- The Statistics group of the Mathematical Institute meets weekly on Mondays, 13.00-13.45.
- Due to the Corona virus, the seminar is held online for an indefinite time.
17-05-2021 | Eni Musta
Cure rate models: a presmoothing estimation approach
Thanks to recent medical advances, many cancer patients are cured. Hence, it is crucial to evaluate a treatment focusing on cure and not only survival prolongation. What makes this problem statistically challenging is the unobserved cure status for patients who, because of censoring, have only been followed for a limited period after treatment. Developed as an extension of classical survival analysis models, cure rate models can assess the cure chances and distinguish curative from life-prolonging effects. In this talk, I will provide an introduction of cure rate models, focusing on the most common logistic-Cox mixture cure model. Two estimation methods will be discussed: the maximum likelihood estimator, implemented in the R package smcure, and an approach based on presmoothing, proposed by Musta, Patilea and Van Keilegom (2020). The latter one is a two-stage procedure that can directly estimate a parametric cure probability without using any model assumption on the survival of the uncured patients. The advantages of this new approach with respect to the MLE will be discussed, supported by some simulation results. Practical use of cure models will be illustrated through applications to osteosarcoma and melanoma data. However, these methods are useful in other fields as well for analysing time-to-event data with an immune proportion of the population, e.g. fertility studies, equipment failure, credit scoring in economics.
24-05-2021 | Elia Biganzoli
31-05-2021 | Amine Hadji