Sheep and goats: separating cervix and corpus uteri from imprecisely coded uterine cancer deaths, for studies of geographical and temporal variations in mortality

Analysing time trends in mortality from cancers of the cervix and corpus uteri using routine data sources (such as the World Health Organistion mortality database) involves two major problems: deaths certified as "uterus, unspecified site", and the presence of a combined category comprising unspecified and corpus uteri cancer deaths. To avoid misleading interpretations, the unspecified and the misclassified data must be incorporated into the analysis to produce rates that allow meaningful comparisons between populations and over time. Reallocation methods based on age- and time-specific distributions of cervix and corpus uteri cancer are applied to the unspecified deaths, while for those in the combined category, the age- and time-specific distributions of unspecified and corpus uteri cancer are considered. Adjustments of the general strategies for reallocation were developed to take into account the different quality of the data. Results from eight European countries with different degrees of coding precision are presented. The reallocation methods bring the cervix and corpus uteri mortality trends more in line with the trends for countries with more precise data while keeping the country-specific characteristics. In addition, the methods ensured the availability of time trends for corpus uteri cancer in women age 50 years and older, which were completely missing without reallocation. We propose generally applicable reallocation methods that allow valid time trend analysis of cervix and corpus uteri cancer mortality using datasets of varying precision. Our results show that any sensible analysis of time trends must involve procedures for correcting for unspecified and misclassified uterine cancer deaths. The modified data are available at http://www-dep.iarc.fr/hmp/reallocation.htm. (C) 2004 Elsevier Ltd. All rights reserved.