ABSTRACT
Clinical trials often involve repeated measurements of variables such as biological markers, cognitive scores or quality of life measures as long as the collection of times of onset of clinical events. In this chapter we review the two approaches for joint modelling a longitunal marker and a time-to-event : the shared random effect joint models and the joint latent class models. We present the various model formulations, the estimation methods and briefly outline model diagnostic. The use of these models in clinical trials are illustrated on several examples either to investigate the surrogacy of the marker or for studying the treatement effect on the marker accounting for possible informative dropout.
