Prostate-Specific Membrane Antigen Expression on PET/CT in Patients with Metastatic Castration-Resistant Prostate Cancer: A Retrospective Observational Study

Visual Abstract

Pr ostate cancer (PC) is the most commonly occurring cancer in men in Europe (1). Approximately 450,000 new cases were diagnosed in 2018, and the age-standardized mortality rate was 19.4 per 100,000 men. Localized PC may be treated with radiotherapy or surgery; however, many patients develop metastatic disease (2). Initial standard treatment for metastatic PC is androgen deprivation therapy (ADT), with or without chemotherapy (2,3). Although initially effective, patients gradually become resistant and ultimately progress to metastatic castration-resistant PC (mCRPC), an aggressive disease with a median overall survival (OS) of approximately 2.5 y (2)(3)(4). Despite multiple available therapies for mCRPC, the optimal treatment sequence or combinations are unknown (5) and there remains a high unmet need for treatments with novel mechanisms of action.
Radionuclide therapies provide targeted delivery of therapeutic radiation to metastatic PC sites and allow for selection of patients likely to benefit (2). Prostate-specific membrane antigen (PSMA) is overexpressed in most PC cells and levels correlate with disease progression, making it a favorable diagnostic and therapeutic target in mCRPC (2,6,7). [ 68 Ga]Ga-PSMA-11 positron emission tomography/computed tomography (PSMA PET/CT) can be used to select patients for [ 177 Lu]Lu-PSMA-617 therapy, a potential treatment option with demonstrated effectiveness in mCRPC (2,8,9).
Monitoring therapy responses in patients with mCRPC treated with novel hormonal therapy, taxanes, and radioligand therapy (RLT) enables clinicians to optimize treatment decisions (5). In PC, there is increasing evidence to support the superiority of PSMA PET/CT over conventional imaging methods and prostate-specific antigen (PSA) serum levels for predicting early response (7,10,11). As such, PSMA is emerging as a promising target for PC imaging (12) and might help avoid the administration of costly therapies that are ineffective or not well tolerated.
Currently, data regarding PSMA expression in patients with mCRPC are limited, and it is unclear how treatments may have an impact. The present study aimed to describe PSMA expression in patients with mCRPC and examine whether PET/CT response as compared with PSA variation is a prognostic indicator for progression-free survival (PFS) and OS.

Study Design, Setting, and Participants
This single-center, retrospective observational cohort study was conducted at the Metropolitan Nuclear Medicine Centre of the S. Orsola-Malpighi University Hospital of Bologna, Italy. The study was based on secondary analysis of patients with mCRPC enrolled in the PSMA-PROSTATA registry study (EudraCT: 2015-004589-27) between March 1, 2016, and October 31, 2020, and who underwent [ 68 Ga]Ga-PSMA-11 PET/CT between January 2016 and October 2019. Inclusion criteria were: age $ 18 y; proven diagnosis of PC; a clinical or biochemical diagnosis of CRPC; and being eligible for second-or subsequent-line therapy. Patients with a history of other tumor diagnosis (i.e., not PC) or with a life expectancy of #6 mo (as assessed by each clinician) were excluded.
The Institutional Ethics Committee approved this retrospective study. All participants included in the study were appropriately informed of the purpose of this study and provided signed written informed consent.

Data Collection and Imaging
Data were collected from medical records at baseline (time of first PET/CT) and during follow-up. Baseline patient characteristics included age, clinical characteristics (Gleason score, pathologic stage, nodal status, tumor burden), treatment history before enrollment in the PSMA-PROSTATA registry, and PSA values (if available within #3 mo before baseline). PSA kinetics were calculated using published methodology (13). During follow-up, treatment-related characteristics were collected.
Radiopharmaceutical usage, PET/CT acquisition, and image interpretation were performed as described previously (14). PET images were acquired in accordance with the Joint European Association of Nuclear Medicine and Society of Nuclear Medicine and Molecular Imaging procedure guidelines for PC imaging (15). First and second (if applicable) PET/CT parameters were collected by an experienced physician evaluating the presence of focal uptake suggestive of prostate disease localization, tumor burden, and SUV max of the most significant lesion or lesions. The maximum-intensity-projection and PET/CT fusion images in axial, coronal, and sagittal slices were assessed at the reporting stage.

Outcomes
The primary outcome was baseline PSMA expression on first PET/CT defined both as SUV max and as the presence of lesions consistent with prostate metastases.
PET/CT response was assessed as responders versus nonresponders by comparing the first with the second PET/CT, as per PSMA PET/CT consensus-based response criteria (16): responders were defined as patients with stable disease, partial response, or complete response; nonresponders were defined as patients with progressive disease. PSA variation (11) between baseline and second PET/CT was assessed as PSA decrease and PSA increase from baseline. PFS was defined as time to PSA recurrence or evidence of radiologic progression. PFS and OS were calculated starting from the date of the second PET/CT until the date of last visit, death, or end of the study period (i.e., October 31, 2020), whichever occurred first.

Statistical Analysis
Continuous data were described using median and interquartile range; minimum and maximum values (i.e., range) were also reported in some instances. Categoric data were summarized as absolute and relative frequencies. Statistical significance was considered for a P value of less than 0.05.
The overall proportion of patients with PSMA expression on first PET/CT was calculated and reported with binomial 95% CI. SUV max was compared among response and different treatmentrelated variable groups using the nonparametric Kruskal-Wallis test (.2 groups) or the Wilcoxon-Mann-Whitney test (2 groups); Benjamini and Hochberg correction was applied for multiple comparisons. The relationship between SUV max and PSA parameters (serum level, doubling time, and velocity) was evaluated using Spearman correlation.
PET/CT response was reported for patients who underwent a second PET/CT. Response groups (responders vs. nonresponders) were compared with respect to treatment-related variables, baseline SUV max , and PSA level variation from baseline to second evaluation using the Wilcoxon-Mann-Whitney test for continuous data and the x 2 test or the Fisher exact test for categoric variables, as appropriate. Concordance between PSA variation and PET/CT response was assessed with Cohen's k-coefficient.
In patients who underwent a second PET/CT, PFS and OS analyses were conducted to assess whether PSA variation and PET/CT response were significant predictors. Kaplan-Meier curves were constructed and compared using the log-rank test. A multiple Cox regression model was then estimated to assess whether PSA variation and PET/CT response remained significant after adjustment for age, number of therapy lines, and SUV max /PSA baseline value.

Data Sharing
Qualified researchers may request data from Amgen clinical studies. Complete details are available at: https://www.amgen.com/science/ clinical-trials/clinical-data-transparency-practices/clinical-trial-datasharing-request/.

Baseline Patient Disposition and Clinical Characteristics
A total of 1,012 individuals were enrolled in the PSMA-PROSTATA registry between March 2016 and October 2020. Of these, 160 men with CRPC met the study eligibility criteria and were included in the analysis (Fig. 1). The median age was 72 y (range, 67-77 y), median Gleason score was 8 (range, 7-9), most patients (n 5 120/160 [75%]) had undergone radical prostatectomy, and 10% had undergone external beam radiation therapy ( Table 1). The median time from primary radical treatment to first PET/CT was 6.1 y (range, 2.9-12.2 y). About half of the patients (49.4%; n 5 79/160) had received $1 life-prolonging therapy before enrollment. There was no association between time from radical treatment to baseline PET/CT and number of prior systemic life-prolonging therapies. The median PSA level at first PET/CT was 11.7 ng/mL (interquartile range [IQR], 2-68 ng/mL), median PSA doubling time was 5.2 mo (IQR, 2.9-10.6 mo), and median PSA velocity was 6.8 ng/nL/mo (IQR, 2.5-23.6 ng/nL/mo).

Baseline PSMA Expression
PSMA expression at first PET/CT was confirmed in 152 of 160 (95.0%) patients (95% CI, 90.4-97.8); these patients were classified as having mCRPC. The distribution of metastatic sites in patients with mCRPC is shown in Supplemental Table 1. PSMA expression at first PET/CT varied by site of relapse or metastasis ( Table 2); SUV max was significantly higher in metastasis involving bone than in relapse or metastasis of other sites (P-adjusted 5 0.023 vs. nodes; P-adjusted 5 0.003 vs. prostate bed relapse; P-adjusted 5 0.047 vs. visceral). SUV max was significantly lower in patients with prostate bed relapse than in those with node lesions (P-adjusted 5 0.023).
Of the 79 of 160 patients who received systemic therapy before their first PET/CT, 78 (98.7%) expressed PSMA; among the 81 patients who did not receive prior systemic therapy (excluding ADT), 74 (91.4%) expressed PSMA (between-group difference P 5 0.075). Baseline SUV max was significantly higher in patients who received systemic treatment before first PET/CT than in those who did not (P 5 0.009; Table 3). There was no significant difference in SUV max by type of last systemic treatment received before first PET/CT.

Correlation of Baseline PSMA Expression and PSA Parameters
SUV max at first PET/CT was significantly and positively associated with baseline serum PSA levels (Spearman r, 0.377; P , 0.001) and PSA velocity (Spearman r, 0.294, P , 0.001), but not with PSA doubling time (Spearman r, 20.071; P 5 0.373). When analyzed according to last systemic treatment received before first PET/CT, positive associations were observed for SUV max at first PET/CT and baseline serum PSA levels in subgroups who received abiraterone/ enzalutamide or no systemic treatment (P 5 0.011 and P , 0.001, respectively). There was no association in subgroups who received docetaxel/cabazitaxel or palliative/ [ 223 Ra]Ra-NaCl/PSMA-RLT (a-/b-emitter prostate-specific membrane antigenradioligand therapy).

Baseline PSMA Expression According to PET/CT Response and PSA Variation
Overall, 70 patients underwent a second PET/CT scan: 45 patients (64.3%) were nonresponders and 25 (35.7%) were responders (Supplemental Table 2). There was no significant difference in median time from first to second PET/CT scans between nonresponders and responders (8.5 Table 3). Similarly, nonresponders had numerically lower median SUV max on first PET/CT than responders in those who received the same therapy before and after first PET/CT (P 5 0.064). In patients who received different therapy before and after first PET/CT, there was no significant difference in median SUV max between nonresponders and responders (P 5 0.568). Considering the last treatment type before second PET/CT, there were no significant differences in SUV max between nonresponders and responders in any of the treatment subgroups (abiraterone/enzalutamide, docetaxel/cabazitaxel/chemotherapy or palliative/[ 223 Ra]Ra-NaCl/PSMA-RLT).
Patients with a PSA decrease between first and second PET/CT had significantly higher median SUV max on first PET/CT versus patients with a PSA increase (30. 4 Table 3). This was particularly evident in patients who received different treatment before and after first PET/CT (P 5 0.039). Differences were also observed in the subgroup who received docetaxel/cabazitaxel/ chemotherapy before second PET/CT (n 5 15, P 5 0.068).
There was a significant difference in PSA change between nonresponders and responders at second PET/CT (P , 0.001). The median change in PSA between first and second PET/CT was 146% (IQR, 15.6-463) in nonresponders and 256.9% (IQR, 24.6 to 216.6) in responders. Analysis of concordance showed a 78.6% agreement between PET/CT response and PSA variation, significantly higher than expected from random chance (Cohen's k 5 0.553, P , 0.001; Supplemental Fig. 1). However, 5 of 70 patients (7.1%) were responders according to second PET/CT but reported an increase in PSA, and 10 of 70 (14.3%) were nonresponders according to second PET/CT but reported a decrease in PSA.
PFS was significantly different between PET/CT response groups (log-rank test P 5 0.005; Fig. 2A). After adjusting for SUV max at first PET/CT, age, and number of therapy lines, PET/CT nonresponders showed an increased risk of progression compared with responders (hazard ratio [HR], 3.0 [95% CI, 1.4-6.7]; P 5 0.006). PFS was also significantly different between PSA variation groups (i.e., increase vs. decrease from baseline; log-rank test P 5 0.031; Fig. 2B). After adjusting for PSA at baseline, age, and number of therapy lines, patients with a PSA increase between first and second PET/CT had a higher risk of progression than did those with a decrease (HR, 2.1 [95% CI, 1.0-4.7]; P 5 0.059).  *SUV max . 2.1. † ADT was not considered. IQR 5 interquartile range; PET/CT 5 positron emission tomography/computed tomography; PSMA 5 prostate-specific membrane antigen; PSMA-RLT 5 a-/b-emitter prostate-specific membrane antigen-radioligand therapy; SUV max 5 maximized standardized uptake value.
When PET/CT response and PSA variation groups were combined, PET/CT nonresponders with an increase in PSA had a significantly higher risk of disease progression than did PET/CT responders with a decrease in PSA (HR, 3.4 [95% CI, 1.4-8.0]; P 5 0.006; Fig. 2C). PET/CT nonresponders with a decrease in PSA were also at higher risk of progression than were PET/CT responders with a decrease in PSA (HR, 2.8 [95% CI, 1.0-8.0]; P 5 0.050). There was no difference in progression risk in PET/CT responders with an increase in PSA compared with PET/CT responders with a decrease in PSA (HR, 1.8 [95% CI, 0.5-7.0]; P 5 0.394). PET/CT nonresponders appeared to have lower OS than responders (Fig. 3A); however, the difference between the 2 Kaplan-Meier curves was not significant (P 5 0.180). There was no difference in OS between patients with a decrease in PSA and those with an increase (P 5 0.932; Figure 3B).

DISCUSSION
The present retrospective analysis of a large, single-center registry examined PSMA expression in patients with CRPC. Monitoring therapy response is important for treatment decisions in patients with mCRPC, and previous evidence suggest that PET/CT may aid in predicting early response to therapy (7,10,11,20,21); however, data are limited. Findings in this study suggest that PSMA expression on PET/CT could be more prognostic than PSA parameters for PFS and might be a promising tool for guiding clinical decisions in patients with advanced PC.
Consistent with the literature, most patients (95%) in our analysis had PSMA expression at baseline, as determined by PET/CT imaging (6). Interestingly, PSMA expression was higher in patients with prior systemic treatment, although this was irrespective of the number and type of therapy line. Higher PSMA expression in these patients may be due to more advanced and aggressive disease. The type of prior treatment did not influence PSMA expression; of particular note, there was no difference in PSMA expression in the 18% of patients last treated with abiraterone or enzalutamide before first PET/CT compared with patients last treated with other systemic therapies. Studies have shown that enzalutamide may affect the expression of PSMA on the PC cell surface early after treatment initiation (from 14 to 25 d) (22,23). Conversely, PSMA variations in patients treated with abiraterone or enzalutamide were not observed when treated over a longer period (87-110 d) (24), suggesting that the upregulation of PSMA expression after abiraterone or enzalutamide is transient. These findings are in line with the absent effect of abiraterone or enzalutamide on PSMA expression in our study, although further research is needed given the small number of patients. Variations in PSMA expression were observed according to the site of relapse or metastasis, and the highest SUV max occurred in metastasis involving the bone. It should be noted that the study included both patients who did and patients who did not undergo radical prostatectomy, which may explain why a relatively high proportion of patients had prostate bed relapse (21%). As expected, we found that PSMA expression at first PET/CT correlated with PSA levels and velocity, but not doubling time. We also confirmed the general relationship understood to exist between high PSMA expression and advanced stage disease (6).
However, higher PSMA expression may also correlate with treatment response; in a retrospective study conducted in patients with mCRPC who had received [ 177 Lu]Lu-PSMA-617, higher PSMA expression was associated with longer OS, longer PFS, and higher PSA variation (25). In our analysis, PET/CT nonresponders had numerically lower SUV max on first PET/CT than responders. The lack of statistical significance may relate to the low number of patients who underwent a second PET/CT. There was also no significant difference in SUV max between nonresponders and responders when analyzed by last treatment received before second PET/CT. Patients with an increase in PSA levels had significantly lower SUV max at baseline than patients with a PSA decrease, with good agreement between PET/CT response and PSA variation. These findings are consistent with another retrospective study that demonstrated correlations between SUV max and PSA response in patients with mCRPC (21), suggesting that PSMA expression on PET/CT may be a predictive marker of treatment response. This could potentially enable better patient selection for therapies targeting PSMA; patients with lower expression at baseline are less likely to respond to further lines of therapy, possibly due to more aggressive and undifferentiated disease.
Previous studies have demonstrated a higher accuracy of PET/ CT in patients with CRPC compared with biochemical response and other conventional methods, supporting its utility as a reliable parameter to predict response to systemic treatment for mCRPC (7,10,11,20,21). Although 1 study reported that the performance of PET/CT was not superior to conventional imaging in differentiating progressive disease from response to treatment, this may be due to the small number of patients involved (20). Our analysis suggests that PET/CT might be more reliable than PSA for predicting response to therapy; however, our findings were not statistically significant given the small sample size.
Response at first PET/CT and PSA decrease from baseline were both significant prognostic factors for PFS. The combined analysis suggested that PET/CT response may be a more significant prognostic factor than PSA variation. In line with recommendations from the Prostate Cancer Clinical Trials Working Group 3 (26), this suggests that therapy should not be discontinued based only on PSA variation. As PSA may not always predict response to therapy, PET/CT may be a more reliable option for early prediction; however, the burden of disease or response to therapy may be underestimated if the timing of PET/CT is not optimal (27). Consensus is needed on the appropriate point at which to repeat PET/CT.
In the current analysis, we did not observe any relation between PET/CT response or PSA variation and OS in patients with mCRPC. A retrospective study in patients with mCRPC reported similar results, with no correlations observed between PET parameters and OS (21); however, the findings may be explained by the limited number of patients included in these analyses.
As with all retrospective single-centered studies, our findings may not be representative of the general population with mCRPC.
The retrospective design and consequent number of patients lost to follow-up also mean that associations between PET/CT response and OS should be interpreted with caution. Further, assessment of response to therapy was not possible in patients without a second PET/CT. A further limitation is the heterogeneity of the cohort of enrolled patients, in terms of therapy management, baseline characteristics, and enrollment. Only those with a suspicion of progression were included, and consequently, a substantial proportion of patients did not undergo PSMA PET/CT during the study period; however, the use of established imaging protocols implemented by experienced operators is a strength of the study.
Finally, in advanced PC, pre-and postdiagnosis management can vary as there are no precise guidelines on the order and duration of second-line therapies, and serious adverse effects may be experienced, particularly by elderly patients and those with comorbidities. This complicates the interpretation of data on the efficacy and usefulness of diagnostic investigations. The ability of PSMA PET/CT to detect recurrence at an earlier stage of disease suggests greater opportunities for life-prolonging treatment; however, given the often indolent clinical course of recurrent PC, the potential benefits of earlier, aggressive therapeutic intervention in patients with limited recurrence will need to be weighed carefully against the risk of associated toxicities and quality of life impairment (28). CONCLUSION Our findings suggest that PSMA expression on PET/CT may be a predictive marker of treatment response in patients with mCRPC regardless of ongoing systemic therapy at the time of PET/CT. The data also suggest that PET/CT response is a more significant prognostic factor for PFS than PSA variation; however, larger studies are warranted to confirm these findings and to further explore PSMA expression in relation to patient survival.

DISCLOSURE
This study was funded by Amgen Inc. (Thousand Oaks, California, USA), which paid the fee to publish the article's open access and played a role in the study design, data collection and analysis, decision to publish, and preparation of the manuscript. Karly S. Louie was employed by Amgen Ltd. at the time the study was conducted, holds stocks in Amgen, and is an employee of and holds stock options/shares in BioMarin Pharmaceutical Inc. Michael Groaning is an employee of and holds stock in Amgen. Stefano Fanti has consulted for AAA, Amgen, Astellas, AstraZeneca, Bayer, GE, Janssen, Novartis, Sofie, and Telix and has received research funding and travel support from AAA, Amgen, Astellas, AstraZeneca, Bayer, GE, Janssen, Novartis, Sofie, and Telix. Funding for medical writing support for this article was provided by Amgen Ltd (Uxbridge, UK). No other potential conflict of interest relevant to this article was reported.