There is increasing evidence that dietary nutrient intake is a major determinant in the etiology of many cancers. The accuracy of dietary nutrient intake measurement is compromised by a number of factors, including nutrient estimation error introduced by allowing nutrient concentrations to deviate from known values by pre-specified amounts. The practice at the Nutrition Coordinating Center (NCC), as elsewhere, is to express these pre-specified maximum allowable food composition errors, or tolerances, per 100 grams of food. We hypothesize that nutrient tolerances expressed per food serving as opposed to per 100 grams of food will more clearly reflect the nutritional significance of food composition errors, and thus will substantially improve the accuracy of dietary nutrient intake measurement. To test this hypothesis, the existing and the proposed nutrient tolerance methods will be evaluated by simulating the occurrence of food composition errors for two nutrients of interest to cancer researchers, beta-carotene and dietary fiber, in nationally representative food consumption data. During this simulation, the magnitude of the nutrient estimation errors will be limited using each of the two nutrient tolerance methods, and the resultant impact on accuracy of dietary nutrient intake measurement will be determined. Findings from this research will be useful for evaluating the accuracy of existing food composition data, improving the quality of future data, and contributing to the development of a general model of error in the measurement of dietary nutrient intake.
