Refractive Errors and Educational Outcomes in Children - A Systematic Review

25 studies (16 cross-sectional or other non-randomized analytical, 4 cohort, 5 randomized), selected from 765 records / 165 full texts assessed

Preschool and school-aged children and adolescents across Asia, Europe, North America and Africa; PubMed, Embase and Scopus searched from inception to January 2025

Narrative systematic review with no quantitative pooling (the authors cite heterogeneity in exposure definitions, outcome measures and reporting as precluding meta-analysis). Twenty of the 25 included studies are non-randomized, so the review's overall conclusion rests mostly on confounded observational work. Its own risk-of-bias appraisal is weak and it says so: formal GRADE "was considered; however, overall certainty was summarized narratively"; 5 of 25 studies were judged high risk of bias, 1 moderate-to-serious, 1 low-to-some-concerns, and the remaining 18 "some concerns" - i.e. not a single study was rated cleanly low risk. Because nothing is pooled, every number in this record is a narrative restatement of a primary study, not an independent estimate.

Reports that uncorrected hyperopia in preschoolers is consistently associated with poorer early literacy and attention outcomes, and that observational school-based studies link reduced habitual visual acuity to lower attainment. On the randomized evidence it reports gains of approximately 0.1 SD in some cluster-randomized trials (Wang 2017, Ma 2014, rural China) and notes effects were sensitive to adherence, timing of outcome assessment, implementation intensity and outcome measurement - which independently corroborates the effect magnitude found in the primary trials. TWO CAUSAL NULLS the earlier extraction omitted, both cutting against the review's own headline: Bruce 2016, a UK cluster randomized trial, "found improved spectacle wear but no clear improvement in overall academic attainment"; and Jiang's mendelian randomization "did not support a clear causal effect of moderate hyperopia on years of education", which the authors present as a caution "against interpreting all observational associations as direct causal effects". Note also that the review does NOT report a fadeout pattern - an earlier version of this record claimed it did. What it reports is cross-sectional sensitivity to when the outcome was assessed and to how long follow-up ran, which is not the same claim and must not be cited as independent corroboration of fadeout.

High for the review as a whole. Sixteen of 25 studies are cross-sectional; children with uncorrected refractive error in these samples differ from their peers in family income, parental education and health-seeking behaviour, all of which carry passive gene-environment correlation with achievement. The authors themselves name confounding as the main risk of bias. The included mendelian randomization study is the one design in the review that addresses this directly, and it returns a null.

Downgraded from `replicated` on full-text read. The review's own randomized subset does not agree with itself: Wang 2017 and Ma 2014 (cluster RCTs, rural China) and Nie 2020 (randomized field experiment) report positive academic effects, but Bruce 2016 (UK cluster trial) found improved spectacle WEAR with no clear improvement in academic attainment, and Jiang's mendelian randomization did not support a causal effect of moderate hyperopia on years of education. Three positive randomized results, one randomized null and one causal-design null is `mixed`, not `replicated`.