Researchers at the University of Tokyo have used an artificial intelligence model to show that insulin resistance, estimated from standard health checkup data, is a risk factor for 12 types of cancer in more than 500,000 UK Biobank participants.
The model, called AI-IR (artificial intelligence–derived insulin resistance), predicts how resistant a person’s cells are to insulin using nine routine clinical measurements collected during typical medical exams. Direct tests of insulin resistance usually require specialized procedures available only in advanced diabetes clinics, so AI-IR effectively substitutes for these tests at population scale. In earlier work, the tool outperformed body mass index (BMI), metabolic syndrome criteria and other common markers in predicting diabetes, which gave the team confidence to test its link to cancer.
Doctors often rely on BMI as a stand‑in for insulin resistance, but BMI can misclassify people who are obese but metabolically healthy or lean individuals who still have insulin resistance. By combining nine clinical parameters into a single score, AI-IR captures insulin resistance that BMI alone cannot explain and showed strong agreement with directly measured insulin resistance in validation datasets. The researchers argue this makes AI-IR a more precise way to identify people whose metabolic status puts them at higher cancer risk.
Using AI-IR on about half a million UK Biobank participants, the team found that people flagged as insulin‑resistant by the model had higher incidence of 12 different cancer types, even when they did not yet have diabetes. The analysis provides the first population‑scale evidence that insulin resistance itself, not just obesity or diabetes diagnoses, is an independent risk factor for multiple cancers. Elevated AI-IR scores were also associated with greater risks of diabetes and cardiovascular disease, underscoring insulin resistance as a common thread in chronic illness.
Because AI-IR runs on standard health checkup data, it could be deployed widely to identify high‑risk individuals and prioritize screening for diabetes, heart disease and cancer without requiring special tests or clinic visits. The model was validated in independent cohorts from the United States and Taiwan before being applied to the UK Biobank, suggesting it may be robust across populations. The research team is now studying how genetic differences shape insulin‑resistance‑related cancer risk, aiming to connect large‑scale human data with molecular biology to develop strategies to prevent or reverse insulin resistance.
