Breakfast skipping and academic achievement at 8-16 years: a population study in South Australia

28,651 students in grades 5, 7 and 9, aged 8-16, in South Australian government schools; 2021 Wellbeing and Engagement Collection linked to May 2021 NAPLAN, with multiple imputation for missing outcome and covariate data

Australian schoolchildren in a well-fed high-income population. 32.3% reported never skipping breakfast, 57.6% sometimes (1-6 days a week), 10.1% always (never eating breakfast).

Cross-sectional population study linking self-reported habitual breakfast consumption to NAPLAN standardized literacy and numeracy scores, with Poisson regression (log link, robust variance, clustered on school) estimating relative risk of falling below the national minimum standard, adjusted for student, family and community-level covariates. Large and unusually well-measured on the outcome side - NAPLAN is a census-administered national standardized test, not teacher report - but the exposure is a single self-reported item and the design is purely cross-sectional, so reverse causation is unexcluded. Two things sharpen the confounding read. (1) Adjustment matters a great deal: unadjusted RRs are 2.41 (numeracy) and 2.13 (reading) for always-skippers and fall to 1.78 and 1.63 once covariates enter - roughly 45% of the excess risk is explained by MEASURED confounders alone, which makes residual confounding by unmeasured ones the default expectation rather than a stretch. (2) The adjustment set includes self-reported overall health, sadness, worries and sleep quality alongside SES and parental education. Those are plausibly mediators or downstream consequences of the same latent family disorganisation, so the adjusted RR is not a clean confounder-adjusted estimate in either direction. The authors nonetheless conclude that "breakfast consumption plays an important role in academic success" and recommend policy on that basis - which is precisely the inference this database exists to refuse.

Students who always skipped breakfast had elevated risk of low achievement on all five tests, highest for numeracy (adjusted RR 1.78, 95% CI 1.64 to 1.94) and reading (RR 1.63, 95% CI 1.49 to 1.77), then grammar (1.58), spelling (1.37) and writing (1.25). Sometimes-skippers sat at intermediate risk (numeracy 1.36, reading 1.32, grammar 1.32, spelling 1.19, writing 1.14). CORRECTION TO OUR EARLIER READING: covariate adjustment does NOT barely attenuate the association - it removes roughly 45% of the excess risk on the two headline outcomes (numeracy 2.41 to 1.78; reading 2.13 to 1.63; grammar 2.00 to 1.58; spelling 1.74 to 1.37; writing 1.57 to 1.25). The authors read the surviving association as robustness; the more economical reading is that a handful of crude measured covariates absorbed half the gradient in a design that can never measure the heritable and behavioural ones, and the raw contrasts are exactly what a common cause looks like (always-skippers are 46% grade 9 versus 21% of never-skippers, 48% report poor or fair health versus 11%, and 18.9% have a parent with a bachelor degree versus 42.9%).

High. Breakfast skipping in adolescents indexes household routine, parental supervision, chronotype and family disorganisation, all substantially heritable and all independently predictive of achievement. Adjustment for measured SES cannot remove passive gene-environment correlation, no within-family comparison is attempted, and the design cannot exclude reverse causation from disengagement to both skipping and low scores. That measured covariates already absorb about 45% of the excess risk is evidence the confounding channel is live, not evidence it has been closed.

The bare association is replicated many times over - it is the pooled finding of nutr-seura-2025-breakfast-skipping-observational-meta (24 studies, 172,988 participants, OR 2.08). This study adds sample size and outcome quality, not design quality. Note the paper own comparison set is inconsistent - an Australian LSAC analysis using the SAME NAPLAN outcomes found a difference on numeracy only and nulls on reading, spelling, grammar and writing, and a US study of 21,400 children aged 5-15 found no association with standardized reading, maths or science at all.