Screening for Ectasia Risk Prior to LASIK: Role of Enhanced Approach
Purpose
To develop and compare different functions to accurately identify ectasia risk (susceptibility) prior to LASIK using clinical, topometric (front curvature indices) and corneal tomography (3-D shape, including curvature, thickness profile and elevation) data.
Methods
In this retrospective study, different algorithms were developed using linear regression analysis to best separate the preoperative status of 266 eyes (141 patients) with stable LASIK outcomes (minimal follow-up of 24 months) and 60 eyes (46 patients) that developed ectasia after LASIK. Clinical data (age, flap thickness, ablation depth, manifest refraction, central thickness and subjective topographic classification) was retrieved from all cases along with Pentacam (Oculus) data. The approaches considered different categories of data: LDA1-clinical, LDA2-clinical and topometric and LDA3-clinical and tomographic. Receiver Operating Characteristic (ROC Curve) and Area Under the ROC (AUC) were used to verify and compared (De Long) accuracy.
Results
The mostly accurate single parameter to distinguish between the groups was the BAD-D (sensitivity = 85%; specificity = 83.83%; AUC = 0.925). There were significant differences between the AUCs for LDA1, LDA2 and LDA3 (DeLong´s Method, p<0.05), which were respectively 0.958, 0.971 and 0.995. LDA3, named as Enhanced Ectasia Susceptibility Score (EESS) obtained 100% of sensitivity and 94.74% of specificity to distinguish post-LASIK ectasia cases from stable LASIK cases.
Conclusion
The performances of screening algorithms were significantly improved by objective data, including topometric indices along with subjective classification of curvature maps. Tomographic data significantly augments accuracy. Artificial intelligence strategies should be considered to optimize accuracy in diagnosis, using conscious and validated combinations of parameters. The EESS represents a significant improvement for detecting ectasia risk prior to LASIK.