IOL Power Formula Prediction Accuracy at the Extremes of Axial Length: A Prospective Outcome Verification Study

Overview

This single-center, prospective, cross-sectional study evaluates how accurately eight modern intraocular lens (IOL) power calculation formulas predicted the postoperative refractive result in eyes with extreme axial length after cataract surgery. Patients previously operated at the study center who had a very short (axial length \< 22.0 mm) or very long (\> 26.0 mm) eye and received a spherical monofocal IOL are invited to a single, non-invasive outcome visit. At this visit, standardized manifest refraction, optical biometry, corneal tomography, slit-lamp examination and, in long eyes, macular OCT are performed. The measured postoperative spherical equivalent is compared with the residual refraction that each of the eight formulas had predicted for the actually implanted IOL power, separately in short and long eyes. The study aims to identify which formula predicts the postoperative outcome most accurately at the extremes of axial length, where formula choice is most consequential and least well studied.

Modern IOL power formulas achieve high refractive predictability in average-length eyes but deteriorate at the extremes of axial length, where effective lens position modelling diverges most between formulas. A preceding retrospective study at this center quantified the formula-dependent variability of IOL power recommendations at the axial length extremes using fully standardized biometric inputs; however, planning discrepancy alone does not indicate which formula's recommendation yields the better clinical outcome. This prospective outcome verification study closes that gap by linking a standardized postoperative assessment to the pre-existing, pseudonymized preoperative planning dataset. Design: single-center, prospective, cross-sectional, observational (paired within-eye) study. From the retrospective source cohort (approximately 927 short and 769 long eyes; approximately 1,000-1,200 patients; index surgeries January 2022-December 2025), a stratified random sample of 200 patients is drawn (110 short stratum, 90 long stratum; 55:45 ratio) and invited to a single outcome visit. Each eligible eye contributes up to eight paired, formula-specific prediction errors that differ only by the calculation formula, while the postoperative reference (manifest refraction) and all preoperative inputs are held constant. Eyes are stratified a priori into a short (AL \< 22.0 mm) and a long (AL \> 26.0 mm) stratum and analyzed separately. Formulas evaluated: Barrett Universal II, Cooke K6, EVO 2.0, Hill-RBF 3.0, Hoffer QST, Kane and PEARL-DGS (via the ESCRS IOL calculator), and the stand-alone Ladas Super Formula 2.0. Outcome visit (single visit, all non-invasive standard ophthalmic techniques, no study-mandated therapy, medication or surgery): uncorrected and best-corrected distance visual acuity, IOLMaster 700 optical biometry, CASIA2 and Pentacam AXL corneal tomography, autorefraction and standardized subjective manifest refraction (primary outcome measure), slit-lamp examination with posterior capsule opacification grading, and macular OCT (obligatory in the long stratum, recommended in the short stratum). Primary analysis: for each formula, the spherical equivalent prediction error (SE-PE = measured postoperative spherical equivalent minus the residual spherical equivalent predicted for the actually implanted IOL power step) is computed and the absolute error (\|SE-PE\|) compared globally across the eight formulas using the Friedman test, with post-hoc pairwise Wilcoxon signed-rank tests (Holm-corrected, 28 formula pairs), separately in each axial length stratum. The primary accuracy metric per formula is the median absolute prediction error. All data are processed in pseudonymized form and linked via a study ID and laterality, in accordance with the Declaration of Helsinki (2024), ICH-GCP, and the GDPR. The study is reported per STROBE and is registered on ClinicalTrials.gov prior to the start of recruitment.