Pooled cohort risk equations developed by the American College of Cardiology and the American Heart Association accurately estimate atherosclerotic cardiovascular disease (ACVD) risk, according to an analysis published online ahead of print March 29 in JAMA. The equations thus may be an appropriate guide for clinicians who must decide whether to recommend statins for particular patients.
“The formulas worked well—the rates of heart attack and stroke observed were similar to those predicted by the formulas,” said Paul Muntner, PhD, Professor of Epidemiology at the University of Alabama at Birmingham School of Public Health. “Additionally, participants who were predicted to have high risk were the ones most likely to have heart attacks and strokes, while those who were predicted to have a low risk had low incidence of heart attacks or strokes.”
The equations were part of cholesterol treatment guidelines published in November 2013 in the Journal of the American College of Cardiology. They were designed to help physicians identify which patients to treat and which patients may not benefit from treatment. Among the concerns that the medical community raised about the guidelines was that the risk equations would overestimate the number of people who would have a heart attack or stroke and thus result in overtreatment.
Applying the Equations to REGARDS Data
Dr. Muntner and colleagues examined data for participants enrolled in the Reasons for Geographic and Racial Differences in Stroke (REGARDS) study to assess the risk equations’ performance. The researchers focused on 10,997 individuals without clinical ACVD or diabetes, with a low-density lipoprotein cholesterol level between 70 and 189 mg/dL, and who were not taking statins. At baseline, the REGARDS researchers conducted computer-assisted telephone interviews to collect information on participants’ age, race, sex, smoking status, comorbid conditions, and use of antihypertensive and antidiabetes medications. Health professionals conducted in-home examinations. Investigators spoke by phone with participants every six months to assess new-onset stroke and coronary heart disease events. Dr. Muntner and colleagues defined the outcome for their primary analyses as the first ACVD event (ie, nonfatal or fatal stroke, nonfatal myocardial infarction, or death resulting from coronary heart disease), which was consistent with the definition used to derive the pooled cohort risk equations.
Predicted and Observed Incidences Were Similar
Participants’ mean age was approximately 62. Of the total population, 37.6% were African American and 40.7% were male. Approximately 35% of participants lived in the stroke belt, and approximately 15% were current smokers. Individuals with higher 10-year predicted ACVD risk were more likely to be older, African American, male, current smokers, and were more likely to take an antihypertensive medication.
For participants in the overall REGARDS population with a 10-year predicted ACVD risk of less than 5%, observed and predicted five-year incidence rates were 2.2 and 2.0, respectively, per 1,000 person-years. In higher 10-year predicted ACVD risk strata, five-year observed risk was lower than predicted risk. For people with predicted risk of 10% or greater, the observed and predicted risks were 12.6 and 17.8, respectively.
Calibration of the equations was better, and overestimation of risk was reduced, among participants for whom statin treatment should be considered based on ACVD risk. Most of the overestimation occurred for participants with a 10-year predicted ACVD risk of 10% or greater. In addition, Hosmer–Lemeshow χ2 analysis indicated good calibration among women, African Americans, and Caucasians. The pooled cohort risk equations performed similarly in the Stroke Belt and in the remainder of the continental United States.
The observed and predicted five-year ACVD incidences per 1,000 person-years were 1.9 and 1.9, respectively, for participants with a 10-year predicted ACVD risk of less than 5%; 4.8 and 4.8, respectively, for individuals with predicted risk of 5% to less than 7.5%; 6.1 and 6.9, respectively, for participants with predicted risk of 7.5% to less than 10%; and 12.0 and 15.1, respectively, for participants with predicted risk of 10% or greater. Among participants with Medicare-linked data, the observed and predicted five-year ACVD incidence per 1,000 person-years were 5.3 and 4.0, respectively, for participants with a predicted risk of less than 7.5%; 7.9 and 6.4, respectively, for participants with predicted risk of 7.5% to less than 10%; and 17.4 and 16.4, respectively, for participants with predicted risk of 10% or greater.
The Guidelines Could Improve the Use of Statins
The authors concluded that the risk equations demonstrated good discrimination and were well calibrated in the population for which they were designed to be used. “We think this is important because there are millions of patients who may benefit from taking statins, and doctors need to identify these patients while not prescribing treatment for patients who may receive little benefit,” said Dr. Muntner.