Prediction of Cycloplegic Refractive Error based on Non-Cycloplegic Measurements in Chinese Children aged 3 to 17 years

Abstract Introduction Cycloplegic refraction remains the gold standard for pediatric refractive assessment, yet its implementation faces challenges in large-scale studies. This study aimed to develop a multivariable model predicting cycloplegic spherical equivalent refraction (SER) using non-cycloplegic measurements in Chinese children. Methods This retrospective study included 28901 children aged 3–17 years, who were allocated to development (n = 23121) and validation (n = 5780) datasets. Ocular biometric parameters were assessed with the IOL Master 500 optical biometer. Cycloplegic SER was measured using a Topcon autorefractor after 1% cyclopentolate. A prediction model was derived and validated, with performance evaluated by R2, correlation coefficient, mean error (ME), mean absolute error (MAE), and clinical accuracy proportions (predictions within ± 0.50 D/±1.00 D). Results The prediction model, incorporating age, gender, height, axial length, and axial length/corneal radius of curvature ratio, achieved R2 = 0.854 (development) and 0.858 (validation), with ME = 0.00 ± 0.72 D and MAE = 0.56 ± 0.46 D in the validation dataset. Clinical accuracy proportions (± 0.50 D/±1.00 D) were 53.86% and 84.91%, respectively. Optimal performance was observed in children aged 3–10 years and those with cycloplegic SER between + 0.50 and + 2.00 D. Conclusions This prediction model based on simplified non-cycloplegic parameters provides reasonably accurate cycloplegic SER estimates in young children with low refractive error. It holds potential for epidemiological surveys in resource-limited settings and clinical management of children with cycloplegia contraindications. Clinical trial number not applicable.