Effects of iron supplementation on cognitive development in school-age children: Systematic review and meta-analysis

13 RCTs in the systematic review, 12 in the meta-analysis; five databases searched to April 2021 and re-run to October 2022. Pooled samples by domain - intelligence n = 3,105 (11 studies); attention and concentration n = 339 (3 of 5 studies); memory n = 718 (5 studies); school achievement n = 2,272 (5 studies).

School children aged 6-12. Ten of thirteen trials are from middle-income countries (India, Indonesia, Thailand, Egypt, South Africa), one from a low-income country (Mali), and two from high-income countries (Croatia, UK); the included trials are predominantly from settings with high baseline anaemia prevalence. Subgroup analyses separate children anaemic, non-anaemic, iron-deficient and iron-replete at baseline - this is the applicability-critical contrast.

Systematic review and meta-analysis restricted to RCTs of iron supplementation in 6-12 year olds with cognitive outcomes; random-effects pooling in RevMan 5.4; Cochrane RoB2 used for risk of bias (six trials high RoB, four some concern, three low). Domains pooled separately - intelligence, attention and concentration, memory, and school achievement. Graded C rather than B because the constituent trials are small, mostly unregistered, mostly single-site, use heterogeneous outcome batteries, and three of thirteen had no placebo control at all; heterogeneity on the headline intelligence pool is I2 = 89%. THREE THINGS THE HEADLINE NUMBERS HIDE, all of them in the paper and all of them absent from our original extraction. (1) RISK OF BIAS DESTROYS THE INTELLIGENCE RESULT. Pooled across everything, iron moves intelligence 0.46 SD; restricted to studies at low or moderate risk of bias it is 0.25 SD (95% CI -0.05 to 0.54, P = 0.10), and the authors state flatly that "there was no evidence of effect of iron supplementation on the intelligence of school-age children in studies with low RoB." Per METHODOLOGY effect-size discipline the bias-corrected estimate is the one to report. (2) IT ALSO DOES NOT SURVIVE REMOVING CO-SUPPLEMENTS - excluding the trials that gave iron alongside folic acid or vitamin C leaves 0.20 SD (-0.04 to 0.44, P = 0.10). (3) LONGER SUPPLEMENTATION IS WORSE, NOT BETTER - trials running four or more months give 0.21 SD (-0.02 to 0.43, P = 0.07) while the short trials carry the significant effect, which is the fadeout signature rather than a dose-response. PUBLICATION BIAS WAS PROMISED AND NEVER REPORTED: the Methods say "publication bias was assessed using Funnel plot asymmetry and Eggen [sic] test", and no funnel plot, Egger statistic or p-value appears anywhere in the paper - a material reporting gap given the small-study composition and I2 = 89%. INTERNAL INCONSISTENCIES between abstract and Results section, noted so future readers do not chase them: attention and concentration is 0.44 (0.07 to 0.81), P = 0.02 in the abstract but 0.45 (0.09 to 0.80), P = 0.01 in Results; memory is 0.44 (0.21 to 0.67) in the abstract but 0.45 (0.04 to 0.69) in Results (the abstract CI is the one consistent with its own point estimate); school achievement is 0.06 (-0.15 to 0.26), P = 0.56 in the abstract but 0.00 (-0.21 to 0.21), P = 1.00 in Results. Both versions of the achievement result are null, so nothing downstream turns on it.

Iron supplementation improves intelligence-test scores (SMD 0.46, 95% CI 0.19 to 0.73, P < 0.001), attention and concentration (SMD 0.44-0.45, 95% CI 0.07 to 0.81, P = 0.02) and memory (SMD 0.44, 95% CI 0.21 to 0.67, P < 0.001) - but has NO effect on school achievement (SMD 0.06, 95% CI -0.15 to 0.26, P = 0.56; 0.00, -0.21 to 0.21 in the Results section). THE HEADLINE IS A DEFICIENCY EFFECT AND A BIAS EFFECT, NOT A NOOTROPIC ONE. Children ANAEMIC AT BASELINE gain markedly: intelligence SMD 0.79 (0.41 to 1.16, P = 0.001, I2 = 40%) and memory SMD 0.47 (0.13 to 0.81, P = 0.006, I2 = 0%). Children NOT anaemic at baseline gain nothing at all: intelligence SMD 0.01 (-0.10 to 0.12, P = 0.89, I2 = 0%) and memory SMD -0.02 (-1.01 to 0.97, P = 0.97); iron-replete children 0.09 (-0.10 to 0.29, P = 0.35). And the pooled intelligence effect collapses to non-significance among low- and moderate-risk-of-bias trials (0.25, -0.05 to 0.54, P = 0.10), among trials without folic acid or vitamin C co-supplementation (0.20, -0.04 to 0.44, P = 0.10), and among trials lasting four months or more (0.21, -0.02 to 0.43, P = 0.07). The effect is a deficiency correction that shows up on IQ-style tests, does not survive bias adjustment in unselected samples, and never reaches the outcome schools actually care about.

None internally - randomized placebo-controlled trials. The anaemia gradient (0.79 SD in anaemic children, 0.01 SD in non-anaemic children with I2 = 0%) is the strongest single piece of evidence in this topic that the mechanism is repletion of a deficiency rather than nutritional enhancement.

The meta itself has not been re-run by another group, but its two load-bearing results are echoed elsewhere in this archive and in the literature it cites. Falkingham et al. 2010 likewise found no effect of iron on scholastic ability regardless of baseline anaemic status, and Fiani 2025 (nutr-fiani-2025-iron-non-anemic-meta) reaches the same conclusion from the other direction - the apparent cognitive effect in non-anaemic samples disappears once iron-deficient participants are removed. The anaemic-versus-non-anaemic gradient is therefore replicated; the pooled headline number is not.