In population-based studies performed by cancer registries, the available data often do not allow the identification of specific causes of death. It is thus difficult to differentiate between the effects of age and gender on the general mortality of the population and on the cancer-related mortality and to completely eliminate confounding due to age and gender. In order to overcome these difficulties, Esteve proposed to include in a single additive model the disease-related mortality hazard l _c and the general population mortality hazard l _e. In our study, the general population mortality hazard is estimated based on published age and gender specific mortality rates in the general French population. The disease-related mortality hazard is estimated from the data at hand, assuming a proportional hazards model, similar to the Cox model. As a consequence the relative survival model proposed by Esteve imposes the proportional hazard assumption for all covariates so that for each covariate the estimated effect b _j is assumed to remain constant. The only difference with the Cox model is that although in the Cox model there is no restriction on the baseline hazard, in the Esteve 's model the baseline hazard is expressed as a step function of the follow-up time t. Accordingly, t _k represents the baseline mortality hazard in the k-th time segment.