Causal effects of later-eating rhythm on adiposity in children through the comparison of two cohorts in the UK and China: a cross-cohort study

Later-eating rhythm (LER) refers to a later timing, greater energy intake, and higher meal frequency in the evening. The role of childhood LER in obesity development is emerging, but most evidence is cross-sectional. Cross-context comparison allows the improvement of causal inference in observational studies by comparing cohorts with different confounding structures. This method is applied to assess the causal effects of LER on adiposity, by exploring the likelihood of residual confounding due to socioeconomic status.In this cross-cohort analysis, we used ongoing birth cohort data from the UK Avon Longitudinal Study of Parents and Children (ALSPAC) established in 1991, and the nationally representative China Health and Nutrition Survey (CHNS) collected in 1989-2011. Children with available data at age 7 years were eligible. We applied indices of inequality for assessing confounding structure by regressing LER/adiposity on the standardised score of socioeconomic status (SES) in each cohort. We used multivariable linear and binary logistic regressions to model cross-sectional and prospective associations between LER at 7 years of age and body-mass index (BMI) at ages 7 and 9 years in both cohorts. Analyses were adjusted by age, sex, ethnicity, residency, and socioeconomic status. We used a p value for the Cochrane Q-test obtained from meta-analysis to test for heterogeneity between cohorts.We analysed data from 4019 children (2170 [54·0%] female; 1849 [46.0%] male) in ALSPAC and 1749 (788 [45·1%] female; 961 [54.9%] male) in CHNS. The associations between SES and LER or adiposity differed between ALSPAC and CHNS (SES and energy intake for evening main meal: b=1·81 [95% CI 0·81 to 2·81] vs -3·02 [-4·76 to -1·27]; SES and frequency of evening snacks: odds ratio [OR]=0·51 [95% CI 0·41-0·63] vs 5·71 [3·54-9·22]; SES and BMI: b=-0·42 [-0·65 to -0·18] vs 1·29 [0·75 to 1·84]). Positive associations between frequency of evening snacks and BMI were seen in both cross-sectional and longitudinal analyses in both cohorts (mean change of BMI with 1 day increase of consuming evening snacks b=0·09 [0·02 to 0·15]; 0·13 [0·03 to 0·22] kg/m2 per day in ALSPAC, and b=0·11 [-0·07 to 0·28]; 0·30 [0·07 to 0·52] kg/m2 per day in CHNS). No associations were found for energy intake. p values for heterogeneity ranged from 0·107 to 0·932.Both cohorts showed consistent results despite varied dietary cultures and SES patterning of LER or adiposity. Energy intake in the evening or night was not associated with adiposity, whereas evening snacking was. More recent, high-quality cohorts are warranted to enhance the strength of the conclusions.None.