Cancer mortality was analysed in 3282 elderly subjects aged ≥ 65 years from 2 cohorts of general population having different life-style patterns. They took part in the CASTEL (CArdiovascular STudy in the ELderly), a 12-year lasting prospective Italian study. The aim of the present analysis was to identify the items able to influence cancer mortality. A biochemical profile and a questionnaire on lifestyle were collected. Continuous items were averaged and compared with analysis of variance, frequencies with the Pearson's χ2 test. Mortality was recorded yearly for 12 years from the Registrar's Office and causes of death double-checked by consulting medical case sheets and family doctors' files. The influence of items on mortality was evaluated with the Cox multivariate analysis. Relative risk (RR) of each item was adjusted for confounders. Age, gender, tobacco smoking, the presence of respiratory symptoms, low body mass index in males, serum alanine transaminase (ALT) and alkaline phosphatase (ALP), as well as the town of residence, were powerful predictors of cancer mortality. In the entire population, 12-year overall mortality was 49.4%, cardiovascular 22.8%, and neoplastic 11%; the latter was higher in males than in females (15.7% vs. 7.9%, p < 0.00001). In subjects with respiratory symptoms neoplastic mortality was 11.6% (RR: 1.47) vs. 9.7% in those without symptoms (p < 0.01). Subjects with very low cholesterol (≤ 178 mg/dl), those with high uric acid (≥ 8.7 mg/dl) and males with low body mass index (≤ 22.7 kg/m2) has an increased risk of cancer mortality. RR of cancer mortality increased with increasing ALT or ALP. It was ∼1 in those having ALT and ALP between 9 and 41.2U/I, 1.41 in those exceeding this latter level and < 1 in those below 9U/I. RR of ALP had a similar trend, the best protective cut-off value being <106 and the worst one > 177U/I. When both serum enzymes were simultaneously raised, RR of cancer mortality increased to 2.84.