Interim Sample Size Recalculation for Observational Studies

• Activity Number: 605
• Type: Contributed
• Date/Time: Thursday, August 7, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Sergey Tarima*+ and Peng He and Tao Wang and Aniko Szabo
• Companies: Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin and Medical College of Wisconsin
• Keywords: sample size recalculation ; logistic regression ; observational data

Abstract

Internal pilot designs allow re-estimation of the sample size at the interim analysis using available information on nuisance parameters. In general, this affects the Type I and II error rates. We propose and investigate a method based on resampling the whole design at the interim analysis, starting with sample size recalculation at the observed interim analysis values of nuisance parameters, and finishing with the decision to accept or reject the null hypothesis. This internal resampling is performed under both the null and the alternative hypotheses allowing the estimation of the bias of the type I error and power. Finally, the bias corrected error rates are used in the original sample size calculation procedure to obtain an updated sample size. We explore the proposed resampling approach under a set of simulation scenarios and compare it with several previously published internal pilot designs. Interim sample size recalculation for logistic regression was explored via Monte-Carlo simulations. It shows the use of the proposed methodology to a wide class of statistical estimators. Illustrative example highlights the benefits of our approach for logistic regression analysis.