School feeding programs for improving the physical and psychological health of school children experiencing socioeconomic disadvantage (Cochrane Review)
Kristjansson E, Dignam M, Rizvi A, Osman M, Magwood O, Olarte D, Cohen JF, Krasevec J, Grover T, Labelle PR, Garner JA, Janzen L, Rossiter S, Dewidar O, Shea B, Welch V, Wells GA · 2025
grade Bmeta-analysisindependentmixed
Sample
40 studies (13 RCTs - 12 cluster-randomized, 1 individually randomized - and 27 non-randomized studies); more than 91,885 students. Meta-analytic estimates rest on 2-6 cluster-RCTs each.
Population
Socioeconomically disadvantaged primary and secondary students aged 5-19. 34 of 40 studies are from low- and middle-income countries. Only one high-income-country study contributed to any outcome, and it was a non-randomized breakfast-club study.
Design
Update of the Kristjansson school-feeding review, searched to Nov 2023 with a Nov 2024 RCT top-up. Eligible interventions provided free or reduced-price school food supplying at least 3% of daily energy and 10% of daily protein, versus no school feeding. RoB 2 for RCTs, SMART for non-randomized studies, GRADE for critical outcomes. The headline estimates below are restricted to cluster-RCTs. Heterogeneity in context, outcome and population is high (I-squared 54-81% on the education outcomes).
Key findings
In LMICs, school feeding improves math achievement slightly (SMD 0.14, 95% CI 0.06 to 0.23, I2 = 68%, 6 cluster-RCTs, 5,587 participants, HIGH-certainty evidence) but may have little to no effect on reading (SMD 0.02, 95% CI -0.06 to 0.11, I2 = 54%, 3 cluster-RCTs, 3,417 participants, low certainty). Enrollment rises slightly (+3.44 pp, 95% CI 0.83 to 6.04, high certainty) but attendance does not (MD 0.17 pp, 95% CI -2.64 to 2.97, low certainty). Height-for-age +0.06 z (moderate certainty) and weight-for-age +0.08 z (moderate certainty). Evidence for high-income countries is essentially absent: a single non-randomized breakfast-club study with very uncertain attendance findings. Equity subgroup analyses by sex and SES were all null and underpowered.
Genetic confound
Low for the RCT-based estimates. But the review is 85% LMIC, so its positive findings speak to populations with real caloric and micronutrient shortfalls and cannot be transported to well-fed children. The review itself identifies almost no high-income-country randomized evidence.
Effects
| Outcome | Metric | Value | Measure | Timing | Vs | Horizon | Class |
|---|---|---|---|---|---|---|---|
| Math achievement, LMIC (domain skills) | SMD | 0.14 (95% CI 0.06 to 0.23), I2 = 68%, 6 cluster-RCTs, n = 5,587, GRADE high | mixed | end of intervention | business-as-usual | end-of-treatment | domain-skill |
| Reading achievement, LMIC (domain skills) | SMD | 0.02 (95% CI -0.06 to 0.11), I2 = 54%, 3 cluster-RCTs, n = 3,417, GRADE low - null | mixed | end of intervention | business-as-usual | end-of-treatment | domain-skill |
| School enrollment (attainment) | MD, percentage points | +3.44 pp (95% CI 0.83 to 6.04), GRADE high | standardized | end of intervention | business-as-usual | end-of-treatment | attainment |
| School attendance (behavioral) | MD, percentage points | 0.17 pp (95% CI -2.64 to 2.97), I2 = 81%, GRADE low - null | standardized | end of intervention | business-as-usual | end-of-treatment | behaviour |
| Height-for-age (health) | MD in z-score | +0.06 (95% CI 0.03 to 0.09), GRADE moderate | standardized | end of intervention | business-as-usual | end-of-treatment | health |
| Weight-for-age (health) | MD in z-score | +0.08 (95% CI 0.05 to 0.12), GRADE moderate | standardized | end of intervention | business-as-usual | end-of-treatment | health |
| High-income-country educational outcomes | any | one non-randomized breakfast-club study, very uncertain; no RCT evidence | unknown | n/a | unclear | unclear | behaviour |
Cited by
- Breakfast, school meals, and micronutrients — what feeding children actually buysmixedconf: mediumgc: low