Hormone therapy use and duration with postoperative radiotherapy for recurrent prostate cancer: an individual patient data meta-analysis.

# Hormone therapy use and duration with postoperative radiotherapy for recurrent prostate cancer: an individual patient data meta-analysis *The Lancet 2026* Articles Hormone therapy use and duration with postoperative radiotherapy for recurrent prostate cancer: an individual patient data meta-analysis Amar U Kishan*, Yilun Sun*, Christopher C Parker, Paul Sargos, Matthew R Sydes, Sylvie Chabaud, Meryem Brihoum, Tahmineh Romero Kalbasi, Michael L Steinberg, Luca F Valle, Kekoa Taparra, Jonathan E Shoag, Jorge A Garcia, Jason R Brown, Matthew B Rettig, Adam E Singer, Robert E Reiter, Scott Eggener, Wayne Brisbane, Soumyajit Roy, Nicholas G Zaorsky, Angela Y Jia, Ting Martin Ma, Nicholas G Nickols, Jason A Efstathiou, Osama Mohamad, James J Dignam, Wendy F Seiferheld, Alan Pollack, Howard M Sandler, Paul L Nguyen, Pascal Pommier, Daniel E Spratt Summary Background Adding hormone therapy to definitive radiotherapy in localised prostate cancer improves overall survival, Lancet 2026; 407: 1059–71 but whether it similarly improves overall survival in the context of postoperative radiotherapy (PORT) after radical Published Online prostatectomy is unclear. Herein, we report an individual patient data (IPD) meta-analysis of randomised trials aimed February 26, 2026 at quantifying the benefit of adding hormonal therapy to PORT. https://doi.org/10.1016/ S0140-6736(26)00137-6 This online publication has Methods This was an IPD meta-analysis that identified randomised, phase 3 trials of PORT with or without hormone been corrected. The corrected therapy. A systematic literature search of MEDLINE, Embase, trial registries, the Web of Science, Scopus, and relevant version first appeared at conference proceedings was done on Dec 15, 2024. IPD were available via the MARCAP consortium. The primary thelancet.com on outcome was overall survival. Meta-analyses evaluated the benefit of adding hormone therapy, short-term hormone March 12, 2026 therapy (4–6 months), or long-term hormone therapy (24 months) to PORT. Tests for interaction based on pre-PORT See Comment page 1030 prostate-specific antigen (PSA) and duration of hormone therapy were evaluated and non-linear associations between *Contributed equally pre-PORT PSA and overall survival were modelled. This study was done under the master protocol of the MARCAP Department of Radiation Consortium (PROSPERO registration CRD42019134376). Oncology (Prof A U Kishan MD, L F Valle MD, Prof M L Steinberg MD, Findings IPD were available for six randomised trials including 6057 patients with a median follow-up of 9·0 years K Taparra MD, (IQR 7·2–10·7 years). Adding hormone therapy to radiotherapy did not significantly improve overall survival (hazard Prof N G Nickols MD), ratio [HR] 0·87, 95% CI 0·76–1·01, p=0·06). There was no significant interaction between hormone therapy duration Department of Medicine Statistical Core and this effect (p =0·17), although there was a significant interaction with pre-PORT PSA greater than 0·5 ng/mL interaction (T R Kalbasi PhD), Department versus 0·5 ng/mL or less (p =0·02). For all pre-PORT PSA values, the upper bounds of the 95% CI of the HR for of Urology (Prof R E Reiter MD, interaction overall survival crossed 1·0 for patients randomly assigned to PORT with or without short-term hormone therapy Prof S Eggener MD, (n=3938). For patients randomly assigned to PORT with or without long-term hormone therapy (n=1088), the W Brisbane MD), Department of Medicine (Prof M B Rettig MD, upper bounds of the 95% CI for overall survival HR fell below 1·0 at PSA greater than 1·6 ng/mL. A E Singer MD), School of Public Health (K Taparra MD) Interpretation Our findings, we believe, provide the strongest level of evidence to date suggesting there might be no University of California, Los Angeles, Los Angeles, CA, meaningful overall survival benefit to adding hormone therapy, either short-term or long-term hormone therapy, to USA; Department of Radiation PORT for PSA 0·5 ng/mL or less, with no apparent difference in efficacy for short-term versus long-term hormone Oncology (Y Sun PhD, therapy. There is an unmet need to identify biomarkers to predict potential hormone therapy benefit. N G Zaorsky MD, S Roy MD, A Y Jia MD, Prof D E Spratt MD), Department of Urology Funding National Institutes of Health. (J E Shoag MD), Department of Medicine (Prof J A Garcia MD, Copyright © 2026 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license. J R Brown MD), University Hospitals Seidman Cancer Introduction necessitating postoperative radiotherapy (PORT).2 Given Center, Case Western Reserve University School of Medicine, Definitive radiotherapy is a standard treatment option for the benefit of hormone therapy with definitive Cleveland, OH, USA; The Royal men with localised prostate cancer. In patients with radiotherapy, several randomised trials investigated the Marsden NHS Foundation intermediate-risk and high-risk prostate cancer, the benefit of adding short-term hormone therapy or Trust, London, UK (Prof C C Parker MD); The addition of short-term hormone therapy (4–6 months) to long-term hormone therapy to PORT,3–8 but to date Institute of Cancer Research, definitive radiotherapy and the adjuvant prolongation of only one has shown an overall survival benefit.7 The lack London, UK (Prof C C Parker); short-term hormone therapy to long-term hormone of an overall survival benefit to adding hormone therapy MRC Clinical Trials Unit at UCL, therapy (18–36 months) have been shown to improve to PORT was supported by the results of a recent Institute of Clinical Trials and Methodology, University overall survival.1 aggregate study-level meta-analysis of these trials.9,10 College London, London, UK Recurrence after radical prostatectomy, the surgical However, owing to potential heterogeneity in absolute (Prof M R Sydes MSc); Institut removal of the prostate and seminal vesicles as a and relative treatment effects in this diverse population, Bergonié, Bordeaux, France definitive treatment of prostate cancer, can occur, often individual patient data are requisite to assess with more (Prof P Sargos MD); Clinical Articles Research Department, Centre Léon Bérard, Lyon, France Research in context (S Chabaud MSc, Evidence before this study therapy and is the first study, to our knowledge, that Prof P Pommier MD); Unicancer, Paris, France (M Brihoum MSc); Prostate cancer can recur after radical prostatectomy, which can quantitatively and comprehensively evaluates the benefit of Department of Radiation prompt postoperative radiotherapy (PORT) to the prostate bed hormone therapy, short-term, and long-term hormone therapy. Oncology, University of with or without pelvic lymph nodes. Although the addition of It provides high-level evidence that the addition of hormone Washington, Seattle, WA, USA hormonal therapy to definitive radiotherapy for intact prostate therapy to PORT does not significantly improve overall survival, (T M Ma MD); Department of Radiation Oncology, Greater cancer has been shown to improve overall survival, the overall with no significant interaction between the duration of Los Angeles VA, Los Angeles, survival benefit of adding hormonal therapy to PORT for hormone therapy and this effect. However, there was a CA, USA (Prof N G Nickols); prostate cancer remains unclear. There is a similar lack of clarity significant interaction effect based on pre-PORT prostate Department of Radiation with regards to the separate benefit of short-term (4–6 months) specific antigen (PSA), with a meaningful effect potentially Oncology, Massachusetts General Hospital, Harvard versus long-term (24 months) hormone therapy. We searched seen for men with pre-PORT PSAs greater than 0·5 ng/mL. An Medical School, Boston, MA, MEDLINE, Embase, trial registries, and major conference exploratory analysis of non-linear associations suggested that USA (Prof J A Efstathiou MD); proceedings to retrieve randomised trials comparing PORT the addition of short-term hormone therapy to PORT would Department of Radiation Oncology, MD Anderson Cancer versus PORT plus hormone therapy with available survival data. not have a significant effect on overall survival for any pre- Center, Houston, TX, USA We did a systematic literature search in the English language PORT PSA, while the addition of long-term hormone therapy to (O Mohamad MD); Department using MEDLINE (Jan 1, 1966, to Dec 15, 2024), Embase PORT might significantly improve overall survival for very high of American University of (Jan 1, 1982, to Dec 15, 2024), trial registries (Cochrane Central pre-PORT PSA levels. Beirut, Beirut, Lebanon (O Mohamad); Department of Register of Controlled Trials and ClinicalTrials.gov), the Web of Implications of all the available evidence Public Health Sciences, Science, Scopus, and major urology and oncology conference Our findings provide the strongest evidence to date suggesting University of Chicago, Chicago, proceedings (Jan 1, 1990, to Dec 15, 2024) using the following IL, USA (Prof J J Dignam PhD); that there might be no meaningful overall survival benefit to search terms: “randomized AND prostate” AND (“androgen NRG Oncology Statistics and adding any form of hormone therapy, whether short-term or Data Management Center, deprivation” OR “hormone therapy”) AND (“radiotherapy” long-term, to PORT for men with PSA of 0·5 ng/mL or less. Philadelphia, PA OR “radiation”) AND (“surgery” OR “prostatectomy”). Furthermore, even long-term hormone therapy would only be (Prof J J Dignam, We identified six trials with detailed overall survival and W F Seiferheld MS); Department expected to increase overall survival for men with pre-PORT metastasis-free survival data. The six trials reached differing of Radiation Oncology, Miller PSA levels that are significantly higher than those routinely School of Medicine and conclusions about the benefit of adding hormone therapy in encountered in a modern PORT setting (ie, >1·6 ng/mL). Sylvester Comprehensive general as well as the specific benefits of short-term and long- Biomarkers are needed to further guide the use of hormone Cancer Center, University of term hormone therapy. Miami, Miami, FL, USA therapy with PORT. (Prof A Pollack MD); Added value of this study Department of Radiation This meta-analysis included individual patient data from Oncology, Cedars-Sinai Medical Center, Los Angeles, CA, USA six published trials comparing PORT to PORT plus hormone (Prof H M Sandler MD); Department of Radiation Oncology, Brigham and Women’s Hospital and Dana granularity any covariable and subgroup effects on literature review. This study was reported in accordance Farber Cancer Institute, overall survival. with the PRISMA guidelines.11 Our statistical analysis Boston, MA, USA Leveraging the Meta-Analysis of Randomised trials in plan and a completed checklist is provided in the (Prof P L Nguyen MD); Cancer of the Prostate (MARCAP) Consortium, which appendix (pp 3–8).12 A systematic literature search was Department of Radiotherapy, Institut Curie, Paris, France has allowed a rigorous and detailed evaluation of the done in the English language by use of MEDLINE (Prof P Pommier) benefit of hormone therapy for definitive radiotherapy,1 (1966–2025), Embase (1982–2025), trial registries Correspondence to: we sought to do an individual patient data meta-analysis (Cochrane Central Register of Controlled Trials and Prof Amar U Kishan, Department (Post Operative Salvage or Adjuvant Radiotherapy and ClinicalTrials.gov), the Web of Science, Scopus, and of Radiation Oncology, anti-androgen therapy for men with prostate cancer; major urology and oncology conference proceedings University of California, Los Angeles, CA 90095, USA POSEIDON) of all published randomised, phase 3 trials (1990–2025) to retrieve relevant studies. Controlled aukishan@mednet.ucla.edu of PORT with or without hormone therapy. We vocabulary was leveraged for studies involving humans See Online for appendix hypothesised that the addition of hormone therapy to by use of the following terms: “randomised” postoperative radiotherapy would not improve overall AND “prostate” AND (“androgen deprivation” survival, but the benefit might be modulated on the basis OR “hormone therapy”) AND (“radiotherapy” of pre-PORT PSA levels. OR radiation”) AND (“surgery” OR “prostatectomy”). The search was done on Dec 15, 2024. We had Methods two prespecified exclusion criteria: trials that did not Search strategy and selection criteria collect distant metastasis or survival data or were single To identify all relevant randomised, phase 3 trials centre. The search was done by a single investigator investigating the addition of hormone therapy to PORT, (AUK) and verified by another (DES). Our search or the prolongation from short-term to long-term identified six trials eligible for analysis (figure 1 and hormone therapy with PORT, we did a systematic appendix pp 10–12).3–8 All six trials were available through 1060 Articles the MARCAP consortium and were included in this The secondary objectives were to assess the impact of meta-analysis.3–8 each treatment intensification method on rates of The MARCAP consortium has been described in detail metastasis-free survival (MFS), event-free survival, and previously.1 Briefly, it contains individual patient data biochemical recurrence. For all endpoints, time-to-event from randomised, clinical trials run through multiple was calculated from the date of randomisation to the date collaborative groups. The requirement for ethics approval of the first qualifying event, or censored at the date of last was waived by the University of California Los Angeles follow-up. Event-free survival was defined as the and University Hospitals Cleveland Medical Center composite time to biochemical recurrence, clinical institutional review boards. progression, initiation of salvage treatment, or death from any cause. Although biochemical recurrence Data analysis definitions varied across the included eras, they were Data extraction and preparation for each individual trial consistent within each trial; therefore, to account for this were done by the relevant governing body. The data heterogeneity, biochemical recurrence was analysed requested included patient age, pathological T category, strictly within the context of the trial-stratified models pathological Gleason score, PSA before radiotherapy, described below, with death from any cause as a surgical margin status, presence of seminal vesicle competing risk. invasion, presence of extracapsular extension, and allocated treatment. Outcomes data requested included Statistical analysis time to event information and event status for Outcomes and analyses were defined a priori in a study- biochemical, local, regional, and distant failure, death, specific statistical analysis plan. This plan was discussed and cause of death. Updated survival status and time to and finalised by the co-principal investigators and last follow-up were requested from the trialist groups. statisticians before data harmonisation and analysis. Data were shared with the coprincipal investigators This study was done under the master protocol of the (AUK and DES) and statisticians (YS and TRK) of the MARCAP Consortium (PROSPERO registration MARCAP consortium (secured at UH Seidman Cancer CRD42019134376). The full statistical analysis plan, Center, Cleveland, OH, and UCLA, Los Angeles, CA, including the prespecified hypotheses, subgroup USA). Data underwent intensive harmonisation and categorisations, and methodology, is available in the integrity checks consistent with best-practice frameworks, appendix (pp 3–4). including external consistency checks, whereby data Median follow-up time and IQR were calculated by use were cross-referenced against primary publications to of the reverse Kaplan–Meier method. Between-trial replicate baseline characteristics and primary outcome heterogeneity was assessed by use of a two-stage random- results; internal consistency and plausibility checks for effects meta-analysis based on the generic inverse logical date sequencing, variable ranges, and missing variance method. Heterogeneity was quantified by use data patterns; and assessment of randomisation balance.13 of τ² (estimated via REML), I², and Cochran’s Q, No noteworthy integrity issues were identified. alongside prediction intervals. The primary meta-analysis focused on the effect of adding hormone therapy to radiotherapy, irrespective of hormone therapy duration, versus radiotherapy alone. A test for interaction of the treatment effects and overall 288 records identified through database searching survival was done between trials that used short-term hormone therapy (ie, 4–6 months) or long-term hormone 280 after duplicates removed therapy (ie, 24 months), as well as subset meta-analyses within each treatment subgroup. A specific exploratory 280 screened analysis comparing short-term hormone therapy versus long-term hormone therapy with PORT was done, restricted to the subset of trials that directly randomly 254 excluded assigned patients to these distinct durations (RADICALS-HD).5,8 For the short-term hormone therapy 26 full-text articles assessed for eligibility meta-analysis, the two groups in NRG/RTOG 0534 that included hormone therapy were combined for comparison against the PORT alone group.6 Metrics for between-trial 20 excluded, with reasons heterogeneity were calculated and summarised in the (20 randomisation did not involve hormone therapy) appendix (p 13). Outcomes 6 studies included in meta-analysis The primary objective of this study was to establish the impact of treatment intensification on overall survival. Figure 1: Trial profile Articles The primary efficacy analysis used a one-stage meta- established from the estimated cumulative incidence analysis framework. Data from all patients were function. analysed by use of a univariable Cox proportional Subgroup analyses were done within strata defined by hazards model (for overall survival, MFS, and event-free randomisation to with or without short-term hormone survival) or a univariable Fine–Gray subdistribution therapy versus with or without long-term hormone hazard model (for biochemical recurrence) assuming a therapy, age (<60 years vs 60–64 years vs 65–69 years common treatment effect, stratified by trial. This vs >70 years), pathological Gleason score (6 vs 7 vs 8–10), stratification allows the baseline hazard function to vary pre-PORT PSA (≤0·20 ng/mL vs 0·21–0·50 ng/mL between trials, thereby absorbing heterogeneity in event vs 0·51–1·00 ng/mL vs >1·00 ng/mL), surgical margin definitions, follow-up duration, and era-specific risks; status (negative vs positive), seminal vesicle invasion treatment effects were thus estimated by comparing (yes vs no), and extracapsular extension (yes vs no) with patients strictly with concurrent controls within the formal testing for interaction effects by use of separate same trial stratum. Notably, the original RADICALS-HD multivariable models for each covariate of interest. short versus none4 and RADICALS-HD long versus Each model included the treatment, the specific short8 publications included patients who were covariate, and a treatment-by-covariate interaction randomly assigned in both the respective two-way term. For these models, PSA was analysed as both a cohorts (short vs none and long vs short continuous variable and as a categorical variable with 1:1 randomisations) as well as patients randomly the aforementioned four strata. For categorical and assigned in the RADICALS-HD 1:1:1 three-way ordinal subgroups, we used a global test of interaction randomisation.5 For the present POSEIDON meta- (simultaneous test of effects) to detect any heterogeneity analysis, the availability of individual patient data in treatment effect across strata without assuming a allowed the ability to treat the randomisations as linear trend. As a sensitivity analysis to assess potential completely distinct trials (ie, patients were not double- aggregation bias, we also did two-stage meta-analyses counted).14 For the short-term hormone therapy use of within-trial interaction effects. Interaction meta-analysis, across trials of PORT with or without coefficients and their corresponding variance– short-term hormone therapy, all patients receiving covariance matrices were estimated within each trial PORT alone were pooled together to form one group independently and subsequently pooled by use of and those receiving PORT with short-term hormone random-effects models to strictly isolate within-trial therapy were pooled together to form the comparator treatment-subgroup contrasts. group. Four trials contributed to the PORT with or On the basis of previous work suggesting that without short-term hormone therapy meta-analysis, all pre-PORT PSA is both a prognostic and a potentially of which predominantly used gonadotropin releasing predictive biomarker for the benefit of long-term hormone agonist-based hormone therapy; details on the hormone therapy in this context,17 the effect of included trials are shown in the appendix (pp 10–12). pre-PORT PSA was explored further. First, to assess for Two trials contributed to the PORT with or without any interaction between pre-PORT PSA and the effect of long-term hormone therapy meta-analysis: RADICALS hormone therapy on MFS and overall survival, three-way, which used 24 months of androgen multivariable models adjusted for age, pathologic deprivation therapy with gonadotropin releasing Gleason score, and pathological T stage were done on hormone agonist medications for most patients the basis of PSA as a continuous variable, PSA with the (425 [86%] of 492 patients) and also included a third aforementioned categorical separation (≤0·20 ng/mL randomised group evaluating 6 months of hormone vs 0·21–0·50 ng/mL vs 0·51–1·00 ng/mL therapy plus radiotherapy (with the no-hormone therapy vs >1·00 ng/mL), and PSA with a binary categorisation group serving as the common control),5 and RTOG 9601, (≤0·5 ng/mL vs >0·5 ng/mL). Second, absolute risk which used high dose non-steroidal androgen receptor reductions at 10 years for distant-metastasis-or-all-cause- antagonist monotherapy with bicalutamide.7 mortality events (as a composite endpoint) and for Overall survival and MFS were reported by use of the all-cause mortality events were calculated as the HR with 95% CIs from stratified univariable Cox differences in event risks, estimated with the regression models along with absolute differences in Kaplan–Meier method at 10 years between the treatment Kaplan–Meier event probability estimates at 10 years.15,16 groups, correspondingly. The numbers-needed-to-treat The Grambsch–Therneau test was applied to identify (NNTs) for these same endpoints for the addition of violation of the proportionality assumption of the Cox hormone therapy to radiotherapy, and separately for the regression models. For each treatment group, 10-year addition of short-term hormone therapy and long-term restricted mean survival time and 10-year restricted hormone therapy, were calculated as the inverse of the mean time lost were calculated to supplement relative absolute risk reduction at 10 years. Third, natural cubic risks. Restricted mean survival time was established splines modelling non-linear associations between from the Kaplan–Meier estimate of the survival pre-PORT PSA and overall survival and MFS were function, whereas restricted mean time lost was developed for the addition of hormone therapy to 1062 Articles radiotherapy, by use of stratified Cox models. The Role of the funding source number of knots was selected from grid search ranging The funders of the study had no role in study design, from 1 to 5 on the basis of the lowest Akaike Information data collection, data analysis, data interpretation, or Criterion. Splines were adjusted for pathological Gleason writing of the report. score, age, and pathological T stage. A previous meta- analysis had suggested a surrogate threshold effect for Results MFS on overall survival in localised prostate cancer, Overall, data were available for six randomised trials such that interventions for which the upper 95% CI of including 6057 patients with a median follow-up of the hazard ratio (HR) for MFS falls below 0·81 would be 9·0 years (IQR 7·2–10·7), of whom 5026 contributed to expected to show a significant benefit in overall survival.18 the primary comparison of PORT with versus without Missing data for baseline covariates were minimal hormone therapy. No data integrity issues were (70 548 (24·7%) 674 (24·0%) 420 (25·1%) 541 (23·9%) 128 (23·6%) 133 (24·4%) Prostate-specific 0·3 (0·2–0·6) 0·3 (0·2–0·6) 0·3 (0·2–0·5) 0·3 (0·2–0·5) 0·5 (0·3–0·9) 0·5 (0·3–0·9) antigen, ng/mL Pathological Gleason score 6 390 (17·6%) 510 (18·1%) 270 (16·1%) 384 (17·0%) 120 (22·7%) 126 (23·7%) 7 1507 (68·0%) 1882 (67·0%) 1188 (71·0%) 1562 (69·0%) 319 (60·4%) 320 (60·2%) 8–10 305 (13·8%) 404 (14·4%) 216 (12·9%) 318 (14·0%) 89 (16·9%) 86 (16·2%) Pathological tumour stage 2 962 (43·4%) 1311 (46·9%) 791 (47·4%) 1122 (49·8%) 171 (31·6%) 189 (34·7%) 3a 833 (37·6%) 970 (34·6%) 623 (37·3%) 758 (33·6%) 210 (38·7%) 212 (39·0%) 3b 366 (16·5%) 439 (15·7%) 241 (14·4%) 337 (14·9%) 125 (23·1%) 102 (18·8%) 4 3 (0·1%) 9 (0·3%) 3 (0·2%) 6 (0·3%) 0 3 (0·6%) Surgical margin status Negative 878 (39·8%) 1182 (42·2%) 726 (43·6%) 1030 (45·7%) 152 (28·0%) 152 (27·8%) Positive 1328 (60·2%) 1619 (57·8%) 938 (56·4%) 1225 (54·3%) 390 (72·0%) 394 (72·2%) Seminal vesicle invasion No 1811 (82·7%) 2325 (84·1%) 1406 (84·9%) 1888 (84·7%) 405 (75·7%) 437 (81·8%) Yes 380 (17·3%) 438 (15·9%) 250 (15·1%) 341 (15·3%) 130 (24·3%) 97 (18·2%) Extracapsular extension No 1221 (56·7%) 1685 (61·0%) 924 (57·0%) 1387 (62·4%) 297 (55·9%) 298 (55·3%) Yes 931 (43·3%) 1078 (39·0%) 697 (43·0%) 837 (37·6%) 234 (44·1%) 241 (44·7%) Data are n (%). Table: Patient characteristics Articles 75 25 Number at risk (censored) Radiotherapy Radiotherapy + hormone therapy Figure 2: Survival effect of adding hormonal therapy to post-prostatectomy radiotherapy Kaplan–Meier survival curves showing overall survival impact of adding hormonal therapy to post- prostatectomy radiotherapy In the overall cohort (A) and in patients with pre-PORT PSA ≤0·20 ng/mL (B), 0·21–0·50 ng/mL (C), 0·51–1·00 ng/mL (D), and >1·00 ng/mL (E). (F) Kaplan–Meier survival curve for metastasis-free survival. PSA=prostate-specific antigen. Six randomised comparisons across five randomised shown in figure 2. There was no clear evidence that the trials evaluated the addition of hormone therapy to addition of hormone therapy to radiotherapy improved PORT (5026 patients, median follow-up of 9·10 years overall survival (HR 0·87, 95% CI 0·76–1·01, p=0·06), [IQR 7·30–10·97]; table). Survival curves for the with 10-year overall survival rates of 83·6% addition of hormone therapy to PORT in the overall (95% CI 81·8–85·4%) with PORT alone versus 84·3% population, as well as in subgroups based on pre-PORT (95% CI 82·7–86·0) with the addition of hormone values of up to 0·20 ng/mL, 0·21–0·50 ng/mL, therapy. For patients with PSA up to 0·2 ng/mL 0·51–1·00 ng/mL, and greater than 1·00 ng/mL are (HR 1·14, 95% CI 0·83–1·57) and PSA 0·21–0·50 ng/mL 1064 )%( lavivrus llarevO A 75 Radiotherapy Radiotherapy+hormone therapy Hazard ratio: 0·87 (95% CI 0·76–1·01, p=0·06) 10-year estimates Radiotherapy: 83·6% (81·8–85·4) 25 Radiotherapy+hormone therapy: 84·3% (82·7–86·0) 0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12 2216 2170 2146 2105 2044 1965 1864 1636 1400 1055 729 453 273 (0) (34) (43) (60) (87) (125) (186)(369) (582)(898)(1194)(1440)(1605) 2810 2755 2718 2664 2604 2522 2344 2034 1671 1250 841 519 291 (4) (45) (66) (90) (117) (167) (300)(566)(884)(1261)(1638)(1934)(2152) )%( lavivrus llarevO B PSA ≤0·20 ng/mL Hazard ratio: 1·14 (95% CI 0·83–1·57, p=0·43) 10-year estimates Radiotherapy: 89·9% (87·4–92·5) Radiotherapy+hormone therapy: 87·4% (84·4–90·4) 695 677 671 658 643 624 592 514 422 310 200 100 46 (0) (16) (19) (24) (34) (47) (71) (141) (222) (330)(438) (531) (583) 826 813 810 791 774 758 705 604 471 327 191 90 26 (1) (9) (9) (18) (25) (35) (77) (172) (298) (430)(560) (655) (714) 75 25 0 1 2 3 4 5 6 7 8 9 10 11 12 Number at risk (censored) Radiotherapy Radiotherapy + hormone therapy )%( lavivrus llarevO C PSA 0·21–0·50 ng/mL 75 Hazard ratio: 0·94 (95% CI 0·74–1·19, p=0·70) 10-year estimates 25 Radiotherapy: 84·9% (82·2–87·7) Radiotherapy+hormone therapy: 84·5% (81·8–87·2) 0 1 2 3 4 5 6 7 8 9 10 11 12 876 864 856 842 819 790 751 659 566 408 285 173 96 (0) (9) (11) (18) (25) (38) (62) (135) (224) (370) (483)(584) (655) 1092 1067 1046 1029 1010 975 913 778 640 474 318 186 97 (1) (23) (36) (46) (56) (77) (125) (237) (359) (509) (651) (770) (858) )%( lavivrus llarevO D PSA 0·51–1·00 ng/mL Hazard ratio: 0·72 (95% CI 0·54–0·96, p=0·02) 10-year estimates Radiotherapy: 76·4% (71·3–81·6) Radiotherapy+hormone therapy: 80·9% (77·1–84·8) 346 338 332 327 317 298 280 248 229 186 131 92 67 (0) (5) (7) (9) (13) (21) (28) (51) (70) (104) (148) (181) (203) 527 522 515 506 490 472 432 387 325 260 189 136 94 (1) (3) (8) (9) (15) (27) (58) (93) (137) (191) (257) (307) (347) 75 25 Number at risk (censored) Radiotherapy Radiotherapy + hormone therapy )%( lavivrus llarevO E 75 PSA >1·0 ng/mL 50 Hazard ratio: 0·69 (95% CI 0·48–0·98, p=0·03) 10-year estimates Radiotherapy: 72·4% (66·1–78·7) 25 Radiotherapy+hormone therapy: 80·5% (75·2–85·7) 0 1 2 3 4 5 6 7 8 9 10 11 12 0 1 2 3 4 5 6 7 8 9 10 11 12 Time (years) 236 232 228 222 213 202 191 174 151 126 96 73 56 2216 2147 2090 2028 1952 1856 1748 1539 1303 968 663 411 239 (0) (2) (4) (5) (8) (12) (17) (25) (40) (61) (84) (101) (114) (0) (34) (43) (59) (85) (118) (177) (337) (528) (823)(1083)(1300)(1451) 290 282 277 271 264 252 232 208 187 152 120 90 63 2810 2750 2696 2614 2529 2426 2238 1927 1580 1171 778 480 261 (0) (7) (10) (11) (15) (21) (31) (50) (67) (98) (124) (150) (174) (4) (45) (66) (90) (117) (166) (292) (543)(844)(1198)(1549)(1820)(2025) )%( lavivrus eerf-sisatsateM F Hazard ratio: 0·79 (95% CI 0·70–0·89, p<0·001) 10-year estimates Radiotherapy: 74·1% (72·0–76·3) Radiotherapy+hormone therapy: 77·9% (76·0–79·8) Time (years) Articles Incidence/n % (95% CI) Years (95% CI) Hazard ratio p value Interaction (95% CI) p value Radiotherapy Radiotherapy + 10-year absolute 10-year RMST benefit hormone therapy benefit of of radiotherapy + radiotherapy + hormone therapy hormone therapy Randomisation 0·17 With or without 218/1674 270/2264 −0·3 (−3·1 to 2·6) 0·1 (−0·1 to 0·2) 0·93 (0·77 to 1·11) 0·39 short-term hormone therapy With or without 153/542 129/546 2·9 (−2·1 to 7·8) 0·2 (0·0 to 0·4) 0·79 (0·63 to 1·00) 0·049 long-term hormone therapy Age 0·80 <60 48/468 43/601 2·5 (−1·3 to 6·3) 0·1 (−0·1 to 0·3) 0·74 (0·48 to 1·13) 0·15 60−64 76/562 85/677 0·3 (−4·0 to 4·6) 0·1 (−0·1 to 0·3) 0·93 (0·68 to 1·28) 0·62 65−69 106/638 127/858 −1·2 (−5·8 to 3·4) 0·1 (−0·1 to 0·3) 0·90 (0·69 to 1·16) 0·39 70+ 141/548 144/674 1·8 (−4·3 to 8·0) 0·1 (−0·2 to 0·3) 0·88 (0·69 to 1·11) 0·32 Gleason 0·75 6 or lower 68/390 73/510 1·6 (−3·5 to 6·8) 0·1 (−0·1 to 0·3) 0·93 (0·66 to 1·29) 0·60 7 223/1507 245/1882 0·1 (−2·8 to 3·0) 0·1 (0·0 to 0·2) 0·91 (0·75 to 1·09) 0·30 8−10 76/305 80/404 1·0 (−6·8 to 8·8) 0·2 (−0·1 to 0·5) 0·76 (0·55 to 1·04) 0·09 PSA (ng/mL) 0·08 ≤0·20 69/695 87/826 −2·5 (−6·4 to 1·4) 0·0 (−0·2 to 0·1) 1·14 (0·83 to 1·57) 0·42 0·21−0·50 133/876 147/1092 −0·4 (−4·3 to 3·4) 0·1 (−0·1 to 0·2) 0·94 (0·74 to 1·19) 0·70 0·51−1 92/346 100/527 4·5 (−1·9 to 10·9)0·2 (−0·1 to 0·5) 0·72 (0·54 to 0·96) 0·02 >1 72/236 59/290 8·1 (−0·1 to 16·3)0·3 (0·0 to 0·7) 0·69 (0·48 to 0·98) 0·03 Margin 0·54 Negative 135/878 160/1182 −0·4 (−4·3 to 3·5) 0·0 (−0·2 to 0·1) 0·89 (0·71 to 1·13) 0·38 Positive 235/1328 238/1619 1·6 (−1·5 to 4·8) 0·2 (0·0 to 0·3) 0·85 (0·71 to 1·02) 0·08 Seminal vesicle invasion 0·20 No 275/1811 316/2325 −0·5 (−3·1 to 2·1) 0·0 (−0·1 to 0·1) 0·94 (0·80 to 1·10) 0·42 Yes 92/380 81/438 4·9 (−2·1 to 11·9)0·3 (0·1 to 0·6) 0·75 (0·55 to 1·02) 0·07 Extracapsular extension 0·61 No 212/1221 225/1685 1·5 (−1·8 to 4·8) 0·1 (0·0 to 0·2) 0·84 (0·70 to 1·02) 0·07 Yes 156/931 167/1078 0·5 (−3·3 to 4·3) 0·1 (0·0 to 0·3) 0·88 (0·70 to 1·10) 0·24 0·25 0·50 0·75 1 1·5 Favours radiotherapy + hormone therapy Favours radiotherapy alone Figure 3: Forest plot showing effect of adding hormonal therapy to post-prostatectomy radiotherapy on overall survival, stratified by subgroups of interest n=number at risk The with or without short-term hormone therapy subgroup includes data from GETUG-AFU 16, NRG/RTOG 0534, RADICALS-HD (two-way), and RADICALS-HD (three-way). The with or without long-term hormone therapy subgroup includes data from RTOG 9601 and RADICALS-HD (three-way). For RADICALS-HD (three-way), the control group contributes to the estimation of the HR in both subgroups to represent the distinct randomisations. The interaction p value comparing treatment effects between with or without hormone therapy and with or without-hormone therapy randomisations was calculated via a split control approach19 to avoid unit-of-analysis error; subgroup-specific estimates use the full common control group to maximise precision for clinical interpretation. PSA=prostate-specific antigen. RMST=restricted mean survival time. (HR 0·94, 0·74–1·19), HRs did not indicate an overall therapy and pre-PORT PSA. Notably, when PSA was survival benefit from hormone therapy. In contrast, analysed as a continuous variable, the interaction for patients with PSA 0·51–1·00 ng/mL (HR 0·72, overall survival was significant in the two-stage model 95% CI 0·54–0·96) and PSA >1·00 ng/mL (HR 0·69, (p=0·04), whereas the one-stage stratified model had 0·48–0·98) showed a reduction in mortality risk from previously yielded a non-significant result (p=0·19). the addition of hormone therapy (nominal p<0·05), The binary interaction at a PSA threshold of 0·5 ng/mL consistent with the significant interaction observed in also remained robust in the two-stage framework the binary analysis. On univariable analysis, the (p=0·03). interaction term based on categorical binning For patients randomly assigned to with or without pre-PORT PSA into these four strata was not significant long-term hormone therapy, the hazard ratio for overall (p =0·08). Using a two-stage multivariate meta- survival favoured the addition of hormone therapy interaction analysis to strictly isolate within-trial contrasts and (HR 0·79, 95% CI 0·63–1·00), whereas those randomly mitigate potential aggregation bias, we confirmed a assigned to with or without short-term hormone therapy significant interaction between the addition of hormone did not (HR 0·93, 0·77–1·11; figure 3). However, the Articles 0 –1 –2 –3 0 0·20·40·60·81·01·21·41·61·82·02·22·42·62·83·03·23·43·6 Figure 4: Natural cubic splines showing associations between pre-post-prostatectomy radiotherapy PSA and effect on overall survival and metastasis-free survival from adding hormonal therapy to radiotherapy Cubic splines show associations between ln(HR) and pre-post-prostatectomy radiotherapy PSA in the context of overall survival (A, C, E) and metastasis-free survival (B, D, F). In all panels, the superimposed grey dashed vertical line with a triangle indicates the pre-PORT PSA at which ln(HR)=0, or the hazard ratio=1. In panels B, D, and F, the additional superimposed red dashed vertical line with a circle indicates the pre-PORT PSA at which the upper bounds of the 95% CI for the ln(HR)=−0·21, or upper bounds for the HR=0·81. This is the ICECaP threshold for an effect size of improving metastasis-free survival that would translate to a subsequent significant improvement in overall survival. In panels E and F, an additional green triangle on the x-axis indicates the pre-PORT PSA at which the upper bound of the 95% CI for ln(HR) equals 0 (ie, upper bound of HR=1·0). PSA=prostate-specific antigen. ICECaP=Intermediate Clinical Endpoints in Cancer of the Prostate. 1066 )lavivrus eerf-sisatsatem ;RH(nl F Treatment effect: radiotherapy+long-term hormone therapy vs radiotherapy 95% CI; n=1033 observations 0 ICECaP threshold –1 –2 –3 0 0·20·40·60·81·01·21·41·61·82·02·22·42·62·83·03·23·43·6 Pre-radiotherapy PSA (ng/mL) )lavivrus llarevo ;RH(nl 2 0 –1 –2 0 0·2 0·4 0·6 0·8 1·0 1·2 1·4 1·6 1·8 2·0 2·2 2·4 2·6 E Treatment effect: radiotherapy+long-term hormone therapy vs radiotherapy 95% CI; n=1033 observations Pre-radiotherapy PSA (ng/mL) )lavivrus eerf-sisatsatem ;RH(nl D Treatment effect: radiotherapy+short-term hormone therapy vs radiotherapy 95% CI; n=3819 observations 1 –1 ICECaP threshold –2 –3 Pre-radiotherapy PSA (ng/mL) )lavivrus llarevo ;RH(nl 0 –1 –2 –3 0 0·20·40·60·81·01·21·41·61·82·02·22·42·62·83·03·23·43·6 C Treatment effect: radiotherapy+short-term hormone therapy vs radiotherapy 95% CI; n=3819 observations 0·2 0·4 0·6 0·8 1·0 1·2 1·4 1·6 1·8 2·0 2·2 2·4 2·6 Pre-radiotherapy PSA (ng/mL) )lavivrus eerf-sisatsatem ;RH(nl B Treatment effect: radiotherapy+hormone therapy vs radiotherapy 95% CI; n=4855 observations 1 ICECaP threshold –1 –2 –3 –4 0 0·20·40·60·81·01·21·41·61·82·02·22·42·62·83·03·23·43·6 Pre-radiotherapy PSA (ng/mL) )lavivrus llarevo ;RH(nl A Treatment effect: radiotherapy+hormone therapy vs radiotherapy 95% CI; n=4855 observations Pre-radiotherapy PSA (ng/mL) Articles interaction p value between randomisation to with or The NNTs to prevent one death at 10 years by adding without short-term hormone therapy versus with or hormone therapy to radiotherapy were negative for PSA without long-term hormone therapy was not significant values between 0·0 and 0·5 ng/mL, reflecting point (p =0·17). Results were similar on multivariable estimates that numerically favoured PORT alone; these interaction analysis. Notably, patients enrolled in with or without estimates are imprecise and should be interpreted long-term hormone therapy trials had higher cautiously. For patients with pre-PORT PSAs of pre-PORT PSAs and more adverse pathological features 0·51–1·00 ng/mL and greater than 1·00 ng/mL, the NNTs than those enrolled in with or without short-term were 22 and 12, respectively. Adjusted cubic hormone therapy trials. splines evaluating non-linear associations between 75 25 0 1 2 3 4 5 6 7 8 9 10 11 12 Time (years) 761 747 730 707 684 645 607 486 384 266 154 86 36 (0) (10) (13) (15) (23) (27) (50) (151)(239)(343)(441)(504)(555) 762 745 730 717 706 689 632 525 403 275 178 86 38 (0) (11) (16) (21) (22) (25) (58) (145)(255)(365)(455)(539)(587) Figure 5: Exploratory analyses of survival effect of adding short-term and long-term hormonal therapy to post-prostatectomy radiotherapy (A, B) Kaplan–Meier survival curves showing overall survival and metastasis-free survival effect of adding short-term hormonal therapy to post-prostatectomy radiotherapy. (C, D) Kaplan–Meier survival curves showing overall survival and MFS effect of adding long-term hormonal therapy to post-prostatectomy radiotherapy. (E, F) Kaplan–Meier survival curves showing overall survival and MFS effect of prolonging short-term hormone therapy to long-term hormone therapy with post-prostatectomy radiotherapy. )%( lavivrus eerf-sisatsateM F 75 Hazard ratio: 0·89 (95% CI 0·68–1·16, p=0·37) Hazard ratio: 0·89 (95% CI 0·68–1·16, p=0·37) 10-year estimates 10-year estimates 25 Radiotherapy+short-term hormone therapy: 82·0% (78·6–85·5) Radiotherapy+short-term hormone therapy: 71·9% (67·9–75·9) Radiotherapy+long-term hormone therapy: 84·6% (81·3–87·8) Radiotherapy+long-term hormone therapy: 78·1% (74·5–81·7) 0 1 2 3 4 5 6 7 8 9 10 11 12 Time (years) Number at risk (censored) Radiotherapy + short-term 761 749 741 732 719 698 661 529 424 289 174 98 43 hormone therapy (0) (10) (13) (15) (23) (27) (50) (167)(259)(378)(486)(557)(614) Radiotherapy + long-term 762 747 737 730 722 714 659 549 425 292 189 92 40 hormone therapy (0) (11) (16) (21) (22) (25) (58) (156)(272)(388)(486)(575)(624) )%( larivrus llarevO 75 Hazard ratio: 0·79 (95% CI 0·63–1·00, p=0·049) 10-year estimates 25 Radiotherapy: 79·8% (76·2–83·4) Radiotherapy+long-term hormone therapy: 82·7% (79·3–86·1) 0 1 2 3 4 5 6 7 8 9 10 11 12 Number at risk (censored) Radiotherapy 542 536 531 519 503 483 463 428 388 347 313 269 219 (0) (2) (3) (6) (10) (19) (26) (47) (78) (110)(130)(159)(201) Radiotherapy + long-term 546 541 532 522 514 494 480 445 402 365 328 292 239 hormone therapy (0) (1) (7) (9) (11) (23) (30) (54) (84) (107)(134)(161)(204) E Radiotherapy+short-term hormone therapy Radiotherapy+long-term hormone therapy )%( larivrus llarevO C 75 Radiotherapy Radiotherapy+long-term hormone therapy 25 0 1 2 3 4 5 6 7 8 9 10 11 12 542 530 519 502 483 460 440 403 358 315 278 236 187 (0) (2) (3) (6) (9) (16) (23) (43) (70) (101)(120)(146)(182) 546 539 528 517 505 480 464 425 384 342 302 266 212 (0) (1) (7) (9) (11) (23) (30) (53) (83) (105)(130)(156)(196) )%( lavivrus eerf-sisatsateM 75 25 0 1 2 3 4 5 6 7 8 9 10 11 12 1674 1617 1571 1526 14691396 1308 1136 945 653 385 175 52 (0) (32) (40) (53) (76) (102)(154)(294)(458)(722)(963)(1154)(1269) 2264 2211 2168 2097 20241946 1774 1502 1196 829 476 214 49 (4) (44) (59) (81) (106)(143)(262)(490)(761)(1093)(1419)(1664)(1829) D Hazard ratio: 0·74 (95% CI 0·60–0·91, p=0·004) 10-year estimates Radiotherapy: 70·8% (66·8–74·9) Radiotherapy+long-term hormone therapy: 76·5% (72·7–80·3) )%( lavivrus eerf-sisatsateM B 75 Hazard ratio: 0·93 (95% CI 0·77–1·11, p=0·39) Hazard ratio: 0·82 (95% CI 0·71–0·95, p=0·006) 10-year estimates 10-year estimates 25 Radiotherapy: 85·1% (83·0–87·2) Radiotherapy: 75·3% (72·8–77·9) Radiotherapy+short-term hormone therapy: 84·8% (82·9–86·7) Radiotherapy+short-term hormone therapy: 78·5% (76·4–80·7) 0 1 2 3 4 5 6 7 8 9 10 11 12 Number at risk (censored) Radiotherapy 1674 1634 1615 1586 1541 1482 1401 1208 1012 708 416 184 54 (0) (32) (40) (54) (77) (106)(160)(322)(504)(788)(1064)(1281)(1404) Radiotherapy + short-term 2264 2214 2186 2142 2090202818641589 1269 885 513 227 52 hormone therapy (4) (44) (59) (81) (106)(144)(270)(512)(800)(1154)(1504)(1773)(1948) )%( larivrus llarevO A Radiotherapy Radiotherapy+short-term hormone therapy Articles pre-PORT PSA concentration and overall survival and MFS negative for PSA values of 0·2 ng/mL or less, indicating are shown in figure 4A–B. For both endpoints, the upper point estimates that numerically favoured PORT alone bounds of the 95% CI for the HR cross 1·0 (ie, ln[HR] but with considerable uncertainty, and were 45 for crosses 0) for all pre-PORT PSAs. Consistent with this, for patients with pre-PORT PSAs of 0·21–0·50 ng/mL, MFS the Intermediate Clinical Endpoints in Cancer of the 16 for patients with pre-PORT PSAs of 0·51–1·00 ng/mL, Prostate (ICECaP) surrogate threshold effect of 0·81 was and 12 for patients with pre-PORT PSAs greater than crossed by the upper bounds of the 95% CI (ie, upper 1·00 ng/mL. NNTs for death-or-metastasis events are bounds of ln[HR] crossed −0·21) for all pre-PORT PSAs. shown in the appendix (p 16). Subgroup analyses are The addition of hormone therapy was associated with shown in the appendix (p 25). On spline analysis, the improved MFS (HR 0·79, 95% CI 0·70−0·89, p<0·001), upper bounds of the 95% CI for the HR crossed 1·0 for with 10-year MFS rates of 77·9% (95% CI 76·0−79·8%) both endpoints: at PSAs of at least 1·6 ng/mL for overall versus 74·1% (72·0–76·3%) with the addition of survival and at least 1·2 ng/mL for MFS (figure 4E–F). hormone therapy (figure 2F). On univariable analysis, The ICECaP threshold for the MFS effect having a the interaction term based on binning pre-PORT PSA subsequent effect on overall survival was met for PSA into the aforementioned four strata was not significant values of at least 2·0 ng/mL (figure 4F). (p =0·1; appendix pp 20−21). Results were similar An exploratory analysis evaluated the benefit of interaction on multivariable analysis or with binary PSA binning or prolonging short-term hormone therapy to long-term treating PSA as a continuous variable. NNTs to prevent hormone therapy with PORT; this analysis used the one death-or-metastasis event at 10 years were negative two-way and three-way RADICALS trials and included for pre-PORT PSA values of up to 0·2 ng/mL, and were 1523 patients with a median follow-up of 8·89 years 21 for PSA for pre-PORT PSA values of 0·21−0·50 ng/mL, (IQR 7·03–10·05; appendix pp 14–15). The prolongation 14 for those of 0·51−1·00 ng/mL, and seven for those of of hormone therapy did not improve overall survival greater than 1·00 ng/mL (appendix p 16). (HR 0·89, 95% CI 0·68–1·16, p=0·37), but did improve Four randomised trials evaluated the addition of short- MFS (HR 0·76, 95% CI 0·61–0·95, p=0·02; term hormone therapy to radiotherapy (3938 patients, figure 5E–F). Subgroup analyses and estimated NNTs median follow-up 8·80 years [IQR 7·13−10·10]; table). for death and death-or-metastasis are shown in the On spline analysis, the upper bounds of the 95% CI for appendix (pp 16, 27). the HR crossed 1·0 for overall survival and MFS endpoints, as well as crossing the surrogate threshold Discussion effect for MFS (figure 4C−D), for all pre-PORT PSAs. To our knowledge, this is the first individual patient data The addition of short-term hormone therapy to meta-analysis of randomised trials evaluating the effect radiotherapy did not improve overall survival (HR 0·93, of adding hormone therapy to PORT in prostate cancer, 95% CI 0·77−1·11; figure 5A), but improved MFS with several immediately clinically relevant results. First, (HR 0·82, 95% CI 0·71−0·95; figure 5B). the addition of hormone therapy to PORT does not The NNTs to prevent one death at 10 years by adding clearly provide relative benefits in overall survival, short-term hormone therapy to radiotherapy were negative particularly for patients receiving radiotherapy at for PSA values between 0·0 and 0·5 ng/mL, again pre-PORT PSA values of 0·5 ng/mL or less, with a small reflecting point estimates that numerically favoured absolute benefit. Second, the treatment effects observed PORT alone and should be interpreted cautiously, and from short-term and long-term hormone therapy appear were 18 for patients with pre-PORT PSAs of more similar than different, with a possible exception in 0·51–1·00 ng/mL and 12 for patients with pre-PORT PSAs men with pre-PORT PSA levels greater than 1·6 ng/mL. greater than 1·00 ng/mL. NNTs to prevent death-or- Finally, although adding hormone therapy does appear to metastasis events are shown in the appendix (p 16). improve MFS, the magnitude of this benefit was small, Subgroup analyses are shown in the appendix (p 23). implying a small effect on overall survival until Two randomised trials evaluated adding long-term pre-PORT PSA levels exceed 2·0 ng/mL. Thus, overall, hormone therapy to PORT (1088 patients, median follow- the POSEIDON results do not support the up 12·34 years [IQR 9·37–13·86]; table). Both trials routine addition of hormone therapy to PORT for included a comparison of 24 months of hormone therapy patients receiving adjuvant or early secondary (previously with PORT versus PORT alone. The addition of long-term called salvage radiotherapy) and suggests that meaningful hormone therapy to PORT was associated with improved benefits in overall survival would primarily manifest for overall survival (HR 0·79, 95% CI 0·63–1·00, p=0·049) patients receiving very late secondary radiotherapy and MFS (HR 0·74, 95% CI 0·60–0·91, p=0·004), (PSA>1·6–2·0 ng/mL). corresponding to 10-year overall survival estimates Our results complement the recent study-level of 82·7% (79·3–86·1) versus 79·8% (76·2–83·4; DADSPORT meta-analysis that evaluated the effect of figure 5C–D). adding hormone therapy to PORT. We show similar The NNTs to prevent one death at 10 years by adding effect sizes for overall survival and MFS with the addition long-term hormone therapy to radiotherapy were of hormone therapy to PORT.10 However, the DADSPORT 1068 Articles authors explicitly noted that “collection and reanalysis of higher preponderance of adverse risk features and higher individual patient data may help clarify who might pre-PORT PSA levels for patients enrolled in trials of benefit”.10 The availability of such data for POSEIDON PORT with or without long-term hormone therapy. Thus, addressed this need by enabling granular analysis of the short-term hormone therapy, if hormone therapy is used, non-linear interaction of pre-PORT PSA with treatment appears to result in similar overall survival to long-term effect, identifying potential interaction thresholds missed hormone therapy on the basis of the available randomised by aggregate trend tests, and standardising covariate data. definitions across trials. The results from POSEIDON have notable contrast Specifically, we present important analyses of non- with the findings pertaining to the benefit of adding linear associations between pre-PORT PSA and the effect hormone therapy with definitive radiotherapy for of adding hormone therapy. For both the overall cohort prostate cancer.1 There are several potential explanations and for men in trials evaluating PORT with or without for this finding. Hormone therapy is known to be a hormone therapy, the upper bounds of the 95% CI for potent radiosensitiser and, in the definitive management the HR crossed 1·0 for all evaluated PSA levels for both of prostate cancer with radiotherapy, improves overall overall survival and MFS. For MFS, the upper bounds of survival.21,22 In the postoperative setting, the addition of the 95% CI for the HR not only always crossed 1·0 but hormone therapy might not meaningfully improve local also always crossed 0·81, the surrogate threshold effect control, and does not clearly offer a meaningful survival limit defined by ICECaP for an MFS effect likely to benefit. This hypothesis is supported by two negative translate into a subsequent overall survival effect.18 randomised trials of dose-escalation in postoperative Therefore, the observed MFS benefits would not be radiotherapy.23,24 Alternatively, the benefit of adding expected to translate to overall survival benefits. For hormone therapy to PORT at high PSA levels in the long-term hormone therapy specifically, the upper postoperative setting could be explained by the treatment bounds of the 95% CI for the overall survival HR (albeit temporarily) of occult metastatic disease given the crossed 1·0 for pre-PORT PSAs of at least 1·6 ng/mL, established correlation of metastatic disease on prostate- and for MFS, the ICECaP threshold was passed for specific membrane antigen positron emission pre-PORT PSAs of at least 2·1 ng/mL. This suggests that tomography (PSMA PET–CT) with pre-PORT PSA. long-term hormone therapy might provide survival This meta-analysis is not without limitations. First, the benefits only when treating at very high pre-PORT PSA analysis includes patients enrolled in trials over a 17-year levels. This group of patients might also derive benefit period (1998–2015) during which definitions of from an intermittent hormone therapy approach that biochemical recurrence varied, and both stage migration includes second generation androgen receptor and Gleason grade migration occurred.25,26 However, our antagonists.20 The NNT analysis also underscores caution use of models stratified by trial ensures that comparisons with use of hormone therapy at lower PSA levels: for are made strictly against concurrent controls subject to the pre-PORT PSA of 0·5 ng/mL or lower, the NNTs to same definitions, and given the consistency of the trial- prevent one death at 10 years with adding either hormone specific results and the lack of significant heterogeneity, therapy or short-term hormone therapy are less than 0, this is unlikely to have substantially impacted our findings. suggesting that, on average, the addition of hormone Second, the majority of patients in the analysis received therapy or short-term hormone therapy might be hormone therapy in the form of androgen deprivation harmful. In contrast, at pre-PORT PSA >0·5 ng/mL, the therapy delivered via gonadotropin releasing hormone NNTs to prevent death at 10 years are 22 with the addition agonist medications; the major exception was the use of of hormone therapy, indicative of an absolute overall androgen receptor antagonist monotherapy in survival benefit of ≥5%. Taken together, these data NRG/RTOG 9601,7 which contributed 70% of the patients suggest that for patients receiving early secondary (760/1088 patients) in the radiotherapy with or without radiotherapy (pre-PORT PSA ≤0·5 ng/mL), there is hormone therapy subgroup. Since androgen receptor generally no survival advantage to adding hormone antagonist monotherapy is less effective than traditional therapy, with benefits emerging only for pre-PORT PSA androgen deprivation therapy,27 this might be expected to greater than 0·5 ng/mL. dilute or obscure the results. However, tests for One might be tempted to conclude that the addition of heterogeneity were non-significant. Molecular imaging, long-term hormone therapy provided greater benefit such as PSMA PET–CT, has greatly improved sensitivity than short-term hormone therapy given that the 95% CI for identifying occult disease at the time of post- for overall survival did not cross 1·0 for long-term prostatectomy recurrence.28,29 These scans were not hormone therapy, but did for short-term hormone generally available during the time interval in which the therapy. However, there was no evidence of an interaction trials included in the POSEIDON analysis were done. In on the basis of hormone therapy duration, nor was there addition, detailed information on baseline comorbidities, an overall survival benefit on meta-analysis comparing PSA doubling times, and testosterone recovery, which short-term versus long-term hormone therapy. The might affect outcomes, was not available.30 Although our observed results are probably the result of the overall analysis was designed to focus on overall survival primarily, Articles and MFS secondarily, earlier endpoints such as methodology, project administration, resources, supervision, writing— biochemical recurrence might be of great impact for original draft. AUK, YS, TRK, and DES had access to all the data patients. These endpoints, however, are inherently more reported in the study. The corresponding author had final responsibility for the decision to submit for publication. variable between trials, which had varying definitions of Declaration of interests biochemical recurrence and varying protocols for follow- AUK reports grant support from the NIH (P50CA09213 and up. Additionally, a larger impact on overall survival with 1R37CA292795) and the Department of Defense (PC210066); contracts longer follow-up cannot be ruled out. Further, in the from Novartis, Janssen, Lantheus, Varian Medical Systems, and ViewRay RADICALS-HD randomisations,4,5,8 more patients in the Systems; consulting fees from Lantheus, Varian Medical Systems, Novartis, and Janssen; honoraria from Janssen, Boston Scientific, Varian with or without hormone therapy randomisation had Medical Systems, and Lantheus, and low-value stock in MiraDx and secondary radiotherapy versus adjuvant radiotherapy than Alethian AI. CCP reports consulting fees from Novartis and Janssen and in the three-way randomisation, probably owing to participation on a data safety montoring board or advisory board for different perceived risks before enrolment. Finally, in our Telix. MRS reports previous emplyoment with the UKRI-MRI; grant suppport from Astellas, Janssen and Sanofi-Aventis; honoraria from primary assessment of interactions, we used standard Eisai, Eli-Lily, and Janssen; support for travel from the Health Research one-stage models that do not explicitly separate within-trial Board of Ireland, the Trials Methodology Network in Ireland, and the from between-trial interaction effects. To address concerns National Cancer Grid in India; and previous participation in several data regarding potential aggregation bias, we did sensitivity monitoring committees in an unpaid capacity. LFV reports grants from Bristol Myers Squibb Foundation Winn CDA, National Institutes of analyses using a two-stage approach that isolates within- Health SPORE Career Enhancement Program, Mike Slive Foundation trial interactions. The key finding that the significant for Prostate Cancer Research, The Parker Institute for Cancer interaction between pre-PORT PSA and treatment effect Immunotherapy, and the US Department of Veteran Affairs; and on overall survival was confirmed in the two-stage analysis consulting fees from the Dedham Group and Health Advances. JES reports grants from the National Institutes of Health and (binary PSA p=0·03 vs p=0·02 in one-stage), and overall Congressionally Directed Medical Research Programs; consulting fees results were consistent across approaches (appendix from Boston Scientific and Engaged MD; honoraria from Onclive; and a pp 18–19). leadership or fiduciary role in the American Urological Association. In summary, the POSEIDON meta-analysis from the JAG reports consulting fees from AstraZeneca, BeiGene USA, Dava Oncology, GlaxoSmithKline, MJH Life Sciences, Pfizer, and Urogen MARCAP consortium is, to our knowledge, the first Pharma. JRB reports honoraria from Ipsen and Merck; support for individual patient data meta-analysis of randomised attending meetings or travel for Johnson & Johnson; and participation in trials evaluating the use and prolongation of hormone data safety monitoring boards or advisory boards for AstraZeneca, Pfizer, therapy with postoperative radiotherapy in prostate and Johnson & Johnson. MBR reports consulting fees from Immune Bio; honoraria from Bayer and Johsnon&Johnson; and a patent for cancer. We provide what we believe to be the strongest Inhibitors of the N-terminal domain of the androgen receptor. level of evidence to date suggesting that the addition of AES reports consulting fees from Adaptyx Biosciences; honoraria from hormone therapy to postoperative radiotherapy might OncLive, ScientiaCME, Targeted Oncology, Eisai; and participation on a not benefit men receiving early salvage radiotherapy at data safety monitoring board or advisory board for Aveo Oncology, Bristol Myers Squibb, Eisai, and Exelixis. SE reports consulting fees with pre-PORT PSA of 0·5 ng/mL or less. Although robust Janssen. SR reports grant support from the Prostate Cancer Foundation; subgroup-specific benefits could not be broadly honoraria from Specialty Network; and stock with Merck, Pfizer, Eli Lilly, identified, pre-PORT PSA appears to be integral, with and Johnson & Johnson. NGZ reports grants from the American Cancer patients who have particularly elevated pre-PORT PSA Society—Tri State CEOs Against Cancer Clinician Scientist Development Grant, CSDG-20-013-01-CCE (2020–) and the Department of Defense levels (>1·6–2·0 ng/mL) appearing to derive the greatest (2020–). AYJ reports grants from Novartis and honoraria from Novartis. potential benefits from the addition of hormone therapy. JAE reports grants from Blue Earth Diagnostics; consulting fees from Notably, in this group of patients, adding short-term Blue Earth Diagnostics, Boston Scientific, and Clarity Pharmaceuticals; honoraria from Elekta, IBA, Genentech, UpToDate, Pfizer, and Astellas; hormone therapy might be as beneficial as adding participation on a data safety monitoring board or advisory board for long-term hormone therapy. Predictive biomarkers are Myovant Sciences, Janssen, Johnson & Johnson, Bayer Healthcare, urgently needed to improve hormone therapy Progenics Pharmaceuticals, Pfizer, Gilead, Lantheus, Angiodynamics, personalisation. NRG GU006 (NCT03371719) aims to Bioprotect, MDxHealth, and Boston Scientific; and leadership or fiduciary roles as a board member for the Massachusetts Prostate Cancer address this unmet need by testing whether a Coalition, American College of Radiology, and Radiation Oncology transcriptionally defined biomarker, PAM50, can identify Institute; as well as functioning as the Co-Chair of the NCI GU Steering a differential treatment effect of adding short-term Committee. OM reports honoraria from Bayer. JJD reports funding from hormone therapy with apalutamide to PORT. National Cancer Institute and US National Institutes of Health. AP reports a leadership role as the co-chair emeritus of the GU Contributors translational research programme at NRG oncology. HMS reports AUK contributed to conceptualisation, funding acquisition, leadership roles as a member of the Board of Directors for ASTRO and investigation, methodology, project administration, resources, Speed of Light, The ASTRO Foundation. PLN reports grants from Bayer, supervision, and writing—original draft. YS contributed to Astellas, and Janssen; consulting fees from Boston Scientific, Augmenix, conceptualisation, data curation, formal analysis, methodology, Janssen, Blue Earth, MDxhealth, AIQ, Novartis, and Bayer; a ledership validation, visualisation, and writing—original draft. AUK and YS role in NRG Oncology as the GU Chair; and stock or stock options in directly accessed and verified the underlying data. CCP, PS, MRS, SC, Nonocan, Reversal Therapeutics, and Strategen Bio. DES reports grant MB, TRK, MLS, LFV, KT, JAE, S, JAG, JRB, MBR, AES, RER, SE, WB, funding from the National Institutes of Health (U01); consulting fees SR, NGZ, AYJ, TMM, NGN, JAE, OM, JJD, WFS, AP, HMS, PLN, and from Boston Scientific; and honoraria from AstraZeneca, Astellas, Bayer, PP contributed to data curation, investigation, project administration, Blue earth, Janssen, Novartis, and Pfizer. All other authors declare no resources, supervision, and writing—review and editing. competing interests. DES contributed to conceptualisation, funding acquisition, investigation, 1070 Articles Data sharing 12 Moher D, Liberati A, Tetzhaff J, Altman DG. Preferred reporting Individual patient data obtained via data sharing agreements were used items for systematic reviews and meta-analyses. The PRISMA in this analysis. Researchers interested in obtaining de-identified patient Statement. PLoS Med 2009; 6: 1000097. data with a data dictionary from this study can contact the corresponding 13 Hunter KE, Aberoumand M, Libesman S, et al. The Individual author via email. This might require a signed data access agreement. Participant Data Integrity Tool for assessing the integrity of randomised trials. Res Synth Methods 2024; 15: 917–39. Acknowledgments 14 Simmonds MC, Higgins JP, Stewart LA, Tierney JF, Clarke MJ, AUK acknowledges charitable donations to funding from the Chapgier, Thompson SG. Meta-analysis of individual patient data from Bershad, De Silva, McCarrick, Halsted, and Moredock-Barnett Families. randomized trials: a review of methods used in practice. Clin Trials We thank Stephane Guerif, Stephane Supiot, and Igor Latorzeff for their 2005; 2: 209–17. invaluable contributions. This manuscript was prepared with data from 15 Cox DR. Regression Models and Life-Tables. the Data Archive of the National Cancer Institute’s (NCI’s) National J R Stat Soc Series B Stat Methodol 1972; 34: 187–202. Clinical Trials Network (NCTN). Data were originally collected from 16 Dinse GE, Lagakos SW. Nonparametric estimation of lifetime and clinical trial NCT number (NCT00567580—Prostate radiation therapy or disease onset distributions from incomplete observations. short-term androgen deprivation therapy and pelvic lymph node Biometrics 1982; 38: 921–32. radiation therapy with or without prostate radiation therapy in treating 17 Dess RT, Sun Y, Jackson WC, et al. Association of presalvage patients with a rising prostate specific antigen after surgery for prostate radiotherapy PSA levels after prostatectomy with outcomes of long- cancer). All analyses and conclusions in this manuscript are the sole term antiandrogen therapy in men with prostate cancer. responsibility of the authors and do not necessarily reflect the opinions JAMA Oncol 2020; 6: 735–43. or views of the clinical trial investigators, the NCTN, or the NCI. 18 Xie W, Ravi P, Buyse M, et al. Validation of metastasis-free survival as a surrogate endpoint for overall survival in localized prostate References cancer in the era of docetaxel for castration-resistant prostate 1 Kishan AU, Sun Y, Hartman H, et al, and the MARCAP cancer. Ann Oncol 2024; 35: 285–92. Consortium group. Androgen deprivation therapy use and duration 19 Rücker G, Cates CJ, Schwarzer G. Methods for including with definitive radiotherapy for localised prostate cancer: information from multi-arm trials in pairwise meta-analysis. an individual patient data meta-analysis. Lancet Oncol 2022; Res Synth Methods 2017; 8: 392–403. 23: 304–16. 20 Shore ND, Luz MA, De Giorgi U, et al. improved survival with 2 Eastham JA, Heller G, Halabi S, et al. Cancer and Leukemia enzalutamide in biochemically recurrent prostate cancer. Group B 90203 (Alliance): radical prostatectomy with or without N Engl J Med 2025; 394: 563–575. neoadjuvant chemohormonal therapy in localized, high-risk 21 Polkinghorn WR, Parker JS, Lee MX, et al. Androgen receptor prostate cancer. J Clin Oncol 2020; 38: 3042–50. signaling regulates DNA repair in prostate cancers. Cancer Discov 3 Carrie C, Magné N, Burban-Provost P, et al. Short-term androgen 2013; 3: 1245–53. deprivation therapy combined with radiotherapy as salvage 22 Spratt DE, Evans MJ, Davis BJ, et al. Androgen receptor treatment after radical prostatectomy for prostate cancer upregulation mediates radioresistance after ionizing radiation. (GETUG-AFU 16): a 112-month follow-up of a phase 3, randomised Cancer Res 2015; 75: 4688–96. trial. Lancet Oncol 2019; 20: 1740–49. 23 Qi X, Li HZ, Gao XS, et al. Toxicity and biochemical outcomes of 4 Parker CC, Clarke NW, Cook AD, et al, and the RADICALS dose-intensified postoperative radiation therapy for prostate cancer: investigators. Adding 6 months of androgen deprivation therapy to results of a randomized phase III trial. Int J Radiat Oncol Biol Phys postoperative radiotherapy for prostate cancer: a comparison of 2020; 106: 282–90. short-course versus no androgen deprivation therapy in the RADICALS-HD randomised controlled trial. Lancet 2024; 24 Ghadjar P, Hayoz S, Bernhard J, et al, and the Swiss Group for 403: 2405–15. Clinical Cancer Research (SAKK). Dose-intensified versus conventional-dose salvage radiotherapy for biochemically recurrent 5 Parker CC, Clarke NW, Cook AD, et al, and the RADICALS prostate cancer after prostatectomy: The SAKK 09/10 randomized Investigators. Randomised trial of no, short-term, or long-term phase 3 trial. Eur Urol 2021; 80: 306–15. androgen deprivation therapy with postoperative radiotherapy after radical prostatectomy: results from the three-way comparison of 25 Fossati N, Passoni NM, Moschini M, et al. Impact of stage RADICALS-HD (NCT00541047). Eur Urol 2024; 86: 422–30. migration and practice changes on high-risk prostate cancer: results from patients treated with radical prostatectomy over the last 6 Pollack A, Karrison TG, Balogh AG, et al. The addition of androgen two decades. BJU Int 2016; 117: 740–47. deprivation therapy and pelvic lymph node treatment to prostate bed salvage radiotherapy (NRG Oncology/RTOG 0534 SPPORT): an 26 Weiner AB, Etzioni R, Eggener SE. Ongoing Gleason grade international, multicentre, randomised phase 3 trial. Lancet 2022; migration in localized prostate cancer and implications for use of 399: 1886–901. active surveillance. Eur Urol 2014; 66: 611–12. 7 Shipley WU, Seiferheld W, Lukka HR, et al, and the NRG Oncology 27 Jia AY, Spratt DE. Bicalutamide monotherapy with radiation therapy RTOG. Radiation with or without antiandrogen therapy in recurrent for localized prostate cancer: a non-evidence-based alternative. prostate cancer. N Engl J Med 2017; 376: 417–28. Int J Radiat Oncol Biol Phys 2022; 113: 316–19. 8 Parker CC, Kynaston H, Cook AD, et al, and the RADICALS 28 Armstrong WR, Kishan AU, Booker KM, et al. Impact of prostate- investigators. Duration of androgen deprivation therapy with specific membrane antigen positron emission tomography/ postoperative radiotherapy for prostate cancer: a comparison of computed tomography on prostate cancer salvage radiotherapy long-course versus short-course androgen deprivation therapy in management: results from a prospective multicenter randomized the RADICALS-HD randomised trial. Lancet 2024; 403: 2416–25. phase 3 trial (PSMA-SRT NCT03582774). Eur Urol 2024; 86: 52–60. 9 Spratt DE. Further randomized data confirming minimal benefit 29 Holzgreve A, Armstrong WR, Clark KJ, et al. PSMA-PET/CT from the addition of hormone therapy to postoperative radiotherapy. Findings in patients with high-risk biochemically recurrent prostate Eur Urol 2024; 86: 431–33. cancer with no metastatic disease by conventional imaging. JAMA Netw Open 2025; 8: e2452971. 10 Burdett S, Fisher DJ, Tierney JF, et al, and the DADSPORT Meta- analysis Collaboration. Duration of androgen suppression with 30 Ong WL, Romero T, Roy S, et al, and the MARCAP Consortium postoperative radiotherapy (dadsport) for nonmetastatic prostate Investigators. Testosterone recovery following androgen cancer: a collaborative systematic review and meta-analysis of suppression and prostate radiotherapy (TRANSPORT): a pooled aggregate data. Eur Urol 2025; 88: 277–90. analysis of five randomized trials from the Meta-Analysis of Randomized Trials in Cancer of the Prostate (MARCAP) 11 Stewart LA, Clarke M, Rovers M, et al, and the PRISMA-IPD consortium. Eur Urol 2025; 87: 49–57. Development Group. Preferred reporting items for systematic review and meta-analyses of individual participant data: the PRISMA-IPD Statement. JAMA 2015; 313: 1657–65. --- [PDF原文](https://sci-net.xyz/storage/7932541/a609ed10227ca317f92b369c0bdf0f96918358f51c9dc6c16ae04f683751dbcf/Hormone-therapy-use-and-duration-with-postoperative-radiotherapy-for-recurrent-prostate-cancer.pdf) DOI: 10.1016/S0140-6736(26)00137-6