Plasma Proteomic Dynamics Preceding Glaucoma Reveal a 15-Year Pre-Diagnostic Window: Causal Insights and Predictive...

To characterize the temporal dynamics of plasma proteomic changes preceding glaucoma diagnosis, identify causal proteins, and evaluate their predictive utility for early detection.

We conducted a prospective cohort study of 45,850 UK Biobank participants without baseline glaucoma, followed for a median of 16.26 years. Plasma levels of 2920 proteins were measured using the Olink Explore 3072 platform. Cox proportional hazards models identified proteins associated with incident glaucoma. Mendelian randomization (MR) using cis- protein quantitative trait loci established causal relationships. Temporal trajectories were modeled using LOESS regression, and time-stratified machine learning models were developed to assess predictive performance.

During follow-up, 977 incident glaucoma cases were identified. After comprehensive adjustment, 136 proteins were significantly associated with glaucoma risk (false discovery rate < 0.05), with EDA2R showing the strongest association (hazard ratio [HR] = 1.21, 95% confidence interval [CI] = 1.16-1.25, P = 2.99 × 10-17). MR analysis identified four causal proteins: LTBP2 (odds ratio [OR] = 1.52, 95% CI = 1.36-1.71, P = 1.07 × 10-12), NRP2 (OR = 0.85, 95% CI = 0.78-0.92, P = 4.73 × 10-5), TNFSF13, and HAVCR1-implicating extracellular matrix remodeling and immune dysregulation in disease pathogenesis. Proteomic dysregulation commenced 12 to 15 years before clinical diagnosis, with three distinct temporal patterns identified. Time-stratified predictive models incorporating these signatures achieved an area under the curve of 0.803 (95% CI = 0.772-0.837) for up to two-year prediction, a 14.65% improvement over demographic models.

This study reveals a 15-year window of detectable plasma proteomic dysregulation preceding glaucoma diagnosis. The identified causal proteins, particularly LTBP2, provide mechanistic insights and represent potential therapeutic targets. The strong predictive performance demonstrates the translational potential of these findings for risk-stratified screening.