Purpose: To assess a new polynomial regression formula integrating the refractive prediction error of the first-operated eye to improve the intraocular lens power calculation of the second eye in cataract surgery.
Setting: Centre Hospitalier Universitaire, Toulouse, France.
Design: Retrospective multicentric dataset study.
Methods: A polynomial regression formula, WeOptimeye2nd (WO2nd), was developed using a machine-learning algorithm trained on a dataset of 534 patients who underwent sequential bilateral cataract surgery. A separate multicentric dataset was used to retrospectively calculate predicted refraction with WO2nd, SRK/T and Barrett Universal II formulas, and 3 other methods of constant factors (CFs) second-eye refinement (CF0.38, CF0.35, and CF0.5). Mean absolute errors (MAEs) and percentage of eyes within ±0.25, ±0.5, and ±1.0 diopter (D) from predicted spherical equivalent were compared between formulas.
Results: The study comprised data on 722 patients. In the overall population, WO2nd had the lowest MAE: 0.339 vs 0.347 (P = .137), 0.340 (P = .956), 0.350 (P = .066), 0.399 (P < .001), and 0.410 (P < .001), with CF0.38, CF0.5, and CF0.35, Barrett II, and SRK/T, respectively. WO2nd had the highest percentage of eyes within ±0.5 D of the predicted refraction, and the difference was statistically significant vs SRK/T and Barrett II formulas but not vs CF0.38, CF0.5, and CF0.35. WO2nd performed the best in axial length (AL) < 22 mm with the lowest MAE and a statistically significant difference vs any other formula.
Conclusions: WO2nd improved the refractive outcome of the second-operated eye and performed well in extreme AL and mean keratometry subgroups.
Copyright © 2021 Published by Wolters Kluwer on behalf of ASCRS and ESCRS.