HARVARD COLLEGE, PRESIDENT & FELLOWS OF

Large prospective studies of diet and cancer have often not supported hypotheses that were based on animal and ecological findings; one possible explanation could be large measurement errors in the assessment of diet. Thus, the National Cancer Institute (NCI) has requested proposals for a series collaborative validation/calibration studies imbedded in existing cohorts using a common design with multiple self-reported and biomarker measures of diet and physical activity including Doubly Labeled Water (DLW), urinary nitrogen (UN) and motion monitors. Thus, we propose to conduct such a study among 750 women in the Nurses? Health Study (NHS) and Nurses? Health Study II (NHSII), which will be of particular value because these cohorts have provided much of the data that have not supported several prominent hypotheses. Also, in these studies we have already collected extensive dietary data using multiple measures of intake and plasma biomarkers of diet; this will allow the evaluation of the validity of dietary assessment methods over a 20-year period. The proposed study design will consist of two Food Frequency Questionnaire (FFQ) administered one year apart; in the year between these measures we will collect four 24-hour recalls using the new NCI automated self-administered 24-hour recall (ASA24), two 7-day dietary records, and biomarkers of diet including one assessment of energy expenditure using doubly-labeled water (DLW, with a replication in 100 women), four 24-hour urinary measures of nitrogen (a measure of protein intake), sodium and potassium, and two fasting blood samples for plasma levels of specific fatty acids, standard lipids, carotenoids, and folate. Physical activity will be assessed twice (together with the FFQ?s) by the modified Paffenbarger physical activity questionnaire used in our studies, four 24-hour recalls using an automated system developed by the NCI, twice by an objective physical activity monitor, and by DLW (subtracting resting metabolic rate). These multiple measurements will allow an evaluation of the error structure associated with dietary and physical activity assessment methods, in particular the influences of variation over time for each method and correlated errors between methods. This will provide information of the relative and absolute validity of the different measures, and inform the interpretation of the published literature on diet and cancer (allowing the correction of observed relative risks and confidence intervals for measurement error) and the design of future validation/calibration studies. Additional analyses will indicate the optimal approaches to adjust nutrient intakes for total energy intake, which is critical for the interpretation of epidemiologic findings.

Clarification of Codes