Abstract MP074: High-sensitivity C-reactive Protein and Family History in the Prediction of Risk for Coronary Heart Disease: The Atherosclerosis Risk in Communities Study
Background: Reynolds Risk Score (RRS) has been proposed as an improved tool for prediction of coronary heart disease (CHD), adding high sensitivity C-reactive protein (hsCRP) and family history of premature coronary heart disease (FH) to traditional risk factors (TRF). RRS was developed in 2 studies of predominantly Caucasian health professionals and has not been validated in other cohorts.
Aim: To assess whether FH and hsCRP improve risk prediction for CHD beyond TRF in the Atherosclerosis Risk in Communities (ARIC) study.
Methods: hsCRP was measured using an immunonephelometric assay in 7240 Caucasian and African-American adults during ARIC visit 4 (1996-1998). FH was defined as parental history of myocardial infarction (MI) before age 60. CHD was defined as hospitalized MI, CHD death or coronary revascularization. Adults with diabetes, known CHD or stroke before visit 4 were excluded. 10-yr risk of incident CHD was estimated using a Cox proportional hazards model and categorized as 0-5%, >5%-10%, >10%-20%, or >20%. Two baseline models were described: model 1, that included age, sex, systolic blood pressure, total and HDL cholesterol, and current smoking; and model 2, which included model 1 + antihypertensive medication use and race. We calculated area under the ROC curve (AUC), Bayes Information Criterion (BIC), Gronnesby & Borgan goodness of fit test, and net reclassification improvement (NRI) for both models with and without log (hsCRP) and FH. Confidence intervals were calculated using bootstrapping. Analyses were stratified by sex.
Results: Over a median of 11.04 years, 430/2945 men (14.6%) and 253/4295 women (5.9%) experienced events. Addition of hsCRP and FH to baseline model 1 was associated with a modest improvement in AUC and BIC (Table). The NRI was not significant. Further, results were attenuated when antihypertensive medication and race were included in the model.
Conclusions: Addition of hsCRP and FH to a model based on TRF resulted in only modest improvements in CHD risk prediction in the ARIC cohort.
|Statistical measure||Model 1*||Model 1* + log(hsCRP) and FH||Model 2†||Model 2† + log(hsCRP) and FH|
|AUC (95% CI)||0.65 (0.62-0.68)||0.66 (0.64-0.69)||0.66 (0.64-0.69)||0.67 (0.65-0.70)|
|GB p value||0.74||0.61||0.69||0.85|
|NRI (95% CI)||---||0.067 (-0.015-0.133)||---||0.035 (-0.012-0.133)|
|AUC (95% CI)||0.68 (0.65-0.72)||0.69 (0.66-0.73)||0.68 (0.65-0.72)||0.69 (0.67-0.73)|
|GB p value||0.98||0.38||0.79||0.37|
|NRI (95% CI)||---||0.033 (-0.039-0.137)||---||0.014 (-0.044-0.128)|
↵* Includes age, total and HDL cholesterol, SBP, current smoking
↵† Includes age, total and HDL cholesterol, SBP, current smoking, antihypertensive medication use and race BIC: Bayes information criterion; AUC: area under the ROC curve; GB: Gronnesby & Borgan goodness-of-fit test; NRI: net reclassification improvement; CI: confidence interval
- © 2012 by American Heart Association, Inc.