Pupillary Light Reflex and Eye Movement Parameters as Objective Measures of Cognitive Decline in Older Adults: A Secondary Analysis of a Multimodal Public Dataset.

Background: Early and objective identification of cognitive decline in aging populations remains a clinical challenge. Pupillary light reflex (PLR) and eye movement parameters represent non-invasive, quantitative biomarkers of autonomic and central nervous system integrity, yet their diagnostic utility for cognitive impairment in community-dwelling older adults, particularly in those with mild or borderline impairment (predominantly GDS-Stage 2), remains underexplored. Methods: This cross-sectional study analyzed 383 community-dwelling older adults (mean age 69.78 ± 6.29 years; 68.7% female). Ten PLR parameters (n = 202 with complete PLR measurements) and ten eye movement parameters were measured. Associations with cognitive decline (deterioration grade, GDS 2-4) were evaluated using Spearman correlation analysis and multivariate linear regression (adjusted for age, sex, BMI, and hypertension). Stratified analyses and ordinal logistic regression sensitivity analysis were performed to assess robustness. FDR correction (Benjamini-Hochberg) was applied for multiple comparisons. Predictive modeling was conducted using ElasticNet regression with 5-fold cross-validation. Results: After FDR correction, resting pupil diameter (ρ = -0.47, q < 0.001), constriction amplitude (ρ = -0.40, q < 0.001), mean constriction velocity (ρ = -0.36, q < 0.001), mean dilation velocity (ρ = -0.36, q < 0.001), and all eye movement velocity parameters (ρ = -0.22 to -0.41, q < 0.001) demonstrated significant negative correlations with cognitive decline. Multivariate regression confirmed resting pupil diameter (β = -0.286, q < 0.001) and constriction amplitude (β = -0.223, q < 0.001) as independent predictors. Sensitivity analysis using ordinal logistic regression yielded consistent results. Predictive modeling yielded modest performance for the primary outcome (PLR-only cross-validated R2 = 0.184), whereas models using eye movement features alone or in combination with PLR features performed near chance (R2 ≤ 0.04) or showed instability, indicating that these parameters are not yet suitable as standalone diagnostic tools. Exploratory analyses of depression and anxiety were limited by floor effects (≥89% zero scores). Conclusions: PLR and eye movement parameters show significant negative associations with cognitive decline in older adults, particularly in a sample skewed toward mild impairment (predominantly GDS-Stage 2). These findings provide preliminary, hypothesis-generating signals that warrant validation in clinical samples with broader cognitive impairment distributions, and these parameters should not yet be considered standalone diagnostic biomarkers.