A comparison of expected number of deaths with the observed is a widely used method with an extensive history. Epidemiologists often study the mortality rate from a particular cancer in an initially healthy population. Using cause-specific mortality rates from a reference population will overestimate the expected number of deaths because it does not take into account the fact that the study cohort is initially cancer free. It is more accurate to calculate the expected number of deaths by explicitly considering the probability of getting cancer in a given year and the probability of dying from it before the end of the follow-up.