Document Type

Article

Publication Title

Canadian Journal of Ophthalmology

Abstract

Objective: Amid growing adoption of AI in ophthalmology, Hill-Radial Basis Function (Hill-RBF) 3.0 formula may improve refractive target prediction across diverse populations, transforming patient care. This systematic review and meta-analysis (PROSPERO: CRD42021231150) evaluated the predictive performance of the Hill-RBF 3.0 against other traditional and AI-based intraocular lens (IOL) power calculation formulas.

Design: Systematic review and meta-analysis.

Methods: EMBASE, MEDLINE, CINAHL, and Web of Science were searched on March 17, 2025. Eligible studies reported refractive or cataract surgeries using Hill-RBF 3.0 and at least 1 comparable formula (Barrett Universal II, EVO 2.0, Haigis, Hoffer Q, Hoffer QST, Holladay I, Ladas Super Formula AI, Kane, Olsen, Pearl-DGS, or SRK/T). Standardized mean difference (SMD) in mean absolute error (MAE) and odds ratio (OR) of achieving postoperative refraction within ±0.25, ±0.50, and ±1.00 diopters (D) were assessed. Meta-analyses were performed.

Results: Of 130 records screened, 31 studies were included qualitatively and 28 quantitatively. Hill-RBF 3.0 showed significantly lower MAE (SMD [95% CI]) compared to Haigis (-0.12 [-0.21 to -0.02]; p = 0.01), Holladay-1 (-0.08 [-0.15 to -0.01]; p = 0.04), and SRK/T (-0.07 [-0.14 to 0.00]; p = 0.04). Odds (odds ratio [OR; 95% CI]) of achieving ±0.50 D were higher versus Haigis (1.09 [1.01-1.19]; p = 0.03), Hoffer Q (1.11 [1.04-1.19]; p = 0.002), Holladay-1 (1.10 [1.02-1.18]; p = 0.02), and SRK/T (1.08 [1.01-1.16]; p = 0.02). Similarly, Hill-RBF outperformed these formulas within ±0.25 D, except in comparison with Holladay-1.

Conclusions: Hill-RBF 3.0 achieved comparable or superior predictive accuracy over traditional and AI-based formulas, enhancing clinical workflows, refractive outcomes, and patient care.

DOI

10.1016/j.jcjo.2026.01.005

Publication Date

2-12-2026

ISSN

1715-3360

Download

Share

COinS