Con preghiera di diffusione tra tutti i possibili interessati, scusandomi per invii multipli. Cordialmente, Giacomo Aletti

-------------- Si comunica che in data 28 Ottobre 2014, ore 11, presso la Sala di Rappresentanza del Dipartimento di Matematica dell'Università degli Studi di Milano

La dott.ssa Giuliana Cortese, del dipartimento di Statistica dell'Università di Padova, terrà un seminario dal titolo

"Dynamic predictions via competing risks regression models with application to chemotherapy in breast cancer"

Abstract In breast cancer, the risk for cardiotoxicity due to chemotherapy typically increases with the cumulative dose of treatment over time. Therefore, it is of interest to predict the cumulative risk for cardiotoxicity in a dynamic way, while controlling the competing risk for mortality. For this purpose, we consider a competing risks regression model with two events, cardiotoxicity and death. A second aim is to predict optimal cumulative doses along a given treatment time that keep the cumulative risk for cardiotoxicity below a certain threshold (e.g. <5%).

We face these prediction problems with different regression models for competing risks, accounting for all the time-dependent information in the data. For example, some covariates are assumed to have time-varying effects, and the cumulative dose was included as (internal) time-dependent covariate into the models. A landmark analysis was used to obtain dynamic predictions.

In this applied context, we compare the competing risks approach based on cause-specific hazards and the direct regression models that allow finding a one-to-one relationship between cumulative risks and covariates. These models were then employed to predict optimal cumulative doses at a sequence of landmark time points.

To control and correct also for the increased risk of dying, we finally predict optimal cumulative doses when a illness-death model is considered. Implementation in R is also discussed.