Whole-body scans improved type 2 diabetes risk prediction
After clinical screening, an imaging model reduced false alarms by 43.0% while retaining 97.3% of correctly flagged future diabetes cases.
Our read
Biomarkers- Studied:
- Cohort study of 46,129 UK Biobank participants without diabetes, against clinical screening alone
- What changed:
- Adding imaging to clinical screening reduced false alarms by 43.0% while retaining 97.3% of correctly flagged future diabetes cases
- The catch:
- Calibration—whether predicted diabetes probabilities matched observed rates—was not reported in the supplied text
Editor of record: Boris Djordjevic · AgingResearch.News
Among 46,129 UK Biobank participants without diabetes at scanning, a retrospective cohort study assessed whether whole-body magnetic resonance imaging could predict later type 2 diabetes. During follow-up, 600 participants developed the disease. Automated image analysis measured organ sizes and the proportions of fat in organs and muscles.
Eight imaging measures independently predicted later diabetes after adjustment for other factors. They included a liver volume measure, pancreatic fat proportion and fat proportion in the buttock muscles. A computer model trained on the imaging measures improved risk classification beyond the Leicester risk assessment score, an established clinical risk score. The authors suggested that muscle fat proportion may better reflect metabolic risk than muscle volume in areas rich in fat between muscles.
Why it matters
Type 2 diabetes is an age-related disease involving several organs. The work addresses whether measuring organ size and fat distribution can help identify risk before diagnosis.
Caveats
Predictive performance in this UK Biobank cohort does not establish how well the model would work in other populations. The findings concern people without diabetes at scanning, not those already diagnosed.
Comments
Be the first to comment.