In medical and epidemiolgical studies there are sometimes two types of response variables, a longitudinal variable that is measured periodically over time and a binary event that may be subject to censoring. The two response variables are very likely associated with each other and also with time-independent variables. In this talk I will describe types of joint models that have been proposed to analyse data of this type. Maximum likelihood, Bayesian and other estimation methods have been proposed. There are a wide variety of uses of joint models, these include correcting for bias in longitudinal and survival analysis due to informative dropout, predicting future longitudinal and event times for individuals, assessing whether the longitudinal variable could be an auxiliary variable or a surrogate endpoint. Illustration of the uses of joint model in prostate cancer research will be given.