This study developed and evaluated a scalable model for predicting insulin resistance using consumer smartwatch data, demographics, and routine blood biomarkers, with homeostatic model assessment of insulin resistance used as the ground-truth measure. In the WEAR-ME cohort of 1,165 adults, insulin resistance was associated with higher fasting glucose, body mass index, HbA1c, triglycerides, resting heart rate, and c-reactive protein, and with lower high-density lipoprotein cholesterol, daily step count, albumin/globulin ratio, and heart rate variability. Notably, 20% of normoglycemic participants were classified as insulin resistant, suggesting that insulin resistance can be present before HbA1c rises into a diabetic range. Models combining wearables, demographics, and routine biomarkers outperformed models using any single data source alone, and a wearable foundation model further improved prediction by capturing complex patterns in activity, sleep, heart rate, and physiological rhythms.

The best-performing models showed strong predictive performance in both the original cohort and an independent validation cohort, supporting the potential value of wearables as a scalable screening tool to identify people who may benefit from confirmatory fasting insulin testing and earlier lifestyle intervention. The researchers also developed an large language model (LLM)-based “insulin resistance literacy and understanding” agent that used predicted insulin resistance status, wearable data, and biomarker results to generate personalized metabolic-health explanations and recommendations. Endocrinologist reviewers generally rated the agent’s responses as more comprehensive, trustworthy, and personalized than a base LLM, although interpretation of some wearable and blood biomarker data still needs refinement. Overall, the study suggests that combining wearable signals with routine clinical data could help detect insulin resistance earlier and support more personalized diabetes-risk prevention.

Reference: Metwally AA, Heydari AA, McDuff D, et al. Insulin resistance prediction from wearables and routine blood biomarkers. Nature. 2026;652:451-461. doi:10.1038/s41586-026-10179-2.

This Swedish nationwide cohort study followed children ages 7 to 17 with obesity who entered treatment and compared those with metabolically healthy obesity (MHO), those with metabolically unhealthy obesity (MUO), and matched peers from the general population. Nearly half of the pediatric obesity cohort met criteria for MHO at treatment initiation, meaning they did not have high blood pressure, impaired fasting glucose, elevated alanine aminotransferase, high triglycerides, or low high-density lipoprotein cholesterol. By age 30, children with MHO had lower cumulative risk of type 2 diabetes, hypertension, and dyslipidemia than children with MUO, but their risk was still substantially higher than that of the general population.

The study also found that a favorable response to lifestyle-based obesity treatment, defined as a body mass index z-score reduction of at least 0.25 units, was associated with lower long-term risk of type 2 diabetes, hypertension, and dyslipidemia. Importantly, this protective association was similar in children with MHO and MUO, suggesting that children who appear metabolically healthy still benefit from obesity treatment. The authors conclude that metabolically healthy obesity in childhood should not be viewed as low-risk or as a reason to delay treatment, because obesity itself remains linked to higher cardiometabolic risk in young adulthood.

Reference: Putri RR, Danielsson P, Hagman E, Marcus C. Long-term cardiometabolic outcomes in children with metabolically healthy and unhealthy obesity. JAMA Pediatr. 2026;180(7):750-757. doi:10.1001/jamapediatrics.2026.0343.

This National Health and Nutrition Examination Survey-based cross-sectional study examined the relationship between weekday sleep duration, weekend catch-up sleep, and estimated glucose disposal rate, a marker of insulin sensitivity. Among 23,475 adults, weekday sleep duration showed an inverted U-shaped relationship with estimated glucose disposal rate (eGDR), with the most favorable eGDR observed at approximately 7.32 hours of weekday sleep. Below that threshold, each additional hour of weekday sleep was associated with higher eGDR, suggesting better insulin sensitivity. Above that threshold, longer sleep was associated with lower eGDR, particularly among women, adults aged 40 to 59 years, and individuals with obesity.

Weekend catch-up sleep appeared to have a nuanced role. In people sleeping less than 7.32 hours on weekdays, modest weekend catch-up sleep of up to about 2 hours was associated with higher eGDR, while excessive catch-up sleep greater than 2 hours was linked to worse eGDR patterns. The authors suggest that too much weekend recovery sleep may disrupt circadian rhythms and contribute to metabolic dysregulation, while modest catch-up sleep may help offset insufficient weekday sleep. Because the study was cross-sectional and relied on self-reported sleep, the findings should be interpreted as associations rather than proof of causation, but they support the clinical value of discussing consistent, adequate sleep as part of metabolic and diabetes care.

Reference: Fan Z, Wei R, Chen T, et al. Association of weekday sleep duration and estimated glucose disposal rate: the role of weekend catch-up sleep. BMJ Open Diabetes Res Care. 2026 Mar 3;14(2):e005692. doi: 10.1136/bmjdrc-2025-005692.