Objective: For proper interpretation of results from epidemiological studies that use food-frequency questionnaires (FFQs), it is necessary to know the relationship between reported intakes from the FFQ and true usual intake. In this paper, we report a calibration study conducted to investigate the performance of the FFQ used in a cohort study, the Canadian Study of Diet, Lifestyle and Health.
Methods: Over a 1-year period, 151 men and 159 women completed a full set of questionnaires including a self-administered baseline FFQ, three 24-hour diet recalls administered by telephone, and a second FFQ self-administered subsequently. The association between the nutrient estimates derived from the FFQs and the diet recalls was evaluated by calculating deattenuated Pearson's correlation coefficients.
Results: FFQs estimated mean daily nutrient intakes higher than the diet recalls. When the log-transformed and energy-adjusted nutrient intakes from the average of three 24-hour recalls were compared against the baseline FFQ, the following deattenuated correlations were obtained in men and women, respectively: total energy 0.44 and 0.32, total fat 0.64 and 0.68, saturated fat 0.68 and 0.70, dietary fibre 0.65 and 0.44, vitamin E 0.32 and 0.37, vitamin C 0.40 and 0.37, beta-carotene 0.34 and 0.29, alcohol 0.74 and 0.67, caffeine 0.81 and 0.76, with a median correlation of 0.49 and 0.53. Correlations between the second FFQ and diet recalls were similar. The correlations between the two FFQs as a test of reliability had a median value 0.64 for men and 0.63 for women for selected nutrients.
Conclusions: The study suggests that the FFQ method gives acceptable levels of nutrients or food component estimates, as assessed by this calibration study against diet recalls, when limited to energy-adjusted and deattenuated values.