Factors associated with the calibration of a food frequency questionnaire for the Bambuí Project, Brazil.

BACKGROUND: The present study aimed to calibrate a food frequency questionnaire (FFQ) using one 24-h recall (R24) associated with foodstuff replicas. METHODS: Calibration was performed by linear regression. Of a probability sample of 150 individuals (>18 years) of the town of Bambui, 98 completed the study. RESULTS: The mean age of the participants was 43.8 years and 63.3% were women. The calibration of the FFQ included sex for the estimation of iron, protein and zinc; and an interaction between sex and R24 was significant for vitamin A and cholesterol models. Age contributed to the vitamin C, polyunsaturated fatty acids, proteins and zinc models, and an interaction between age and R24 was significant for protein and zinc. CONCLUSIONS: Calibration coefficients and many interactions in linear regression models revealed important differences between dietary methods, indicating the complexity encountered when combining two methodologies for estimating food intake.