[Using capture-recapture models to estimate transition rates between states in interval-censored data]

We introduce capture-recapture models, survival models developed in animal population biology to estimate survival / transition rates between sites within a population. We then show how to use these models in epidemiology or clinical research, to estimate survival / transition rates between stages of a disease (e.g. cancer, AIDS...), when patients enter a longitudinal study with imperfect follow-up resulting in interval-censoring. This method yields unbiased estimates, since the compliance (defined here as the probability of visiting the doctor at the planned date) is modelled jointly. We apply the method to cancer data, and show that "time" (elapsed since entry of a patient in the study) affects compliance in these data.