J Biopharm Stat. 2008;18(4):631-8; discussion 439-45. doi: 10.1080/10543400802071311.
Journal of biopharmaceutical statistics
Michael J Dallas
PMID: 18607795 DOI: 10.1080/10543400802071311
Upon termination of a clinical trial that uses interim evaluations to determine whether the trial can be stopped, a proper statistical analysis must account for the interim evaluations. For example, in a group-sequential design where the efficacy of a treatment regimen is evaluated at interim stages, and the opportunity to stop the trial based on positive efficacy findings exists, the terminal p-value, point estimate, and confidence limits of the outcome of interest must be adjusted to eliminate bias. While it is standard practice to adjust terminal statistical analyses due to opportunities to stop for "positive" findings, adjusting due to opportunities to stop for "negative" findings is also important. Stopping rules for negative findings are particularly useful when monitoring a specific rare serious adverse event in trials designed to show safety with respect to the event. In these settings, establishing conservative stopping rules are appropriate, and therefore accounting for the interim monitoring can have a substantial effect on the final results. Here I present a method to account for interim safety monitoring and illustrate its usefulness. The method is demonstrated to have advantages over methodology that does not account for interim monitoring.
