Multivariable Model to Predict Toric Ortho-K Lens Prescription

This is the authors' abstract. We don't add key points for this paper.

OBJECTIVE: Orthokeratology (Ortho-K) effectively controls myopia progression. Toric Ortho-K lenses are used for moderate-to-high astigmatism, but predictive models with multiple corneal parameters are limited. This study aimed to identify key predictors for toric lens prescription in Ortho-K. METHODS: A retrospective analysis was conducted on 506 patients undergoing Ortho-K treatment, incorporating 15 parameters such as age, sex, refractive error, corneal astigmatism, and flat eccentricity. Using both univariate and multivariate models, significant predictors were identified using logistic regression and refined with backward stepwise regression, evaluated by receiver operating characteristic curve analysis. RESULTS: The 4-variable model (corneal astigmatism [∆K], steep eccentricity [Steep e], Sagittal Height Difference at 8 mm [SD8], and Corneal Cylindrical to Spherical [C/S] ratio) achieved an area under the curve (AUC) of 0.92 (sensitivity 89.7%, specificity 79.5%). Receiver operating characteristic analysis validated thresholds: ∆K>1.50 D (AUC 0.87), SD8 greater than 34 μm (AUC 0.82), flat eccentricity greater than 0.68 (AUC 0.73), and C/S ratio greater than 0.59 (AUC 0.63). CONCLUSION: This study developed the 4-variable model for toric Ortho-K lens fitting, identifying corneal astigmatism and SD8 as key predictors for detecting limbus-to-limbus astigmatism. Steep and flat eccentricities offered predictive and clinical insights, respectively.