Statistical problems in measuring cancer prevalence

Cancer prevalence is an important measure of the burden of the disease in a population and on the health care system. it is usually defined as the proportion of diseased people of age x, at a given calendar time. The main source of data to estimate cancer prevalence are cancer registries (CRs). One of the main problems in estimating prevalence from such data is due to insufficient lenght of the observation period. Incident cases diagnosed before the beginning of CR recording activity, and still alive at the time of interest, are lost to the collection of prevalence. A method for correcting this bias was proposed, consisting on the estimation of a completeness index, based on modelled incidence and survival by age and time since diagnosis. Different patterns of the index by age and periods of observation, are discussed for selected cancer sites, and for possible variations in survival and incidence levels. The purpose of this work is to investigate the correction factor reliability when applied to different scenarios of prognosis. Applications to Italian data from CRs of Varese and Parma, and to the data from Connecticut Tumor Registry are briefly reported and discussed, to show the applicability of the index.