Study Finds Thigh-Length Home Test Can Identify Diabetes and Prediabetes Risk - Video

AI-powered home test may detect hidden diabetes using a simple thigh measurement

Researchers have developed an artificial intelligence (AI)-based home screening tool that may accurately identify people with undiagnosed type 2 diabetes or prediabetes using simple measurements that can be taken at home—including the unexpected measurement of thigh length. The findings were published in the Journal of Clinical Epidemiology.

The new tool, called MEDWACS, was designed to help identify people who may have diabetes but have not yet sought medical testing. Researchers say early detection is important because type 2 diabetes can remain unnoticed for years while increasing the risk of heart disease, kidney failure, vision loss, and other serious complications.

The AI model was developed using more than 30 years of health data from the U.S. National Health and Nutrition Examination Survey (NHANES). Researchers analyzed nearly 3,700 possible health variables before selecting seven simple measurements that people can collect at home using a tape measure, bathroom scale, and blood pressure monitor.

These include age, sex, body mass index (BMI), blood pressure, and thigh length. The researchers were initially surprised that thigh length emerged as an important predictor.

According to the research team, thigh length may reflect nutritional status during early childhood, which can influence diabetes risk later in life. In addition, the thighs contain the body's largest muscle groups, which play a major role in removing glucose from the bloodstream. Shorter thighs generally indicate lower muscle mass available to help regulate blood sugar.

The researchers validated MEDWACS using independent datasets from the United States and South Korea. The tool accurately identified people with prediabetes or previously undiagnosed type 2 diabetes and performed as well as—or better than—other established diabetes risk assessment methods.

Although the model has not yet been validated using Danish health data, the researchers believe its strong performance across different populations suggests it could be useful more broadly.

REFERENCE: Yoo D, Maggiore U, Jolliet O; Enhancing prediabetes and diabetes detection through a machine learning-enabled self-assessment approach; Journal of Clinical Epidemiology, 2026; DOI: 10.1016/j.jclinepi.2026.112266