Modelling intervention effects after cancer relapses.

This article addresses the problem of incorporating information regarding the effects of treatments or interventions into models for repeated cancer relapses. In contrast to many existing models, our approach permits the impact of interventions to differ after each relapse. We adopt the general model for recurrent events proposed by Peña and Hollander, in which the effect of interventions is represented by an effective age process acting on the baseline hazard rate function. To accommodate the situation of cancer relapse, we propose an effective age function that encodes three possible therapeutic responses: complete remission, partial remission, and null response. The proposed model also incorporates the effect of covariates, the impact of previous relapses, and heterogeneity among individuals. We use our model to analyse the times to relapse for 63 patients with a particular subtype of indolent lymphoma and compare the results to those obtained using existing methods.