To compare, specifically by age group, proxy-reported food group estimates obtained from the food frequency section of the Children's Eating Habits questionnaire (CEHQ-FFQ) against the estimates of two non-consecutive 24 h dietary recalls (24-HDR).

Estimates of food group intakes assessed via the forty-three-food-group CEHQ-FFQ were compared with those obtained by a computerized 24-HDR. Agreement on frequencies of intakes (equal to the number of portions per recall period) between the two instruments was examined using crude and de-attenuated Pearson's correlation coefficients, cross-classification analyses, weighted kappa statistics (κw) and Bland–Altman analysis.

Kindergartens/schools from eight European countries participating in the IDEFICS (Identification and prevention of Dietary- and lifestyle-induced health EFfects In Children and infantS) Study cross-sectional survey (2007–2008).

Children aged 2–9 years (n 2508, 50·4 % boys).

The CEHQ-FFQ provided higher intake estimates for most of the food groups than the 24-HDR. De-attenuated Pearson correlation coefficients ranged from 0·01 (sweetened fruit) to 0·48 (sweetened milk) in children aged 2–<6 years (mean = 0·25) and from 0·01 (milled cereal) to 0·44 (water) in children aged 6–9 years (mean = 0·23). An average of 32 % and 31 % of food group intakes were assigned to the same quartile in younger and older children, respectively, and classification into extreme opposite quartiles was ≤12 % for all food groups in both age groups. Mean κw was 0·20 for 2–<6-year-olds and 0·17 for 6–9-year-olds.

The strength of association estimates assessed by the CEHQ-FFQ and the 24-HDR varied by food group and by age group. Observed level of agreement and CEHQ-FFQ ability to rank children according to intakes of food groups were considered to be low.