Prediction Model For Treatment Outcomes 3 Years After Corneal Crosslinking For Keratoconus

To investigate the preoperative predictors affecting visual acuity and maximum keratometry at 3 years after corneal crosslinking (CXL) for keratoconus (KC) and develop a prediction model.

Retrospective cohort study.

 The study included 68 KC patients (100 eyes) who were treated with CXL and had at least 3 years of follow-up. We collected the preoperative data as predictors including age, UDVA, CDVA, cylinder, SE, K1, K2, CA, eccentricity, Kmean, Kmax, CCT, TCT, and BAD-D. Postoperative changes in corrected distance visual acuity (Delta CDVA) and Kmax (Delta Kmax) were considered the primary outcomes. Univariate linear regression analysis was used to determine the correlation between the main results and the predictors. Then multivariate linear regression analysis was used to find independent influencing factors and establish prediction models.

UDVA, CDVA and Kmax remained stable at 3 years follow-up (p >0.05). Multivariate analysis showed that worse preoperative CDVA (ß coefficient -0.668, 95% CI -0.858 to -0.478, p <0.001) and lower preoperative Kmean (ß coefficient 0.018, 95% CI 0.010 to 0.025, p <0.001) could be associated with a greater improvement in CDVA after CXL. A smaller preoperative eccentricity (ß coefficient 8.896, 95% CI 2.235 to 15.560, p=0.01) and a higher preoperative Kmean (ß coefficient -1.264, 95% CI -1.672 to -0.855, 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.

CXL is a highly effective treatment for KC. The preoperative factors CDVA and Kmean were able to predict visual acuity 3 years after CXL. And preoperative predictors eccentricity and Kmean could predict Kmax 3 years after CXL.