Confidence Intervals for Measures of Interaction

Abstract

Cigarette smoking has been associated with adenomatous polyps of the large bowel but not with increased risk of colorectal cancer. Giovannucci et al recently proposed a hypothesis to explain this inconsistency. A key testable aspect of the hypothesis is that smoking in the distant past increases the risk of large polyps. Questions also remain about the association between colorectal polyps and consumption of alcohol. To address these issues, we examined data from 488 cases with adenomatous polyps and 488 controls. Subjects were members of a prepaid health plan in Los Angeles who had a sigmoidoscopy in 1991–1993. As expected, the adjusted odds of polyps in current smokers compared with never-smokers was increased [odds ratio = 2.43; 95% confidence interval (CI) = 1.56–3.79]. For those who had smoked in the distant past (for example, 30 or more pack-years before 20 years ago), the adjusted odds of an adenoma ≥1 cm, relative to nonsmokers, was 0.88 (95% CI = 0.23–3.42). The adjusted odds of polyps in those consuming ≥46 gm per day of alcohol compared with nondrinkers was 1.50 (95% CI = 0.72–3.13). Although imprecise, these data do not support the hypothesis that past smoking increases the risk of large polyps, but our results indicate a weak association between alcohol use and risk of adenomatous polyps.