Estimating net survival using a life-table approach

Cancer registries are often interested in estimating net survival (NS), the probability of survival if the cancer under study is the only possible cause of death. Pohar Perme, Stare, and Esteve (2012, Biometrics 68: 113-120) proposed a new estimator of NS based on inverse probability weighting. They demonstrated that existing estimators of NS based on relative survival were biased, whereas the new estimator was unbiased. The new estimator was developed for continuous survival times, yet cancer registries often have only discrete survival times (for example, survival time in completed months or years). Therefore, we propose an approach to estimation for when survival times are discrete. In this article, we describe the stnet command for life-table estimation of NS, adapting the Pohar Perme estimation approach to life-table estimation. Estimates can be made using a period or hybrid approach in addition to the traditional cohort (or complete) approach, and age-standardized survival estimates are available.