Background: Inconsistent associations between eating frequency (EF) and adiposity may be mainly due to measurement errors of EF.
Objective: This cross-sectional study examined the association of EF with body mass index (BMI) and waist circumference (WC), by focusing on the confounding of energy misreporting and the effect of exclusion of underreporters (URs).
Design: Dietary intake was assessed using a 7-day weighed dietary record in 1487 British adults aged 19-64 years. EF was calculated based on all eating occasions (EF(all)), after excluding those providing no energy (EF(energy)), and after excluding those providing <210 kJ of energy (EF(⩾210 kJ)). Energy misreporting was assessed as reported energy intake divided by estimated energy requirement (EI:EER).
Results: The mean values (1st and 99th percentiles) of EF(all), EF(energy) and EF(⩾210 kJ) were, respectively, 7.8 (3.1, 15.3), 7.2 (2.9, 12.7), and 5.6 (2.3, 10.7) times/day in men and 7.6 (3.0, 13.9), 6.7 (2.7, 12.1), and 4.8 (1.9, 9.1) times/day in women. In the univariate analyses of the entire male population, EF(⩾210 kJ), but not EF(all) and EF(energy), was inversely associated with BMI and WC. After full adjustment (including EI:EER), all three measures of EF were positively associated with BMI and WC. In the univariate analyses of the entire female population, all three measures of EF were inversely associated with BMI and WC. After full adjustment, EF(⩾210 kJ) was positively associated with BMI and WC while EF(all) and EF(energy) showed null associations. When URs (EI:EER <0.665) were excluded, the multivariate analyses showed that EF(all) and EF(energy) were positively associated with BMI in men while EF(⩾210 kJ) was positively associated with BMI and WC in both sexes.
Conclusions: We showed positive associations of EF with BMI and WC. Adjustment for EI:EER and the exclusion of URs, as well as definitions of EF, radically affected the results of the analysis.