Abstract
BACKGROUND: Glaucoma remains the leading cause of irreversible blindness worldwide. The development of predictive, preventive, and personalized medicine (3PM) strategies in the area is essential to address high inter-individual heterogeneity in glaucoma progression, in order to effectively protect stratified patients against disease progression. AIM: This study aims to develop and validate a personalized, multimodal predictive modeling framework that integrates structural, functional, and vascular biomarkers for individualized risk stratification of progression rates in primary open-angle glaucoma (POAG). METHODS: Patients with POAG at varying stages were monitored for at least 36 months and underwent comprehensive multimodal evaluation, including structural optical coherence tomography (OCT), OCT angiography (OCT-A), automated perimetry, and biomechanical assessments. Predictive modeling was performed using Ranked Partial Least Squares Discriminant Analysis (Ranked PLS-DA). Model performance and variable importance were established through Procrustes Cross-Validation and optimization procedures. RESULTS AND DATA INTERPRETATION IN THE FRAMEWORK OF 3PM: The final models included up to 27 parameters in early-stage POAG and 20 in advanced disease, leading to high prognostic accuracy (AUC up to 0.90) for classifying slow, moderate, and rapid rates of glaucoma progression. Feature importance analysis demonstrated that different biomarkers dominate at different disease stages: RNFL thickness, peripapillary microvascular dropout, parafoveal vascular density and corneal hysteresis in early POAG, while age, ganglion cell complex thickness, specific macular thickness measures, and peripapillary perfusion parameters were most predictive in advanced stages. CONCLUSIONS AND 3PM-RELEVANT OUTLOOK: The proposed innovation utilizes multimodal predictive disease modeling that supports accurate risk stratification, personalized glaucoma management and individualized protection against disease progression. Successful clinical application requires initial profiling, regular model recalibration, and adaptive treatment strategies - altogether leading to improved visual outcomes in stratified patients and leveraging resources used.