Exploring Relative Survival

Project Team

With survival analysis, the most widely-used model is Cox's (1972) proportional hazard model. However, cancer control organizations and researchers using cancer registry data most often fit relative survival models.  These models estimate survival in a hypothetical world where cancer is the only cause of mortality.  This language arouses curiosity.  We wondered whether it was used in the original 1961 cancer monograph and whether it is used now as simple shorthand because at no time in the past or in the future will we experience a world where cancer is the only cause of mortality.  Cancer control agencies probably use the resulting estimates because it permits them to compare control effects across time, countries and groups.  However, an unrealistic hypothetical world is one thing and an idealized world is another.  This project asks whether it is feasible to use a counterfactual analysis in this area of research.