EXTENSIONS OF THE COX MODEL TO ANALYSE MULTIPLE TIME-TO-EVENT DATA

In longitudinal studies, it is very usual that a subject can experience several events, which can result from the occurrence of multiple-type events or the repetition of a single-type event (known as recurrent events). This kind of data arises in a wide variety of settings such as in biomedical studies on relapses of a certain disease. Over the past four decades, several extensions of the Cox model have been suggested to analyse multiple time-to-event data. Some of the most applied models were proposed by: Prentice, Williams and Peterson (PWP); Andersen and Gill (AG); Wei, Lin e Weissfeld (WLW); and Lee, Wei and Amato (LWA). The aim of this seminar is to carry out a brief, but careful, description of these four models. Moreover, another goal is to provide useful and practical guidelines to support the choice of the most appropriate model for each situation. Bearing this in mind, it will be illustrated how these models are implemented in R statistical software through the analysis of bladder cancer recurrences data available as a part of the survival package. The R code for fitting each model is also provided.