Researchers at Mohamed bin Zayed University of Artificial Intelligence have developed a groundbreaking artificial intelligence system capable of predicting cardiovascular mortality risks up to twelve years before symptoms manifest. The innovative technology, named GluFormer, leverages continuous glucose monitoring data to identify subtle metabolic patterns that conventional blood tests consistently overlook.
Published in the prestigious journal Nature, the study co-led by Professor Eran Segal demonstrates how wearable glucose monitors—typically utilized by diabetes patients—can uncover hidden health dangers years before clinical signs emerge. The AI model analyzed over 10 million glucose measurements collected from 10,812 participants, most without diabetes, tracking readings at 15-minute intervals to capture comprehensive metabolic patterns.
GluFormer achieved remarkable predictive accuracy, identifying 69% of cardiovascular deaths within its highest-risk category while recording zero fatalities in the lowest-risk group during the study period. ‘Traditional blood tests act like a single still frame,’ Professor Segal explained. ‘GluFormer analyzes the entire feature film of your metabolic life.’
The research holds particular significance for the UAE, where recent reports indicate approximately 40% of both adults and children face obesity-related health challenges, contributing to increased prevalence of diabetes, heart disease, and cancer.
The AI system functions by detecting ‘risk trajectories’—patterns revealing how the body manages energy during daily activities, including post-meal responses and sleep metabolism. Rather than focusing on isolated readings, GluFormer assesses continuous glucose dynamics to generate personalized risk forecasts. Surprisingly, 40% of participants classified as ‘normal’ through traditional fasting glucose tests exhibited patterns consistent with prediabetes when monitored continuously.
For predictive purposes, participants only required 10-14 days of continuous glucose monitoring. In comparative assessments, GluFormer outperformed HbA1c—the current clinical standard—by identifying 66% of future diabetes cases among prediabetic individuals. Beyond cardiovascular risk, the model successfully forecasted indicators related to visceral fat accumulation, kidney function, liver health, and lipid profiles years in advance.
An enhanced version incorporating dietary data alongside glucose readings demonstrated improved prediction accuracy for over 90% of participants regarding eating patterns. The long-term vision involves creating a ‘digital twin’ of individual metabolism—a virtual model simulating how lifestyle modifications might influence future health outcomes.
While the scientific validation is complete, Professor Segal notes that widespread clinical implementation requires additional trials and healthcare infrastructure upgrades to manage continuous data streams. Given UAE’s direct involvement in this pioneering research, local institutions are optimally positioned to participate in subsequent validation studies as predictive medicine advances toward practical application.