点击显示 收起
Lipid Metabolism Laboratory, Jean Mayer USDA Human Nutrition Research Center on Aging, Tufts University, 711 Washington Street, Boston, MA 02111, E-mail: eschaefer{at}hnrc.tufts.edu
Dear Sir:
Food-frequency questionnaires are useful in ranking individual intakes to describe relations between diet and disease in large epidemiologic studies. The use of food-frequency questionnaires to estimate absolute nutrient intakes is frequent in research, although food-frequency questionnaires have not been validated against an appropriate standard (1–4). Validation of food-frequency questionnaires (comparison with an observed diet of chemically defined composition for 1 y) is not possible given the expense of such an undertaking and the difficulty in finding subjects to participate in such a study. Thus, we chose to investigate the efficacy of the food-frequency questionnaire during a controlled metabolic study in which biochemical evaluation (serum lipid profiles) was concurrent (5). We provided subjects with 3 diets composed of common and recognizable foods, but that varied greatly in diet composition, in 6-wk intervals. Triplicate samples of each diet were chemically analyzed to provide the most accurate standard by which to compare or validate food-frequency intake estimates. We then asked our subjects to report their dietary intakes by using the Willett food-frequency questionnaire at the end of each 6-wk period, with instructions to base responses according to their intake during this 6-wk period.
Willett pointed out that our methods were "artificial." Although our methods did not simulate a free-living situation, we argue that in real life many persons consume meals in restaurants and do not know how foods are prepared. Furthermore, it is typical for one person in a family to be responsible for food preparation, whereas other members remain unaware of the meal preparation or type of ingredients used. Moreover, the reference standard for assessing the validity of any diet instrument must be, in our view, a chemically defined diet. Such an assessment has never been carried out by Willett with his instrument.
Willett also pointed out that subjects were provided all of their food, and that this does not replicate a subject's usual involvement in developing knowledge of his or her long-term dietary pattern. We were not asking individuals to report their long-term dietary intakes, but rather to report on the diet that they were consuming over the previous few weeks. Finally, our subjects were not blinded to the study phase and were aware of the fat content being altered in each phase of the study. They were aware that items were reduced in total fat but were unaware of the types of fats used in cooking. Because the subjects were given their food, this should have been a best-case scenario for the food-frequency questionnaire. We know of no more valid method than assessing an instrument in comparison with a chemically defined diet.
We agree that dietary assessment tools, whether they involve food-frequency questionnaires or food diaries, are subject to reporting biases. In our study, however, the blood lipid profiles of the subjects were consistent with their actual dietary intakes and were accurately assessed by 3-d food records. Our subjects may have been free of bias in their reporting because they were given all of their food during the study and their body weight and blood lipids reflected compliance with the diets provided. Our subjects had no reason to falsely report their dietary intakes.
We agree that the subjects were asked to report on their dietary intake over a 6-wk period as opposed to a 1-y period, which is the recommended time for using the food-frequency questionnaire. We also agree that our sample was small (19 subjects), but we believe that studying few subjects carefully is more valuable than is studying many subjects with a nonefficacious instrument. Our findings indicate that food-frequency questionnaires do not accurately estimate absolute dietary intakes. We found that our 3 diets, composed of very different amounts of total fat (15%, 26%, and 35% of energy), were estimated to be very similar by the food-frequency questionnaire (22%, 27%, and 28%, respectively, of energy). The food-frequency questionnaire could not reliably distinguish a high-fat diet from a low-fat diet in our study. Our data are consistent with the view that 3-d food records provide a much more accurate assessment of dietary macronutrient intake than do food-frequency questionnaires, which in our view can be highly misleading.
REFERENCES