The effect of sugar on behavior or cognition in children: a meta-analysis
Wolraich ML, Wilson DB, White JW · 1995
grade Bmeta-analysisindependentreplicated
Sample
16 reports meeting inclusion criteria, yielding 23 within-subject design studies
Population
Children, including samples selected for attention-deficit disorder and for parent-reported sugar sensitivity.
Design
Meta-analysis restricted to studies that (1) administered a known quantity of sugar, (2) used an artificial-sweetener placebo, (3) blinded subjects, parents AND research staff, and (4) reported statistics permitting effect-size computation. That inclusion filter makes this a meta-analysis of blinded placebo-controlled challenge experiments, which is why it grades B rather than C. Effects were computed separately across 14 measurement constructs rather than pooled into one number. The within-subject design and the small per-study samples are the main limitations.
Key findings
Across all 14 measurement constructs the mean effect sizes ranged from -0.14 (direct behavioural observation) to +0.30 (academic tests), and the 95% confidence interval for EVERY one of the 14 included zero. Sugar does not affect the behaviour or cognitive performance of children. The authors attribute the strength of parental belief to expectancy and common association. They explicitly do not rule out a small effect or an effect in some subset of children. This is a genuine debunking and one of the cleanest in the nutrition-and-behaviour literature.
Genetic confound
None - blinded within-subject challenge experiments; each child is their own control.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Behaviour and cognition, all 14 measurement constructs | weighted mean effect size per construct | range -0.14 to +0.30; 95% CI includes zero for all 14 | mixed | acute, post-challenge | business-as-usual | end-of-treatment | behaviour |
| Direct behavioural observation | weighted mean effect size | -0.14, CI includes zero | researcher-designed | acute | business-as-usual | end-of-treatment | behaviour |
| Academic tests | weighted mean effect size | +0.30, CI includes zero | mixed | acute | business-as-usual | end-of-treatment | domain-skill |
Cited by
- Breakfast, school meals, and micronutrients — what feeding children actually buysmixedconf: mediumgc: low