Abstract 009: Incremental Predictive Utility Of Subclinical Disease Measures For Cardiovascular Disease Risk

Abstract

Introduction: The Framingham Risk Score (FRS) has been widely used to predict cardiovascular disease (CVD) risk. However, a comparison of the incremental prognostic utility of different subclinical disease (SubDz) measures is not clear. Methods: We evaluated participants aged 40-79 years (mean age 55 years, 56% women) from the Framingham Offspring (Exam 8, 2005-2008) and Third Generation cohorts (Exam 1, 2002-2005), free of CVD and diabetes, with data on coronary artery calcium (CAC, n=2497), and two measures of target organ damage: urine albumin-to-creatinine ratio (UACR, n=4011) and left ventricular mass (LVM, n=3770). We categorized FRS: <10%, 10-19%, and ≥20% and defined high CAC as CAC≥100, microalbuminuria (MA) as UACR ≥25mg/g in men and ≥35mg/g in women, and left ventricular hypertrophy (LVH) as LVM/body surface area>115 g/m 2 (men) and >95g/m 2 (women). We created 6 cross-classified groups: FRS <10%-No SubDz; FRS <10% + SubDz; FRS 10-19%-No SubDz; FRS 10-19% + SubDz; FRS ≥20%-No SubDz; and FRS ≥20% + SubDz. We related the groups to CVD risk using Cox regression adjusting for age, sex, and cohort and plotted Kaplan-Meier curves to display CVD cumulative incidence by each SubDz cross-classified group. Results: Over a median follow-up of 12 years, 7% of participants developed CVD. Comparing FRS 10-19%-No SubDz and FRS 10-19% + SubDz to FRS <10%-No SubDz (referent), we observed hazards ratios (95% CI) for CVD of 1.68 (0.99-2.83) and 6.50 (3.64-11.61) for high CAC; 1.33 (0.95-1.85) and 2.15 (1.10-4.18) for MA; and 1.43 (0.99-2.07) and 2.18 (1.33-3.57) for LVH. Each SubDz measure predicted CVD risk incrementally over the FRS. In a sub-sample with all three SubDz measures, the model c-statistic with FRS only was 0.725, increasing to 0.773, 0.726, and 0.728 when adding CAC, MA, and LVH, respectively. Conclusion: Presence of a high CAC score outperformed other measures of target organ damage (MA or LVH) for predicting CVD risk, regardless of FRS. Additional studies of larger multi-ethnic samples are warranted to confirm our findings.