Social Disparities in Cancer Survival: Methodological Considerations

Population-based cancer registry data are invaluable tools for cancer epidemiology. However, they present some challenges for the statistical analysis to derive cancer-specific quantities useful for patients, carers and policymakers. In this chapter, we present these challenges as well as some of the statistical methods used and developed in this context, referred to as ‘relative survival’ methods. We start by detailing the main assumption behind the relative survival methods. We present the non-parametric estimation of ‘net survival’, which can be interpreted as the survival probability of cancer patients once other causes of death have been removed. Then, we detail the general principles of fitting hazard-based regression models for modelling the excess mortality hazard. Finally, we present an extension of regression models for the excess mortality hazard when there is a hierarchical structure in the data, which proves to be useful when the main aim is to investigate socioeconomic disparities in cancer survival. We illustrate the methods using real data for 15–75-year-old men diagnosed with a lip–oral cavity–pharynx cancer between 1997 and 2010 in a French region.