Final overall survival and safety analyses of the phase III PSMAfore trial of [¹⁷⁷Lu]Lu-PSMA-617 versus change of androgen receptor pathway inhibitor in taxane-naive patients with metastatic castration-resistant prostate cancer

Abstract

Background:

In PSMAfore, [¹⁷⁷Lu]Lu-PSMA-617 (¹⁷⁷Lu-PSMA-617) prolonged radiographic progression-free survival (rPFS) in taxane-naive patients with metastatic castration-resistant prostate cancer (mCRPC), with a favourable safety profile, versus a change in androgen receptor pathway inhibitor (ARPI). We report the final overall survival (OS) analysis and updated safety data.

Patients and methods:

PSMAfore (NCT04689828) was an open-label, international, phase III trial. Patients with prostate-specific membrane antigen (PSMA)-positive mCRPC who had experienced disease progression once on a previous ARPI and were candidates for ARPI change were randomized 1 : 1 to ¹⁷⁷Lu-PSMA-617 or ARPI change to abiraterone or enzalutamide. Crossover from ARPI change to ¹⁷⁷Lu-PSMA-617 was allowed after centrally confirmed radiographic progression. Endpoints included rPFS (primary), OS (key secondary), and safety (secondary).

Results:

Patients were randomized to ¹⁷⁷Lu-PSMA-617 or ARPI change (n = 234 each): 141/234 participants (60.3%) randomized to ARPI change crossed over (75.4% of those with centrally confirmed radiographic progression). The median OS was 24.48 months [95% confidence interval (CI) 19.55–28.94 months] with ¹⁷⁷Lu-PSMA-617 versus 23.13 months (95% CI 19.61–25.53 months) with ARPI change [hazard ratio (HR) 0.91, 95% CI 0.72–1.14, P = 0.20] based on the intention-to-treat (ITT) principle; the crossover-adjusted OS HR by inverse probability of censoring weighting modelling was 0.59 (95% CI 0.38–0.91). For ¹⁷⁷Lu-PSMA-617 versus ARPI change, exposure-adjusted incidences of grade ≥3 and serious treatment-emergent adverse events were 60.8 versus 85.1 and 32.5 versus 49.9 per 100 patient-treatment years, respectively. Dry mouth occurred in 135/227 participants (59.5%; 2/227 grade ≥3) and anaemia in 62/227 (27.3%; 14/227 grade ≥3) in the ¹⁷⁷Lu-PSMA-617 arm.

Conclusions:

OS analyses did not show a statistically significant difference between the ¹⁷⁷Lu-PSMA-617 and ARPI arms based on the ITT principle; results were likely confounded by the high rate of crossover. The safety profile of ¹⁷⁷Lu-PSMA-617 was favourable with no new safety signals identified.

INTRODUCTION

Patients with metastatic castration-resistant prostate cancer (mCRPC) who experience disease progression on a prior androgen receptor pathway inhibitor (ARPI) may receive a second-line ARPI, ^1–5 despite guidelines recommending taxane-based chemotherapy.^6–8 Second-line ARPI is not the most active treatment in this disease setting^9,10 but is a common sequence of therapy in some countries for patients with mCRPC who wish to defer or avoid the toxicities of chemotherapy or who are ineligible for other nontaxane options (e.g. poly[ADP-ribose] polymerase inhibitors^11,12). Patients with mCRPC who cannot easily receive taxanes because of age or comorbidities¹³ are candidates by default for a second ARPI.¹⁴

[¹⁷⁷Lu]Lu-PSMA-617 (¹⁷⁷Lu-PSMA-617) is a prostate-specific membrane antigen (PSMA)-targeted radioligand therapy approved for the treatment of PSMA-positive mCRPC in the post-taxane setting based on the demonstration in the VISION trial of prolonged radiographic progression-free survival (rPFS), overall survival (OS), and time to worsening in health-related quality of life (HRQoL) versus standard of care.^15–18 The randomized phase III PSMAfore trial of ¹⁷⁷Lu-PSMA-617 in taxane-naive patients previously treated with an ARPI and considered appropriate to delay taxane-based chemotherapy also met its primary endpoint of rPFS.^19,20 In the primary analysis, ¹⁷⁷Lu-PSMA-617 prolonged rPFS with a median of 9.30 months [95% confidence interval (CI) 6.77-not estimable (NE)] versus 5.55 months (95% CI 4.04–5.95 months) with ARPI change (hazard ratio [HR] 0.41, 95% CI 0.29–0.56).²⁰ At the time of third interim OS analysis with more mature data (median time from randomization to data cut-off 24.11 months), the updated median rPFS was 11.60 months (95% CI 9.30–14.19 months) with ¹⁷⁷Lu-PSMA-617 versus 5.59 months (95% CI 4.21–5.95 months) with ARPI change (HR 0.49, 95% CI 0.39–0.61).^{20 177}Lu-PSMA-617 had a favourable safety profile, with a lower incidence of grade 3–5 adverse events (AEs) compared with ARPI change, and no new safety signals.²⁰ Most endpoints favoured ¹⁷⁷Lu-PSMA-617 over ARPI change, including time to worsening in HRQoL, time to symptomatic skeletal events, objective response rates, and prostate-specific antigen response.^20,21 No difference in OS based on the intention-to-treat (ITT) principle was observed between arms at the third interim data cut-off. Of participants randomized to ARPI change, 57% (78% who had centrally confirmed radiographic progression) had crossed over to receive ¹⁷⁷Lu-PSMA-617,²⁰ but the impact of crossover on OS was not reported. Overall, the results indicated that ¹⁷⁷Lu-PSMA-617 is an effective alternative to ARPI change for second-line therapy in patients with mCRPC.^19,20

We now present the final OS analyses of PSMAfore based on the ITT principle, crossover-adjusted OS analyses, updated safety outcomes, and in-depth analyses of exposure-adjusted safety and key safety topics of interest.

METHODS

Study design

PSMAfore (clinicaltrials.gov: NCT04689828) was an open-label, international, randomized phase III trial of ¹⁷⁷Lu-PSMA-617 versus ARPI change in taxane-naive patients with PSMA-positive mCRPC.

The study was conducted at 74 sites in North America and Europe in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. The protocol was approved by independent ethical review boards at all participating sites and all participants provided written informed consent.

Patients

Eligibility criteria have been published.²⁰ Participants were adults who had PSMA-positive mCRPC and experienced disease progression once on abiraterone, enzalutamide, darolutamide, or apalutamide as their most recent anti-cancer treatment and were candidates for a change of ARPI. Eastern Cooperative Oncology Group performance status of 0–1 and adequate organ function were required. Exclusion criteria included previous treatment with chemotherapy (except six or fewer cycles of taxane therapy in the adjuvant or neoadjuvant settings ≥12 months ago), immunotherapy (except sipuleucel-T), or systemic or hemibody radiotherapy (≤6 months ago). Known candidates for alternative standard therapies based on presence of genomic alterations were ineligible. PSMA positivity required at least one PSMA-positive metastatic lesion with uptake greater than liver parenchyma by [⁶⁸Ga]Ga-PSMA-11 positron emission tomography/computed tomography and no exclusionary PSMA-negative lesions.²⁰

Interventions

Participants were randomized 1 : 1 to ¹⁷⁷Lu-PSMA-617 [7.4 GBq (200 mCi) ± 10% once every 6 weeks for six cycles] or ARPI change (abiraterone or enzalutamide at treating physician’s discretion). Randomization was stratified by setting of previous ARPI (hormone-sensitive versus castration-resistant prostate cancer) and pain symptoms (score 0–3 versus >3 on item 3 of the Brief Pain Inventory—Short Form²²). Participants randomized to ARPI change could cross over to receive ¹⁷⁷Lu-PSMA-617 if eligibility criteria were met (Supplementary Methods, available at https://doi.org/10.1016/j.annonc.2025.07.003), including radiographic disease progression confirmed by blinded independent central review according to the Prostate Cancer Working Group 3—modified Response Evaluation Criteria in Solid Tumours v1.1.²³ Computed tomography or magnetic resonance imaging and ^99mTc bone scans were carried out every 8 weeks for 24 weeks, then every 12 weeks until radiographic disease progression. Clinical chemistry and haematology laboratory parameters were assessed at screening, weekly during cycle 1, at weeks 1, 3, and 5 for cycles 2–6, then every 12 weeks until end of treatment. Follow-up visits occurred within 1 week after discontinuation of study treatment and 1 month later for safety. Long-term follow-up visits were carried out every 12 weeks until the first of death, loss to follow-up, or withdrawal of consent.

Endpoints

The primary endpoint was rPFS. The key secondary endpoint was OS, defined as time from randomization to death from any cause. Safety and tolerability were secondary endpoints. A full list of prespecified endpoints has been published.²⁰

Statistical analyses

The study was powered for the primary and key secondary endpoints (Supplementary Methods, available at https://doi.org/10.1016/j.annonc.2025.07.003).²⁰ Details of the timings of the interim OS analyses have been published²⁰; the final OS analysis was to occur after ~297 deaths, providing 80% power to detect a HR of 0.72 (one-sided α = 0.025). The prespecified OS analysis method was based on the ITT principle. Supplementary OS analyses to correct for the effect of crossover could include, but were not limited to, inverse probability of censoring weighting (IPCW)²⁴ and rank-preserving structure failure time (RPSFT)²⁵ and were selected according to the appropriateness of the data for the assumptions of the methods (Supplementary Methods, available at https://doi.org/10.1016/j.annonc.2025.07.003). All data are from the final OS analysis, except time to soft tissue progression, radiographic progression after crossover (rPFS2), and time to chemotherapy, which are from the third interim OS analysis (previously unpublished).

Time-to-event endpoint, OS, was assessed using a log-rank test. Treatment effect estimates, expressed as HRs (95% CI), were estimated using stratified Cox proportional hazards models, and medians (95% CI) were estimated using the Kaplan—Meier method.

Analysis sets are defined in the Supplementary Methods, available at https://doi.org/10.1016/j.annonc.2025.07.003.

Safety analyses

Safety analyses were conducted in the safety analysis set. The Common Terminology Criteria for Adverse Events v5.0 and Medical Dictionary for Regulatory Activities v27.1 were used for assessment and grading of AEs. To account for differences in exposure between arms, incidence rates for treatment-emergent AEs (TEAEs) were adjusted for the number of patient-treatment years (PTY), calculated as number of events per 100 PTY = (TEAE incidence [n]/PTY) × 100, where PTY is the sum of exposure duration for all participants. Safety topics of interest are groupings of AEs (by preferred term) of scientific and medical concern specific to ¹⁷⁷Lu-PSMA-617. AEs collected during long-term follow-up are outlined in Supplementary Methods, available at https://doi.org/10.1016/j.annonc.2025.07.003.

RESULTS

Participants

Of 585 patients screened, 468 were randomized between 15 June 2021 and 7 October 2022 to ¹⁷⁷Lu-PSMA-617 (n = 234) or ARPI change (n = 234) (Supplementary Figure S1, available at https://doi.org/10.1016/j.annonc.2025.07.003). Data related to the primary analysis of rPFS²⁰ and the second²⁶ and third²⁰ interim OS analyses have been published.The data cut-off for the final OS analysis was 1 January 2025. The median time from randomization to this date was 34.27 months (min-max 26.9–42.2 months). At the time of the final OS analysis, 141/234 participants (60.3%) randomized to ARPI change had crossed over to receive ¹⁷⁷Lu-PSMA-617, representing 75.4% of the 187 who had centrally confirmed radiographic progression. Of those who did not crossover, 40/93 (43.0%) did not have centrally confirmed radiographic progression, 46/93 (49.5%) had centrally confirmed radiographic progression, and 7/93 (7.5%) died. The median time from randomization to crossover was 7.66 months (min-max 1.9–36.8 months). An additional nine participants randomized to ARPI change who did not cross over as per study protocol received ¹⁷⁷Lu-PSMA-617 outside of the trial but were not accounted for in the crossover-adjusted OS analyses.

Participant baseline characteristics were generally well balanced between the randomized treatment arms and comparable to those of the crossover group (Supplementary Table S1, available at https://doi.org/10.1016/j.annonc.2025.07.003).²⁰

Efficacy

Overall survival.

At the time of the final OS analysis, 299 deaths were reported in 142/234 participants (60.7%) in the ¹⁷⁷Lu-PSMA-617 arm and 157/234 (67.1%) in the ARPI change arm; deaths were related to prostate cancer in 115/142 (81.0%) and 130/157 (82.8%), respectively. Among the participants who crossed over, 90/141 (63.8%) died: 79/90 (87.8%) deaths were related to prostate cancer. In the safety set, there were four on-treatment deaths in the ¹⁷⁷Lu-PSMA-617 arm (none considered treatment related) and 10 in the ARPI change arm (one considered treatment related: cerebrovascular accident) (Supplementary Table S2, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Median OS was 24.48 months (95% CI 19.55–28.94 months) with ¹⁷⁷Lu-PSMA-617 and 23.13 months (19.61–25.53 months) with ARPI change (HR 0.91, 95% CI 0.72–1.14, P = 0.20) based on the ITT principle (Figure 1). Crossover-adjusted OS HRs by IPCW and RPSFT are shown in Table 1. The crossover-adjusted OS HR for ¹⁷⁷Lu-PSMA-617 versus ARPI change was 0.59 (95% CI 0.38–0.91) using the full IPCW model, and across the other prespecified IPCW models constructed as part of sensitivity analyses, HRs ranged from 0.54 to 0.62 (Table 1).

Figure 1. Intention-to-treat analysis of overall survival (key secondary efficacy endpoint).

ARPI, androgen receptor pathway inhibitor; CI, confidence interval; no., number; PSMA, prostate-specific membrane antigen.

Table 1.

Crossover-adjusted overall survival by IPCW and RPSFT

	177Lu-PSMA-617 (n = 234)	ARPI change (n = 234)
IPCW-adjusted OS hazard ratio (95% CI)a
Model 1
	0.56 (0.35–0.88)
Model 2
	0.54 (0.32–0.89)
Model 3 (full model)
	0.59 (0.37–0.94)
Model 4
	0.60 (0.37–0.96)
Model 5
	0.60 (0.38–0.96)
Model 6
	0.55 (0.36–0.86)
Model 7 (full model)
	0.59 (0.38–0.91)
Model 8
	0.62 (0.42–0.94)
RPSFT-adjusted OS hazard ratio (95% CI)
	0.84 (0.55–1.28)

ARPI, androgen receptor pathway inhibitor; BICR, blinded independent central review; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; IPCW, inverse probability of censoring weighting; PSA, prostate-specific antigen; OS, overall survival; PSMA, prostate-specific membrane antigen; rPD, radiographic progressive disease; RPSFT, rank-preserving structural failure time.

^aCovariates used for weight estimation and included in weighted Cox model: Model 1 = rPD confirmed by BICR, ECOG performance status, and randomized treatment; Model 2 = rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as covariate), symptomatology (as covariate), and randomized treatment; Model 3: age, stage at diagnosis, previous ARPI medication, previous radiation, presence of liver metastases, bone involvement, PSA level, initial Gleason score, lactate dehydrogenase level, alkaline phosphatase level, haemoglobin level, lymphocyte count, neutrophil count, time since prostate cancer diagnosis, rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as covariate), symptomatology (as covariate), and randomized treatment; Model 4: stage at diagnosis, previous radiation, alkaline phosphatase level, haemoglobin level, time since prostate cancer diagnosis, rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as covariate), symptomatology (as covariate), and randomized treatment; Model 5: alkaline phosphatase level, haemoglobin level, time since prostate cancer diagnosis, rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as covariate), symptomatology (as covariate), and randomized treatment; Model 6: rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as stratification factor), symptomatology (as stratification factor), and randomized treatment; Model 7: age, stage at diagnosis, previous ARPI medication, previous radiation, presence of liver metastases, bone involvement, PSA level, initial Gleason score, lactate dehydrogenase level, alkaline phosphatase level, haemoglobin level, lymphocyte count, neutrophil count, time since prostate cancer diagnosis, rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as stratification factor), symptomatology (as stratification factor), randomized treatment; Model 8: alkaline phosphatase level, haemoglobin level, time since prostate cancer diagnosis, rPD confirmed by BICR, ECOG performance status, previous ARPI use setting (as stratification factor), symptomatology (as stratification factor), and randomized treatment.

Other efficacy endpoints.

Previously unpublished results for time to soft tissue progression, rPFS2, and time to chemotherapy are shown in Supplementary Figures S2-S4, available at https://doi.org/10.1016/j.annonc.2025.07.003, respectively.

Postdiscontinuation therapies.

After discontinuation, 130/234 participants (55.6%) in the ¹⁷⁷Lu-PSMA-617 arm and 121/234 (51.7%) in the ARPI change arm received chemotherapy: docetaxel was the most common agent (Supplementary Figure S5 and Table S3, available at https://doi.org/10.1016/j.annonc.2025.07.003). After discontinuation of crossover ¹⁷⁷Lu-PSMA-617, 79/141 participants (56.0%) received chemotherapy (Supplementary Table S3, available at https://doi.org/10.1016/j.annonc.2025.07.003). Median duration, number of cycles, and cumulative dose of postdiscontinuation docetaxel were similar between arms; the numbers for discontinuation of docetaxel owing to AEs and dose interruption or dose reduction were low and similar between arms (Supplementary Table S3, available at https://doi.org/10.1016/j.annonc.2025.07.003). Toxicity from study treatment was an uncommon reason for participants in the ¹⁷⁷Lu-PSMA-617 arm not to receive postdiscontinuation antineoplastic medication (n = 1 each for haematologic AEs and dry mouth); poor performance status was the reason in 11/234 participants (4.7%) in the ¹⁷⁷Lu-PSMA-617 arm and 20/234 (8.5%) in the ARPI change arm.

Safety

Exposure.

The median duration of exposure was 8.41 months (min-max 0.36–11.63 months) in the ¹⁷⁷Lu-PSMA-617 arm, 6.54 months (min-max 0.03–39.39 months) in the ARPI change arm, and 7.16 months (0.49–11.04 months) in the crossover group (Supplementary Table S4, available at https://doi.org/10.1016/j.annonc.2025.07.003). Median cumulative dose of ¹⁷⁷Lu-PSMA-617 was 42.39 GBq (min-max 7.0–45.4 GBq) and 36.47 GBq (6.9–46.0 GBq) in the ¹⁷⁷Lu-PSMA-617 arm and crossover group, respectively (Supplementary Table S4, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Treatment-emergent adverse events.

Safety data from the third interim OS analysis have been published.²⁰ Updated safety data are shown in Supplementary Table S5 and Supplementary Results, available at https://doi.org/10.1016/j.annonc.2025.07.003. Incidences of TEAEs among the participants exposed to randomized or crossover ¹⁷⁷Lu-PSMA-617 were similar (Supplementary Table S6, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Exposure-adjusted safety.

The exposure-adjusted incidence of any-grade TEAEs was higher with ¹⁷⁷Lu-PSMA-617 (1513.1 per 100 PTY) than ARPI change (894.7); however, grade 3 or higher TEAEs (¹⁷⁷Lu-PSMA-617, 60.8; ARPI change, 85.1) and any-grade serious TEAEs (¹⁷⁷Lu-PSMA-617, 32.5; ARPI change, 49.9) were less frequent with ¹⁷⁷Lu-PSMA-617 (Table 2). The incidence of TEAEs leading to discontinuation was similar between arms (¹⁷⁷Lu-PSMA-617, 7.7 per 100 PTY; ARPI change, 6.4); however, TEAEs leading to dosage adjustment (¹⁷⁷Lu-PSMA-617, 5.2; ARPI change, 22.1) or interruption (¹⁷⁷Lu-PSMA-617, 19.6; ARPI change, 27.8) were less frequent with ¹⁷⁷Lu-PSMA-617. The most common exposure-adjusted TEAEs in the ¹⁷⁷Lu-PSMA-617 arm were dry mouth, nausea, and asthenia (Table 2).

Table 2.

Exposure-adjusted safety during randomized treatment period (safety analysis set)

Participants with treatment-emergent adverse events, n (incidence per 100 patient-treatment years)	177Lu-PSMA-617 (n = 227)		ARPI change (n = 232)
	Any grade	Grade ≥3 Any grade	Grade ≥3 Any grade	Grade ≥3
Any	224 (1513.1)	80 (60.8)	228 (894.7)	114 (85.1)
Any treatment-related	203 (649.1)	31 (21.2)	148 (190.2)	28 (16.1)
Serious	46 (32.5)	41 (28.4)	76 (49.9)	70 (45.7)
Serious treatment-related	9 (5.9)	8 (5.2)	6 (3.2)	6 (3.2)
Leading to deatha	4 (2.5)	4 (2.5)	5 (2.7)	5 (2.7)
Leading to death, treatment-related	0	0	1 (0.5)	1 (0.5)
Leading to adjustment of doseb	8 (5.2)	3 (1.9)	36 (22.1)	7 (3.8)
Leading to interruption of treatment	29 (19.6)	13 (8.5)	45 (27.8)	20 (11.3)
Leading to discontinuation of treatment	12 (7.7)	7 (4.5)	12 (6.4)	9 (4.8)
Occurring with incidence >10 per 100 patient-years in 177Lu-PSMA-617 arm
Dry mouth	135 (164.1)	2 (1.3)	6 (3.2)	0
Nausea	74 (64.6)	0	29 (16.4)	1 (0.5)
Asthenia	73 (59.1)	2 (1.3)	67 (48.4)	8 (4.4)
Anaemia	62 (46.5)	14 (9.1)	45 (27.7)	16 (8.7)
Fatigue	54 (42.2)	1 (0.6)	59 (39.7)	4 (2.1)
Constipation	50 (37.4)	1 (0.6)	33 (19.6)	0
Decreased appetite	49 (35.9)	0	44 (25.3)	2 (1.1)
Arthralgia	46 (33.2)	0	54 (36.8)	1 (0.5)
Diarrhoea	38 (27.6)	0	23 (13.1)	1 (0.5)
COVID-19	37 (25.9)	1 (0.6)	29 (17.4)	1 (0.5)
Back pain	31 (21.2)	3 (1.9)	46 (28.6)	6 (3.2)
Vomiting	26 (17.8)	0	12 (6.5)	0
Bone pain	20 (13.4)	1 (0.6)	18 (9.9)	2 (1.1)
Dysgeusia	19 (12.9)	0	6 (3.2)	0
Headache	19 (12.7)	0	11 (6.1)	0
Oedema peripheral	19 (12.9)	0	29 (17.4)	0
Thrombocytopenia	18 (11.8)	5 (3.2)	7 (3.8)	2 (1.1)
Pain in extremity	17 (11.4)	1 (0.6)	27 (15.6)	1 (0.5)

Data are for participants who underwent randomization and received at least one dose of ARPI or administration of ¹⁷⁷Lu-PSMA-617 per their randomized treatment.

ARPI, androgen receptor pathway inhibitor; COVID-19, coronavirus disease; PSMA, prostate-specific membrane antigen.

^aIn the ¹⁷⁷Lu-PSMA-617 arm: COVID-19 pneumonia, cardiac arrest, intestinal ischaemia, and sepsis (all n = 1; all considered not related to study drug by the investigator); in the ARPI change arm: cardiac arrest, coma, dyspnoea, multiple organ dysfunction syndrome (all n = 1; all considered not related to study drug by the investigator), and cerebrovascular accident (n = 1; considered related to study drug by the investigator).

^bDose adjustment is the action taken in relation to the adverse event.

Select safety topics of interest.

A complete list of treatment-emergent safety topics of interest at the third interim OS analysis has been published²⁰; data for select topics at the final OS analysis are shown in Table 3.

Table 3.

Incidence of dry mouth, haematologic adverse events, and renal toxicity during randomized treatment period (safety analysis set)

Safety topic, n (%)	177Lu-PSMA-617 (n = 227)		ARPI change (n = 232)
	Any grade	Grade ≥3	Any grade	Grade ≥3
Dry mouth	138 (60.8)a	2 (0.9)	6 (2.6)a	0
Dry mouth	135 (59.5)	2 (0.9)	6 (2.6)	0
Mucosal dryness	5 (2.2)	0	0	0
Salivary hyposecretion	2 (0.9)	0	0	0
Dry throat	1 (0.4)	0	0	0
Lip dry	1 (0.4)	0	0	0
Haematologic adverse eventsb	89 (39.2)	30 (13.2)	55 (23.7)	20 (8.6)
Anaemia	62 (27.3)	14 (6.2)	45 (19.4)	16 (6.9)
Thrombocytopenia	18 (7.9)	5 (2.2)	7 (3.0)	2 (0.9)
Neutropenia	13 (5.7)	3 (1.3)	2 (0.9)	1 (0.4)
Lymphocyte count decreased	12 (5.3)	9 (4.0)	3 (1.3)	1 (0.4)
White blood cell count decreased	9 (4.0)	1 (0.4)	3 (1.3)	1 (0.4)
Platelet count decreased	6 (2.6)	2 (0.9)	2 (0.9)	0
Leukopenia	5 (2.2)	2 (0.9)	0	0
Lymphopenia	4 (1.8)	1 (0.4)	0	0
Neutrophil count decreased	3 (1.3)	0	3 (1.3)	1 (0.4)
Febrile neutropenia	2 (0.9)	2 (0.9)	0	0
Haemoglobin decreased	2 (0.9)	0	1 (0.4)	0
Pancytopenia	1 (0.4)	0	1 (0.4)	0
Renal toxicity	14 (6.2)	2 (0.9)	19 (8.2)	8 (3.4)
Blood creatinine increased	10 (4.4)	0	7 (3.0)	0
Acute kidney injury	4 (1.8)	2 (0.9)	10 (4.3)	6 (2.6)
Renal failure	2 (0.9)	1 (0.4)	3 (1.3)	2 (0.9)
Blood urea increased	1 (0.4)	0	0	0

Data are for participants who underwent randomization and received at least one dose of ARPI or administration of ¹⁷⁷Lu-PSMA-617 per their randomized treatment.

Safety topics of interest are groupings of treatment-emergent adverse events (by Medical Dictionary for Regulatory Events v27.1 terms).

Treatment-emergent adverse events were those occurring upon or after the start of randomized treatment and up to 30 days after the last dose of ARPI or up to 41 days after the last administration of ¹⁷⁷Lu-PSMA-617.

Adverse events were coded using Common Terminology Criteria for Adverse Events v5.0 grading and Medical Dictionary for Regulatory Events v27.1 terms.

ARPI, androgen receptor pathway inhibitor; PSMA, prostate-specific membrane antigen.

^aGrade 2 dry mouth: 18 participants (7.9%) in ¹⁷⁷Lu-PSMA-617 arm; no participants in ARPI change arm.

^bHaematologic adverse events represents the myelosuppression safety topic of interest grouping of treatment-emergent adverse events.

With ¹⁷⁷Lu-PSMA-617, the median time to onset of any-grade dry mouth event was 4.17 months (95% CI 2.83–5.49 months) and median time to resolution was 1.12 months (min-max 0.0–27.2 months) (Table 4). Dry mouth was recovered/resolved in 67/138 participants (48.6%), recovering/resolving in 11/138 (8.0%), and ongoing in 59/138 (42.8%): grade 3 dry mouth was experienced by two participants and reported as recovered/resolved or recovering/resolving (Supplementary Table S7, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Table 4.

Time to onset and resolution of dry mouth, haematologic adverse events, and renal toxicity (safety analysis set)

Safety topic of interest	177Lu-PSMA-617 (n = 227)		ARPI change (n = 232)
Dry moutha, n (%)	138 (60.8)		6 (2.6)
Time to first occurrence, months, median (95% CI)b	4.17 (2.83–5.49)		NE (NE-NE)
Hazard ratio (95% CI)b		33.55 (14.80–76.04)
Time to resolution, months, median (min-max)c,d	1.12 (0.0–27.2)		2.89 (2.9–2.9)
Haematologic adverse eventse, n (%)	89 (39.2)		55 (23.7)
Time to first occurrence, months, median (95% CI)b	11.40 (11.40-NE)		NE (25.10-NE)
Hazard ratio (95% CI)b		1.86 (1.31–2.63)
Time to resolution, months, median (min-max)c,d	1.77 (0.2–18.0)		0.92 (0.1–9.2)
Renal toxicityf, n (%)	14 (6.2)		19 (8.2)
Time to first occurrence, months, median (95% CI)b	NE (NE-NE)		NE (NE-NE)
Hazard ratio (95% CI)b		0.70 (0.35, 1.40)
Time to resolution, months, median (min-max)c,d	0.71 (0.0–16.2)		0.20 (0.1–20.7)

Data are for participants who underwent randomization and received at least one dose of ARPI or administration of ¹⁷⁷Lu-PSMA-617 per their randomized treatment.

Safety topics of interest are groupings of treatment-emergent adverse events (by Medical Dictionary for Regulatory Events v27.1 terms).

Adverse events were coded using Common Terminology Criteria for Adverse Events v5.0 grading and Medical Dictionary for Regulatory Events v27.1 terms.

Time to occurrence analysis censors those without at least one event.

ARPI, androgen receptor pathway inhibitor; CI, confidence interval; NE, not estimable; PSMA, prostate-specific membrane antigen.

^aDry mouth includes dry mouth, mucosal dryness, salivary hyposecretion, dry throat, and lip dryness.

^bEstimated using KaplaneMeier method based on all participants; participants without an event are censored.

^cCalculated as numerical median based on participants with at least one event.

^dResolution of an event means the event has an end of event date and outcome of ‘recovered/resolved’ or ‘recovered/resolved with sequelae’. Analysis carried out in participants with at least one event [for the ¹⁷⁷Lu-PSMA-617 and ARPI arm, respectively: n = 67/138 (48.6%) and 1/6 (16.7%) for dry mouth; n = 30/89 (33.7%) and 25/55 (45.5%) for haematologic adverse events; and n = 12/14 (85.7%) and 13/19 (68.4%) for renal toxicity].

^eHaematologic adverse events represents the myelosuppression safety topic of interest grouping of treatment-emergent adverse events, which includes anaemia, thrombocytopenia, neutropenia, white blood cell count decreased, lymphocyte count decreased, platelet count decreased, leukopenia, lymphopenia, neutrophil count decreased, febrile neutropenia, haemoglobin decreased, and pancytopenia.

^fRenal toxicity includes blood creatinine increased, acute kidney injury, renal failure, and blood urea increased.

The median time to onset of any-grade myelosuppression event (herein referred to as haematologic AE) was 11.40 months (95% CI 11.40 months-NE) with ¹⁷⁷Lu-PSMA-617 but was not reached (95% CI 25.10-NE) in the ARPI change arm (Table 4). Median time to resolution was 1.77 months (min-max 0.2–18.0 months) with ¹⁷⁷Lu-PSMA-617 (Table 4).

There was one event of treatment-related bone marrow failure leading to death outside the TEAE assessment window in a participant who had received five cycles of ¹⁷⁷Lu-PSMA-617.

Among the participants who received 5 or 6 administrations of ¹⁷⁷Lu-PSMA-617, the incidence of haematologic AEs was higher at 1–4 administrations [48/165 participants (29.1%)] than at 5–6 administrations [37/165 (22.4%); Figure 2A]. Overall, in participants with >20 bone metastases by bone scan at baseline, the incidence of any-grade and grade 3 or higher haematologic AEs was higher than in those with 0, 1–5, or 6–20 bone metastases at baseline. A breakdown of haematologic AEs by number of bone metastases is shown in Figure 2B. The incidence of participants requiring blood product transfusions was low in both arms (Supplementary Table S8, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Figure 2. Incidence of haematologic TEAEs by number of ¹⁷⁷Lu-PSMA-617 administrations and number of bone metastases at baseline.

(A) Proportion of participants with haematologic TEAEs occurring at 1–4 versus 5–6 administrations of ¹⁷⁷Lu-PSMA-617, among participants who received 5 or 6 administrations ¹⁷⁷Lu-PSMA-617 (n = 165). (B) Proportion of participants with haematologic TEAEs according to number of bone metastases at baseline (0^a: ¹⁷⁷Lu-PSMA-617 n = 29, ARPI change n = 30; 1–5: ¹⁷⁷Lu-PSMA-617 n = 80, ARPI change n = 84; 6–20: ¹⁷⁷Lu-PSMA-617 n = 69, ARPI change n = 84; or >20/superscan: ¹⁷⁷Lu-PSMA-617 n = 49, ARPI change n = 34). Adverse events were coded using Common Terminology Criteria for Adverse Events v5.0 grading and Medical Dictionary for Regulatory Events v27.1 terms.

^aParticipants with a nonevaluable number of lesions are counted in the ‘0 bone metastases at baseline’ subgroup.

^bIncludes haemoglobin decreased.

^cIncludes lymphocyte count decreased.

^dIncludes neutrophil count decreased (panel A and B) and febrile neutropenia (panel B only).

^eIncludes platelet count decreased.

^fIncludes white blood cell count decreased. ARPI, androgen receptor pathway inhibitor; No., number; PSMA, prostate-specific membrane antigen; TEAEs, treatment-emergent adverse events.

The median time to onset of any-grade renal toxicity event was not reached in either arm, and median time to resolution was 0.71 months (min-max 0.0–16.2 months) with ¹⁷⁷Lu-PSMA-617 (Table 4). The median estimated glomerular filtration rate during treatment remained stable among the participants in the ¹⁷⁷Lu-PSMA-617 safety set (Supplementary Figure S6, available at https://doi.org/10.1016/j.annonc.2025.07.003). During long-term follow-up, 2.2% of participants in the ¹⁷⁷Lu-PSMA-617 arm and 2.8% in the crossover group experienced a renal toxicity event (Supplementary Table S9, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Second primary malignancies were uncommon, being reported in 4/227 participants (1.8%) in the ¹⁷⁷Lu-PSMA-617 arm and 3/232 (1.3%) in the ARPI change arm; none were considered treatment related and there were no reports of myelodysplastic syndrome in either arm (Supplementary Results, available at https://doi.org/10.1016/j.annonc.2025.07.003). During long-term follow-up, 1/227 participants (0.4%) in the ¹⁷⁷Lu-PSMA-617 arm had a second primary malignancy (Supplementary Table S9, available at https://doi.org/10.1016/j.annonc.2025.07.003).

Opioid use.

The proportion of participants requiring use of opioid analgesics during randomized treatment was low (Supplementary Results, available at https://doi.org/10.1016/j.annonc.2025.07.003).

DISCUSSION

Patients with mCRPC who experience disease progression on a prior ARPI may receive a subsequent ARPI as their next line of treatment.^1–5 In PSMAfore, the primary endpoint of rPFS, a registrational endpoint,²⁷ was improved and time to worsening in HRQoL was delayed with ¹⁷⁷Lu-PSMA-617 versus ARPI change.^20,21 In this final OS analysis, there was no statistically significant difference in OS between the treatment arms based on the ITT principle. The IPCW crossover-adjusted OS analyses suggested that crossover confounded the OS analysis. The overall safety profile of ¹⁷⁷Lu-PSMA-617 at the final OS analysis of PSMAfore was favourable and similar to that reported at the third interim OS analysis²⁰ with no new safety signals identified.^{15,20,28–31}

The rate of crossover was high in PSMAfore, with 60.3% of participants who were randomized to ARPI change (75.4% of those with centrally confirmed radiographic progression) crossing over. Therefore, the majority of participants in the comparator arm also received the investigational drug, and rather early in their time on study after a median of 7.66 months. Selection bias and unmeasured variables may have further affected outcomes in the crossover group because this was a post-randomization event, and these factors likely confounded the ITT analysis of OS. After correction for crossover using the IPCW method, the OS HR for ¹⁷⁷Lu-PSMA-617 versus ARPI change was 0.59 (95% CI 0.38–0.91) in the full model (including all prespecified variables as covariates), and the HRs for all other constructed IPCW models favoured ¹⁷⁷Lu-PSMA-617 with 95% CIs that excluded 1. IPCW assumes there are no unmeasured confounders. Although this assumption cannot be verified, evaluation of measured covariates selected based on clinical relevance, along with findings from other studies in similar disease areas,³² provides support for the appropriateness of this assumption.²⁴ Other variables that could have influenced patient outcomes, such as PSMA expression, are not accounted for in the correction models.^33,34 The RPSFT method for crossover correction was also assessed; the common treatment assumption was, however, likely not met and the method fails to adjust for confounding in the context of overlapping ITT survival curves.^25,35 The IPCW analysis identified and estimated an OS benefit that otherwise would not have been detected using an unadjusted ITT or RPSFT analysis.

Another potential confounding factor, and a limitation of the study, is that patients with known genomic alterations (e.g. in the DNA damage response [DDR] pathway^36,37) that confer eligibility for alternative standard therapies were excluded. In particular, homologous repair recombination gene alterations are known to be present in up to 30% of patients with mCRPC,³⁸ but genomic testing was not required in PSMAfore so the prevalence of these gene alterations in the study population was not known. There is conflicting evidence on whether such genetic alterations affect sensitivity to ¹⁷⁷Lu-PSMA-617,^39–42 but it is possible that the benefits of ¹⁷⁷Lu-PSMA-617 in the PSMAfore population could be underestimated compared with a population of patients with a higher proportion of DDR alterations. Finally, the trial may have been underpowered to detect an OS difference between arms: the statistical power for OS was limited to 80%, which may have been further reduced by crossover. Rates of postdiscontinuation taxanes were comparable between arms, so it is unlikely that therapies received after discontinuation of the study confounded the OS analysis. Other limitations of the trial have previously been discussed, including the inability to draw conclusions on optimal sequencing of ¹⁷⁷Lu-PSMA-617 because of the study design.²⁰

There was no indication that toxicity from study ¹⁷⁷Lu-PSMA-617 prevented use of subsequent therapies, with only two instances of this being given as the reason for participants not receiving postdiscontinuation antineoplastic therapy. Over half of participants in the ¹⁷⁷Lu-PSMA-617 (55.6%) and ARPI change (51.7%) arms received postdiscontinuation chemotherapy, and 56.0% of those who crossed over received chemotherapy after crossover ¹⁷⁷Lu-PSMA-617. The exposure to postdiscontinuation docetaxel and rate of discontinuation owing to AEs was also comparable between treatment arms. Efficacy and safety outcomes following subsequent antineoplastic therapies were not recorded so it is not possible to provide conclusions on the impact of ¹⁷⁷Lu-PSMA-617 on participants’ response to these therapies.

Exposure-adjusted safety analyses did not reveal any unexpected safety signals; the incidence of toxicities that have previously been associated with ¹⁷⁷Lu-PSMA-617, including dry mouth, nausea, and haematologic AEs,^15,29,31,43 remained higher with ¹⁷⁷Lu-PSMA-617 than ARPI change. The overall exposure-adjusted incidence of dry mouth in the ¹⁷⁷Lu-PSMA-617 arm (59.5%; incidence per 100 PTY 164.1) was higher than previously observed in the VISION study (38.8%¹⁵; incidence per 100 PTY 75.1).²⁸ However, most events were grade 1 or 2, did not lead to dosage modifications, and were either resolved or resolving at the time of the final OS analysis. It is possible that, with increased familiarity with ¹⁷⁷Lu-PSMA-617-associated AEs, reporting of dry mouth was more robust in PSMAfore than in VISION. Additionally, participants in PSMAfore received a median of six cycles compared with five cycles in VISION.¹⁵ There was a higher incidence of haematologic AEs with ¹⁷⁷Lu-PSMA-617 versus ARPI change, but the rate of high-grade events and participants requiring management with blood product transfusions was low overall and comparable between arms.

Any-grade and grade 3 or higher haematologic AEs, particularly anaemia, were more common in participants with a higher burden of bone disease at baseline, a known prognostic factor in mCRPC,⁴⁴ and this pattern was observed across both treatment arms. It is conceivable that the mechanism of action of ¹⁷⁷Lu-PSMA-617 may result in high bone marrow irradiation in patients with numerous bone metastases. Using ¹⁷⁷Lu-PSMA-617 earlier in the disease course^45,46 may reduce the risk of severe haematologic toxicity compared with later use, supported by the observation that the incidence of grade 3 or high haematologic toxicity was higher in VISION²⁸ than PSMAfore.

Urinary excretion is the main route by which ¹⁷⁷Lu-PSMA-617 is cleared from the body,⁴⁷ which may increase the risk of renal toxicity.⁴⁸ Rates of renal toxicity were low and comparable between treatment arms in PSMAfore, and may have therefore been related to factors such as disease progression or underlying comorbidities rather than ¹⁷⁷Lu-PSMA-617 treatment. These results are consistent with previous reports of the renal safety of ¹⁷⁷Lu-PSMA-617 in VISION.^28,49 Renal toxicity resolved quickly in the ¹⁷⁷Lu-PSMA-617 arm (median 0.71 months), and the incidence of renal toxicity was also low during long-term follow-up, although the analyses of renal toxicity may be confounded by concomitant medications or disease progression. Long-term safety data should be considered descriptive for participants exposed to ¹⁷⁷Lu-PSMA-617 and comparisons between arms cannot be made because of differential AE reporting during the long-term follow-up period. Longer term follow-up data are needed for a fuller picture of renal sequelae related to ¹⁷⁷Lu-PSMA-617 treatment.

Radiation therapy has been associated with second primary malignancies;⁵⁰ these are not commonly reported with ¹⁷⁷Lu-PSMA-617, however, and studies reporting on longer follow-up confirm low incidences.⁵¹ There were no instances of second primary malignancies related to ¹⁷⁷Lu-PSMA-617 treatment in PSMAfore, but longer term follow-up is needed. All participants in active therapy or follow-up upon closure of the trial will be eligible for a further long-term follow-up study (NCT05803941),⁵² which will provide insights into the long-term safety outcomes of patients receiving ¹⁷⁷Lu-PSMA-617.

Conclusion

In the final OS analysis of PSMAfore, OS was not statistically significantly different between the ¹⁷⁷Lu-PSMA-617 and ARPI change arms based on the ITT principle. The IPCW crossover-adjusted analyses suggested that crossover confounded the OS analysis. An in-depth safety analysis revealed that the safety profile of ¹⁷⁷Lu-PSMA-617 was acceptable and consistent with previous reports,^20,28 with no new safety signals identified.⁵³ Taken together with the previous observations from PSMAfore that ¹⁷⁷Lu-PSMA-617 prolonged rPFS, time to prostate-specific antigen progression, and time to worsening in HRQoL and pain,^20,21 these results further support ¹⁷⁷Lu-PSMA-617 as an alternative treatment option to ARPI change for patients with PSMA-positive mCRPC who have experienced disease progression once after a previous ARPI and are considered appropriate to delay taxane-based chemotherapy.

DATA SHARING

Novartis is committed to sharing with qualified external researchers, access to patient-level data and supporting clinical documents from eligible studies. These requests are reviewed and approved by an independent review panel on the basis of scientific merit. All data provided is anonymized to respect the privacy of patients who have participated in the trial in line with applicable laws and regulations. This trial data availability is according to the criteria and process described on www.clinicalstudydatarequest.com.