Prognostic role of early prostate specific antigen changes after [177Lu]Lu‐PSMA radioligand therapy of metastasized prostate cancer: A meta‐analysis

Approximately 10%–20% of prostate cancers progress to metastatic and castration‐resistant forms (mCRPC). Radioligand (RLT) therapy with [177Lu]Lu‐prostate‐specific membrane antigen (PSMA) is an emerging treatment for metastasized mCRPC and its efficacy is assessed not only but also by prostate specific antigen (PSA) measurement after 12 weeks or more after treatment. Our aim was to evaluate the role of early PSA measurement after RLT in predicting overall survival (OS) of mCRPC patients.


| INTRODUCTION
Prostate cancer (PCa) is the most common malignancy in men worldwide, with a higher incidence in the elderly population 1 and in Western countries, where it remains the second most common malignancy after lung cancer. 2 It represents an economic burden on the healthcare system of about 8.5 billion euros per year in Europe and $2800/ month per patient after diagnosis in USA. 3,4 Moreover, its costs increase eight-fold with the use of newly approved therapeutics. 5 While most PCa are diagnosed at a localized stage and treated with curative intent by local therapies such as surgery or radiotherapy, approximately 10%-20% of cases progress to metastatic castration-resistant forms (mCRPC) within 3 years from the diagnosis 6 and a significant number of patients die each year. 7 Patients with mCRPC are currently treated with systemic therapy consisting of second generation hormonal therapy, chemotherapy or, in selected cases, poly-ADP ribose polymerase inhibitors (PARPi). 5,8,9 Despite significant improvements in treatment options, the 5-year survival rate for mCRPC patients is approximately 30%, 10 underscoring the need for alternative approaches. Prostate-specific membrane antigen (PSMA) is an II-type enzymatic transmembrane protein that is overexpressed in 90%-100% of prostate cancer cells 11 with heterogeneous patterns related to tumours' characteristics. In particular, reduced PSMA expression was reported in smaller prostate cancers and cancers with infiltrative growth pattern and/or WHO/ISUP Grade 2 pattern, respectively. 12,13 Notably, PSMA expression is maximally elevated in hormone-resistant and metastasized tumours, suggesting that PSMA expression is correlated with disease progression. 14 Positron emission tomography with radiolabeled PSMA have been approved and are currently used to stage newly diagnosed high-risk prostate cancer, to assess patients with biochemical recurrence and to evaluate the response to treatment of advanced cancers, respectively. [15][16][17] Notably, the subsequent introduction of several [ 177 Lu]Lu-targeted PSMA ligands such as [ 177 Lu] Lu-PSMA-617 and [ 177 Lu]Lu-PSMA-I&T, PSMA-based radioligand therapy (RLT) has gained prominence for the treatment of metastasized mCRPC 18,19 and has been approved by the Food and Drug Administration and the European Medicines Agency 20,21 after the results of the phase III VISION trial demonstrated that the Lu-PSMA-617 treatment prolongs progression free survival (PFS) and OS in mCRPCP heavily pretreated PSMA positive mCRPC. 22 Moreover, the phase II TheraP trial has show great efficacy of Lu-PSMA-617 in improving PFS compared to cabazitaxel treatment, but failed to demonstrate OS improvements. 23 Overall, [ 177 Lu]Lu-PSMA has been shown to be effective in reducing prostate specific antigen (PSA) levels in almost 70% of patients, 24 with a favourable toxicity profile and mild side effects. 5,25,26 Nevertheless, the expression of PSMA in healthy organs such as the small intestine, the central nervous system and especially the salivary and lacrimal glands could induce potential serious side effects. 27 Then, protecting the kidneys, salivary and lacrimal glands is relevant to minimize side effects and preserve the quality of life of treated patients. 27 Changes of PSA often are related to the response to [ 177 Lu]Lu-PSMA RLT. 28 Regardless of the specific [ 177 Lu]Lu-PSMA ligands applied, a PSA decline from baseline (any rate) was observed in 68%-75% of patients and a PSA decrease >50% in 34%-46% of patients after therapy, 29-31 both corresponding to a positive impact on patients' outcomes. On the other hand, an increase in PSA value by ≥25% predicts poor response to therapy and shorter OS. 32 Despite such positive results, however, both sampling intervals and interpretation criteria for PSA after [ 177 Lu]Lu-PSMA RLT are still discussed. Generally, PSA is measured 12 weeks after treatment at the earliest according to Prostate Cancer Working Group 3 (PCWG3) recommendations. 7,33 However, such recommendations are mostly derived from patients under chemotherapy with taxanes, while recent studies suggest earlier PSA assessments can suitably differentiate RLT responders from non-responders, respectively. 28,[34][35][36] Therefore, to provide robust evidence, the present systematic review and meta-analysis was performed to evaluate the role of PSA measurement after initial [ 177 Lu]Lu-PSMA cycles to early assess the response to treatment and predict patients' outcomes.

| MATERIALS AND METHODS
This work was conducted according to PRISMA guidelines for systematic reviews and meta-analyses of prognostic studies. 37

| Eligibility criteria
Peer-reviewed research articles were considered. Eligible studies were selected according to the following criteria: (1) randomized clinical trials (RCTs), prospective or retrospective studies; (2) studies including patients with mCRPC treated with [ 177 Lu]Lu-PSMA; (3) studies reporting overall survival (OS) and prostate-specific antigen (PSA); (4) studies analysing PSA decline or increase after 1-2 cycles of [177Lu]Lu-PSMA. Studies with ˂30 patients with mCRPC or for which it was impossible to extract or convert valid data, repeated publications or publications with overlapping patients, and studies focused only on patients with ECOG ≥4 were excluded.

| Search strategy
A systematic search strategy was carried out on PubMed, Web of Science and Scopus from January to December 2022 without time and language restrictions using the following keywords: ([177Lu]Lu-PSMA OR Lutetium-177 OR PSMA) AND (metastatic castration-resistant prostate cancer OR mCRPC) AND (outcome OR predictor). Title/abstract screening and duplicates removal were performed by two independent reviewers (MC and GP). Full texts of the remaining potentially relevant articles, those that met inclusion and exclusion criteria, were retrieved and screened by other two independent reviewers (MLG and LG). Any disagreement was discussed until a decision was made by consensus. The final eligibility of each study was assessed, and the reasons for exclusion were recorded. Two authors (MLG and LG) executed the definitive article selection. When there was disagreement, a third experienced reviewer (KH) was consulted to achieve a consensus.

| Data extraction
Two reviewers independently extracted data from included studies according to the CHARMS-PF checklist 38 and reported them in a datasheet. Any disagreement was resolved by consensus. Study authors were not contacted to obtain unpublished data. Collected data included: (1) study characteristics: first author, year, country, observation period; (2) sample size; (3) patients' mean (or median) age; (4) PSA assessment criteria; (5) [177Lu]Lu-PSMA cycles (total and median for patient); (6) Lu-PSMA Activity (median cumulative or median injected per cycle);

| Risk of bias-Quality assessment
The quality of prognostic studies (QUIPS) tool was used to assess the risk of bias in included studies. 39 The tool comprises six domains (study participation, study attrition, . Heterogeneity was also investigated through the Galbraith plot: specified analysis was performed to evaluate the impact of these possible sources of heterogeneity. A cumulative meta-analysis was also reported to track the accumulation of evidence per year. Finally, a sensitivity analysis (one study removed) was conducted to investigate the impact of each study on the total results of the overall prevalence. Given the a priori heterogeneity among studies, a comparison between fixed and random models was not performed. Publication bias was assessed according to Cochrane guidelines. All analyses were performed through STATA17 (StataCorp., College Station).

| Search results and studies characteristics
The search strategy identified 590 articles from databases (PubMed: 78, Web of Science: 32; Scopus: 480) ( Figure 1). After exclusion of duplicates (n = 92), articles without abstract (n = 14) and articles classified ineligible by automation tools because of out-of-topic or systematic reviews (n = 143), 341 articles were screened by title and abstract. Thirty-eight articles met the inclusion criteria and were subsequently screened by full text. Of them, 13 were excluded mainly because no PSA data were reported after early [177Lu]Lu-PSMA cycles; six studies were out-of-topic, two had sample size <30 patients and one included patients with ECOG ≥4. Eleven additional studies were retrieved by citation searching: five were excluded because they were considered partial or complete overlapping samples. Other two studies did not report data about PSA changes after early [177Lu] Lu-PSMA cycles while a third article included a sample <30 patients, respectively. Finally, nine studies from databases and three from citation searching were included in our present study 28 Confounding factors were reported in many works, but a proper evaluation of their impact was not always deemed.  49 The studies reported a total of 177Lu-PSMA cycles ranging from 164 32 to 1493, 42 with a median of 3 cycles for patients. Median follow-up period ranged from 2 months 34 to 23 months. 28 Four hundred and eighty-nine patients were alive at the end of the analysis while this data was not clarified in three studies. 28,35,47 More than 1000 patients had a baseline ECOG performance score ranging between 0 and 1 while the median Gleason's score was 8 for most studies. PSA ranged from a median of 58 ng/mL (range: 0.51-1594 ng/mL) 32 to 361 ng/mL (IQR: 80-755 ng/ mL). 50 Almost 90% of patients in the included studies presented bone metastasis; approximately 70%-80% presented lymph node metastasis; lung and liver metastasis were reported in about 10%-20% of patients, while other kind of metastases (e.g. brain metastases) where reported in ˂10% of patients. Complete description of the characteristics of included studies are reported in Table 1 Figure 4A,B). Finally, from the cumulative meta-analysis, the hazard ratio increased over years, stabilizing in 2022 (HR: 0.39, 95% CI: 0.31-0.50) ( Figure 5). The leave-one-out meta-analysis showed several studies with more significant influence when compared with other studies on the estimation of the overall effect size ( Figure 6). Publication bias are reported in Figure 7.

| PSA fluctuations
Four studies involving a total of 397 patients examined PSA fluctuations after 1-2 [177Lu]Lu-PSMA cycles. 28,32,34,35 Ahmadzadehfar et al. showed that 5 out of 10 patients who did not show a PSA response after 1 cycle, did not respond to the second and third cycles, while five showed a PSA decrease after the second and third cycles. 34 In Bulbul et al. only 3 out of 45 patients with initial PSA increase reported stable or PSA decline after last cycle. 32 Similarly, Gafita et al. showed that only three of 36 patients with an initial PSA increase reported a stable or ≥30% PSA decline. 28  For each study, the navy-blue square represents its effect size centred at the point estimate of the effect size, while the horizontal line represents the confidence interval. A larger area of the square indicates a higher weight of the study. The green diamond below represents the overall effect size, centred at the overall effect size and its width is the width of the overall confidence interval. The first row on the left shows the heterogeneity measures, the second the homogeneity test and the third the overall effect size test (null hypothesis: effect size = 0).
with initial PSA increase, presented subsequent significant PSA increases, sustained throughout the observation period, with a very low late response rate. 35

| DISCUSSION
In recent years, [177Lu]Lu-PSMA therapy has been shown to be an effective treatment that prolongs overall survival in mCRPC patients with a good safety profile. To date, several studies, mostly retrospective, have addressed the role of PSA changes after multiple [177Lu]Lu-PSMA cycles in predicting OS and PFS proving that any PSA decline has a positive impact while an increase in PSA value by ≥25% is a significant predictor of poor response to therapy and shorter OS. 29,32,51 In contrast, the prognostic role of the early PSA response is less understood. From our findings, almost 50% of mCRPC patients treated with [177Lu] Lu-PSMA experienced a PSA after 1-2 cycles, with a significantly longer OS compared to stable or increased PSA levels (HR: 0.39, 95% CI: 0.31-0.50). About 30% of patients showed a ≥50% decrease in PSA resulting in a better OS F I G U R E 3 (A) Galbraith Plot for heterogeneity (shaded region = 95% CI)-The navy blue circles represent the included studies. The precision of the studies increases as you move to the right on the x-axis, so studies that are close to the y-axis have low precision. The green reference line (y = 0) represents the ""no effect"" line, while the red line is the regression line through the origin. If there is no substantial heterogeneity, about 95% of the studies are within the 95% CI range (shaded area) (B) Forest plot without outliers, HR: 0.43 (95% CI: 0.36;0.52). For each study, the navy-blue square represents its effect size centred at the point estimate of the effect size, while the horizontal line represents the confidence interval. A larger area of the square indicates a higher weight of the study. The green diamond below represents the overall effect size, centred at the overall effect size and its width is the width of the overall confidence interval. The first row on the left shows the heterogeneity measures, the second the homogeneity test and the third the overall effect size test (null hypothesis: effect size = 0). (HR: 0.69, 95% CI: 0.57-0.83) compared to lower reduction rates. Overall, a positive linear relationship between higher PSA decline rates and OS is likely, but data on the optimal threshold for PSA decline remain sparse. Although the majority of the included studies adopted the recommended threshold of 50%, some other authors argued that lower thresholds (e.g. ≥20.87% or 30%) should be more accurate to assess the response to therapy. 28,48 Transient PSA elevations, occurring in 1% 28 -17% 52 of patients, were examined in a few included studies. Transient PSA increases, known as the "flare phenomenon" (i.e. increase of ~25% in the first 6 weeks, followed by a decrease below baseline at 12 weeks), may hamper the interpretation of PSA kinetics and should be carefully investigated. 28,35 The quality of the included studies was good, but their retrospective study design increased the risk of F I G U R E 5 (A) Cumulative Metaanalysis of the overall survival analysis for any PSA decline-The analysis shows a trend in the overall hazard ratio over time; (B) PSA decline ≥50%-The analysis shows a slight variation in hazard ratio over time.

F I G U R E 6
One-leave-out analysis-For each study, the effect size shown corresponds to a total effect size calculated without that study.  42 In contrast, patients with bone metastases previously treated with Radium-223 therapy showed a longer OS compared to patients without Ra-223 therapy, possibly due to an existing long-acting synergistic effect of both treatments in patients with bone metastases. 42 Although one of the aims of many studies was to identify and evaluate potential prognostic factors for assessing PSA response after 177Lu-PSMA, the lack of consistent inclusion/exclusion criteria between studies prevented an assessment of the impact of pre-therapeutic biomarkers (e.g. alkaline phosphatase, haemoglobin) or other relevant factors (e.g. prior therapy) on early PSA response. Finally, even when PSA responders had a significantly longer OS compared to non-responders, some PSMA nonresponders still showed a better OS than patients who did not receive PSMA therapy at all. 34 This meta-analysis has several limitations. First, the included studies were retrospective and single arm. Moreover, injected activities were heterogenous among the included studies and frequently ˂7.4 GBq currently approved. Indeed, a lot of the analysed data comes from Germany and the "early" pre-VISION days of PSMA-RLT and was collected outside of clinical studies (application was done under compassionate use mechanism). At this time very often 6 GBq were used as standard activity due to two reasons: the cycles were only 6 weeks apart and to err on the safe side less activity was given. However, since the VISION days the standard activity applied is now 7.4 GBq also throughout Germany. Overall, the use of lower activities may have reduced the response rates but is unlikely that major differences in PSA kinetic were introduced. Second, the aim of this meta-analysis was to assess the prognostic role of a positive response to [177Lu]Lu-PSMA: in-depth assessments of the toxicity profile of 177Lu-PSMA only from the 1-2 cycles of treatment and the role of subsequent cycles in determining OS were not considered. Third, we would not have been able to evaluate HR by comparing OS between patients who had a PSA decline of ≥50% with that of patients who had a decline of ˂50%.

| CONCLUSIONS
A positive PSA response after just 1-2 cycles of [177Lu] Lu-PSMA RLT is a valuable prognostic factor in mCRPC patients, regardless of the percentage of PSA decline. Early responders are likely to have a further decline in PSA and a favourable clinical response. Stable/increasing levels of PSA reduce the likelihood to have a clinically significant response. However, caution is needed, in discontinuing [177Lu]Lu-PSMA RLT in mCRPC patients with early PSA elevation before further data will be provided.

AUTHOR CONTRIBUTIONS
Luca Giovanella conceived the study. Material preparation and data collection were performed by all authors. Maria Luisa Garo performed data analysis. Luca Giovanella and Ken Herrmann drafted the first version of the manuscript. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

CONFLICT OF INTEREST
Luca Giovanella received institutional research funding and personal fees from Roche Diagnostics, outside the submitted work. Ken Herrmann reports personal fees from Bayer, personal fees and other from Sofie Biosciences, personal fees from SIRTEX, non-financial support from ABX, personal fees from Adacap, personal fees from Curium, personal fees from Endocyte, grants and personal fees from BTG, personal fees from IPSEN, personal fees from Siemens Healthineers, personal fees from GE Healthcare, personal fees from Amgen, personal fees from Novartis, personal fees from YMabs, personal fees from Aktis Oncology, personal fees from Theragnostics, personal fees from Pharma15, outside the submitted work. All other authors reported no conflicts of interest.
F I G U R E 7 Publication Bias-Funnel plot represents the scatter plot of the log odds ratios against their standard errors-No relevant asymmetry was found.