Dietary inflammation and childhood adiposity: Analysis of individual participant data from six birth cohorts

Background & aims

Childhood adiposity and inflammation impact long-term health. However, associations between dietary inflammation and childhood adiposity are unclear. We investigated if more pro-inflammatory diets are associated with greater adiposity in early-, mid-, and late-childhood.

Methods

We pooled individual participant data (IPD) from 13,978 children in six European birth cohorts in the ALPHABET consortium: Lifeways Cross-Generation Cohort Study (Lifeways), the Randomised cOntrol trial of LOw glycaemic index diet during pregnancy study (ROLO), the Avon Longitudinal Study of Parents and Children (ALSPAC), the Southampton Women's Survey (SWS), the Polish Mother and Child Cohort (REPRO_PL), and The Generation R Study (Generation R). Dietary inflammation was determined using the Children's Dietary Inflammatory Index (C-DII™). Adiposity-related outcomes included BMI z-score (primary outcome), abdominal circumference, skinfolds, fat-mass- and fat-free-mass-indices (secondary outcomes). Two-stage random effects IPD meta-analysis (IPD-MA), with adjusted linear and logistic regression models, was conducted. Quantile regression (QR) examined C-DII associations with BMI z-score percentiles.

Results

Median, 25th and 75th percentile C-DII scores trended upwards from early 0.18 (−0.65, 1.03) to late-childhood 0.51 (−0.40, 1.49). Pooled QR revealed positive C-DII associations across BMI z-score percentiles, particularly in late-childhood unadjusted β (95 % CI) 75th (0.075 (0.046, 0.105), p < 0.001); 85th (0.077 (0.045, 0.108), p < 0.001); and 95th (0.051 (0.011, 0.091), p = 0.01). Adjusted cohort-specific QR identified contrasting associations at early-childhood (ALSPAC and SWS) and late-childhood (Generation R). Pooled adjusted IPD-MA showed C-DII associations with late-childhood obesity [OR (95 % CI) 0.89 (0.81, 0.97), p = 0.01].

Conclusions

C-DII associations across BMI z-score distribution varied by cohort, quantile, and time-point, with some potentially explained by adiposity rebound, reverse causation and questionnaire response biases, highlighting insights not evident with linear regression.