LAUSR

PURPOSE Screening of early-stage keratoconus using auto-keratometer parameters. DESIGN Evaluation of a screening approach METHODS: Participants: We enrolled at six major centers in Japan, 123 eyes of 123 patients with Amsler-Krumeich classification stage 1 (age < 50 years, average 26.36 ± 8.68 years; 84/39 male/female) and 205 eyes of 205 healthy subjects (average 26.20 ± 7.34 years, 139/66 male/female). PROCEDURES Participants were divided 2:1 into a prediction and application group. In the prediction group, a multivariate logistic regression analysis was performed with keratoconus diagnosis as the dependent variable, and auto-keratometer parameters including average K, steep K, flat K, astigmatism, and astigmatic axis (no, with-the-rule, against-the-rule, and oblique) as independent variables. The diagnostic probability determined by regression analysis was defined as the keratometer keratoconus index (KKI). The cutoff value was determined from the receiver operating characteristic (ROC) curve. This prediction equation was evaluated in the application group. MAIN OUTCOME MEASURES Accuracy of the prediction equation for discriminating keratoconus from normal eyes. RESULTS The selected explanatory variables were steep K (partial regression coefficient [β]: 1.284, odds ratio [OR]: 3.610), flat K (β: -0.618, OR: 0.539), and with-the-rule astigmatism (β: -3.163, OR: 0.042). The area under the ROC curve of KKI was 0.90, which was significantly better than individual parameters (p< 0.001). The sensitivity and specificity values in the verification group were 85.0% and 86.7%, respectively. CONCLUSIONS Although the sensitivity/specificity was not very high, the new prediction equation using auto-keratometer-derived parameters enabled better discrimination of early-stage keratoconus than the isolated parameters.