Abstract
Purpose :
To estimate the impact of ethnicity and race on time to detect progression by standard automated perimetry (SAP) and optical coherence tomography (OCT) in subjects with glaucoma.
Methods :
This was a retrospective cohort study with patients from the Bascom Palmer Eye Institute. Data from 47,022 SAP tests (5,405 eyes) and 25,540 OCT tests (4,136 eyes) followed for 8.6±4.2 years were included. 30% of subjects self-identified as Hispanic/Latino, 70% Not Hispanic/Latino; 20% Black, 80% White. Standard deviation of the residuals from linear regression models for OCT retinal nerve fiber layer thickness (RNFL) and SAP mean deviation (MD) over time was used as a measure of variability. Residuals distributions were used in computer simulations to estimate SAP MD and OCT RNFL trajectories, under different assumptions about baseline disease and rate of change. Time to detect progression was obtained for the simulated tests and compared between groups.
Results :
Significantly larger SAP variability over time was noted in Hispanic compared to non-Hispanic subjects (1.81±1.46 vs. 1.52±1.10dB; P<0.001), and Black compared to White subjects (1.80±1.30 vs 1.57±1.22dB; P<0.001). For OCT, variability was larger in non-Hispanic compared to Hispanic subjects (2.2±1.4 vs. 2.1±1.2μm; P=0.021), and Black compared to White subjects (2.3±1.5 vs. 2.1±1.3μm; P=0.001). Overall, larger variability led to delayed SAP progression detection in all simulated scenarios: by 1.3 (95%CI, 0.7-1.9) years in Hispanic compared to non-Hispanic subjects, and by 1.5 (95%CI, 0.9-2.1) years in Black compared to White subjects for baseline MD of -10dB and slope of -0.25dB/y. Differences in time to detection of progression with OCT were small and generally less than 1 year for all comparisons between races and ethnicities.
Conclusions :
Variability was higher in perimetric testing for Black subjects compared to White subjects, and for Hispanic compared to non-Hispanic subjects, leading to potential delays in detection of progression in minority groups. Such differences are likely related to systemic biases in test administration procedures and deserve further investigation.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.