Handling of Missing Data for Composite Endpoints in Clinical Trials
View Presentation View Presentation
*Hui Quan, Sanofi  Ji Zhang, Sanofi 

Keywords: Missing at Random, ITT, EM algorithm, MLE

Composite endpoints are often used in clinical trials in order to increase the study power through the reduction of endpoint variability or the increase of the number of events. Just as any study endpoints, missing data can occur in the components of a composite endpoint. If a patient has missing data in some of the components but not all the components, this patient may have incomplete data but partial data for the composite endpoint. The commonly used naïve methods may completely discard the patient from the analysis. To be consistent with the intention-to-treat principle, partial data from the patient should be included in the analysis. In this presentation, we discuss approaches for the analysis of composite endpoints with missing data in components focusing on binary endpoints. The main idea is to first derive the probabilities of all possible study outcomes and then to construct the overall rate for the composite endpoint. Simulation results are presented to compare the proposed approach with two naïve methods. A trial example is used to illustrate the application of the approach.