Submitted to: American Journal of Epidemiology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: May 25, 2005
Publication Date: August 24, 2005
Citation: Paul, D.R., Rhodes, D.G., Kramer, M.H., Baer, D.J., Rumpler, W.V. 2005. Validation of a food frequency questionnaire by direct measurement of habitual ad libitum food intake. American Journal of Epidemiology 162:1-9 Interpretive Summary: The Food Frequency Questionnaires (FFQ) is one method commonly used to assess habitual food intake. One limitation with them is the known measurement error that is commonly reported. In order to reduce measurement error, investigators apply correction factors based on total energy intake or expenditure. The problem with that approach is that the "true" habitual intake is generally not known, so the degree of measurement error and the effect of the correction is difficult to understand. In this study, we allowed 12 men to consume meals ad libitum for 16 weeks, then asked them to report what they ate based on the FFQ. As expected, there was significant measurement error. The application of correction factors did not improve the measurement in all cases. It appears as though the best way to correct FFQ data in this population is to use an individualized correction factor based on energy expenditure for carbohydrate and protein intake, and an individualized correction factor based on energy expenditure and body weight for fat intake. Scientists will benefit from this study because it will provide the basis for research focusing on correction factors that improve the interpretation of food intake data. Once the appropriate correction factors have been developed, nutritionists and the public will benefit from the improved data gleaned from FFQs.
Technical Abstract: Background: Food Frequency Questionnaires (FFQ) are a common technique used to assess habitual food intake. Although FFQs are thought to be susceptible to measurement error, the amount of error and the effectiveness of correction factors are unknown. Objective: To detect measurement error in FFQ predicted macronutrient intakes, and quantify the effectiveness of correction factors based on energy expenditure and/or body weight. Design: Twelve men consumed an ad libitum, cafeteria-style diet for 16 weeks. At the end of the study period, subjects reported their food intakes with a FFQ (Block 98). Subjects were also dosed with doubly labeled water (DLW) and maintained physical activity logs (PA). Results: Absolute and relative macronutrient intakes were not well predicted by the FFQ. Although not statistically significant, the application of single, group mean correction factors (for DLW or PA) tended to reduce the variance in carbohydrate intake and increase the variance in fat and protein intakes (all P > 0.05). The application of individual subject corrections reduced the variance for carbohydrate and protein intakes (P < 0.05). The only procedure that significantly reduced the variance in fat intake was the combination of body weight and DLW (P < 0.05). Conclusions: Although the results of this study may only be applicable this relatively small and heterogeneous group of subjects, absolute and relative macronutrient intakes were not well predicted by the FFQ. However, the application of different correction factors based on energy expenditure and/or body weight tend to reduce measurement error.