|
Abstract:
|
Clinical trials use composite outcomes to overcome many statistical issues, including power and inflation of type I error. However, there are problems with the interpretation of composite outcomes, in terms of possible heterogeneity of treatment effect on the components of these composites. If significant heterogeneity is found the composite outcome may be invalidated/inappropriate for use. However, tests for heterogeneity are known to lack power. The purpose of this paper is to compare the power of different tests to detect heterogeneity of treatment effect across components of composite binary outcomes. We conducted a simulation for a typical trial in cardiovascular disease to test the power of 5 different models detect treatment heterogeneity among components of a composite. Power is shown for variations in heterogeneity degree, component numbers/balance, correlation, and sample size.
|
