Efficacy of 0.01% atropine for myopia control in children: an artificial intelligence-assisted multivariate Bayesian meta-analysis

作者：Like Zhang, Xiao Chen, Xiujing Deng, Xiaobing Wang, Rui Chen · 发表于：Frontiers in Medicine · 年份：2026 · DOI：10.3389/fmed.2026.1752902 · 研究领域：Ophthalmology and Visual Impairment Studies、Corneal surgery and disorders、Retinopathy of Prematurity Studies

Background: Low-concentration atropine (0.01%) has been widely investigated as a safe intervention for myopia control in children, yet its true efficacy remains uncertain due to inconsistent trial findings and high heterogeneity. Methods: An artificial intelligence (AI)-assisted systematic review and Bayesian multivariate meta-analysis of randomised controlled trials (RCTs) was performed to evaluate 0.01% atropine in children. An AI pipeline was employed for literature screening, deduplication and structured data extraction. The primary outcomes were annualised changes in spherical equivalent refraction (SER, D/year) and axial length (AL, mm/year). Bayesian joint models synthesised SER and AL effects, explored heterogeneity with meta-regression and assessed the probability of clinically meaningful benefits. Results: The 17 RCTs included demonstrated that 0.01% atropine significantly reduced AL (mean deviation [MD] = 0.04 mm/year; 95% credible interval [CrI]: -0.02-0.06), whereas the effect on SER was modest and heterogeneous (MD = 0.07 D/year; 95% CrI: -0.04 to 0.18). Meta-regression indicated attenuated efficacy in older children, in cohorts with longer baseline AL and in more recent or Western trials. The probability of achieving clinically meaningful thresholds (≥0.25 D/year SER; ≥0.10 mm/year AL) was 68 and 72%, respectively. Sensitivity analyses confirmed robustness of results, though small-study bias was likely. Conclusion: The results revealed that 0.01% atropine confe...