Subar et al. Respond to "A Further Look at Dietary Questionnaire Validation" and "Another Perspective on Food Frequency Questionnaires"

Abstract

We thank Drs. Willett (1) and Block (2) for their thoughtful comments. Our goal in creating the Diet History Questionnaire was to improve the accuracy of food frequency questionnaires. Table 7 of our paper (3) indicates that, in comparing food frequency questionnaires with recalls, we were successful but that the improvement was small. Although we should continue to improve food frequency questionnaires, we agree with Dr. Willett that perhaps there is a ceiling on their ability to accurately measure diet. For surveillance and public health, knowledge of absolute intakes is essential, as indicated by Dr. Block. Such information allows evaluation of progress toward meeting national dietary recommendations. Dr. Willett asserts that, for nutritional epidemiology, energy adjustment is essential because of the need to reflect dietary composition. Are we sure, however, that energy-adjusted nutrients are the only appropriate biologic exposures for nutritional epidemiology and that every nutrient exposure necessarily relates to energy intake? Using energy-adjusted nutrients allows one to assess the effect of substituting one nutrient in relation to another. We agree that this substitution effect is important. However, so too is the addition effect, which represents changes in absolute intakes. Knowledge of both the substitution and addition effects contributes to a more complete understanding of diet-disease associations (4). Further, although energy adjustment improves correlations between food frequency questionnaires and recalls/records, there is a wealth of literature showing that energy is poorly measured in all self-reported dietary instruments, including recalls/records (5). Therefore, how can we be sure that energy adjustment really controls for true energy intake and that estimated associations between disease incidence and energy-adjusted nutrients do not suffer from residual confounding? Studies like the National Cancer Institute's Observing Protein and Energy Nutrition Study, a dietary biomarker-based validation study with doubly labeled water and urinary protein, sodium, and potassium, will provide data to address this question.