We performed a case-control study in Lazio, a region in central Italy characterized by high levels of indoor radon, Mediterranean climate and diet. Cases (384) and controls (404) aged 35-90 years were recruited in the hospital. Detailed information regarding smoking, diet and other risk factors were collected by direct interview. Residential history during the 30-year period ending 5 years before enrollment was ascertained. In each dwelling, radon detectors were placed in both the main bedroom and the living room for 2 consecutive 6-month periods. We computed odds ratios (ORs) and 95% confidence intervals (CIs) for time-weighted radon concentrations using both categorical and continuous unconditional logistic regression analysis and adjusting for smoking, diet and other variables. Radon measurements were available from 89% and 91% of the time period for cases and controls, respectively. The adjusted ORs were 1.30 (1.03-1.64), 1.48 (1.08-2.02), 1.49 (0.82-2.71) and 2.89 (0.45-18.6) for 50-99, 100-199, 200-399 and 400+ Bq/m(3), respectively, compared with 0-49 Bq/m(3) (OR = 1; 0.56-1.79). The excess odds ratio (EOR) per 100 Bq/m(3) was 0.14 (-0.11, 0.46) for all subjects, 0.24 (-0.09, 0.70) for subjects with complete radon measurements and 0.30 (-0.08, 0.82) for subjects who had lived in 1 or 2 dwellings. There was a tendency of higher risk estimates among subjects with low-medium consumption of dietary antioxidants (EOR = 0.32; -0.19, 1.16) and for adenocarcinoma, small cell and epidermoid cancers. This study indicates an association, although generally not statistically significant, between residential radon and lung cancer with both categorical and continuous analyses. Subjects with presumably lower uncertainty in the exposure assessment showed a higher risk. Dietary antioxidants may act as an effect modifier.