Predictive Accuracy of Intraocular Lens Power Calculation Formulas for Cataract Surgery in Keratoconus: A Systematic Review and Network Meta-Analysis

Background: Keratoconus (KCN) alters corneal curvature, complicating intraocular lens (IOL) power calculations for cataract surgery and increasing the risk of refractive errors. Traditional IOL formulas, designed for normal corneas, often yield suboptimal results, necessitating the evaluation of keratoconus-specific formulas. Objective: This study compares the predictive accuracy of IOL power calculation formulas in KCN patients undergoing cataract surgery, focusing on mean prediction error (MPE), mean absolute error (MAE), and the proportion of eyes achieving refractions within ±0.50 D and ±1.00 D. Methods: A systematic search of MEDLINE, Embase, and Cochrane databases identified seven studies (530 eyes). A Bayesian network meta-analysis (NMA) ranked formulas, including Barrett True-K (measured/predicted), Kane Keratoconus, Barrett Universal II, SRK/T, Holladay 1 and 2, Haigis, Higgins, and Hoffer Q. Subgroup analyses classified KCN severity using Amsler-Krumeich criteria. Results: Barrett True-K Measured and Kane Keratoconus demonstrated the highest predictive accuracy and consistently ranked among the best-performing formulas overall. Barrett True-K Predicted showed favorable performance in mild keratoconus, whereas SRK/T performed comparatively well in moderate disease. Conventional formulas, particularly Hoffer Q and Holladay formulas, showed lower predictive accuracy, especially in more advanced keratoconus. Overall, Barrett True-K and Kane Keratoconus achieved the highest proportions of eyes within ±0.50 D and ±1.00 D of target refraction. Conclusion: Barrett True-K and Kane Keratoconus demonstrated superior accuracy, while traditional formulas underperformed, especially in severe KCN. Keratoconus-specific formulas should be prioritized. The ranking probabilities of the including formulas are exploratory and should be interpreted cautiously due to limited and heterogeneous data. Future research should include larger sample sizes and standardized biometry for improved clinical decision-making.