Comparison of patient-specific computational modeling predictions and clinical outcomes of lasik for myopia

Abstract

PURPOSE. To assess the predictive accuracy of simulation-based LASIK outcomes. METHODS. Preoperative and 3-month post-LASIK tomographic data from 20 eyes of 12 patients who underwent wavefront-optimized LASIK for myopia were obtained retrospectively. Patient-specific finite element models were created and case-specific treatment settings were simulated. Simulated keratometry (SimK) values and the mean tangential curvature of the central 3 mm (Kmean) were obtained from the anterior surfaces of the clinical tomographies,and computational models were compared. Correlations between Kmean prediction error and patient age,preoperative corneal hysteresis (CH),and corneal resistance factor (CRF) were assessed. RESULTS. The mean difference for Kmean between simulated and actual post-LASIK cases was not statistically significant (-0.13 ± 0.36 diopters [D],P = 0.1). The mean difference between the surgically induced clinical change in Kmean and the model-predicted change was -0.11 ± 0.34 D (P = 0.2). Kmean prediction error was correlated to CH,CRF,and patient age (r ¼ 0.63,0.53,and 0.5,respectively,P < 0.02),and incorporation of CH values into predictions as a linear offset increased their accuracy. Simulated changes in Kmean accounted for 97% of the variance in actual spherical equivalent refractive change. CONCLUSIONS. Clinically feasible computational simulations predicted corneal curvature and manifest refraction outcomes with a level of accuracy in myopic LASIK cases that approached the limits of measurement error. Readily available preoperative biomechanical measures enhanced simulation accuracy. Patient-specific simulation may be a useful tool for clinical guidance in de novo LASIK cases.