FDA Provides Guidance of Using Multiple Endpoints in Clinical Trials

A new drug can affect multiple aspects of the disease it is intended to treat, and hence clinical trials are designed to evaluate multiple efficacy endpoints compared to the controls. However, evaluating multiple endpoints in a clinical trial should involve appropriate statistical adjustments to account for the existence of multiple comparisons, called the multiplicity effect. FDA’s final guidance on the topic lists various statistical approaches to address the multiplicity effect in trials using multiple endpoints to support claims of overall efficacy. Most diseases cannot be characterized based on a single endpoint. Developers also evaluate multiple aspects of drug effects to de-risk their clinical development programs. The standard practice is to group the various endpoints in a clinical trial into Primary, Secondary and Exploratory endpoints. The Primary endpoints are those that provide definitive conclusions regarding the effectiveness of the drug, and the Secondary endpoints are those that extend understanding of the primary benefit. The primary and secondary endpoints require statistical adjustments to address multiplicity effects as they could together define the overall benefits of the drug compared to the control. The Exploratory endpoints generally are not used to support definitive conclusions and hence do not require statistical adjustments. Between the Primary and Secondary endpoints, trials with multiple Primary endpoints create greater risk of the multiplicity effect. There are four scenarios for using multiple endpoints in clinical trials: Co-Primary endpoints where more than one endpoint are together critical for demonstrating efficacy, Multiple Primary endpoints where more than one endpoints independently demonstrate efficacy, Composite endpoints where multiple clinical outcomes are combined for a single primary endpoint, and multi-component Primary endpoints which involve within-patient combination of two or more efficacy components. FDA recommends specific statistical approaches for each scenario and presents eight statistical methods to address the multiplicity effect. Many of the approaches included in the finalized guidance documents have been raised in previous FDA discussion but the guidance provides an excellent primer for clinical trial statisticians looking for a confirmation of FDA acceptable approaches.