A Hybrid Bootstrap Approach to Interim Analyses of Time-to-Event Data Using Kaplan-Meier Methods

• Activity Number: 471
• Type: Contributed
• Date/Time: Wednesday, August 1, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Kevin Lawson*+ and Aaron C Camp and Austin L Hand
• Companies: PPD and PPD and PPD
• Keywords: Bootstrap ; Bayesian ; Oncology ; interim ; survival ; censoring

Abstract

Efficacy analyses in oncology clinical trials commonly include time-to-event endpoints analyzed using Kaplan-Meier methods. Decisions are frequently required prior to completion of the study. Bootstrapping is a well-known method for drawing inferences from partial data, but is problematic when the maximum event time is small or when right-censoring is large. An alternative method is to use a parametric bootstrap procedure by sampling from a hypothesized or fitted parametric distribution. Here, we present a hybrid bootstrap procedure that uses three strategies to combine the empirical distribution with a conditional parametric distribution fit with observed data. The bootstrap will sample from this mixture distribution under 3 schema for selecting the weights. We implement this method at several timepoints using data from 3 oncology trials. The results of the hybrid bootstrap, including the 95% bootstrap confidence intervals on the median time-to-event and the probability that the median time-to-event will exceed a predetermined value, are compared to the Bayesian posterior distribution of the survival function and to the actual median time-to-event.