Predicting factors for the efficacy of cross-linking for keratoconus.

Purpose: To evaluate predictors for success in corneal crosslinking (CXL) for keratoconus in a large cohort and extended follow-up.

Design: A retrospective study based on a prospectively built database.

Methods: Participants underwent CXL for keratoconus from 2007 to 2018. Statistical analysis was performed for patients with at least 1-year follow-up. We analyzed effects of CXL type (Epithelium-on or Epithelium-off and Accelerated (9mW/cm2@10min) or Standard (3mW/cm2@30min)) and pre-operative factors including age, gender, baseline LogMAR visual acuity (LogMARpre), maximal corneal power (Kmaxpre), pachymetry, refractive and topographic cylinders, spherical equivalent (SEpre), mean corneal power (MeanK) and follow-up time on outcome measures. The outcome measures were the final change of Kmax (Delta Kmax) and the final change in LogMAR visual acuity (Delta LogMAR). A more negative Delta Kmax or Delta LogMAR represents a favorable effect of crosslinking.

Results: 517 eyes had Kmax results, and 385 eyes had LogMAR results with more than one year follow-up. These eyes were included in the study. The mean follow-up time was 2.29 years. Mean Kmax decreased from 54.07±5.99 diopters to 52.84±5.66 diopters (p<0.001), and Mean LogMAR decreased from 0.28±0.20 to 0.25±0.21 (p<0.001). Non-accelerated epithelium-off CXL resulted in greater flattening of Kmax when compared with other protocols. Visual acuity improvement was similar when comparing different CXL protocols. Multivariate analysis showed four factors associated with negative Delta Kmax: high Kmaxpre, high SEpre, high MeanKpre, and non-accelerated procedure. Multivariate analysis showed three factors associated with negative Delta LogMAR: high LogMARpre, high SEpre, and Low MeanKpre. After excluding corneas with Kmaxpre >65 D or Pachymetry<400 microns, multivariate analysis showed that high Kmaxpre, high SEpre, and non-accelerated CXL were associated with negative Delta Kmax while high LogMARpre and high SEpre were associated with negative Delta LogMAR.

Conclusion: CXL for keratoconus is a highly effective treatment, as evident by its effects on the outcome measures: Delta Kmax and Delta LogMAR. CXL was more successful in eyes with high Kmaxpre, high SEpre, and high LogMARpre, which express disease severity. The non-accelerated epithelium-off protocol was associated with greater flattening of corneal curvature but did not show a better effect on visual acuity as compared to the other CXL protocols.