Reproducibility and relative validity of a semiquantitative food frequency questionnaire in European preschoolers: The ToyBox study

Objectives: The aim of this study was to examine the reproducibility and relative validity of a semiquantitative food frequency questionnaire (FFQ) in assessing food group estimates. Methods: Food group estimates were assessed via a 37-item FFQ and a 3-d food record (FR). Pearson's correlation coefficients for log-transformed values were calculated to assess the reproducibility and Spearman's rank correlation coefficients for log-transformed values were calculated to assess the validity. Kindergartens from six European countries participated in the preparatory substudies of the ToyBox intervention study; data from preschool children 4 to 6 y of age (n =196, reproducibility study; n = 324, validation study) were obtained. Results: In the reproducibility study, positive Pearson's correlation coefficients for single and aggregated food groups ranged from 0.14 for pasta and rice to 0.90 for cooked vegetables. In the validation study, the FR gave higher estimates of 40 of the 50 food items (single and aggregated) examined compared with those obtained from the FFQ, Positive crude Spearman rank correlation coefficients ranged from 0.01 for total beverages (added sugar) and rice to 0.62 for tea. Corrections for the deattenuation effect did not improve observed correlations. Quartiles and tertiles were calculated for a small number of food groups (N =14) owing to zero consumption in the rest of the groups. Conclusions: Moderately good reproducibility and low-moderate relative validity of the FFQ used in preschool children was observed. Relative validity, however, varied by food and beverage group; for some of the "key" foods/drinks targeted in the ToyBox intervention (e.g., biscuits), the validity was good. The findings should be considered in future epidemiologic and intervention studies in preschool children. (C) 2019 Elsevier Inc. All rights reserved.