or disease recurrence following treatment. Positron emission tomography (PET) is a molecular imaging modality that provides diagnostic information in the setting of malignancy. In the context of prostate cancer, choline-based and fluorodeoxyglucose tracers have been used because of their affinity to prostate cancer [3]. Despite significant advances in these techniques, their diagnostic capability is limited, and they cannot reliably identify local recurrence, lymph node involvement, or soft-tissue deposits [3], [4], and [5].
Prostate-specific membrane antigen (PSMA) is a transmembrane protein with a 707-amino-acid extracellular portion. The PSMA gene (FOLH1) is located on the short arm of chromosome 11. PSMA is expressed in the apical region of prostatic cells, the epithelium surrounding prostatic ducts [6]. Dysplastic changes in the prostate result in the expression of PSMA on the luminal surface of prostatic ducts [7] and [8]. Increasing prostate cancer stage and grade result in higher cell-membrane PSMA expression [9] and [10]. The eventual progression to advanced prostate cancer and castrate resistance corresponds to further increases in PSMA expression [11]. PSMA expression in prostate cancer cell membranes is 100- to 1000-fold that in normal cells [9] and [10]. Thus, PSMA represents a promising target for imaging of prostate cancer. The first imaging agent targeting PSMA was an antibody conjugated to diethylenetriaminepentaacetic acid and radiolabelled with 111In (111In-capromab-pendetide; Prostascint) [12] but this targeted the intracellular epitope and was further limited by single-photon emission computed tomography (SPECT) imaging. The antibody J591 targeting the extracellular domain and labelled with a positron emitter 89Zr [13] was a significant advance, but is limited by slow plasma clearance and slow tumour uptake. Most recently, groups have reported the use of small-molecule PSMA inhibitors labelled with radiotracers including 68Ga [14] and [15], 18F [16] and [17], 89Zr [13], and 99mTc [18] that bind with high affinity to the PSMA receptor.
To date, the use of 68Ga-PSMA has been well reported, with the compound Glu-NH-CO-NH-Lys-(Ahx)-[68Ga]-HBED (68Ga-PSMA-11) developed by the Heidelberg group in Germany. Initial series revealed superior sensitivity and specificity profiles compared to conventional choline-based tracers [19]. We systematically reviewed the literature outlining the use of 68Ga-PSMA PET imaging in advanced prostate cancer. Our aim was to identify predictors of positive 68Ga-PSMA PET and the true sensitivity and specificity of 68Ga-PSMA PET–positive lesions confirmed by histopathology. We also clarify the role of 68Ga-PSMA in primary staging and recurrence staging in prostate cancer.
A systematic review was performed in accordance with Cochrane Collaboration and Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines [20] and [21]. Scientific literature databases (MEDLINE, EMBASE, ScienceDirect, Cochrane Libraries, and Web of Science) were systematically searched in April 2016 using several keywords, including: “prostate” or “prostate cancer” or “prostate neoplasm” or “prostate malignancy”; “positron emission tomography” or “PET”; and “prostate specific membrane antigen” or “PSMA”. The search and article selection were performed by three independent evaluators (M.P., D.W., and D.C.) and any discrepancies were resolved. After screening based on the study title and abstract, the remaining articles were assessed based on the full text and were excluded with reasons when appropriate. Case reports, conference proceedings, editorial comments, and letters to the editor were excluded, as study quality could not be assessed.
Studies evaluating the utility of 68Ga-PSMA PET in detection of metastatic disease in advanced prostate cancer were included for analysis. Study designs considered for inclusion were clinical trials, prospective studies, and retrospective cohorts or comparative series. Studies assessing the diagnostic utility of 68Ga-PSMA PET in prostate cancer staging (before definitive treatment) or staging for recurrent disease (following therapy) were included for assessment. Studies were excluded if 68Ga-PSMA PET was used in assessing primary (prostatic) disease only or a specific visceral metastatic deposit (eg, pulmonary or cerebral metastases). In the current analysis, only studies using the 68Ga-PSMA-11 PSMA surface antigen HBED-CC (PSMA-11) were included for analysis provided the tracer was bound to 68Ga. Studies were excluded if alternative PSMA-bound radiotracers were used (eg, 18F or 99mTc tracers). No language or sample-size restrictions were used. Where duplicate study populations or analyses of repeated data were identified from the literature review, the publication reporting a larger sample size was used for analysis.
The primary outcome was to identify predictors of 68Ga-PSMA PET positivity. Studies that included only 68Ga-PSMA PET–positive studies were excluded, as predictive data were not available for analysis. The secondary outcome measure was the sensitivity and specificity of 68Ga-PSMA PET–positive lesions in advanced prostate cancer. For sensitivity and specificity analysis, only studies that included routine 68Ga-PSMA PET before preplanned lymph node dissection were included. Inclusion of series for which nodal biopsy was performed at the clinician's discretion does not provide meaningful false-negative data and thus does not provide accurate specificity values. An inadequate number of studies robustly assessed prostatic bed recurrence or suspicious bony metastatic disease in the manner outlined above. Therefore, we only assessed sensitivity and specificity values for 68Ga-PSMA PET in lymph nodes.
Studies were assessed for quality using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool [22]. The QUADAS-2 tool primarily assesses four domains: risk of bias in patient selection, index test, reference standard, and the timing of reference test. The degree of applicability of the patient selection, index test, and reference standard, is also assessed. Each paper was scored independently by two evaluators (M.P. and D.C.) and any discrepancies were resolved.
The following information was extracted from each study: sample size, age, indication for PET (primary staging or recurrent disease staging), PSA, previous therapies, initial cancer stage, 68Ga-PSMA PET characteristics, rates of positive PET, and histopathologic correlation data. Rates of 68Ga-PSMA PET positivity were collected and stratified by pre-PET PSA and PSA doubling time (PSAdt) when possible. When histopathologic correlation data were available, numbers of true positives, false positives, true negatives, and false negatives were collected, as appropriate. For studies that included 68Ga-PSMA PET for both primary staging and recurrent cancer staging, the extracted data are displayed separately when available.
Extracted data were collated in Excel 2007 (Microsoft Corporation, Redmond, CA, USA) and analysis was performed using Stata v.12.0SE (College Station, TX, USA). Meta-analysis of proportions was performed using the metaprop command in Stata [23]. A random-effects model was applied using the method of DerSimonian and Laird. Proportions were transformed with the Freeman-Tukey double inverse sine transformation, and confidence intervals (CIs) were calculated with the Score method. This form of transformation is preferred for meta-analysis as it stabilises the variance of the estimates, and studies with zero or one effect size can be included [23] and [24]. Heterogeneity within and between subgroups was assessed with the I2 statistic [25]. Small study effects were assessed with Egger's test and funnel plots were produced for the subgroup meta-analyses (Supplementary material). Significance was set at the 0.05 level. For study samples divided into subpopulations, we used the within-group sample size to calculate the variance used for that subpopulation. We consider this appropriate, as the subgroups are likely to define different clinical cohorts.
In patients undergoing a scan for disease recurrence, we performed a PSA level analysis. Note that for the cohort used by Afshar-Oromieh et al [14], not all patients (91.5%) had disease recurrence; however, given that they were in the overwhelming majority, we included this study in the analysis. Where a range was reported by a study as a PSA category, we took the midpoint of this range as the reference point for the category except for Sachpekidis et al [42], Demirkol et al [31], and Kabasakal et al [37], who reported individual patient-level data and manually calculated the reference point. Note that in the study by Kabasakal et al, only two out of 13 secondary staging patients had PSA <2 ng/ml, so the whole cohort was categorised in the ≥2 ng/ml PSA subpopulation. These points were used to create four PSA subgroups: 0–0.19, 0.20–0.99, 1.00–1.99, and ≥2 ng/ml; hence, categories described by a study were combined in some cases for our analysis. Similar adjustments were made to create two PSAdt subgroups: <6 mo and ≥6 mo (Supplementary material).
Random-effects meta-regression with Freeman-Tukey transformation was performed for study-defined PSA categories where the reference point was <2 ng/ml. We could not assign a valid reference point for categories greater than this as the reporting of PSA means, medians, and ranges was heterogeneous. Values were back-transformed using the equation described by Miller [26].
Summary sensitivity, specificity, and hierarchical receiver operating characteristic curves were created using the midas command [27]. In brief, this program fits a two-level mixed-effects binary regression model to the data, with separate distributions within and between studies.
Using the systematic search strategy outlined in the Supplementary material, 1895 articles were identified, of which 189 were duplicate records and excluded. Of the remaining 1706 records, 1470 were not relevant to the research question. A further 202 were conference abstracts, reviews, letters, and editorials that could not be quality assessed, and thus were excluded. From the remaining 34 articles, 11 were excluded as they did not contain relevant results and five contained duplicate data. One additional study was excluded as patients were enrolled following negative choline-based PET, and thus provided a skewed patient population [28]. This left 18 articles were suitable for assessment [14], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [40], [41], [42], [43], [44], [45], and [46]; a summary of the search strategy is shown in Figure 1. On evaluation of predictive values for positive 68Ga-PSMA PET, two studies included only patients with positive 68Ga-PSMA PET findings and predictive data were not available for analysis [34] and [35]. On evaluation of sensitivity and specificity, 11 articles reported histopathologic correlation, but only five of these were suitable for analysis according to our inclusion criteria [29], [36], [38], [41], and [43].
Of the 16 studies relevant to the meta-analysis, one was prospective and 15 were retrospective in nature. Two studies included outcomes for 68Ga-PSMA PET performed before definitive therapy (primary staging), eight reported outcomes for biochemical recurrence or disease progression staging (secondary staging), and six included mixed groups. Basic details for the studies included are listed in Table 1.
Table 1
Characteristics of the studies included
| Author | Study type (year) | Location | n | Uptake | CT | Inclusion criteria | Median | Median PSA, | Patient characteristics |
|---|---|---|---|---|---|---|---|---|---|
| time | technique | age, yr | ng/ml | ||||||
| (min)a | (range) | (range) | |||||||
| Primary staging | |||||||||
| Badaus [29] | Retrospective patient cohort (2016) | Hamburg, Germany | 30 | NR | NR | Confirmed PCa, high risk, before RP | 62 (44–75) | 8.8 (1.4–376) | High risk of LNM (>20%); subsequent RP and LND |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 8 | 45 and 90 | NR | Confirmed PCa, high risk, for staging | 68 (54–72) | 15.0 (0.3–20) | No formal risk classification; subsequent CTx or local definitive therapy |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 20 | 60 | CE CT | Confirmed PCa, high risk, before RP + LND | 71c (59–80) | 56c (3.3–363) | Intermediate risk 4/20, high risk 16/20; subsequent RP and LND |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 15 | 45–60 | HR NCE | Confirmed prostate cancer, high risk | 64 (50–77) | 19.4 (5.1–70.5) | RP 3/28; RTx: 5/28, ARTx: 2/28, ADT: 12/28 |
| Maurer [38] | Retrospective analysis (2015) | Munich, Germany | 130 | 36–165 | CE (35/130)b | Confirmed PCa, high risk, before RP | 66 (45–84) | 11.6 (0.57–244) | Intermediate risk 42/130, high risk 88/130 (D’Amico); subsequent RP and LND |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 12 | 60 and 180 | Low-dose | Confirmed PCa, high risk, for staging | 70 (64–77) | 17.0 (6–90) | High-risk PCa, no formal classification; subsequent RP and LND or ADT |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 15 | 50–70 | Conventional | Confirmed PCa, high risk, for staging | 69 (54–83) | 7.0 (0.28–45) | Intermediate and high risk (D’Amico); subsequent therapy NR |
| Secondary and mixed staging | |||||||||
| Afshar-Oromieh [14] | Retrospective patient cohort (2015) | Heidelberg, Germany | 319 | 45–70 | NCE | Mixed: 292 with BCR, 27 primary staging | 68 (46–86) | 4.6 (0.01–41395) | RP 226/292; RTx and ARTx 177/292; ADT 86/292 |
| Ceci [30] | Retrospective patient cohort (2015) | Innsbruck, Austria | 70 | 60 | CE | BCR after definitive primary treatment | 67 (38–91) | 1.7 (0.2–32.2) | RP 62/70; RTx 8/70; ARTx 13/70; ADT 20/70 |
| Demirkol [31] | Retrospective patient cohort (2015) | Istanbul, Turkey | 14 | 45 and 90 | NR | Progressive disease or BCR after treatment | 67 (43–86) | 2.5 (0.2–191.5) | RP 9/14; RTx 2/14; ARTx 6/14; ADT 9/14 |
| Dietlein [32] | Retrospective comparative (2015) | Cologne, Germany | 14 | 60 | NCE | BCR after definitive primary treatment | 68 (51–86) | 2.0 (0.17–50) | NR |
| Eiber [33] | Retrospective patient cohort (2015) | Munich, Germany | 248 | 41–74 | CE | BCR after RP | 70 (46–85) | 2.0 (0.2–59.4) | RP 241/248; SRP 7/248; RTx 7/248; ARTx 0/248; ADT 70/248 |
| Herlemann [36] | Retrospective analysis (2016) | Munich, Germany | 14 | 60 | CE | Secondary staging before secondary LND | 63c (50–76) | 5.3c (0.3–18.8) | RP 14/14; RTx 0; ARTx 6/14; ADT 4/14 |
| Kabasakal [37] | Retrospective patient cohort (2015) | Istanbul, Turkey | 13 | 45–60 | HR NCE | BCR after definitive primary treatment | 70 (58–85) | 10.6(0.28–120) | RP 3/28; RTx 5/28; ARTx 2/28; ADT 12/28 |
| Morigi [40] | Prospective trial (2015) | Sydney, Australia | 38 | 45 | NCE | BCR after definitive primary treatment | 68c (54–81) | 1.7c (2.8–20.2) | RP 34/38; RTx 4/38; ARTx 12/38; ADT NR |
| Pfister [41] | Retrospective comparative (2016) | Aachen, Germany | 28 | 45 | CE | Disease progression before salvage LND | 67 (46–79) | 2.4 (0.04–8) | RP 23/28; RTx 3/28; HIFU 2/28; ARTx 16/28; ADT 12/28 |
| Sachpekidis [42] | Retrospective patient cohort (2016) | Heidelberg, Germany | 31 | 80–90 (static CT) | Low-dose NCE, static and dynamic | BCR after definitive primary treatment | 71 (54–77) | 2.0 (0.1–130) | RP 29/31; HIFU 1/31; RTx 1/31; ARTx 11/31; ADT 1/31 |
| Sahlmann [43] | Retrospective patient cohort (2016) | Goettingen, Germany | 23 | 60 and 180 | Low-dose | BCR after definitive primary treatment | 73 (49–78) | 2.4 (0.13–30) | RP, RTx, and ARTx NR; ADT 4/23 |
| Sterzing [44] | Retrospective patient cohort (2016) | Heidelberg, Germany | 42 | 60 min | Low-dose | BCR after RP | 70 (53–83) | 2.8 (0.16–113) | RP 42/42; RTx and ADT NR |
| van Leeuwen [45] | Prospective trial (2016) | Sydney, Australia | 70 | 45 | NCE | BCR after RP, considered for salvage RTx | 62 (57–67) | 0.2 (0.12–0.32) | RP 70/70; RTx 0/70; ARTx 0/70; ADT NR |
| Verburg [46] | Retrospective patient cohort (2016) | Aachen, Germany | 155 | 60 | CE | Not specified | 70 (43–86) | 4.0 (0–2000) | RP 99/155; RTx 45/155; ARTx 57/155; ADT 76/155 |
a The tracer used in all cases was 68Ga-PSMA-11.
b PET was used for 95 patients and magnetic resonance imaging for all 130 patients.
c Mean.
ADT = androgen deprivation therapy; ARTx = adjuvant radiotherapy; BCR = biochemical recurrence; CE = contrast-enhanced; CT = computerised tomography; CTx = chemotherapy; HIFU = high-intensity focused ultrasound; HR = high resolution;, LND = lymph node dissection, LNM = lymph node metastases; NCE = non–contrast-enhanced; NR = not reported; PCa = Prostate cancer; PET = positron emission tomography; PSMA = prostate-specific membrane antigen; RP = radical prostatectomy; RTx = primary radiotherapy; SRP = salvage prostatectomy.
While patient selection was generally acceptable in the studies included, a few studies did not clearly report the inclusion criteria. Furthermore, several studies included mixed patient populations including primary and secondary staging groups, raising concerns regarding applicability. All the studies clearly reported methodology for the index test and were thus not considered a significant source of potential bias. Of the studies included, 11 involved a reference test: histopathologic correlation of 68Ga-PSMA PET–positive lesions. However, in terms of flow and timing, multiple studies were at risk of bias, as many included targeted biopsies of suspicious lesions only. Such articles were excluded from sensitivity and specificity analyses because the false-negative data are not accurate. In total, five studies performed 68Ga-PSMA PET before planned lymph node sampling. Summary findings for the QUADAS-2 appraisal are illustrated in Figure 2.
In total, we analysed 16 studies involving 1309 patients who underwent a 68Ga-PSMA PET scan, of which 926 (70.7%) were positive. In an overall meta-analysis by cohort type, 40% (95% CI 19–64%) of scans were positive for patients undergoing primary staging and 76% (95% CI 66–85%) for those undergoing secondary staging (Fig. 3). There was high heterogeneity between groups and within all subgroups (I2 > 70%). Egger's test for small-study effects did not reach significance (p > 0.10; Supplementary Figs. 1 and 2).
Fig. 3
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by patient cohort type. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
In assessing disease recurrence, PSMA PET positivity increased with the PSA category (Fig. 4). For patients with PSA <0.2 ng/ml, the pooled estimate was 42%, which increased to 58%, 76%, and 95% for the 0.2–0.99, 1.00–1.99, and >2.00 ng/ml PSA subgroups, respectively. There was high heterogeneity within the <0.2 and 1.00–1.99 ng/ml subgroups and very high heterogeneity between subgroups (I2 = 97.4%); the test for small-study effects did not reach significance (Supplementary Fig. 3). In meta-regression analysis (Fig. 5), the predicted positivity was 48% (95% CI 38–57%) for PSA of 0.2 ng/ml, 56% (95% CI 49–64%) for 0.5 ng/ml, and 70% (95% CI 63–76%) for 1.0 ng/ml.
Fig. 4
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by PSA category. ES = effect size; CI = confidence interval. Reference numbers for the studies are as in Table 1.
Fig. 5
Scatterplot of prostate specific antigen (PSA) level versus the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity. The blue line is the meta-regression prediction and shading shows the 95% confidence interval. The size of the circles is related to the inverse of the variance.
A similar finding was observed for PSAdt: the pooled PSMA positivity was 64% for PSAdt ≥6 mo and 92% for PSAdt <6 mo (Fig. 6). There was high heterogeneity between the subgroups (I2 = 84.2%) and evidence of a small-study effect in the long PSAdt subgroup (Supplementary Fig. 4).
Fig. 6
Forest plot of the proportion of 68Ga–prostate-specific membrane antigen positron emission tomography (68Ga-PSMA PET) positivity by prostate-specific antigen doubling time (PSA-dt) category. ES = effect size; CI = confidence interval.
As a sensitivity analysis, we excluded subpopulations with a sample size of less than ten and repeated the meta-analyses. This decreased the point estimate for 68Ga-PSMA PET positivity in the primary staging cohort to 27% (95% CI 15–42%). All other effects were negligible, with differences of ≤2% in the point estimates and CI boundaries for overall positivity and two PSA category proportions.
Of the studies included, 11 reported histopathologic correlation with 68Ga-PSMA PET; however, only five met the aforementioned inclusion criteria (summarised in Table 2). Of the five studies reporting the predictive ability of 68Ga-PSMA PET imaging with respect to histology-proven disease, per-patient and per-lesion analyses were possible for four studies each (Fig. 7). For per-lesion analysis, the summary sensitivity is 80% and specificity is 97%. For per-patient analysis, the summary sensitivity and specificity are both 86%, although the confidence intervals are especially wide because of the low patient numbers.
Table 2
Studies with histopathologic correlation data for 68Ga PSMA PET–positive lesions included in pooled analysis
| Study | Study type | Staging | HP type | Patients | Per patient | Lesions | Per lesion | ||
|---|---|---|---|---|---|---|---|---|---|
| setting | with HP | SS | SP | with HP | SS | SP | |||
| (n) | (%) | (%) | (n) | (%) | (%) | ||||
| [29] | Retrospective cohort | Primary | Extended primary LND | 30 | 33 | 100 | 608 | 64 | 93 |
| [36] | Retrospective analysis | Mixed | Template primary and secondary LND | 34 | 91 | 67 | 71 | – a | – a |
| [38] | Retrospective analysis | Primary | Template primary LND | 130 | 66 | 99 | 734 | 74 | 99 |
| [41] | Retrospective comparison | Recurrence | Template secondary LND | 28 | 100 | 308 | 87 | 93 | |
| [43] | Retrospective cohort | Mixed | Template primary and secondary LND | 17 | – a | – a | 213 | 94 | 99 |
a Not included in the pooled analysis as data points did not meet the inclusion criteria.
HP = histopathology; LND = lymph node dissection; SS = sensitivity; SP = specificity.
68Ga-PSMA PET is a novel targeted imaging modality that may improve the identification of metastatic prostate cancer. Our meta-analysis results identified indication, PSA, and PSA-based kinetics as risk factors for positive imaging. Furthermore, pooled sensitivity and specificity for the available data appear promising compared to alternative imaging modalities for metastatic prostate cancer.
The current study highlights the growing body of evidence supporting the use of 68Ga-PSMA PET for primary staging. In this setting, groups have reported pooled positivity of 40%. Given that the risk of metastatic spread is unlikely in low-risk disease, a majority of the studies included patients with intermediate- and high-risk prostate cancer in accordance with the D’Amico classification [47]. In the primary setting, identification of metastatic disease has a considerable influence on treatment choices and contributes to prognostic estimations. Traditionally, primary staging of lymph nodes is performed using CT or magnetic resonance imaging (MRI), but this relies on pathologic changes in lymph node morphology and size criteria. However, up to 80% of metastasis-involved nodes are smaller than the threshold limit of 8 mm typically used in clinical practice [48]. Thus, there is an inherent need for more sensitive lymph node staging in primary staging of high-risk prostate cancer. Meta-analytical data for the traditional CT and MRI imaging approaches suggest sensitivity of 39–42% and specificity of 82% [48]. These unfavourable detection rates have prompted the introduction of PET imaging augmented with tracers that include 18F-flourdeoxyglucose, 11C-choline, 18F-choline, 18F-flourocholine, and 11C-acetate [49], [50], [51], [52], and [53]. A recent meta-analysis of choline-based tracers for PET/CT revealed sensitivity of 49.2% and specificity of 95% for detection of lymph nodes [5]. Despite improved detection rates, current guidelines do not recommend the use of PET with choline-based tracers for primary staging of lymph nodes [54]. While limited literature is available, the introduction of 68Ga-PSMA PET has resulted in promising detection profiles in the setting of primary staging [55]. Maurer et al [38] performed a retrospective review of 130 consecutive patients undergoing 68Ga-PSMA PET before primary lymphadenectomy in high-risk prostate cancer, with sensitivity of 65.9% and specificity of 98.9%. These values are superior to those for traditional imaging approaches and alternative PET tracers. Thus, in high-risk disease, addition of 68Ga-PSMA PET to traditional approaches could allow for more complete and accurate primary staging compared to current practice and potential improvement in patient care.
To date, the majority of data outlining the utility of 68Ga-PSMA PET are in the setting of secondary staging for biochemical recurrence after definitive therapy. In clinical practice, early detection and highly accurate localisation of disease recurrence are critical, as these may facilitate initiation of subsequent therapies, such as radiotherapy [54]. As with primary staging, traditional imaging approaches and choline-based PET have limited sensitivity and specificity, and there is a need for better accuracy. While there number of studies is limited, the pooled data from our review support the use of 68Ga-PSMA PET in the context of secondary staging. Furthermore, several groups have directly compared 68Ga-PSMA PET and choline-based PET in secondary staging. Bluemel et al [28] reported that 44% of patients had positive 68Ga-PSMA PET following negative 18F-choline PET for biochemical recurrence. Afshar-Oromieh et al [19] compared 18F-flouromethylcholine PET with 68Ga-PSMA PET and demonstrated that the latter yielded superior detection. Similarly, Pfister et al [41] demonstrated superior performance for 68Ga-PSMA PET compared to 18F-fluoroethylcholine PET among patients with biochemical recurrence. While only early results are available, 68Ga-PSMA PET appears to provide superior diagnostic performance compared to CT, MRI, and choline-based PET imaging.
Pooled data from the current study highlight the promising detection rates for 68Ga-PSMA PET in prostate cancer with low PSA or PSAdt. Despite advances in bone scintigraphy and CT imaging, detection of locoregional or distant recurrence with low PSA has been problematic. Therefore, most guidelines recommend imaging when patients are symptomatic or PSA levels are >10 ng/ml [4] and [54]. However, polymetastatic disease may be present at this stage and may limit subsequent treatment options. Hence, there is a critical need for better early detection and localisation of prostate cancer recurrence. The introduction of choline-based PET imaging marginally improved the detection of small lesions with low PSA or PSAdt [56] and [57]. Despite this improvement, choline-based PET/CT is not recommended for patients with recurrent cancer and PSA <1–2 ng/ml [54] and [58]. Compared to choline-based PET, evidence suggests that 68Ga-PSMA PET has better sensitivity in detecting prostate cancer recurrence. On pooled analysis, 68Ga-PSMA positivity was 42% for PSA between 0 and 0.2 ng/ml and 64% for the shorter PSAdt group. The moderate to high heterogeneity within some PSA and PSAdt subgroups should be noted. This is probably because of inherent differences in study populations. Specifically, studies included cohorts with varying proportions of initial definitive therapy, whether radiotherapy or prostatectomy. Furthermore, a majority of studies included patients on current androgen deprivation therapy, for which the precise effect on 68Ga-PSMA PET is yet to be determined. Regardless, the results of the current study illustrate the improved early detection of recurrent prostate cancer compared to traditional radiological measures.
An increasing number of centres are reporting early outcomes for 68Ga-PSMA PET, particularly in Germany and Australia, with results for many series yet to be published [59] and [60]. Despite significant advances, there is considerable scope for further research in PSMA PET. There is a need for more robust sensitivity and specificity data. From the current meta-analysis, much of the histopathologic correlation data available were not suitable for inclusion in our analysis because biopsy was performed according to clinician discretion. Selective lesion biopsy does not provide meaningful false-negative rates or specificity. Several groups have reported the use of 68Ga-PSMA PET for localisation of intraprostatic malignancies, particularly in the context of focal therapies [61]. Furthermore, while PSMA is heavily expressed in advanced prostate cancer, the precise utility of 68Ga-PSMA PET in lower-risk disease is yet to be assessed. Additional advances in 68Ga-PSMA PET imaging, including the use of PET/MRI instead of CT, may improve tumour detection rates [34] and [55]. Maurer et al [38] used PET/MRI in a considerable proportion of their cohort. While their results are in line with studies using PET/CT, the precise effect of PET/MRI in the 68Ga-PSMA setting is not clear.
There are several limitations to the current study. First, a majority of the series used for meta-analysis were derived from small, retrospective, single-institutional studies. In addition, the heterogeneous nature of the patient cohorts, treatment protocols, and study designs included represents a potential limitation of the data available for analysis. Second, of the studies used for pooled sensitivity and specificity data, the majority had a small sample size and assessed patients in the primary staging setting. Additional data in the future will undoubtedly be of benefit for such analyses.
The absence of accurate imaging for detection of small-volume metastases in advanced prostate cancer has prompted the introduction of 68Ga-PSMA PET. The results of the current study suggest that PSA and associated kinetics predict the risk of metastatic disease diagnosed by 68Ga-PSMA PET. This novel imaging modality appears to provide superior sensitivity and specificity compared to alternative techniques. These promising early results for 68Ga-PSMA PET indicate a significant need for further clinical data.
Author contributions: Nathan Lawrentschuk had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lawrentschuk, Perera, Papa.
Acquisition of data: Perera, Papa, Christidis, Wetherell.
Analysis and interpretation of data: Papa, Perera, Hofman.
Drafting of the manuscript: Perera, Papa, Hofman, Murphy, Bolton.
Critical revision of the manuscript for important intellectual content: Murphy, Bolton, Hofman, Lawrentschuk.
Statistical analysis: Papa.
Obtaining funding: None.
Administrative, technical, or material support: None.
Supervision: Bolton, Lawrentschuk, Murphy, Hofman.
Other: None.
Financial disclosures: Nathan Lawrentschuk certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (eg, employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), are the following: None.
Funding/Support and role of the sponsor: None.
Acknowledgments: Marlon Perera is supported by a Royal Australasian College of Surgeons scholarship.
Pooled data appear promising, especially in biochemical recurrence setting, suggesting favorable sensitivity and specificity profiles of 68 Ga-PSMA PET compared to choline-based PET imaging. 68 Ga-PSMA PET may help in localizing prostate cancer recurrence at lower PSA values.