Authors (including presenting author) :
Wan KHN(1)(2), Chan TCY(2)(3)
Affiliation :
(1)Department of Ophthalmology, Tuen Mun Hospital (2)Department of Ophthalmology and Visual Sciences, The Chinese University of Hong Kong (3)Hong Kong Eye Hospital
Introduction :
The advances in techniques in cataract surgery have led to a paradigm shift from it being a rehabilitation procedure to a refractive procedure, allowing postoperative spectacle independence. A substantial deviation from the intended refractive target leading to refractive surprises is a common reason for litigation. Myopia is a global public health problem with a high incidence in the urban parts of Asia. The prevalence of high myopic eyes undergoing cataract surgery will become more common. Furthermore, high myopia is a risk factor for cataract formation, and these individuals develop cataract at a young age. In high myopia, accurate intraocular lens (IOL) prediction remains challenging after cataract surgery; it is not uncommon to encounter refractive surprises in these eyes. The new Hill-Radial Basis Function (Hill-RBF) formula is an artificial neural network formula developed to select the power of an IOL independent of a distinct effective lens position calculation using pattern recognition and data interpolation.
Objectives :
We aim to compare the accuracy and precision of Hill-RBF formula with 5 other validated formulas (Barrett Universal II, Haigis, Hoffer Q, Holladay 1, and SRK/T) in predicting residual refractive error after phacoemulsification in high axial myopic eyes.
Methodology :
127 eyes of 127 patients with axial length (AL) ≥ 26mm were included. The refractive prediction error (PE) was calculated as the difference between the postoperative refraction and the refraction predicted by each formula for the IOL power actually implanted. Standard deviation (SD) of PE, median absolute PE (MedAE), proportion of eyes within ±0.25 diopter (D), ±0.50 D, ±1.00 D of PE were compared. A generalized linear model was used to model the mean function and variance function of the PE (indicative of the accuracy and precision) with respect to biometric variables.
Result & Outcome :
The MedAE and SD of Hill-RBF were lower than that of Hoffer Q, Holladay 1, and SRK/T (p≤0.036), and were comparable to Barrett Universal II and Haigis (p≥0.077). Hill-RBF had more eyes within ±0.25 D of the intended refraction (59.84%) compared to other formulas (p≤0.034) except Barrett Universal II (p=0.472). AL was associated with the mean function or variance function of the PE for all formulas except Hill-RBF. In this study, the precision of Hill-RBF is comparable to Barrett Universal II and Haigis. Unlike the other 5 formulas, its dispersion and the accuracy of the refractive prediction is independent of the AL.
