Biases and adjustments in nutritional assessments from dietary questionnaires

In nutritional epidemiology, it is essential to use Food Consumption Assessment Methods that have been validated and accepted by the international community for estimating food consumption of individuals and populations. This assessment must be made with the highest quality possible so as to avoid, as far as possible, sources of error and confusion in the processes. The qualities that are required in a measurement method are validity and accuracy; validity being the main factor. Lack of validity produces biases, or systematic errors. These can reside in the process of subject selection, or processes of information gathering where the lack of accuracy produces random errors. For many nutrients, the intra-individual variances are due to many factors such as day-of-the-week or season, and could create problems in the data analyses. Adjustments are needed to minimize these effects. Confounding factors may over- or under-state the real magnitude of the observed association, or even alter the direction of the real association. Total energy intake can be a confounding variable when studying a relationship between nutrient intake and disease risk. To control for this effect several approximations are proposed such as nutrient densities, standard multivariate models and the nutrient residual model.