Prediabetes Progression Model Validation in U.S. Cohorts

3,081 participants from the Diabetes Prevention Program cohorts were used to externally validate a machine learning model for diabetes progression. The model achieved an area under the ROC curve of 0.74 for predicting 3-year progression to type 2 diabetes. High-risk individuals had a 67% increased risk of microvascular complications.

Use Cases

• Validate the ML-PR model's discrimination and calibration using performance metrics like the area under the ROC curve and calibration curves.
• Stratify patients into high- or low-risk categories based on ML-PR scores to evaluate differential outcomes in type 2 diabetes incidence.
• Analyze interactions between risk group and intervention (lifestyle modification, metformin) using Cox proportional hazards and logistic regression models.
• Assess the model's clinical utility for predicting microvascular complications and cardiovascular events across risk groups.

Strengths

• Validation uses data from 3,081 participants across two established U.S. cohorts (DPP and DPPOS).
• Model performance is quantified with specific metrics, including an area under the ROC curve of 0.74 (95% CI: 0.71–0.78).
• Analysis includes clinical outcomes like diabetes incidence, microvascular complications, and cardiovascular events.

Limitations

• Dataset is a 14.8 MB document; underlying tabular data for the 3,081 participants is not directly accessible.
• Geographic scope is limited to U.S. cohorts, limiting generalizability to other populations.
• The study found no significant difference in cardiovascular risk across risk groups, limiting the model's utility for that outcome.

Provenance

Source

Diabetes Prevention Program (DPP) and DPP Outcome Study (DPPOS) cohorts.

Collection Method

Clinical study data analyzed for external validation of a prediction model.

Time Range

Covers the follow-up periods of the DPP and DPPOS studies.

Freshness

Last updated March 2026.

Geography

United States