Stratification of individuals without prior diagnosis of diabetes using continuous glucose monitoring

An unmet need for preventing diabetes complications is the early detection of metabolic dysregulation. While glucose dynamics provide high-dimensional insights into metabolic states, extracting comprehensive and interpretable information from such data remains challenging. Here we show that the majority of inter-individual variation in glucose dynamics can be captured by just three features - mean, variance and autocorrelation - each independently associated with diabetes-related measures, even in individuals without a prior diabetes diagnosis. Analysis of continuous glucose monitoring data from 8,025 individuals showed that these three measures explained over 80% of the inter-individual variation in glucose dynamics. These measures outperformed conventional measures, including fasting, mean, and 2-hour postprandial glucose levels, in reconstructing postprandial glucose dynamics. Each feature showed independent associations with vascular or hepatic status. By condensing high-dimensional glucose dynamics into three interpretable features with minimal loss of information, this framework provides a basis for a more accurate diabetes risk assessment.