Abstract 16638: Predictive Validity of Common Cardiovascular Disease Risk Scores Overall and in Relation to Race and Gender
Background: Cardiovascular risk score algorithms are routinely used in clinical practice and research. Whereas multiple approaches exist, there is concern related to accuracy of the algorithms when applied across racial and gender subgroups.
Methods: From the Heart Strategies Concentrating on Risk Evaluation (Heart SCORE) study (n=1,949), baseline risk factors were used to calculate cardiovascular disease (CVD) risk scores using published algorithms for Framingham Risk Score (FRS), Reynolds Risk Score (RRS) and Pooled Cohort Risk Score (PCRS). Discrimination and calibration of each risk score was evaluated in relation to a composite CVD outcome (death/MI/acute ischemic syndrome/stroke/revascularization) over long-term follow-up.
Results: Mean age at baseline was 59 years, 65.7% were female, and 43.8% were of black race. All 3 risk scores were able to discriminate participants at lower vs. higher risk of experiencing the composite CVD outcome. However, the mean predicted 10-year risk of CVD was highly variable across algorithms, and was 6.7% using RRS vs. 13.6% using FRS (rate ratio=0.44, p<0.001). The 10-year predicted risk of CVD was intermediate (9.1%) with use of PCRS. The high variability in predicted 10-year CVD risk scores was particularly evident among black females (p<0.001); RRS=4.6%; RRS=8.8%; FRS=11.0%). In calibration analyses that divided each risk score calculation into quintiles (Q), FRS consistently over-predicted observed rates of CVD, whereas RRS showed relatively strong concordance between predicted and observed rates of the composite CVD outcome (table).
Clinical Implications: The recent (2013) PCRS is endorsed by AHA/ACC guidelines, however, our analysis suggests that the RRS more accurately predicts future risk of CVD in a diverse sample of adults. More comparative research across racial and gender subgroups is needed, and clinicians should be aware of the wide variability in risk estimation when counseling and treating patients.
Author Disclosures: J. Wilson: None. C.A. Lengacher: None. P. Athilingam: None. Y. Huang: None. D.M. Diamond: None. S.E. Reis: None. K.E. Kip: None.
- © 2016 by American Heart Association, Inc.