Clinical Trials
Bayesian Reanalysis Finds High Probability of OS Benefit With [177Lu]Lu-PSMA-617 (Investigational)
Swayamjeet Satapathy and Kunal ChandekarJCO clinical cancer informatics
![Bayesian Reanalysis Finds High Probability of OS Benefit With [177Lu]Lu-PSMA-617 (Investigational)](https://wnjopjzmxekqrxnqlahm.supabase.co/storage/v1/render/image/public/blog-images/ai-generated/ai-curator/2026-11-21/9b8f44bd-e0ea-4b50-a2fa-3583718aff0e.png?width=800&quality=70&resize=contain)
In the PSMAfore trial, [177Lu]Lu-PSMA-617 significantly improved radiographic progression-free survival compared with an androgen receptor pathway inhibitor (ARPI) change in patients with chemotherapy-naïve metastatic castration-resistant prostate cancer (mCRPC). However, evaluating overall survival (OS) was complicated by significant treatment crossover. To address this, investigators performed a Bayesian reanalysis of the final PSMAfore OS data evaluating both intention-to-treat (ITT) and crossover-adjusted inverse probability-of-censoring weighting (IPCW) estimands.
Published summary statistics from the final OS analysis were analyzed using normal likelihoods specified on the log hazard ratio (HR) scale with variances derived from reported confidence intervals. Posterior distributions were derived using conjugate normal-normal updating across skeptical, weakly informative, enthusiastic, and pessimistic priors. Under the IPCW estimand, posterior median HRs were 0.62 to 0.69 (95% CrI, 0.43 to 0.99), with a 0.98 to 0.99 probability of any OS benefit and a 0.79 to 0.91 probability of clinically meaningful OS benefit [P(HR ≤ 0.80)]. Under the ITT estimand, posterior median HRs ranged from 0.89 to 0.92, with 95% CrIs crossing unity and a 0.78 to 0.86 probability of any OS benefit. The probability of severe harm [P(HR ≥ 1.20)] was ≤0.01 across primary analyses, and only strong pessimistic priors substantially attenuated the probability of clinically meaningful benefit.
The authors concluded that Bayesian reanalysis demonstrates a high probability of OS benefit with early [177Lu]Lu-PSMA-617, particularly after adjusting for crossover, indicating that a statistically nonsignificant ITT analysis should not be interpreted as evidence of no survival benefit.
Key Takeaways
- Under the crossover-adjusted inverse probability-of-censoring weighting (IPCW) estimand, posterior median hazard ratios for overall survival ranged from 0.62 to 0.69 (95% CrI, 0.43 to 0.99), yielding a 0.98 to 0.99 probability of any survival benefit [P(HR < 1)] and a 0.79 to 0.91 probability of clinically meaningful benefit [P(HR ≤ 0.80)].
- Under the intention-to-treat estimand, posterior median hazard ratios ranged from 0.89 to 0.92 across primary priors with 95% credible intervals crossing unity, a 0.78 to 0.86 probability of survival benefit [P(HR < 1)], and a probability of severe harm [P(HR ≥ 1.20)] of ≤0.01 across primary analyses.
Read the original sourceJCO clinical cancer informatics
