Dynamic risk prediction for diabetes using biomarker change measurements

Area of Science:

• Biostatistics
• Epidemiology
• Preventive Medicine

Background:

• Dynamic risk models incorporate longitudinal data and disease-free survival for improved health status predictions.
• Static models may not fully capture evolving risk factors over time.

Purpose of the Study:

• To develop and apply a dynamic prediction model for estimating type 2 diabetes mellitus (T2DM) risk.
• To compare the predictive accuracy of dynamic versus static risk models for T2DM.

Main Methods:

• A dynamic landmark model and a static prediction model were employed.
• Predictions of 2-year diabetes-free survival were updated at 1, 2, and 3 years post-baseline.
• Data from 2057 participants in the Diabetes Prevention Program (DPP) study were analyzed.

Main Results:

• The dynamic landmark model showed good prediction accuracy (AUC 0.645–0.752).
• While AUC did not significantly differ from static models, the dynamic model yielded significantly better Brier Scores.
• The dynamic model's Brier Scores were notably lower at 1, 2, and 3 years post-baseline.

Conclusions:

• Dynamic prediction models utilizing repeated risk factor measurements can enhance future health status predictions.
• These models show promise for more accurate risk assessment in conditions like T2DM.