Development of a Combinatorial Ophthalmic Biomarker of Dementia

Abstract

Purpose : To evaluate the diagnostic performance of potential ophthalmic biomarkers in predicting dementia.

Methods : Participants with research criteria diagnosis of normal cognition, mild cognitive impairment (MCI), or dementia were recruited from the University of Washington Alzheimer’s Disease Research Center. Measurements included photopic and mesopic visual acuity (VA), intraocular pressure, pupillometry, fundus photography, optical coherence tomography (OCT), OCT angiography, flavoprotein fluorescence, and cognitive testing batteries. Participants were grouped as impaired (MCI and dementia) or normal cognition in the analyses. Univariate logistic regression models adjusting for age and education were used to evaluate associations of ophthalmic features with cognitive impairment. Variables showing significance (p<0.05) were included for multivariable analysis. Selected univariate and multivariate variable(s) and their corresponding areas under the receiver operating characteristic curves (AUCs) were compared.

Results : A total of 105 participants (47% women) of mean age 75 (range 65-96) were included. Higher mesopic VA (odds ratio [OR] = 0.965, p = 0.021), thicker circumpapillary retinal nerve fiber layer (cpRNFL) (OR = 0.954, p=0.006), and delayed pupillary constriction onset (OR 1.02, p=0.03) were associated with lower odds of impaired cognition on univariate analysis.
The baseline model of age and education achieved an AUC = 0.597. The addition of cpRNFL thickness improved performance the most among all univariate models (AUC = 0.679). Augmenting this model with mesopic VA and pupillary constriction latency improved AUC to 0.722 in predicting cognitive impairment.

Conclusions : Lower cpRFNL thickness was most strongly associated with cognitive impairment in our cohort. Several other visual function and retinal imaging variables were significantly associated with cognitive impairment, supporting the potential role of a combinatorial ophthalmic biomarker for dementia.

This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.

 

Table. Univariate associations between ophthalmic features and risk of impaired cognition

 

Figure. Receiver operating characteristic curves. 1a. Baseline model: age and years of education. 1b. Model 1a with circumpapillary retinal nerve fiber layer. 1c. Model 1a with significant ophthalmic features from univariate analysis. AUC, area under the curve.