We propose to continue and intensify our ongoing study of diet an cancer being conducted among a cohort of U.S. female registered nurses now aged 38 to 63 years. The dietary aspect of this study began in 1980 when a semiquantitative food frequency questionnarie (SFFQ) was completed by 96,000 participants. In an extensive validation study this questionnarie was found to be reproducible, and correlated reasonably with a series of one-week diet records and selected biochemical parameters. The first component of this proposal entails reassessing dietary intake among cohort members in 1984 and 1986 using an expanded, more detailed, and optically scannable SFFQ. The additional data will reduce misclassification of dietary exposures, whether due to measurement error or to true changes in food consumption over time. In addition, these data will provide exposure information closer in time to the diagnosis of cancer, which is likely to be important for a number of nutritional factors postulated to act late in carcinogenesis. The second component of our proposal is a detailed documentation study among 150 participants who will have completed the revised and expanded SFFQ. Each of these women will complete one-week diet records and provide blood for a wide variety of biochemical measures of nutritional status twice during the first year of the study and again after two years. These data, which will provide estimates of short and long-term variability of dietary intake, will quantify any misclassification associated with our SFFQ and allow us to convert relative intake scores into absolute nutrient intakes. This proposed research will greatly strengthen our ability to test a wide variety of hypotheses relating dietary factors (including total, saturated, and polyunsaturated fat, protein, dietary and crude fiber, carotene, preformed vitamin A, vitamins C and E, cholesterol, coffee, and alcohol) to the incidence of cancers of the breast, colon, and lung and malignant melanoma. After four years of follow-up (assuming 80% power and an alpha error of 0.05) relative risks of 1.4, 1.7, and 1.8 for extreme quintiles of nutrient distributions can be detected for cancers of the breast, colon, and lung respectively. These data will also provide exposure information which may potentially be related to other outcomes, including cardiovascular disease.
Showing the most recent 10 out of 53 publications
