Evaluating the prediction accuracy of the Hill-RBF 3.0 formula using a heteroscedastic statistical method

Abstract

Purpose: To evaluate the accuracy of the Hill-RBF 3 formula, with and without direct measurements of total corneal power, using a heteroscedastic statistical method for analysis.

Setting: Department of Ophthalmology, Shaare Zedek Medical Center, Jerusalem, Israel.

Design: Retrospective, consecutive case series.

Methods: Records of consecutive patients who underwent routine cataract surgery between February 2018 and June 2020 were retrospectively reviewed. The prediction accuracy of the Hill-RBF 3.0 formula was compared with that of the Barrett Universal II, Emmetropia Verifying Optical 2.0, Haigis, Hill-RBF 2.0, Hoffer Q, Holladay 1, Holladay 2, Kane, Olsen, and SRK/T formulas, based on biometry measurements by swept-source optical coherence tomography (SS-OCT) with standard keratometry (K), SS-OCT with total keratometry (TK), and an optical low-coherence reflectometer (OLCR). Statistical analysis was applied according to a heteroscedastic statistical method with SD of prediction errors as the main parameter for formula performance.

Results: The study included 153 eyes of 153 patients. The SD values that were obtained by Hill-RBF 3.0 (0.266 to 0.285 diopters [D]) were significantly lower compared with those by Hill-RBF 2.0 (0.290 to 0.309 D), Hoffer Q (0.387 to 0.407 D), Holladay 1 (0.367 to 0.385 D), Holladay 2 (0.386 to 0.401 D), and SRK/T (0.377 to 0.399 D) formulas (P < .036). The prediction accuracy of the Hill-RBF 3.0 was similar across the SS-OCT (K), SS-OCT (TK), and OLCR methods of measurement (P > .51).

Conclusions: The Hill-RBF 3.0 was more accurate than the Hill-RBF 2.0 and older generation formulas and had similar prediction accuracy compared with new generation formulas. The use of TK did not provide significant improvement to its prediction accuracy.