al., 1984;Ferraroni et al., 1989;Bidoli et al., 1992) and cohort studies (Pukkala and Teppo, 1986;Vagero and Persson, 1986;Leon, 1988) showed predominantly positive associations between SES and colon cancer risk. In these studies hardly any adjustment was made for potential confounders. Most of the time, age was included in the analyses, and two studies included some lifestyle characteristics such as smoking (Williams and Horm, 1977) and coffee and alcohol consumption (Ferraroni et al., 1989 (Goldbohm et al., 1994a). Therefore, these dietary factors were omitted from our analyses. Materials and methodsThe cohort studyIn September 1986, The Netherlands Cohort Study (NLCS), investigating various lifestyle variables, sociodemographic indicators and cancer risk, was started. A detailed description of the cohort study design has been reported elsewhere (Van den Brandt et al., 1990a). Briefly, the cohort included 58 279 men and 62 573 women aged 55-69 years at the beginning of the study. The study population originated from 204 municipal population registries throughout the country. Data were collected by means of a self-administered questionnaire. For data analysis the case-cohort approach was used in which cases are derived from the entire cohort, while the personyears at risk are estimated from a random sample of 3500 subjects (subcohort). After the baseline exposure measurement the subcohort was randomly sampled (1688 men and 1812 women) and it has been followed up biennially for vital status information.Follow-up for incident cancer has been established by record linkage with all regional cancer registries in The Netherlands and with a national pathology register (PALGA). The method of record linkage has been described previously (Van den Brandt et al., 1990b). The analysis is restricted to colon cancer incidence in the period from September 1986 to December 1989. In this period, completeness of follow-up was estimated to be 95% ( Van den Brandt et al., 1993). After these 3.3 years of follow-up, 351 colon cancer cases were detected. We excluded self-reported prevalent cancer cases other than skin cancer (n = 28), cases with in situ carcinoma (n = 8), cases without microscopically confirmed diagnosis (n = 2) and sarcoma (n = 1). Therefore 312 incident cases (157 males and 155 females) were available for analysis. Self-reported prevalent cancer cases other than skin cancer were also excluded from the subcohort, with the result that 3346 subjects (1630