You are here
Evaluation of the ACC/AHA CVD Risk Prediction Algorithm Among HIV-Infected Patients
Susan Regan1, James B. Meigs1, Joseph Massaro2, Ralph B. D'Agostino2, Steven Grinspoon1, Virginia A. Triant1
1 Massachusetts General Hospital, Harvard Medical School, Boston, MA, United States. 2 Boston University , Boston, MA, United States.
Background: The 2013 American College of Cardiology (ACC)/ American Heart Association (AHA) cardiovascular disease (CVD) risk prediction algorithm (Pooled Cohorts Equations) has not previously been evaluated in HIV populations.
Methods: Framingham Risk Scores (FRS) and ACC/AHA risk scores were calculated for patients in a longitudinal HIV clinical care cohort during a 3-year interval ending January 1, 2009. Patients were not eligible if they were under age 18, had expired prior to January 1, 2009, were missing relevant data to populate the risk score, or had undergone a relevant outcome event prior to the date of risk score calculation. CVD risk was considered high if 10-year predicted risk of the relevant outcome event was ≥10 percent for FRS and ≥7.5 percent for ACC/AHA. Outcome events were coronary heart disease (CHD) for FRS and atherosclerotic CVD (ASCVD) for ACC/AHA.
Results: The FRS was calculated for 2270 patients, with a median follow-up time of 6.3 years, and the ACC/AHA risk score was calculated for 2152 patients, with a median follow-up time of 6.2 years. Risk scores were discordant in 17 percent of patients, with the ACC/AHA score only predicting high risk in 10 percent and the FRS only predicting high risk in 7 percent. In comparisons of these discordant subgroups, patients classified as high-risk by ACC/AHA but low-risk by FRS were older (median age 56 for ACC/AHA high vs. 48 for FRS high) and more likely to be female (68% vs. 0%), diabetic (52% vs. 6%) and black (22% vs. 12%) but less likely to be smokers (44% vs. 66%) than those low-risk by ACC/AHA and high-risk by FRS. Actual event rates were estimated and compared with predicted rates. As shown in the figure, actual 6-year event rates were similar to 10-year predicted rates for the FRS and were similar to or exceeded predicted rates for the ACC/AHA risk score.
Conclusions: Our findings suggest that CVD risk prediction scores designed for the general population, and particularly the new ACC/AHA risk score, may underestimate risk for HIV-infected patients. Accurate CVD risk prediction is an important component of the long-term management of chronic disease complications in HIV.