Use of Continuous Glucose Monitoring with Machine Learning to Identify Metabolic Subphenotypes and Inform Precision Lifestyle Changes

TL;DR

This paper reviews how continuous glucose monitoring (CGM) and wearable data can reveal metabolic subphenotypes underlying early dysglycemia that are hidden by static glucose thresholds. By combining frequently sampled oral glucose tolerance tests (OGTT) and high-resolution CGM data with machine learning, the authors demonstrate accurate prediction of muscle insulin resistance, beta-cell dysfunction, and incretin deficiency, including in at-home OGTT settings. Real-food CGM challenges link postprandial glycemic responses to underlying physiology, enabling a potential biomarker based on the potato–grape PPGR difference, while wearables show how lifestyle timing (diet, sleep, activity) differentially relates to subphenotypes. The findings support precision lifestyle interventions and targeted prevention strategies that go beyond glycemic control, with implications for personalized diabetes prevention and metabolic health monitoring.

Abstract

The classification of diabetes and prediabetes by static glucose thresholds obscures the pathophysiological dysglycemia heterogeneity, primarily driven by insulin resistance (IR), beta-cell dysfunction, and incretin deficiency. This review demonstrates that continuous glucose monitoring and wearable technologies enable a paradigm shift towards non-invasive, dynamic metabolic phenotyping. We show evidence that machine learning models can leverage high-resolution glucose data from at-home, CGM-enabled oral glucose tolerance tests to accurately predict gold-standard measures of muscle IR and beta-cell function. This personalized characterization extends to real-world nutrition, where an individual's unique postprandial glycemic response (PPGR) to standardized meals, such as the relative glucose spike to potatoes versus grapes, could serve as a biomarker for their metabolic subtype. Moreover, integrating wearable data reveals that habitual diet, sleep, and physical activity patterns, particularly their timing, are uniquely associated with specific metabolic dysfunctions, informing precision lifestyle interventions. The efficacy of dietary mitigators in attenuating PPGR is also shown to be phenotype-dependent. Collectively, this evidence demonstrates that CGM can deconstruct the complexity of early dysglycemia into distinct, actionable subphenotypes. This approach moves beyond simple glycemic control, paving the way for targeted nutritional, behavioral, and pharmacological strategies tailored to an individual's core metabolic defects, thereby paving the way for a new era of precision diabetes prevention.

Figures (8)

• Figure 1: Deep metabolic profiling via gold-standard quantitative tests were used in 13 to evaluate: (1) Muscle insulin resistance via the Insulin Suppression Test (IST), (2) Beta-cell function by calculating the disposition index from an Oral Glucose Tolerance Test (OGTT) using C-peptide deconvolution, (3) Incretin effect using an Isoglycemic Intravenous Glucose Infusion (IIGI) and an OGTT, (4) Hepatic insulin resistance using a validated formula based on OGTT insulin levels and demographic information.
• Figure 2: Robust performance in predicting muscle IR, beta-cell dysfunction, and incretin effect using glucose timeseries compared to currently-used surrogate markers.
• Figure 3: Study design of the at-home OGTT test via CGM to predict muscle IR and $\beta$-cell function. Participants underwent gold-standard testing at the research unit for insulin resistance (Insulin Suppression Test) and $\beta$-cell function (16-point OGTT with C-peptide deconvolution adjusted for SSPG and expressed as DI) as described, as well as two OGTTs administered at home under standardized conditions during which glucose patterns were captured by a CGM within a single 10-day session (Dexcom G6 pro).
• Figure 4: Study design of the standardized meal study. 55 participants consumed 7 different standardized carbohydrate meals and 3 mitigator foods with rice with replicates while wearing CGM. Participants were grouped by their highest response to carbohydrate and measured metabolic subphenotypes.
• Figure 5: Postprandial glycemic responses to different carbohydrates. Left: Mean CGM curves after different meals. The X-axis is time, with the food log consumption time as 0, and the Y-axis is glucose level. Right: Number of participants classified to each spiker type as defined by both delta glucose peak and AUC(>baseline). The X-axis indicates different carbohydrates, and the Y-axis is the number of participants for whom a given carbohydrate produced the highest spike.