Abstract
BACKGROUND: Cardiovascular disease (CVD) is a major cause of mortality in HIV infected patients. Agreement between commonly used risk prediction equations for classification of high-risk individuals is varied in different populations. We aimed to compare the degree of agreement of four CVD risk calculators in a multicenter cohort. METHODS: A cross-sectional study was conducted among adult HIV patients who are followed in five tertiary centers between July 2016 and February 2017. Inclusion criteria were: age 40–74 years, without known CVD and not receiving statins. All necessary information to calculate risk scores were collected during follow-up visits with a standardized form. Web-based tools for each score were used for calculations. Persons were considered at higher risk if 10-year CVD risks ≥20% with FRS-CVD, >10% with SCORE for high-risk countries, >7.5% for ASCVD, and 5 year risk ≥5% with DAD or if they had additional risk factors defined for each score for automatic high-risk stratification. Based on the interpretation of CVD risk, the patients were placed in two categories: low/medium and high/very high. Agreement between scores was assessed by Cohen’s kappa (κ) statistics. RESULTS: Of 667 patients who were active during the study period, CVD scores of 527 HIV-infected patients (82% male) were assessed. Median (interquartile range) age was 48 (43–54) years. Prevalence of CVD risk factors were: 11% family history of early-onset CVD, 50% current smokers, 57% overweight or obese, 22% hypertension, and 8% diabetes mellitus. The prevalence of high CVD scores or risk equivalents was high ranging from 20.3% to 36.3%. The DAD-full, DAD-reduced, ASCVD and SCORE had 83.9%, 85%, 83.5% and 93.2% agreement compared with the FRS-CVD (κ = 0.55, 0.59, 0.61 and 0.80), respectively. European AIDS Clinical Society, European Society of Cardiology, Adult Treatment Panel-III and 2013 American College of Cardiology/American Heart Association guidelines would recommend statin therapy for 35.1%, 21.8%, 31.9% and 36.4% of patients, respectively. CONCLUSION: We found moderate/substantial agreement among risk prediction tools evaluated in this study. Agreement was high for lower scores and at higher ages. Whether those scores accurately estimate risk at population level needs further evaluation. DISCLOSURES: All authors: No reported disclosures.