The use of repeated blood pressure measures for cardiovascular risk prediction: a comparison of statistical models in...

Area of Science:

• Cardiovascular disease research
• Biostatistics
• Epidemiology

Background:

• Cardiovascular disease (CVD) risk prediction models are crucial in primary care.
• Novel biomarkers and baseline predictors enhance model accuracy.
• The utility of repeated measures for common risk predictors remains underexplored.

Purpose of the Study:

• To compare different methods of incorporating repeated systolic blood pressure (SBP) measures into CVD risk prediction models.
• To evaluate the performance of simple summary measures versus complex modeling techniques.
• To assess improvements in model discrimination and calibration.

Main Methods:

• Comparison of six models using Atherosclerosis Risk in Communities (ARIC) study data and simulations.
• Models included: baseline SBP only, last observation carried forward, cumulative mean SBP, ordinary regression calibration, risk-set regression calibration, and joint longitudinal and survival models.
• Evaluation metrics focused on discrimination and calibration.

Main Results:

• The cumulative mean of repeated SBP measures showed modest improvements in discrimination and calibration compared to the baseline-only model.
• Complex methods (ordinary regression calibration, risk-set regression calibration, joint models) offered little additional benefit over the cumulative mean.
• No substantial improvements were observed beyond the cumulative mean approach.

Conclusions:

• Repeated measures of systolic blood pressure can enhance CVD risk prediction models.
• Simple summary measures, like the cumulative mean, are effective and may be preferable to complex modeling techniques.
• Findings suggest practical strategies for improving CVD risk assessment in primary care using routinely collected data.