B0649
Title: Penalised competing risks regression
Authors: Thomas Scheike - University of Copenhagen (Denmark)
Federico Ambrogi - University of Milan (Italy) [presenting]
Abstract: High dimensional data analysis is an important topic in many research fields. For example, biomedical research generates increasing amount of data to characterise patients bio-profiles (e.g. from genomic, imaging, physiological measurements, laboratory tests etc.). In the last decades many forms of penalized regression have been developed, as a modern form of variable selection, to cope with high dimensional settings. The increasing complexity in the characterisation of patients bio-profiles, is added to the complexity related to the prolonged follow-up of patients with the registration of the occurrence of possible adverse events. Although in the last years the number of contributions for coping with high dimensional data in standard survival analysis have increased, the research regarding competing risks is less developed. The aim is to consider how to do penalized regression when considering the crude cumulative incidence. The direct binomial regression model is reformulated in a penalized framework to possibly fit a sparse regression model. The proposed approach is easily implementable using existing high performance software to do either ridge, or lasso or elastic net penalization. Results from simulation studies are presented together with an application to genomic data when the endpoint is progression free survival.
