Evaluation of the Pooled Cohort Equations for Prediction of Cardiovascular Risk in a Contemporary Prospective Cohort

• Published in: The American journal of cardiology Vol. 119; no. 6; pp. 881 - 885
• Main Authors: Emdin, Connor A., Khera, Amit V., Natarajan, Pradeep, Klarin, Derek, Baber, Usman, Mehran, Roxana, Rader, Daniel J., Fuster, Valentin, Kathiresan, Sekar
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 15.03.2017; Elsevier Limited
• Subjects: Aged; Aged, 80 and over; American Heart Association; Cardiovascular; Cardiovascular Diseases - epidemiology; Cohort Studies; Demography; Female; Health risk assessment; Humans; Low density lipoprotein; Male; Middle Aged; Practice Guidelines as Topic; Predictive Value of Tests; Risk Assessment - methods; Risk Factors; Societies, Medical; United States - epidemiology; Veins & arteries; Womens health; United States > US
• ISSN: 0002-9149; 1879-1913; 1879-1913
• DOI: 10.1016/j.amjcard.2016.11.042

Most guidelines suggest a baseline risk assessment to guide atherosclerotic cardiovascular disease (ASCVD) prevention strategies. The American Heart Association/American College of Cardiology Pooled Cohort Equations (PCEs) is one tool to assess baseline risk; however, the accuracy of this tool has been called into question. We aimed to examine the calibration and discrimination of the PCEs in the BioImage study, a contemporary multiethnic cohort of asymptomatic adults enrolled from 2008 to 2009 in the Humana Health System in Chicago, Illinois, and Fort Lauderdale, Florida. Our primary end point was hard ASCVD, defined as cardiovascular death, myocardial infarction, and stroke. A total of 3,635 adults who were not on lipid-lowering therapy at baseline were followed for a maximum of 4.6 years. The mean age was 68.6 years; 2000 (55%) participants were women and 935 patients reported being of non-white race (26%). Although 74 ASCVD events were observed over a median follow-up of 2.7 years, 198 events were predicted by the PCEs. The observed event rate was 7.9 per 1,000 participant-years (95% confidence interval [CI] 6.1 to 9.8), whereas the predicted rate by the PCEs was 21 per 1,000 participant-years (95% CI 20.7 to 21.8). This represents an overestimation of 167% (Hosmer–Lemeshow chi-square = 173; p <0.001). With regard to discrimination, the C-statistic of the PCEs was 0.65 (CI 0.58 to 0.71). In an analysis restricted to 3,080 participants without diabetes mellitus and with low-density lipoprotein cholesterol between 70 and 189 mg/dl, the PCEs similarly overestimated risk by 181% (152 predicted events vs 54 observed events; p <0.001). The PCEs substantially overestimate ASCVD risk in this middle-aged adult insured population. Refinement of existing risk prediction functions may be warranted.