Abstract
Purpose :
We applied an artificial intelligence algorithm to new-onset visual field (VF) loss encountered in 3 US prospective cohorts followed for primary open-angle glaucoma (POAG). The algorithm objectively quantified VF loss into archetypical patterns. We examined self-reported race in relation to the specific archetypes (ATs) of VF loss.
Methods :
Participants aged ≥40 years who without glaucoma and reported eye examinations in the Nurses Health Study (NHS) (n=75,767; 1980 to 2018), NHS2 (n=80,857 women; 1989-2019), and the Health Professionals Follow-up Study (n=36,838 men; 1986-2018) were followed. Information on demographics including race, medical conditions and lifestyle were assessed with biennial questionnaires. Incident POAG cases with reproducible VF loss were confirmed by medical record review. The earliest reliable VF for the worse eye was identified, and archetypal analysis was used to identify the optimal number of VF loss patterns. Each case was classified according to the most dominant VF loss archetype. Multivariable-adjusted relative risks (RRs) for each archetype and 95% confidence intervals (CIs) were estimated using Cox proportional hazards regression on pooled data. Major covariates included cohort, age, family history of glaucoma, census-tract based socioeconomic index, number of eye exams during follow-up, and medical history such as diabetes and hypertension. False discovery rate was used to account for multiple comparisons.
Results :
Mean age was 57±10 years and 81% were women. Of the 1377 POAG cases, 1301 were Caucasian, 36 were of African descent, 21 were Asian, and 19 were Hispanic white. We observed 14 ATs (see Figure), of which 13 reflected function loss. We observed that having African descent compared to being non-Hispanic White was associated with ATs consistent with advanced glaucomatous loss: AT5→RR=2.84 (95%CI=1.26, 6.41); AT8→RR=5.35 (95%CI=1.79,16.0); AT10→RR=3.80 (95%CI=1.27,11.4); AT11→RR=3.79 (95%CI=1.32,10.9); and AT12→RR=11.43 (95%CI=3.77, 34.6). Being Asian compared to being non-Hispanic White was not significantly associated with any AT. Being Hispanic White compared to being non-Hispanic White was significantly associated with AT10 (RR=4.97; 95%CI=1.48,16.7) and AT11 (RR=6.58; 95%CI=2.56,16.9).
Conclusions :
African and Hispanic Americans with POAG were more likley to present with central or advanced VF loss.
This is a 2021 ARVO Annual Meeting abstract.