Effect of outdoor activity on myopia onset and progression in school-aged children in northeast China - the Sujiatun Eye Care Study

3,051 students in the primary-outcome analysis (1,735 intervention, 1,316 control) out of 3,521 eligible (13.3% attrition/exclusion, balanced across arms); cycloplegic subsample n = 391 (214 intervention, 177 control), about 12.8% of the whole sample

Grades 1-5 (aged 6-11) and grades 7-8 (aged 12-14) in urban and rural Sujiatun District, Shenyang, northeast China; 1-year intervention Nov 2012 - Nov 2013

CORRECTED: allocation WAS randomized, contrary to the earlier reading of this record. The methods state "Two nearby primary schools in urban areas with comparable academic level were assigned to the intervention arm or control arm randomly. The two urban junior high schools, two rural primary schools and two rural junior high schools were assigned similarly" - i.e. four matched pairs of schools, each pair randomly allocated. (The abstract's "two primary and two junior high schools" undercounts what the methods paragraph describes.) Registered NCT02271373. The intervention was two extra 20-min compulsory outdoor recesses (the 09:30 and 14:30 breaks stretched from 10 to 30 min), with the school day EXTENDED by 40 min so the outdoor time was added rather than taken from instruction. Why this is still only grade C, per METHODOLOGY's "cluster-randomization errors" clause: only 8 clusters were randomized, and every reported comparison is a plain Student's t-test or chi-square on individual students with NO adjustment for clustering, so all P values are anti-conservative. Arms are markedly unbalanced in size (1,735 vs 1,316) and, more importantly, unbalanced on confounders (see genetic_confound). The whole-sample primary outcome was UNCORRECTED visual acuity on a Chinese standard chart, a poor proxy for refractive development and sensitive to effort and testing conditions; cycloplegic refraction - the outcome that matters - came only from a 12.8% subsample resting on 23 total incident myopia cases (8 vs 15). The authors also concede a contamination risk: intervention and control children could mix outside school.

Mean uncorrected visual acuity was better in the intervention group after 1 year (P < 0.001, repeated-measures MANOVA); the between-group difference was not yet significant at 6 months. In the cycloplegic subsample, new myopia onset was 3.70% vs 8.50% (8 vs 15 children, P = 0.048), refractive change -0.10 +/- 0.65 D/year vs -0.27 +/- 0.52 D/year (P = 0.005), axial elongation 0.16 +/- 0.30 vs 0.21 +/- 0.21 mm/year (P = 0.034), and IOP change -0.05 +/- 2.78 vs 0.67 +/- 2.21 mmHg/year (P = 0.006, which the authors themselves call not clinically significant). NULLS the earlier extraction omitted: the END-OF-STUDY refractive error did NOT differ between arms (-0.93 +/- 1.50 D vs -1.13 +/- 1.67 D, P = 0.202) - only the one-year change did; and corneal curvature and anterior chamber depth changes showed no significant differences. No achievement, cognitive or attainment outcome was collected.

Medium. CORRECTED: parental refraction IS reported, and it is imbalanced - the intervention group had MORE myopic parents (30.3% vs 24.2%, P < 0.001 whole sample; 33.6% vs 23.2%, P = 0.023 in the cycloplegic subsample), plus higher parental education and higher family income (both P < 0.001). All three are established risk factors for childhood myopia, so on the hereditarian reading these imbalances load the dice AGAINST the intervention arm and the protective result is conservative with respect to them - the authors make this argument themselves. The residual worry is the other direction: with only 8 clusters, randomization plainly failed to balance the arms, so other unmeasured catchment differences are equally likely to be present in either direction, and no covariate-adjusted analysis is reported to check.

Directionally consistent with the randomized outdoor-time trials it cites and is cited alongside - Guangzhou (GOAL, Morgan/He) found a ~25% reduction in new myopia cases and axial elongation of 0.29 vs 0.33 mm/year, and the Taiwan recess-outdoors trial found 8.41% vs 17.65% myopia onset. This trial's 3.70% vs 8.50% sits in the same direction with a larger relative reduction (~56%) on far fewer events.