A comparison of food habit and food frequency data as predictors of breast cancer in the NHANES I/NHEFS cohort.

We compared two methods of assessing dietary fat and breast cancer incidence in the first complete follow-up of the National Health Epidemiologic Follow-up Study (NHEFS) cohort. Between 1982 and 1984, 6156 women aged 32-86 y completed the NHEFS survey, which included a 93-item food frequency questionnaire (FFQ). In addition, women answered questions regarding food habits, such as choice of salad dressing, trimming fat from meat, and eating skin on poultry. In the 4 y of follow-up, these women contributed a total of 23,949 person years, during which 53 women reported a breast cancer diagnosis. The rate ratio (RR) and 95% confidence interval (CI) for each quartile of percentage of energy from fat were 1.0, 0.96 (0.5-2.1), 1.0 (0.5-2.2) and 0.98 (0.5-2.1). Thus the breast cancer rates for women in the upper three quartiles, who reported a diet with higher than 30% of energy from fat, were not different from those of women in the lowest quartile of intake (< or = 29.4% energy from fat). In contrast, the "high-fat" response to three of the four food habit questions identified women at increased risk of developing breast cancer: women who used salad dressings other than low fat had a RR and 95% CI of 1.3 (0.7-2.3), women who reported eating the skin on poultry had a RR and 95% CI of 1.7 (0.9-2.9), and women who did not use lean or extra lean ground beef had a RR and 95% CI of 2.2 (1.2-4.0). These food habit questions may be less subject to misclassification than dietary information of fat intake derived from the FFQ. Further investigation is needed to evaluate what is measured by the food habit questions.