In a long term clinical study, the subjects may experience repeated occurre
nces of the same type of events or multiple events of entirely different na
ture during the following period. In this article, we reviewed the three mo
st frequently used extensions to the proportional hazards model for analysi
ng such multivariate failure time data. Specifically, we explored the under
lying hypotheses for each of these extensions. The three procedures were ap
plied to a bladder cancer study.