Multivariate Analysis of the Spatial Distance Between the Corneal Thinnest Point and Maximum Keratometry in...

Background: This study quantified the planar distance (D, mm) between the corneal thinnest point (TP) and maximum keratometry (Kmax) and examined its association with tomographic features and keratoconus severity. Methods: Tomographic data from 140 patients (280 eyes) with bilateral manifest keratoconus were obtained with a Scheimpflug camera (Pentacam HR; Oculus, Wetzlar, Germany). D was calculated from the Cartesian coordinates of TP and Kmax. A multiple linear regression model included simulated anterior average keratometry (SimKavg), anterior and posterior radii of curvature (ARC and PRC), and anterior Q-value (Q-ant). Inference used patient-clustered robust standard errors. Internal validation comprised repeated grouped 10-fold cross-validation and patient-level bootstrap resampling. Performance was assessed using R², root-mean-square error (RMSE), mean absolute error (MAE), and Bland-Altman limits of agreement. Results: The model explained 51.8% of the apparent variance in D (adjusted R² = 0.511; p < 0.001). Apparent RMSE and MAE were 0.568 and 0.428 mm. Mean grouped cross-validated performance was R² = 0.472, RMSE = 0.594 mm, and MAE = 0.443 mm. The out-of-fold bias was -0.005 mm, with 95% limits of agreement from -1.162 to +1.151 mm. ARC, SimKavg, and Q-ant remained significant with patient-clustered inference, whereas PRC did not (p = 0.111). Conclusions: D was associated with keratoconus severity and could be estimated with moderate internal performance from routinely available tomographic variables. Collinearity, fellow-eye dependence, and limits of agreement of approximately ±1.1 mm restrict interpretation of individual coefficients and preclude interchangeability with direct measurement. External and longitudinal validation is required before clinical use.