INTEGRATE pooled phase II/III results are robust to post-progression switching and winner's curse.

Abstract

Background: The INTEGRATE phase 3 (P3) trial in advanced gastric and esophagogastric junction cancer involved pooling overall survival (OS) data with its preceding phase 2 (P2) trial, raising concerns about misalignment due to treatment switching in P2 or the Winner's curse. We evaluated P2 results, adjusted for these opposing effects, against P3 according to the prespecified statistical analysis plan.

Methods: OS estimates were adjusted for treatment switching using rank-preserving structural failure time model (RPSFTM) and inverse probability of censoring weights (IPCW) method. A novel shrinkage approach (NSE) mitigated overestimation from the Winner's curse, and Bayesian prediction (BP) methods predicted P3 outcomes from P2 estimates. A simulation study modelled 10,000 seamless P2/P3 trials to quantify bias in the pooled estimate.

Results: The observed P3 hazard ratio (HR 0.71, 95% CI 0.54-0.93) for OS was more conservative than the adjusted P2 estimates (RPSFTM and NSE: 0.61, 95% CI 0.29-1.29; RPSFTM and BP: 0.59, 95% CI 0.48-0.73; IPCW and NSE: 0.55, 95% CI 0.31-0.99; IPCW and BP: 0.58, 95% CI 0.46-0.72). Simulations indicated negligible bias in the pooled log(HR): -0.011 and 0.005 under the null and alternative hypotheses, respectively.

Conclusion: Adjusting P2 estimates for both treatment switching and the Winner's curse produced point estimates similar to the unadjusted P3 results. A prospective plan to pool trials data under a closed testing procedure may be a reasonable strategy when a recruitment shortfall in P3 is anticipated, provided that potential sources of misalignment are thoroughly assessed.