Standardversus adaptive monitoring procedures: a commentary

Abstract

In the standard approach to designing definitive clinical trials, the primary endpoint and test statistic to be used for the primary analysis are specified before trial initiation. The false positive error rate for the null hypothesis and statistical power to detect the targeted size of treatment effect are also specified. Standard monitoring procedures, such as the group sequential guidelines, enable interim monitoring while maintaining the integrity of this approach. In contrast, adaptive monitoring procedures seek to provide flexibility to modify these pre-specified design features during the course of the trial. However, these procedures have several undesirable properties, including lesser statistical efficiency, reduced interpretability of primary outcome results, basing design changes on unreliable interim estimates of efficacy, risks to the integrity and credibility of the trial, loss of flexibility to use emerging results from external sources to alter key design features, and overemphasis of the importance of statistical significance relative to clinical significance.