Copula-based joint survival modelling of overall survival and disease free times: A study of breast cancer data

Beatriz Dos Santos Lourenço

Key information

Authors:

Beatriz Dos Santos Lourenço (Beatriz Dos Santos Lourenço)

Supervisors:

Giovani Loiola da Silva (Giovani Loiola da Silva)

Published in

October 26, 2020

Abstract

When implementing survival models for two or more times-to-event separately, some information may be getting mislaid, especially if these are correlated. For that reason, here we focus on jointly modelling two times-to-event: overall survival time (OST) and disease free time (DFT), using copula functions to model the dependence between them, under a context of Weibull regression. This work was motivated by a study involving 684 female breast cancer patients from the Portuguese Institute of Oncology of Lisbon (IPO-Lisboa). The copula-based joint survival modelling approach was implemented to the motivating dataset, where several risk factors were considered. The patient clinicopathological staging and DNA ploidy are considered statistically relevant for disease free and overall survival times, whereas patient age effect is only important for the overall survival time of older patients. The hazard of occurring cancer relapse or death, given the patient clinical profile, is different for univariate and bivariate models. In addition, the copula association parameter estimate demonstrates a high association between disease free and overall survival times. That association is influenced by the patient age, i.e., older patients present a stronger association between the times-to-event. The computational results were obtained resorting to the "GJRM" R package.