Real‐World Comparison of Triplet Versus Doublet Therapy in Japanese Patients With High‐Volume Metastatic Hormone‐Sensitive Prostate Cancer: A Propensity‐Weighted Analysis From the YUSHIMA Registry and Exploratory Bayesian Evidence Synthesis

Background To compare the effectiveness of triplet versus doublet therapy for time to castration‐resistant prostate cancer (CRPC) in Japanese patients with high‐volume metastatic hormone‐sensitive prostate cancer and examine Bayesian integration of external phase III trials. Methods We analyzed patients in the YUSHIMA registry with CHAARTED high‐volume metastatic hormone‐sensitive prostate cancer who started first‐line therapy between 2021 and 2024 and received triplet (androgen‐deprivation therapy + docetaxel + androgen receptor pathway inhibitor [ARPI]; n = 36) or ARPI‐based doublet (androgen‐deprivation therapy + ARPI; n = 104). The primary endpoint was time to CRPC. Confounding was addressed using overlap weighting. Weighted Cox models and restricted mean survival time (RMST) were estimated at 12 and 24 months. An exploratory Bayesian power‐prior synthesis combined the YUSHIMA likelihood with reconstructed individual patient data from seven trials. Results Median follow‐up was 23 months; 34 patients progressed to CRPC. Overlap‐weighted Cox analysis showed a statistically nonsignificant difference between triplet and doublet therapy (hazard ratio: 0.81, 95% confidence interval: 0.31–2.12). RMST differences (triplet minus doublet) were 0.21 months (95% confidence interval from −0.64 to 1.15) at 12 months and 0.95 months (95% confidence interval from −1.95 to 3.96) at 24 months. Bayesian synthesis analysis with borrowing strength α = 0.5 revealed posterior mean RMST differences of 1.42 and 3.19 months at 12 and 24 months, respectively. Conclusions In this propensity‐weighted registry analysis, triplet therapy was not demonstrably superior to ARPI‐based doublet therapy for time to CRPC. Exploratory Bayesian synthesis suggests a modest CRPC‐free RMST advantage with triplet therapy, but inferences depend on external evidence and remain hypothesis‐generating.