Bayesian reanalysis to assess consistency of TALAPRO-2 results in the Japanese subgroup with metastatic castration-resistant prostate cancer.

BACKGROUND

The phase III TALAPRO-2 trial showed that talazoparib plus enzalutamide significantly improves radiographic progression-free survival (rPFS) in patients with metastatic castration-resistant prostate cancer. In contrast, the Japanese subgroup (n = 116) demonstrated non-significant results with wide confidence intervals, as expected, because of the limited sample size. We attempted to provide complementary insights using Bayesian methods to quantify the probability of treatment benefit.

METHODS

We performed a Bayesian analysis using published data from the TALAPRO-2 trial. Prior distributions were set across four scenarios based on global trial data: neutral prior (no borrowing), conservative borrowing (prior standard deviation [SD] = 2 × standard error [SE]), moderate borrowing (prior SD = 1.5 × SE), and strong borrowing (prior SD = SE). The posterior distributions for the Japanese subgroup were derived using normal-normal conjugate updating.

RESULTS

The 'treatment effectiveness (hazard ratio [HR] < 1)' probability in Japanese all-comers reached 98% with conservative borrowing and > 99% with moderate and strong borrowing. With moderate borrowing, the posterior mean HR was 0.67 (95% credible interval 0.50-0.89). In the homologous recombination repair-deficient subgroup, all borrowing scenarios excluded HR = 1.0, with P(HR < 1.0) exceeding 99%.

CONCLUSION

Using Bayesian evidence synthesis with prespecified borrowing levels, the Japanese subgroup results were compatible in direction and magnitude with the overall TALAPRO-2 effect. These findings reduce uncertainty around local extrapolation; a dedicated randomized study would be required to establish efficacy independently in Japanese patients.