Accuracy of Topometric Indices for Distinguishing between Keratoconic and Normal Corneas

Abstract

Purpose: To evaluate the sensitivity and specificity of the Pentacam topometric indices derived from the corneal surface curvature to distinguish between normal and keratoconic corneas. Methods: The study consisted of 226 normal corneas from 113 patients and 88 keratoconic eyes from 44 patients. Eyes were defined as keratoconus based on comprehensive ocular examination, including Placido-disk-based corneal topography (Atlas Corneal Topography System; Humphrey, San Leandro, California) and rotating Scheimpflug corneal tomography (Pentacam HR; Oculus, Wetzlar, Germany). Corneal Topometric indices ISV, IVA, KI, CKI, IHA and IHD, along with the TKC (Topometric Keratoconus Classification) score were calculated from the Pentacam HR exam. Statistical analysis were accomplished using BioEstat 5.0 (Instituto Mamiraua, Amazonas, Brazil) and MedCalc 12.0 (MedCalc Software, Mariakerke, Belgium) using unpaired nonparametric Mann Whitney test (Wilcoxon ranked-sum). ROC curves were calculated for each topometric parameter to determine the best cut off values from the significantly different parameters. A logistic regression analysis was performed to provide a combined parameter for optimizing accuracy. Results: Statistical significant differences were found between keratoconic and normal corneas for all topometric indices (Mann Whitney, p < 0.05). There were four false negative cases among the keratoconic cases on the TKC classification (4.54%) and 16 false positive cases among normal (7.08%), so that the sensitivity and specificity of the TKC were 95.54 and 92.92% respectively. The areas under the ROC curves (AUC) for the individual topometric indices varied from 0.843 (CKI) and 0.992 (ISV). The sensitivity and specificity of the most accurate ISV were 97.7 and 96.5% respectively. The calculated parameter from logistic regression had AUC of 0.996, with sensitivity of 97.7% and specificity of 98.7%. Conclusion: Pentacam topometric indices were useful for distinguishing between normal and keratoconic corneas. The TKC classification should be expected to have false positives and negatives and should not be considered alone. TKC had more false positives and false negatives than some individual topometric parameters. A novel combined parameter based on logistic regression analysis may improve accuracy for the diagnosis of keratoconus. Further studies are necessary to evaluate if adding other curvature derived indices is beneficial for the regression analysis, as well as for testing the sensitivity of such parameters for the diagnosis of milder forms of ectasia and for testing correlations with severity of the disease.