Development and validity of a short web-based semi-quantitative Food Frequency Questionnaire applicable in both clinical and research setting: an evolution over time

Background: Assessing dietary intake is valuable both in clinical practice and in research. In research and in clinical practice, long-term habitual dietary intake is most often of interest. Therefore, a web-based semi-quantitative Food Frequency Questionnaire (FFQ) was developed to measure habitual intake of nutrients and foods.

Aim: This study aimed to assess content validity, convergent validity, and reliability of a 32-item semi-quantitative FFQ for adults.

Methods: A total of three different cohorts of Flemish adults were recruited in the past 10 years. The first cross-sectional validation study took place in 2013, consequently in 2019 and 2021. Content validity was assessed in 2019 through a semi-structured cognitive interview. Convergent validity was assessed by examining mean differences, Wilcoxon signed-rank test, Spearman's correlation coefficients (SCC), and Bland-Altman analysis for energy, nutrient, and food group intake compared with a 3-day food record (FR). Additionally, consumers-only analysis was performed together with cross-classification analysis by assessing the ranking capabilities of the FFQ into quartiles and weighted kappa. Reliability was assessed through the evaluation of SCC and intra-class correlation (ICC) of test-retest assessment of the FFQ.

Results: Spearman's correlation coefficient (SCC) for energy and absolute nutrient intake between the FFQ and the FR ranged from 0.02 to 0.54. Compared with absolute macronutrients, higher SCC was found for the majority of the relative macronutrient intake and most food groups. Bland-Altman plots showed improved agreement and decreasing bias between the FFQ and the FR over time. Misclassification of the FFQ for nutrients was acceptable and decreased over time (7.4, 7.5, and 6.8% in 2013, 2019, and 2021, respectively), but weighted kappa remained mostly fair (κ ≤ 0.20). The reliability of the FFQ was good and improved over time (mean SCC of 0.65 and 0.66 p <0.001 in 2013 and 2019).

Conclusion: The short web-based FFQ is an easy, low-cost, and feasible tool with good reliability, low misclassification, and acceptable validity to compare nutrient densities and food group intake at the population level. The measurement of absolute intake remains debatable.