This paper discusses, following Gómez and Lagakos (2011) methodology, to what extent is there a gain in efficiency from adding a component event to a relevant endpoint when the treatment effect on this component is not as strong as on the original relevant endpoint under ideal (independence) circumstances. It presents the bivariate copula model used to overcome the independence assumption and presents the relationship between the components of the asymptotic relative efficiency and a set of interpretable parametres.

Peer Reviewed