TL40: Clinical Experiment and Statistical challenges to analyze and report Composite Endpoints
*Tianhui Zhou, Pfizer 


Some diseases (e.g. Lupus, Rheumatoid Arthritis) can be extremely varied in their signs, symptoms; and affect many different organ system. Composite endpoints can be helpful to capture the heterogeneity. Studies with composite outcomes lend to a higher statistical efficiency because they have a higher event rate; allowing for a smaller sample size, decreased financial investment, and shorter time to complete the study. The selection of composite endpoints has been regarded as "a useful strategy" for registration of pharmaceuticals in the ICH E9 guideline.

Recent article by Cordoba G etc. published in BMJ (2010;341:c3920) showed that the reporting of study with composite endpoints are riddled with multiple problems ranging from illogical construction, lack of consistent reporting, and the ability to pick results to present a clinically significant outcome.

Statistical challenges arise because the individual components are often correlated, with different importance and coming from different types of components such as: efficacy, safety, death, dropout, rescue medications etc. Principles of analysis and reporting composite endpoints will be widely discussed in this session.