In health services planning, in addition to the basic measures of disease occurrence incidence and mortality, other indexes expressing the demand of care are also required to develop strategies for service provision. One of these is prevalence of the disease, which measures the absolute number, and relative proportion in the population, of individuals affected by the disease and that require some form of medical attention. For most cancer sites, cases surviving 5 years from diagnosis experience thereafter the same survival as the general population, so most of the workload is therefore due to medical acts within these first 5 years. This article reports world-wide estimates of 1-, 2–3- and 4–5-year point prevalence in 1990 in the population aged 15 years or over, and hence describes the number of cancer cases diagnosed between 1986 and 1990 who were still alive at the end of 1990. These estimates of prevalence at 1, 2–3 and 4–5 years are applicable to the evaluation of initial treatment, clinical follow-up and point of cure, respectively, for the majority of cancers. We describe the computational procedure and data sources utilised to obtain these figures and compare them with data published by 2 cancer registries. The highest prevalence of cancer is in North America with 1.5% of the population affected and diagnosed in the previous 5 years (about 0.5% of the population in years 4–5 and 2–3 of follow-up and 0.4% within the first year of diagnosis). This corresponds to over 3.2 million individuals. Western Europe and Australia and New Zealand show very similar percentages with 1.2% and 1.1% of the population affected (about 3.9 and 0.2 million cases respectively). Japan and Eastern Europe form the next batch with 1.0% and 0.7%, followed by Latin America and the Caribbean (overall prevalence of 0.4%), and all remaining regions are around 0.2%. Cancer prevalence in developed countries is very similar in men and women, 1.1% of the sex-specific population, while in developing countries the prevalence is some 25% greater in women than men, reflecting a preponderance of cancer sites with poor survival such as liver, oesophagus and stomach in males. The magnitude of disease incidence is the primary determinant of crude prevalence of cases diagnosed within 1 year so that differences by region mainly reflect variation in risk. In the long-term period however different demographic patterns with long-life expectancy in high-income countries determine a higher prevalence in these areas even for relatively uncommon cancer sites such as the cervix. © 2002 Wiley-Liss, Inc.