Abstract
Cancer survival using population-based cancer registration data is usually quantified using age standardized marginal relative survival. The main reason for this is that researchers want to compare survival between population groups with differential other cause mortality rates. However, the interpretation is somewhat awkward; i.e. the probability of surviving if it was impossible to die any cause other than the cancer under study, if the age distribution was that of an common reference population. This awkward interpretation has led to confusion and incorrect interpretation.
I will discuss why relative survival remains the most common measure used in these studies and why other measures have been avoided. I will describe reference-adjusted measures as alternative. Reference adjusted measures aim to enable fair comparison between population groups using all-cause survival and crude probabilities of death by incorporating additional reference expected mortality rates into the estimation process.
I will describe three estimation approaches, non-parametric, using flexible parametric models and some more recent work using pseudo-observations. I will also discuss how choices need to be made regarding what reference mortality rates to use and what age distribution to standardize to.