Background: Smoking is associated with lower body weight, but an increased risk of diabetes in some studies. Because smoking may increase waist-to-hip ratio (WHR), a risk factor for diabetes, we postulated that the relation between smoking and diabetes may be mediated in part by smoking-associated differences in body fat distribution.
Methods: We conducted a cross-sectional analysis of baseline data from 9,435 elderly nonblack women enrolled in the Study of Osteoporotic Fractures. Data were collected by Self-report and physical examination. Linear and logistic models were used to determine the relation of smoking to WHR and prevalence of self-reported diabetes.
Results: Current and past smokers had greater WHRs compared with never smokers. In multivariate models that adjusted for body mass index, the prevalence of diabetes was lower among smokers of < or = 10 cigarettes/day [odds ratio (OR) = 0.57, 95% confidence interval (CI) 0.31-1.03] and higher among smokers of > 10 cigarettes/day (OR = 1.38, 95% CI 0.99-1.92) compared with never smokers. The relation of smoking > 10 cigarettes/day to prevalence of diabetes was slightly attenuated after further adjustment for WHR.
Conclusions: Smoking-associated differences in WHR may mediate, at least in part, the prevalence of diabetes among smokers of > 10 cigarettes/day. The decreased prevalence of diabetes that we observed among smokers of < or = 10 cigarettes/day was not explained by differences in obesity and requires confirmation.