Estimation of Treatment Effect for Clinical Trials with Interim Analyses

• Activity Number: 193
• Type: Contributed
• Date/Time: Tuesday, August 10, 2004 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Michael Lee*+ and Hui Quan
• Companies: Merck & Co., Inc. and Merck & Co., Inc.
• Keywords: early stopping ; sample size re-estimate ; confidence interval ; alpha-adjustment ; estimation bias

Abstract

Interim analyses are commonly seen in clinical trials, especially in long-term studies. Results of an interim analysis can affect the study in several ways. The study can be stopped due to extremely favorable/unfavorable results, or sample size can be redetermined to achieve desired power. Research has been done largely under the framework of sequential and flexible study design. Many methods have been proposed for hypothesis tests that take the interim analysis into account and control the overall type I error rate. However, methods for estimation of treatment effect are not available in many cases. Without adjustment for interim analyses, a point estimate is potentially biased if early stopping boundary is asymmetric. Furthermore, when the trial is stopped early, a confidence interval based on a small level assigned to the interim analysis is too conservative. We propose a method for estimation of treatment effect for use with interim analysis. The proposed method corrects potential bias and produces confidence intervals with adequate coverage rate. We use a parallel study design with interim analyses and simulations for illustration.