Bayesian reanalysis of avelumab maintenance therapy in Asian and Japanese patients with advanced urothelial carcinoma: evidence from the JAVELIN Bladder 100 trial.

BACKGROUND

The JAVELIN Bladder 100 trial established avelumab maintenance as standard care for advanced urothelial carcinoma (UC). However, Asian (n = 147) and Japanese (n = 73) subgroups showed non-significant results with wide confidence intervals, as expected given limited sample sizes. We sought to provide complementary insights using Bayesian methods that quantify probability of treatment benefit.

METHODS

We performed Bayesian reanalysis using published data from JAVELIN Bladder 100. The global trial data (n = 700; HR = 0.76, 95% CI: 0.63-0.91) informed prior distributions across four scenarios: neutral prior (no borrowing), conservative borrowing (prior SD = 2 × Standard Error [SE]), moderate borrowing (prior SD = 1.5 × SE), and strong borrowing (prior SD = SE). Posterior distributions were derived using normal-normal conjugate updating.

RESULTS

For the Asian cohort, posterior mean hazard ratios ranged from 0.754-0.758 across borrowing scenarios, with probability of benefit P(HR < 1) ranging from 85.5% (neutral) to 99.9% (strong). The Japanese cohort showed posterior mean HRs of 0.763-0.828, with P(HR < 1) ranging from 71.9% (neutral) to 99.9% (strong). Under moderate borrowing, both cohorts achieved credible intervals excluding 1.0 (Asian: 0.591-0.966; Japanese: 0.594-0.990). Probability of clinically meaningful benefit P(HR < 0.8) reached 67.5% in Asian and 62.7% in Japanese patients.

CONCLUSIONS

While statistical limitations exist regarding the non-independence of data sources and dominance of global information, the high probabilities of benefit across all analytical scenarios support the use of avelumab in these populations. These findings illustrate both the utility and limitations of Bayesian methods in interpreting underpowered regional subgroups, supporting equitable access to effective cancer therapies.