Abstract
Purpose :
Visual acuity (VA) is measured with the ETDRS standard in clinical trials. Snellen remains the real-world standard, although Snellen VA is typically worse by 6 ~ 12 letters (0.12 ~ 0.24 LogMAR). In this study, we modeled ETDRS-Snellen discrepancies for 163 observers from Kaiser (2009) to enable better translation between the VA standards.
Methods :
We used a VA behavioral function to simulate VA behavior for 27 observers with VA threshold τ from -0.3 to 2.3 LogMAR and VA-dependent range 0.253 + 0.185 VA LogMAR (Zhao et al. 2021). We scored VA behavior using ETDRS and Snellen rules (Kaiser, 2009). From repeated simulations of each observer, we computed the mean and standard deviation (SD) of VA scores and modeled ETDRS-Snellen score pairs using maximum likelihood.
Results :
For the two charts, Figure 1 shows the bias (deviation from identity line) and variability (SD) of VA scores. Small biases observed when τ < 1.0 LogMAR increased significantly as VA worsened: bias > 0.20-0.50 LogMAR as τ > 1.0 LogMAR. Likewise, variability increased for worse VA: SD up to 0.15 for ETDRS and 0.19 for Snellen. Results from each simulated observer were used to construct the probability distributions for ETDRS-Snellen score pairs (Figure 2A). They were used to successfully model all the ETDRS and Snellen scores (Figure 2B) in Kaiser (2009) (p > .05 for all differences between observed scores and model predictions).
Conclusions :
Using a VA behavioral function, we successfully modeled discrepancies between ETDRS and Snellen scores from a large sample of subjects with a wide range of VA scores. The model can be used to translate between VA measured in trials and real world.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.