Background: Our objective was to determine the possible influence that different designs of a food frequency questionnaire might have on food, energy and nutrient intake estimates.
Methods: A population-based survey included 6783 women, 40-70 years old, living in central Sweden. Using a factorial study design, we compared eight different types of questionnaire covering combinations of three factors: increasing/decreasing frequency categories; addition of portion sizes; and addition of non-dietary questions. All questionnaires included the same list of 60 food items. One of the eight questionnaires was mailed to each subject according to a random assignment. The overall response rate was 77%.
Results: Compared with increasing frequencies, decreasing order of frequency categories entailed 3-11% higher estimates of mean intake for 7 of 14 food groups, 4% higher estimates for energy and 3-6% higher estimates for 13 of 18 nutrients. Addition of portion sizes had heterogeneous effects, both on dietary items (e.g. from -30% decrease for eggs to +76% increase for coffee) and on calculated nutrients (from -7% for beta-carotene to +19% for vitamin C). The inclusion of some additional non-dietary questions did not influence the estimated mean intake of any food or nutrient.
Conclusions: The results of this study have implications for the design of questionnaires and for pooled analyses in nutritional epidemiology, when different food questionnaires are used.