Microsimulation model to predict incremental value of biomarkers added to prognostic models

Area of Science:

• Biostatistics
• Cardiovascular Disease Epidemiology
• Biomarker Discovery

Background:

• Prognostic risk models are crucial for predicting disease outcomes.
• Assessing the added value of new biomarkers to existing models is challenging.
• Simulated data offers a potential cost-effective alternative to real-world data for such assessments.

Purpose of the Study:

• To develop and validate a simulation model for evaluating the incremental value of novel biomarkers in cardiovascular disease (CVD) risk prediction.
• To compare the performance of a simulation model with analyses based on actual patient data.

Main Methods:

• Utilized data from 4522 women and 3969 men from the Framingham CVD risk prediction tool.
• Developed a simulation model to mimic the characteristics of the original dataset.
• Assessed biomarker added value using discrimination, measured by the area under the curve (AUC).

Main Results:

• The simulation model closely replicated results obtained from actual data (AUC: 0.800 vs 0.799 in women, 0.778 vs 0.776 in men).
• Biomarkers positively correlated with existing risk factors showed a smaller impact (ΔAUC 0.002-0.024).
• Biomarkers negatively correlated with risk factors demonstrated a stronger effect (ΔAUC 0.026-0.101) compared to no correlation (ΔAUC 0.003-0.051).

Conclusions:

• Simulation models provide a reliable method for assessing the potential value of new biomarkers in prognostic risk models.
• Researchers are advised to utilize simulation models prior to initiating large-scale, expensive studies using actual data.
• This approach can optimize resource allocation in biomarker research for CVD risk prediction.