BACKGROUND: In case-control studies, selection of an appropriate group of controls is a critical step which may affect the outcome of the analysis. METHODS: We studied the differences in reproductive, lifestyle, and anthropometric variables between controls from a hospital-based study and controls from a screening clinic. Odds ratios for breast cancer were calculated using the two types of controls, and the impact on the resulting odds ratio was studied. RESULTS: Some interesting differences in odds ratios obtained with the two sets of controls were found. Among premenopausal screening subjects, the odds ratio for breast cancer did not change across quartiles of body mass,while among hospital subjects, the risk of breast cancer significantly decreased with increasing body mass. For liquor and beer consumption, a three-fold increase in breast cancer risk was observed among premenopausal hospital subjects, whereas no association was found among screening subjects. Among postmenopausal women, a significant decrease in breast cancer risk with ovariectomy was only observed in the hospital-based study. CONCLUSIONS: Our results suggest that in examining the association between body mass, alcohol consumption, or ovariectomy and breast cancer risk, the choice of control group used should be carefully considered.
|Bulletin du Cancer
|Published - Jun 2004