This is the first population-based study carried out in a southern European region to evaluate the risk of a cohort of cancer patients for developing further cancers. The Tuscany Tumour Registry, the Ragusa Cancer Registry and the Cancer Registry of Romagna, three of the 14 population-based cancer registries active in Italy, were involved in the present study. Overall, 19,252 incident cases of cancer of the female breast, and of the colon, rectum, lung and stomach were followed-up for 48 358.3 person-years. Only second metachronous cancers were considered. Controlateral breast cancers were analysed separately. Multiple primaries (MPs) were defined according to the IACR-IACR rules. The observed (O) numbers of MPs were compared with those expected (E) from age-, sex- and registry-specific incidence rates. Overall, 463 MPs were diagnosed (O/E = 0.87, P < 0.001). The O/E ratios for cancers of the colon (O/E = 0.66), rectum (O/E = 0.72) and all sites combined (O/E = 0.78) in males were significantly lower than expected. The deficit of observed MPs was significant during the first period (2-12 months) and increased over time. Patients over 65 years of age had a significant lower risk of MP, whereas young cancer patients had significantly higher risks for all cancers and for female breast cancer. Male lung cancer patients had a significantly reduced O/E ratio for stomach cancer (O/E = 0.21). Rectal cancer patients had reduced risks of developing stomach cancer and tumours of all sites combined and a 3-fold increased risk of kidney cancers. Colon cancer patients had an overall reduction in risk of MPs, but female colon cancer patients had a significantly increased risk for tumours of the ovary and small intestine; no significant results were found for primary stomach cancers. Female breast cancer patients had a significantly increased risk of rectal cancer (O/E = 1.97), and when synchronous and bilateral breast cancers were considered, significant overall increases in risk were seen for all cancer sites (O/E = 1.6) and for rectal (O/E = 2), and especially for breast cancers (O/E = 3). The cohort analysed had a lower risk of developing further independent tumours than the general population. Several artefacts may have biased these results: the exclusion of synchronous cancers greatly reduced the overall MP risk, and the age-related differences may have been due to reduced medical surveillance and diagnostic aggressiveness. We have confirmed the increased risk for kidney cancers in rectal cancer patients and the association between cancers of the colon and ovary. The significantly increased risk for rectal cancer in female breast cancer patients is probably due to hormonal and dietary factors. For female breast cancer patients, controlateral breast cancer represented the highest risk. The increased risk of cancer of the small intestine in patients with colon cancer may be due to overdiagnosis within increased medical surveillance.