Reproducibility, relative validity, and calibration of a food-frequency questionnaire for adults in Greater Metropolitan Porto Alegre, Rio Grande do Sul State, Brazil

This study tested the relative validity and estimated calibration factors of a food-frequency questionnaire (FFQ) with 120 food items in 128 adults. We applied one FFQ and three 24h recalls. Validity was tested by the Bland-Altman method, intraclass correlation coefficient (ICC), classification in quartile categorization, and weighted kappa. Calibration factors were estimated by linear regression, with 24h recall of food intake as the dependent variable and the FFQ values as the independent variable. FFQ underestimated the nutrient intake when compared to 24h recall. The energy-adjusted and de-attenuated ICC ranged from 0.19 (zinc) to 0.82 (calcium). 37% of subjects were classified in the same quartile, with mean kappa of 0.28 (zinc: 0.08; calcium: 0.49). The highest calibration factors were 0.48 (calcium), 0.47 (potassium), and 0.44 (carbohydrates), and the lowest was 0.10 (retinol). FFQ showed satisfactory relative validity and calibrated energy and nutrient intake values similar to 24h recall.