For decades, diabetes diagnosis has relied on measuring blood sugar levels, but this approach misses millions. The rising global prevalence—14% of adults now live with diabetes, up from 7% in 1990—highlights a critical need for better tools.

The danger isn't just the disease itself; it's the silent damage that accumulates over years before diagnosis. Persistently high blood sugar increases the risk of heart disease, stroke, kidney failure, blindness and nerve damage. The earlier the condition is identified, the greater the chance of preventing these complications.

Researchers are exploring new methods to identify diabetes earlier. At Stanford University, Michael Snyder’s team uses continuous glucose monitors (CGMs) to detect hidden metabolic patterns before conventional diagnosis. Their AI-powered algorithm can spot different forms of Type 2 diabetes with around 90% accuracy.

In the UK, Fu Siong Ng and Arunashis Sau at Imperial College London have developed an AI system that analyses electrocardiograms (ECGs) to identify people at higher risk years before blood sugar rises. The tool predicts future risk in diverse populations with around 70% accuracy, offering a scalable solution for early detection.

Both tools could become part of routine preventative health care, allowing individuals to take preventive measures and reducing long-term risks. However, type 1 diabetes poses a different challenge due to its autoimmune nature; by the time blood sugar levels are high enough for diagnosis, significant beta cell damage has likely occurred.