The estimates for the five regions of Africa are those based on the sparsest data- there were 22 (of 53) countries with no data at all (Table 2). Nevertheless, the estimates (for all cancers combined) were not derived using the mortality estimates of WHO (which were based on broad cause of death models). This was because these rates were systematically greater than or equal to the incidence recorded by the local cancer registries in urban areas of the few countries where the cancer registry data were considered of good quality. The country-specific cancer incidence rates (and mortality using the 5-year survival method) were therefore based on data reported by local cancer registries that generally cover the capital city or predominantly urban areas. Due to the local circumstances, under enumeration of cancer cases is a distinct possibility,31 particularly in some of the unpublished datasets that were used. Thus, for example, the estimated age-specific mortality rates from lung cancer for African