Predictive Models of a Future Coronary Artery Calcium Score in a Black Cohort: The GENOA Study

OBJECTIVE: This retrospective study of a Black cohort sought to create predictive models to calculate the probability of a coronary artery calcium score (CACS) about one decade after obtaining cardiovascular risk measures. STUDY DESIGN AND SETTING: Participants (n = 656) in GENOA had CV risk variables measured (1995-2000) and a CACS about one decade later (2009-2011). Using multivariate regression, computer models were written to calculate the probability of a future CACS of zero, ≥ 10, and ≥ 100. ROC values were 0.78, 0.77, and 0.76, respectively. Machine learning models did not perform any better than multivariate regression. RESULTS: Age, height, smoking duration, sex, and hypertension were significant for all three models in predicting a future CACS. Height was inversely related to future CACS, but weight and BMI were not contributory to the models. Lipid-lowering medications and exercise were associated with an increased CACS, the so-called CACS "paradox." CONCLUSION: Predictive models of a future CACS such as these may help in identifying important risk factors for a future CACS. By identifying these risk factors and implementing early modification of CV risk factors, the development of CV disease may be slowed.