Fitting portion sizes in a self-administered food frequency questionnaire.

For epidemiological studies, a simple semiquantitative FFQ was developed to assess the frequency of intake of food items demonstrated with graphically displayed portion sizes. As a validation study, a random sample of 393 participants in the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study completed 2 unannounced 24-h dietary recalls (24HDR) and the FFQ during 1 y. To calculate food and nutrient intakes, we compared the use of fitted portion sizes with the use of predefined portion sizes. Fitted portion sizes were calculated by summing food intakes over the 2 24HDR and dividing the sum by the frequency of intake reported in the FFQ for each FFQ food item, leading to similar mean intakes for FFQ and 24HDR. As predefined portion sizes, amounts that had been used in previous dietary assessments in EPIC-Potsdam were used. Mean intake of 12 food groups was 102% for men or women with fitted portion sizes and 79% for men and 95% for women with predefined portion sizes of intake measured with 2 24HDR. However, deattenuated, energy-adjusted correlation coefficients between FFQ and 24HDR were not better for 19 nutrients by the use of fitted portion sizes, with a mean correlation coefficient of 0.53 for men and 0.56 for women. Mean correlation coefficients for food groups also were similar for fitted and predefined portion sizes. Fitting portion sizes using recent reference data from a random sample of study participants improved the quantitative assessment of food and nutrient intake, but not ranking of study participants, compared with predefining portion sizes based on prior knowledge.