PURPOSE: To compare the accuracy of 12 intraocular lens (IOL) formulas in eyes with axial length (AL) ≥ 28 mm and to identify the anterior segment parameters independently affecting prediction error (PE). METHODS: A total of 415 eyes undergoing phacoemulsification with IOL implantation at Shanghai General Hospital were included in this retrospective observational study. Refractive PEs were calculated for 5 conventional formulas (Haigis, Hoffer Q, Holladay 1, Holladay 2 and SRK/T) with optimization and 7 new-generation formulas (Barrett Universal II BUII, Cooke K6, EVO 2.0, Hoffer QST, Kane, Ladas Super Formula LSF and Pearl-DGS). Predicted postoperative spherical equivalent (SE) was calculated using keratometry (K) and total keratometry (TK) measured by IOLMaster 700. Q values, posterior mean keratometry, corneal volume, anterior chamber volume and angle and back-to-front corneal radius ratio were measured by Pentacam. Stepwise linear regression identified parameters associated with PE. RESULTS: Using the K method, BUII, EVO 2.0, Hoffer QST, Holladay 2, Kane and Pearl-DGS showed the lowest median absolute PE (MedAE = 0.32, 0.31, 0.32, 0.30, 0.30 and 0.32 D) and highest formula performance index (FPI = 0.74, 0.80, 0.69, 0.68, 0.60 and 0.69). BUII, Cooke K6, EVO 2.0, Holladay 2 and Pearl-DGS performed stably across axial groups (p > 0.05). K achieved lower AE than TK for all formulas (p < 0.05), except Hoffer QST (p = 0.497), Holladay 2 (p = 0.140) and Pearl-DGS (p = 0.912), with Kane performing better using TK (p = 0.024). Lens thickness (LT) independently affected PE for Holladay 2 and all new-generation formulas, with greater LT associated with myopic shifts (p < 0.05). CONCLUSIONS: EVO 2.0 formula provided the most accurate predictions with K method. Except for Kane, K outperformed TK. LT independently influences PE and should be considered in future IOL calculations.
Wu et al. (2026) studied this question.