Joint longitudinal and survival-cure models to improve the assessment of being cured in oncology studies

International audienceIn the cancer framework, a number of studies such as certain relating to breast and prostate cancers actually include two kinds of patients: those who will not experience the event of interest (e.g. clinical recurrence, time to progression of cancer…) and are said to be “cured”, and those who will develop the event, and are said to be “susceptible”. However, the cure status is unobserved in (right-)censored patients. While most of the work on cure models focus on the time-to-event for the uncured patients (latency) or on the baseline probability to cure or not (incidence), we focus in this research on the conditional probability of being cured after a medical intervention (surgery, chemotherapy, radiotherapy…) given that a patient or a profile of patients survived until a certain time. An important feature of the cancer studies is the availability of longitudinal measurements collected over the follow-up such as, biomarkers, size of tumor, or quality of life. We therefore consider a joint model for the longitudinal and survival data given a cure fraction. This model includes a mixed model to fit the trajectory of longitudinal measurements and a mixture cure model. Different share latent structures to link both models are compared in order to improve the analysis and the computation of the conditional probability of interest