Cancer prevalence in Italian regions with local cancer registries.
OBJECTIVE: To provide estimates and projections of cancer incidence and prevalence for those Italian regions whose population is partially covered by a cancer registry (CR) and to determine to what extent local CRs can be considered representative of the region, thus improving the potential of the information provided by CRs.
METHODS: A statistical method, MIAMOD (mortality-incidence analysis model), was used to estimate regional cancer incidence and prevalence from regional cancer mortality data and patient survival data recorded by the cancer registries. Estimates of the cancer incidence and prevalence in the various regions have thus been obtained for a number of major cancer sitas. A first and important step in validating the regional estimates has been the comparison of the MIAMOD estimates in the areas covered by the cancer registries with empirical incidence and prevalence observed by CRs, in order to assess the consistency in data, methods and assumptions. Empirical prevalence has been calculated by counting patients with a diagnosis of cancer who were alive on the reference date by PREVAL method. A correction factor has been applied to include patients diagnosed before the period of activity of the registry.
RESULTS: General consistency was found between empirical and estimated (by MIAMOD) incidence and prevalence in the registry areas, which is indicative of the quality and the completeness of all data involved as well as the appropriateness of model choices. The prevalence of all cancers combined for Italian regions with CRs was estimated and projected to the year 2000 as ranging between 1,240 per 100,000 in Sicilia and 2,781 in Emilia-Romagna for men, while for women these figures were 1,765 in Sicilia and 4,019 in Liguria. Comparison of cancer prevalence in CR areas with regional estimates shows quite good consistency for Piemonte, Liguria and Lombardia, which means that the local CRs (of Torino, Genova and Varese, respectively) are representative of their respective regions. Prevalence in Emilia-Romagna appears to be rather well represented by only one, the Parma CR, of the three local CRs. The southern Italian registries of Latina and Ragusa recorded a lower cancer prevalence than was actually estimated in their respective regions.
DISCUSSION: Cancer registries with a longer period of activity showed better agreement between empirical and estimated figures due to the more precise information provided, particularly regarding survival and incidence trends. In conclusion, this work shows the potential of the cancer registries not only to represent their population with respect to cancer morbidity but also as an invaluable tool to extrapolate this information to the larger areas they represent.