Reproducibility of a meal-based food frequency questionnaire. The influence of format and time interval between questionnaires.

OBJECTIVE: To compare the reproducibility and reported level of energy intake obtained using three versions of a meal-based food frequency questionnaire (FFQ) considering, firstly, the influence of FFQ format and, secondly, the influence of time interval between questionnaires.

DESIGN: The study was conducted in two parts. In the first, subjects were randomly allocated to three groups (one for each FFQ) and sent, by mail, the same FFQ to complete on two separate occasions with a time interval of 4-6 weeks. In the second study, the reproducibility of one of the FFQs used in the initial study was compared, after an interval of 3 months, in a further group of people, matched as far as possible for gender, age and socio-economic status of area of residence.

SUBJECTS: The study population was 651 supermarket shoppers from Geelong, a regional centre with a population of 150,000 situated in Victoria, Australia, who had previously responded to an in-store survey about meat purchasing patterns. Of the 651 shoppers, 144 women and nine men (38% of those eligible) in part I and 98 women and two men (45% of those eligible) in part II of the study satisfactorily completed a FFQ on both occasions.

RESULTS: While there were few statistically significant differences in terms of mean nutrient intake and nutrient density between the three FFQ formats, all under-estimated energy intake relative to the minimum estimated energy requirements for a sedentary population. A significant decrease in reported intake of approximately 10% was also observed, regardless of FFQ format used, when the same questionnaire was completed a second time after an interval of 4-6 weeks. In contrast, when the time interval between questionnaires was increased to 3 months, there were few significant differences in intake between the first and second administrations. Moreover those changes in food intake which were significantly different after the longer interval were, in general, consistent with expected seasonal changes in food intake patterns.

CONCLUSION: Under the conditions of our study differences in FFQ format appeared to have less effect on estimates of mean intake than the length of the time interval between questionnaires. Our results suggest that motivation to complete a FFQ is significantly diminished on the second occasion, when the interval between FFQs is only 4-6 weeks and to a lesser extent when it is 3 months. Researchers planning studies which aim to assess short-term changes in food intake by means of a FFQ, for example after an intervention programme, need to be aware of this effect and to determine its magnitude, by assessing the reproducibility of their FFQ over the relevant time-interval prior to the proposed intervention and by including an appropriate non-intervention comparison group in the design of their study if seasonal effects are likely to occur in the course of the study.