AI Tool Predicts Heart Disease Death Risk 12 Years Early Using Blood Glucose Data

Scientists in the United Arab Emirates have created an AI tool that could transform early disease detection. The model, named GluFormer, uses just two weeks of blood glucose readings to predict a person’s risk of dying from heart disease up to 12 years ahead. It also assesses their likelihood of developing diabetes and other metabolic conditions.

Type 2 diabetes often goes unnoticed for years, as symptoms develop slowly and may not appear until serious damage has occurred. By the time a diagnosis is made, the disease may have already harmed the heart, kidneys, or blood vessels. Current risk assessments rely on factors like age, weight, family history, and blood sugar levels, but these methods often miss subtle biological shifts.

Researchers at the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) designed GluFormer to fill this gap. The AI analyses metabolites—molecules in the blood that reflect metabolic health—before blood sugar levels rise enough for a diabetes diagnosis. In testing, 66% of participants flagged as high-risk by GluFormer later developed diabetes, compared to only 7% in the lowest-risk group. The study also found that diet and lifestyle choices influence metabolites linked to type 2 diabetes far more than those unrelated to the condition. This insight could help shape targeted prevention plans and treatments. With roughly one in nine adults worldwide living with diabetes—and over 90% of cases being type 2—the model offers a clearer way to identify at-risk individuals early.

GluFormer outperforms standard tests like HbA1c in spotting high-risk patients for both diabetes and cardiovascular death. The tool could enable earlier interventions, reducing long-term health complications. Its ability to detect hidden risks years in advance may also guide future research into prevention and treatment strategies.