Relative validity of a computer-based semi-quantitative FFQ for use in the Pelotas (Brazil) Birth Cohort Studies

To assess the relative validity of a FFQ developed for the Pelotas Birth Cohort Studies. Participants completed a ninety-two-food-item FFQ and then answered two 24-h recalls (24HR), one in-person interview and a second one by telephone, administered 14-28 d apart. Median and relative differences of energy, fifteen nutrients and eleven food groups were estimated based on the FFQ and the average of two 24HR. Nutrients were log-transformed and energy-adjusted using residual method. Validity was assessed by crude, energy-adjusted and de-attenuated Pearson and Lin's concordance correlation coefficients. Agreement of quartiles and weighted κ were performed. Differences in energy and nutrient estimations between methods were plotted in Bland-Altman graphs. Pelotas, southern Brazil. Two hundred fifty-four participants randomly selected from the 1993 Pelotas Birth Cohort during the 22-year follow-up (2015). The FFQ overestimated energy and most nutrients and food groups compared with the two 24HR. Energy-adjusted and de-attenuated Pearson correlation coefficients ranged from 0·21 to 0·66. The highest energy-adjusted and de-attenuated concordance correlation coefficients were observed for Ca (0·48), niacin (0·32), Na (0·29), vitamin C (0·28) and riboflavin (0·25). The percentage of nutrients classified into the same and opposite quartiles ranged from 36·5 to 60·3 %, and from 4·8 to 19·1 %, respectively. Weighted κ was moderate for Ca (0·51), beans and legumes (0·50) and milk and dairies (0·49). The FFQ provides a reasonable dietary intake assessment for habitual food consumption. However, the relative validity was weak for specific nutrients and food groups.