Abstract
Our focus is on the joint analysis of longitudinal nonnormal responses and early discontinuation in (pre)-clinical trials. Separate models are fitted to the two series (response and discontinuation) to account for covariate and time effects. The serial dependence and the dependence between response and drop-out are also modeled. This is done using particular dependence functions, called copulas. Copulas are used to create a joint distribution with given marginal distributions. Applications are given for the analysis of heart rate/morbidity in toxicology and pain severity/intake of rescue medications in a trial on migraine. Using copulas, the level of dependence between two variables remains invariant to changes in the marginal distribution of either variable. This proves interesting in modeling the association in a longitudinal setting when responses change over time.
ACKNOWLEDGMENTS
We wish to thank Dr. Roy Tamura from Eli Lilly and Company for his helpful suggestions on the initial version of the manuscript. The authors also thank two referees for their suggestions, which improved the exposition in several places.