|
Abstract:
|
Single-arm phase II trials remain relevant in drug development, requiring smaller sample size compared to randomized phase II trials, and are practical when recruiting small target populations (e.g. molecularly selected patents). Historically, objective response (binary endpoint measuring tumor shrinkage) has been the endpoint of choice in single-arm phase II; however, modern therapies may extend progression-free survival (PFS) and overall survival (OS) without tumor shrinkage. Single-arm trials with PFS endpoints are typically designed using standard methodology assuming right-censored survival data despite PFS being mixed interval-censored, i.e. progression is interval-censored and death is right-censored. Ignoring the interval-censoring nature of progression leads to overestimation of time to progression and potentially introducing bias and inflating type I error in treatment comparisons with historic cohorts. We evaluate and compare methods applied to mixed interval-censored endpoints in terms of type I error and study power.
|
