Continuous glucose monitoring–derived uGMI appears to predict incident diabetic retinopathy more accurately than HbA1c in adults with type 1 diabetes. Published in Diabetologia, the study compared long-term predictive performance of uGMI and HbA1c using data spanning up to seven years prior to retinopathy diagnosis.

The analysis used a previously established longitudinal case–control cohort with extensive continuous glucose monitoring (CGM) records. uGMI showed a markedly stronger relationship with future retinopathy, reflected by a mutual information score of 0.148 compared with 0.078 for HbA1c. ROC analysis supported this finding, with uGMI achieving an AUC of 0.733 versus 0.704 for HbA1c.

Adding both metrics to decision tree models did not enhance prediction, indicating limited complementary value. Machine learning approaches further highlighted the superiority of uGMI, especially within the mid-range HbA1c interval of 54–58 mmol/mol (7.1–7.5%), where retinopathy risk increased sharply.

These results suggest that uGMI offers a modest but meaningful improvement over HbA1c in assessing retinopathy risk, although the two measures do not provide additive predictive benefit when combined.