Are Current CVD Risk Prediction Algorithms Applicable in HIV?
Calibration was consistently poor; recalibration did not significantly improve model fit.
Commonly applied cardiovascular disease (CVD) risk prediction functions, consisting of a composite of various risk factors associated with CVD development, are currently inadequate for estimating heart disease risk in men infected with HIV, according to a study published in Circulation.
Investigators applied CVD risk prediction functions to patients enrolled in the partners HIV cohort, an observational cohort of HIV-infected patients receiving longitudinal HIV care in Boston, Massachusetts.
Three previously established CVD risk prediction models were evaluated: Framingham Heart Study functions for hard coronary heart disease (Framingham CHD, with heart coronary disease defined as myocardial infarction or coronary death), the 2008 Framingham global CVD function (Framingham ASCVD), and the 2013 American College of Cardiology/American Heart Association ASCVD function (ACC/AHA ASCVD).
Comparison Cox regression analyses were performed to examine the accuracy of each function for predicting CVD risk. Shared risk factors in all functions included age, diabetes, smoking, blood pressure, total cholesterol, and high-density lipoprotein levels.
During a median follow-up of 4.4 years, researchers studied the CVD risk prediction functions in a total of 1280 HIV-infected men. A total of 80 (6.3%) ASCVD events occurred during follow-up, representing a 5-year incidence rate of 16.7 per 1000 person-years. There was moderate to poor discrimination for the Framingham CHD, ACC/AHA, and Framingham ASCVD functions as demonstrated by low c statistics (0.68, 0.65, and 0.67, respectively).
The observed risk performed better than predicted risk for each function except for >7.5% predicted risk in the Framingham hard CHD function. All functions were deemed to be inadequate fits for the HIV cohort as assessed by the calibration χ2 statistics (Framingham CHD function [χ2 = 13.6, P =.019, 5 degrees of freedom]; ACC/AHA function [χ2 = 23.9, P =.001], 7 degrees of freedom; Framingham ASCVD function [χ2 = 24.6, P =.0004, 6 degrees of freedom]).
Although the approximately 5-year follow-up period may have been long enough to provide convincing results of the risk functions in this cohort, a longer-term follow-up may be necessary to increase the number of events analyzed. Additionally, because women were not included in the partners cohort, the findings from this study can only be generalizable to men infected with HIV.
Despite the study's current limitations, the findings indicate that application of standard “CVD risk functions in clinical practice may not result in accurate CVD risk assessment and that reliance on such measures may fail to identify high-risk HIV-infected individuals who would benefit from aggressive cardiovascular risk modification.”
Triant VA, Perez J, Regan S, et al. Cardiovascular risk prediction functions underestimate risk in HIV infection [published online February 14, 2018]. Circulation. doi:10.1161/circulationAHA.117.028975