Predicting the onset of myopia in children and adolescents using artificial intelligence: A systematic review and meta-analysis.

BACKGROUND

Artificial intelligence (AI) models have been used to predict myopia onset, but results are inconsistent. This systematic review and meta-analysis aimed to evaluate their performance.

METHODS

We searched PubMed, EMBASE, Web of Science, and Cochrane Library up to January 2026 for studies using AI to predict myopia onset. A random‑effects model pooled area under the curve (AUC). Risk of bias was assessed with PROBAST+AI, and evidence certainty with GRADE.

RESULTS

Seven studies (24 models) were included. At one‑year follow‑up, the pooled AUC for the seven studies (optimal models) was 0.95 (95%CI:0.93-0.96, I²=89.2%), and for all 24 models it was 0.85 (95%CI:0.82-0.89, I²=99.2%). Traditional machine learning (pooled AUC=0.85) and deep learning (0.86) showed comparable discrimination, while prospective studies (0.96) outperformed retrospective ones (0.89). Heterogeneity was high in all analyses. GRADE quality was "low".

CONCLUSIONS

AI models show promise for predicting adolescent myopia, but current studies suffer from high bias risk, lack of external validation, and regional imbalance. Their clinical utility remains uncertain, and future multicenter, externally validated cohort studies are needed.