a1 Nutritional Epidemiology Group, University of Leeds, Leeds, LS2 9JT, UK
a2 Institute of Human Nutrition, University of Southampton, Southampton, SO16 6YD, UK
A review of the literature concerning the design, utilisation and validation of food-frequency questionnaires (FFQ) has been carried out using a semi-systematic approach to obtaining, reviewing and extracting data from articles. Databases were searched from 1980 to 1999. The present review identified 227 validation (from 1980 to September 1999) and 164 utilisation (for 1998 only) studies. A number of design issues have been evaluated through the present review. These include: the need to consider how portion sizes have been described, self-defined giving higher mean correlations; how an FFQ was administered, interviewer-administered giving higher mean correlations for some nutrients; how many items to include on an FFQ, those with the largest number of items having higher correlations. Validation techniques were described. Most validation studies involved comparing an FFQ against another dietary assessment method; only 19 % compared an FFQ to a biomarker. Measurement differences were most commonly assessed by correlation coefficients as opposed to other more appropriate methods. Mean correlation coefficients were highest for Ca and fat, and lowest for vitamin A and vegetables. The utilisation studies showed that FFQ were most commonly used in cross-sectional surveys, with ninety-three of the FFQ being designed to be disease-specific. The present review results were presented to a group of experts and a consensus arrived at concerning the development, validation and use of FFQ. Recommendations derived from the consensus arising from the literature review are presented as an appendix to the present paper.