A new study using an artificial intelligence tool suggests insulin resistance (IR) is associated with higher risk of multiple cancers, and can be detected even in people with normal BMI. Published February 16 in Nature Communications, the research found people the AI flagged as insulin-resistant had about a 25% higher overall risk of developing 12 different cancers, and a 134% greater risk of uterine (endometrial) cancer.
The model, called AI-IR, was developed by teams at the University of Tokyo and Taichung Veterans General Hospital. It uses nine routinely collected inputs — age, sex, race, body mass index (BMI) and five common blood tests (including hemoglobin A1C) — and applies machine learning to predict insulin resistance. The investigators chose these variables because primary care clinicians commonly record them, which could make real-world implementation feasible if the tool is validated for clinical use.
AI-IR identified elevated risks not only for cancer but also for diabetes and heart disease. The six cancers most strongly linked to predicted insulin resistance were uterine, kidney, esophagus, pancreas, colon and breast. Six additional cancers showed weaker but statistically significant associations: renal pelvis, small intestine, stomach, liver and gallbladder, leukemia, and bronchial and lung cancers.
The strongest association was with uterine cancer. That relationship is consistent with prior evidence connecting endometrial cancer to excess weight and metabolic dysfunction, but AI-IR added predictive power beyond BMI alone. Even after adjusting for body weight, the model flagged some patients at increased uterine cancer risk, suggesting metabolic dysfunction itself may raise risk independent of body size. Mechanistically, insulin resistance could promote tumor development through altered hormonal and growth-factor signaling.
A notable advantage of the AI-IR approach is its ability to detect metabolic dysfunction in people with normal BMI who might be considered low risk by standard screening. The study found the model identified elevated lung and bronchial cancer risk independently of BMI. Outside experts highlighted that normal-weight individuals with high body fat or high insulin levels — a phenotype sometimes called normal-weight obesity — can have substantially increased cancer risk and can be missed by BMI-based screening alone.
The model was trained on datasets from U.S. and Taiwanese populations and validated on nearly 400,000 participants in the United Kingdom. Because the UK validation cohort was predominantly of European ancestry, the authors note limitations in generalizing results to all ethnic groups and emphasize the need for broader validation before clinical adoption.
AI-IR is not yet available for clinical use, but the markers it relies on — for example hemoglobin A1C and body fat percentage — are measurable in routine care. The study and commentators suggest practical steps clinicians and patients can take now to assess and reduce metabolic and cancer risk:
– Measure body composition: ask about body fat percentage. DEXA scans are most accurate; bioelectrical impedance (BIA) scales are an accessible option. Monitor trends rather than a single reading; many experts suggest aiming for total body fat below roughly 30%, though individual targets vary.
– Rethink bloodwork: routine hemoglobin A1C testing for diabetes screening can also indicate metabolic risk. An A1C in the 5.5–5.7% range or higher may signal elevated cardiometabolic — and possibly cancer — risk worth discussing with a clinician.
– Improve lifestyle factors: more physical activity, a healthier diet, and weight loss when appropriate reduce insulin resistance and lower risk for diabetes, heart disease and multiple cancers.
Researchers and oncologists view this work as an important step toward more personalized cancer risk prediction that moves beyond BMI to include metabolic drivers. If further validated and implemented, tools like AI-IR could help identify at-risk patients earlier and guide interventions to prevent diabetes, cardiovascular disease and cancer.
