Prediction model for treatment outcomes 3 years after corneal cross-linking for keratoconus

Abstract

Purpose: This study aimed to identify preoperative factors that predict visual acuity and Kmax 3 years after corneal cross-linking (CXL) in patients with keratoconus (KC), and to develop a prediction model.

Methods: We enrolled 68 patients with KC and followed up on 100 eyes that received CXL for at least 3 years. Preoperative data, including age, UDVA, CDVA, cylinder, SE, and the parameters of tomography including Kmax were collected as predictors. The primary outcomes were changes in CDVA (Delta CDVA) and Kmax (Delta Kmax) postoperatively. Univariate and multivariate linear regression were used to identify the correlation between the primary outcomes and predictors and establish prediction models.

Results: Both CDVA and Kmax remained stable from baseline to 3 years after CXL: from 0.25 ± 0.18 to 0.22 ± 0.20 (P = 0.308) and from 58.70 ± 9.52 D to 57.02 ± 8.83 D (P = 0.187), respectively. Multivariate analysis showed that worse preoperative CDVA (ß coefficient - 0.668, P < 0.001) and lower preoperative Kmean (ß coefficient 0.018,P < 0.001) were associated with greater improvement in CDVA after CXL. A smaller preoperative eccentricity (ß coefficient 8.896, P = 0.01) and a higher preoperative Kmean (ß coefficient - 1.264, P < 0.001) predicted a more flattening of postoperative Kmax. The prediction model for CDVA (R² = 0.43) and Kmax (R² = 0.37) could accurately estimate treatment outcomes.

Conclusions: CXL is highly effective in halting or preventing further progression of KC. The preoperative factors CDVA and Kmean were able to predict visual acuity changes 3 years after CXL. And preoperative eccentricity and Kmean could predict Kmax changes 3 years after CXL.

Keywords: Corneal cross-linking; Keratoconus; Multivariate analysis; Prediction model.