Validity and reproducibility of a semi-quantitative food frequency questionnaire in Spanish preschoolers — The SENDO project

Introduction: currently, it is important to determine whether food frequency questionnaires (FFQ) are valid tools to collect information on usual diet in children. Objective: we evaluated the reproducibility and validity of the semi-quantitative FFQ used in a Spanish cohort of children aged 4-7 years. Methods: to explore its reproducibility, parents filled a 138-item FFQ at baseline (FFQ-0) and then one year later (FFQ-1). To explore its validity, the FFQ-1 was compared with four weighed 3-day dietary records (DRs) that were used as standard of reference. To estimate associations we calculated deattenuated Pearson's correlation coefficients to correct for season-to-season variability, and the Bland-Altman index. We also calculated the weighted kappa index and assessed participant's gross misclassification across quintiles. We analyzed data from 67 (for reproducibility) and 37 (for validity) children aged 4-7 years old, recruited by the pilot study of the SENDO project. Results: regarding reproducibility, we found mean Bland-Altman indexes of 0-10.45 % for nutrients and 1.49 %-10.45 % for foods. The adjusted r ranged between 0.29 and 0.71, and between 0.27 and 0.74 for nutrients and foods, respectively. Regarding validity, we found mean Bland-Altman indexes of 0 %-16.22 % and 0 %-10.81 % for nutrients and for food groups, respectively. The deattenuated r ranged between 0.38 and 0.81 for nutrients, and between 0.53 and 0.68 for foods. The weighted kappa index for agreement across quintiles ranged from 54.1 to 85.1 for nutrients, and from 55.4 to 78.4 for food groups. Conclusions: our results showed acceptable levels of both reproducibility and validity, and that the ad-hoc developed FFQ is a valid tool for assessing usual diet in Spanish preschoolers.