Refractive prediction errors of Barrett KCN, Kane KCN, SRK/T, and a customized approach in moderate-to-severe keratoconus after cataract surgery

To compare the refractive prediction accuracy of Barrett KCN, Kane KCN, SRK/T, and a customized ray-tracing-based approach in eyes with moderate-to-severe keratoconus undergoing cataract surgery. This ambispective cohort study included 34 eyes with moderate (n = 28) or severe (n = 6) keratoconus. Preoperative biometric and tomographic data were obtained from medical records, and postoperative manifest refraction was measured ≥ 1 month after surgery. Prediction error was defined as actual postoperative spherical refraction minus formula-predicted refraction for the implanted intraocular lens. Accuracy was assessed using signed prediction error, mean and median absolute prediction errors, and proportions within ± 0.25, ± 0.50, ± 1.00, and ± 2.00 D. The customized approach estimated total corneal power using ray tracing through anterior and posterior corneal surfaces within the central 4-mm zone and incorporated these values into the Barrett formula. The customized ray-tracing-based approach was used to calculate the IOL power and contributed to the selection of the implanted IOL. In the 33-eye complete-case cohort, the mean signed prediction errors were 1.33 ± 1.46 D for Barrett KCN, 0.55 ± 0.73 D for Kane KCN, 1.19 ± 1.59 D for SRK/T, and −0.02 ± 0.54 D for the customized approach (Friedman P < 0.001). The corresponding mean absolute prediction errors were 1.59, 0.72, 1.29, and 0.37 D, respectively (Friedman P < 0.001). With the customized approach, 18/33 (54.5%), 25/33 (75.8%), 30/33 (90.9%), and 33/33 (100%) of eyes were within ± 0.25, ± 0.50, ± 1.00, and ± 2.00 D, respectively. The customized ray-tracing approach demonstrated the lowest refractive prediction error and highest accuracy. However, these preliminary findings require confirmation in a larger, independent prospective study.