AIMS AND BACKGROUND: Cancer prevalence in a population, defined as the proportion - or the number - of people who were diagnosed with a cancer during their lives and are still alive at a given date, is a crucial indicator for heath care planning and resource allocation. Long-term population-based cancer registries (CR) are the appropriate tools to produce prevalence figures, which, however, are scarcely available. This paper contains a review up to 1999 of the published data world-wide (reports and articles) on cancer prevalence: including measured and estimated figures. MATERIALS AND METHODS: Data on cancer prevalence from CRs are available for the Nordic countries, Connecticut, and Italy. In addition, electronic data are available for the European Union (EU). Data for the Nordic countries were first published in the mid-seventies, reporting the prevalence for 1970. The first data from Connecticut were available 10 years later. Estimates for all EU countries were published by the International Agency for Research on Cancer (IARC) in 1997. In Italy, observed and estimated data on the prevalence of respiratory and digestive tract cancer and breast cancer have been published during the nineties, followed by a systematic analysis for all cancers in 1999. By using information obtained from CRs, cancer prevalence data were calculated directly (observed prevalence) by means of incidence and follow-up information on individual cancer patients, or indirectly (estimated prevalence) by means of mathematical models, which generally use epidemiological information at the aggregate level. RESULTS: Cancer prevalence for all cancers combined (proportions per 100,000 inhabitants) showed values of less than 700 in males and less than 800 in females in 1970 (Finland) to over 2,300 in males and over 3,000 in females in 1992 (Italian registries). With few exceptions, in each country and period considered the cancer sites contributing most to cancer prevalence are lung, colon-rectum, prostate and bladder in males, colon-rectum, breast, uterus (both cervix and corpus) and ovary in females. At present, comparison of measurements from different areas is difficult because there exists no standardized mode of presentation. CONCLUSIONS: In spite of their being potentially useful for health care planning, prevalence data have been produced inconsistently and late by cancer registries, at least in comparison with the systematic availability of incidence and survival statistics. The available data can be compared only to a limited extent due to differences in completeness, in the choice of indicators, in the standard populations, and in the frequency of publication. It would be desirable that in the future data will be produced systematically, with a higher level of standardization compared to the past, and, most importantly, on the same geographic and administrative scale as health-care decision-making.