Predicting the Risk of Developing Type 1 Diabetes Using a One-Week Continuous Glucose Monitoring Home Test With Classification Enhanced by Machine Learning: An Exploratory Study
Authors' conclusion
A machine learning technology, combining a potentially self-administered one-week CGM home test, has the potential to reliably assess the T1D risk.
Study context
- Population
- Healthy relatives of people with T1D; mean ± SD age 23.8 ± 10.5 years, HbA1c 5.3% ± 0.3%, BMI 23.2 ± 5.2 kg/m²; one-week home continuous glucose monitoring (CGM) with three standardized liquid mixed meals
- Participants
- 42 relatives
- Comparison
- High risk (≥2 autoantibodies (Ab)) vs low risk (zero Ab) relatives; CGM-based machine learning risk classifiers
- High riskn = 21
- Low riskn = 21reference
- Outcome
- Classification of high vs low T1D risk (Ab status) by machine learning classifiers using CGM glycemic features, evaluated by AUC-ROC
Methodological detailsTrialNet ancillary study, machine learning, 10-fold cross-validation
- TrialNet ancillary study
- Machine learning
- 10-fold cross-validation
Performance
As reported in the abstract
Classification of low-risk vs high-risk individuals
Best model: linear SVM with Recursive Feature Elimination (RFE) features, AUC-ROC
0.88
Further findings
| Outcome | P value |
|---|---|
| Glycemic features from 2-h post-meal CGM traces, high vs low riskBetween High risk and Low risk · percent time of glucose >180 mg/dL (T180) | .040 |
| Glycemic features from 2-h post-meal CGM traces, high vs low riskBetween High risk and Low risk · glucose range | .035 |
| Glycemic features from 2-h post-meal CGM traces, high vs low riskBetween High risk and Low risk · glucose CV | .028 |
Glycemic features from 2-h post-meal CGM traces, high vs low risk
- T180, glucose range and glucose CV were the only significant differences between groups
Original publication
Predicting the Risk of Developing Type 1 Diabetes Using a One-Week Continuous Glucose Monitoring Home Test With Classification Enhanced by Machine Learning: An Exploratory Study. J Diabetes Sci Technol. Published 9 November 2023. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →