Estimating and Modeling Relative Survival

When estimating patient survival using data collected by population-based cancer registries, it is common to estimate net survival in a relative-survival framework. Net survival can be estimated using the relative-survival ratio, which is the ratio of the observed survival of the patients (where all deaths are considered events) to the expected survival of a comparable group from the general population. In this article, we describe a command, strs, for life-table estimation of relative survival. We discuss three methods for estimating expected survival, as well as the cohort, period, and hybrid approaches for estimating relative survival. We also implement a life-table version of the Pohar Perme (2012, Biometrics 68: 113–120) estimator of net survival, and we describe two methods for age standardization. We also explain how, in addition to net probabilities of death, crude probabilities of death due to cancer and due to other causes can be estimated using the method of Cronin and Feuer (2000, Statistics in Medicine 19: 1729–1740). We conclude this article with discussion and examples of modeling excess mortality using various approaches, including the full-likelihood approach (using the ml command) and Poisson regression (using the glm command with a user-specified link function).