Modeling of the Lower Limits of IOL Prediction Error Resultant from Instrument Precision

Abstract

Purpose : As IOL calculation formulas improve, prediction error approaches a fundamental limit defined by IOL manufacturing tolerance and precision of biometry instruments. In this study, we characterize the contribution of instrument precision to IOL prediction error.

Methods : Fifty-one eyes of 35 subjects (age: 70.9 ± 10.1 years, SEQ: -0.40 ± 0.62 D) who underwent cataract surgery and IOL implantation at Bascom Palmer Eye Institute were included. Individualized eye models were created from post-operative intraocular distances (central corneal thickness (CCT), anterior chamber depth (ACD), vitreous chamber depth (VCD), and IOL thickness (tIOL)) measured with a custom-built extended-depth OCT system (Ruggeri et al, Biomed Opt Exp 2012), and post-operative anterior (Rant cornea) and posterior (Rpost cornea) corneal curvatures measured with an anterior segment biometry system (Pentacam). The IOL was modeled as a symmetric biconvex thick lens. Post-operative refraction was calculated using backwards ray tracing. To simulate measurement variability, random Gaussian-distributed noise was added to eye model parameters based on the repeatability values derived from experimental data: 15, 100, 2, 12, 24, 5 µm for Rant cornea, Rpost cornea, tIOL, VCD, ACD, and CCT, respectively. Post-operative spherical equivalent refraction was estimated over 1000 simulation trials for each subject.

Results : The effect of single biometric parameters on post-operative refraction is depicted in Figure 1. Results from the Monte Carlo simulation are shown in Figure 2. An average variability (95% CI) of 0.27 ± 0.04 D across all subjects was determined.

Conclusions : Instrument repeatability contributes 0.25 to 0.30 D of error to IOL power calculation, representing a fundamental limit independent of IOL calculation formulas.

This is a 2020 ARVO Annual Meeting abstract.

 

Figure 1. Change in estimated post-operative refraction after shifting values for single biometric parameters across a set range; analyses for each subject are shown in different colors.

 

Figure 2. Monte Carlo simulation of uncertainty in post-operative refraction due to Gaussian-distributed noise. Plot shows results from 1000 iterations for each of the 51 eyes with error bars representing mean and 95% CI.