A1135
Title: Single-index mixture cure models: An application to a study of cardiotoxicity in breast cancer patients
Authors: Ricardo Cao - University of Coruna (Spain) [presenting]
Ana Lopez-Cheda - University of A Coruna (Spain)
Beatriz Pineiro-Lamas - Universidade da Coruna (Spain)
Abstract: Standard survival models assume that the event of interest would always happen if there was sufficient follow-up time. However, this is not always realistic. For instance, HER2-positive breast cancer patients usually receive trastuzumab. Although this therapy has antitumor efficacy, it can cause a problem in the heart, known as cardiotoxicity, in some patients. In this context, there will be a fraction of individuals that will never suffer the side effect just because they are not susceptible to it. They are said to be cured in the sense that no matter how long you observe them, they will never experience the final event. To study the time until the cardiotoxicity appears, mixture cure models are appropriate. They allow us to estimate both the probability of being cured and the survival function of the uncured population, depending on some covariates. In the literature, nonparametric estimation of both functions is limited to continuous unidimensional covariates. This important gap is filled by considering multidimensional covariates and proposing a single-index model for dimension reduction. BC-Cardiotox, a dataset related to cardiotoxicity from the University Hospital of A Corua, is constructed and analyzed considering these techniques.