AIM: To construct and validate a diagnostic model for early detection of keratoconus based on parameters in Sirius.

METHODS: The study comprised of 46 early keratoconus eyes(including 20 right eyes and 26 left eyes in 34 patients)and 46 age- and gender-matched normal eyes(including the right eyes of 46 patients)in the prediction group. The predictive index was constructed using LASSO and Logistic regression analyses based on the topographic, pachymetric and aberrometry variables of Sirius. There were 23 early keratoconus eyes categorized as suspected keratoconus cases by Sirius(including 12 right eyes and 11 left eyes in 23 patients)and 23 age- and gender-matched normal eyes(including the right eyes of 23 patients)included in the application cohort. External validation of predictors was performed for the application cohort.

RESULTS: Sirius Keratoconus Index(SKI)was calculated based on the minimum corneal thickness and symmetry index back of Sirius. Highest AUC values were obtained in the prediction group(AUC=0.932)after Logistic regression analysis. The cut-off value of SKI was set at 0.44. Then, the receiver operating characteristic(ROC)curve, calibration plot and nomogram of the diagnostic formula were analyzed for the prediction cohort in detail. Finally, the accuracy of the SKI was evaluated in the application cohort; the sensitivity was 91% and the specificity was 96%.

CONCLUSION: SKI based on minimum corneal thickness and symmetry index back of Sirius is a simple and effective method for early detection of keratoconus in the preoperative screening for refractive surgery.