Corneal radius-specific axial length/corneal radius ratio cut-offs for improved myopia detection in children: a population-based study.

Abstract

To investigate how corneal radius (CR) affects the optimal axial length-to-corneal radius (AL/CR) ratio cut-off for myopia detection and to construct a multivariable model for estimating myopia probability in children and adolescents. A total of 6782 children and adolescents aged 3-18 years from Shanghai, China, were included. Myopia was defined as a cycloplegic spherical equivalent (SE) less than or equal to -0.50 dioptres. Participants were stratified into six CR groups (0.25 mm intervals). Receiver operating characteristic curve analysis was used to determine optimal AL/CR cut-off values for each CR group. Linear regression was performed to model the relationship between CR and optimal cut-offs. A multivariable logistic regression model incorporating CR, AL/CR ratio, age and sex was fitted, and model-based estimates were summarised using contour plots and lookup tables. Mean age was 9.36±3.59 years, and myopia prevalence was 33.8%. Optimal AL/CR cut-offs decreased with increasing CR, from 3.10 (7.08≤CR≤7.33 mm) to 2.93 (8.33<CR≤8.60 mm), with areas under the curve (AUCs) of 0.94-0.97 across groups. The optimal AL/CR cut-off was linearly associated with CR (optimal cut-off=4.073-0.135×CR; R²=0.948; p=0.001). The multivariable model showed high discrimination (AUC=0.97) in this dataset. Age- and sex-specific lookup tables and probability contour plots were generated to summarise model-based estimates. The optimal AL/CR ratio cut-off for myopia detection varies systematically with CR. These findings indicate that CR contributes to variation in AL/CR-based myopia classification. The fitted equation, probability contour plots and lookup tables provide CR-informed reference estimates for interpreting AL/CR in children and adolescents.