A cluster-adjusted rank-based test for a clinical trial concerning multiple endpoints with application to dietary intervention assessment.

Multiple endpoints are often naturally clustered based on their scientific interpretations. Tests that compare these clustered outcomes between independent groups may lose efficiency if the cluster structures are not properly accounted for. For the two-sample generalized Behrens-Fisher hypothesis concerning multiple endpoints we propose a cluster-adjusted multivariate test procedure for the comparison and demonstrate its gain in efficiency over test procedures that ignore the clusters. Data from a dietary intervention trial are used to illustrate the methods.