Development, validation and reproducibility of a food frequency questionnaire to measure flavonoid intake in older Australian adults

AIM: To develop and assess the validity and reproducibility of a food frequency questionnaire (FFQ) to measure total flavonoid intake, and individual flavonoid subclasses, in older adults. METHODS: Retrospective analysis of flavonoid intake in older adults informed the development of a FFQ to measure flavonoid intake and determine the flavonoid subclasses consumed (anthocyanins, flavan‐3‐ols, flavones, flavonols and flavanones). Older adults (n = 42, mean age 75.3 ± 8.6 years) attended two interviews 1 month apart where anthropometrics (height and weight), blood pressure (BP), demographic data and a 93‐item self‐administered FFQ were collected. A 4‐day food record (FR) was randomly administered between the two interview dates, and each food item was assigned a flavonoid and flavonoid subclass content using the United States Department of Agriculture flavonoid database. The criterion validity and reproducibility of the FFQ was assessed against a 4‐day FR using the Wilcoxon signed‐rank sum test, Spearman's correlation coefficient (r), Bland‐Altman Plots and Cohen's kappa. RESULTS: Total flavonoid intake was determined (median intake FFQ = 919.3 mg/day, FR = 781.4 mg/day). Tests of validity indicated that the FFQ consistently overestimated total flavonoid intake compared with the 4‐day FR. There was a significant difference in estimates between the FFQ and the 4‐day FR for total flavonoid intake (Wilcoxon signed‐rank sum P < 0.001; Bland‐Altman plots indicated large bias and wide limits of agreement), but they were well correlated (Spearman's r 0.93, P < 0.001; Cohen's kappa κ = 0.619, P < 0.001). For individual flavonoid subclasses, the tests of validity indicated greater discrepancy compared with 4‐day FR. The FFQ showed high reproducibility for estimating total flavonoid intake (FFQ1vsFFQ2: Wilcoxon signed‐rank sum test, P > 0.05; Spearman's r 0.91, P < 0.001; Bland‐Altman plots visually showed small, non‐significant bias and wide limits of agreement; and Cohen's kappa κ = 0.619, P < 0.001), with a small mean percentage difference (6.7%). For individual flavonoid subclasses, the tests of reproducibility between FFQ1 and FFQ2 showed similarly high reproducibility. CONCLUSIONS: The developed FFQ appears suitable for satisfactorily ranking individuals according to total flavonoid intake. The FFQ shows limitations for estimating absolute total flavonoid intake and intake of flavonoid subclasses in comparison to a 4‐day FR in terms of overestimating intake. Refinement and further validation of this tool may be required.