Tuesday, March 3, 2015

ACC/AHA and Framingham calculators overestimate cardiovascular risk

- Kenny Lin, MD, MPH

After more than a decade of titrating medications to low density lipoprotein cholesterol targets, family physicians who have transitioned to the 2013 American College of Cardiology / American Heart Association cholesterol treatment guideline now base treatment decisions on a patient's estimated 10-year risk for a cardiovascular event. Although it endorsed the ACC/AHA guideline last year, the American Academy of Family Physicians expressed concern that the guideline's new risk calculator had not been validated in contemporary U.S. populations and could potentially overestimate risk compared to the venerable Framingham calculator.

An analysis published in the Annals of Internal Medicine compared predicted risk scores from the ACC/AHA calculator and four other cardiovascular risk calculators (three derived from the Framingham Heart study) to actual cardiovascular events observed in the Multi-Ethnic Study of Atherosclerosis (MESA), a diverse cohort of adults recruited from six U.S. communities in 2000 to 2002 and followed for ten years. The authors found that both the ACC/AHA and Framingham-derived risk calculators overestimated cardiovascular risk by 37 to 154 percent in men and 8 to 67 percent in women. The Reynolds Risk Score, which includes a measurement of high-sensitivity C-reactive protein, was the most accurate at predicting cardiovascular risk in the MESA cohort, underestimating events by 3 percent overall.

A previous AFP editorial criticized the ACC/AHA guideline for recommending a statin for primary prevention in patients with 7.5 percent 10-year cardiovascular risk, noting that personal estimates of potential benefits of statin therapy relied on a calculator with a "nontrivial margin of error." Nontrivial, indeed. The Annals analysis found that men in the MESA cohort with a calculated ACC/AHA risk score of 7.5 to 10 percent had an actual event rate of only 3 percent; and just over 5 percent of women with a similar risk score experienced cardiovascular events.

Although statins appear to reduce the risk of future cardiovascular events by the same relative proportions in high-risk and low-risk populations, lower-risk patients will experience lower absolute benefits that may not be outweighed by the inconvenience, expense, and potential side effects of therapy. Compounding this problem, a recent systematic review found that most patients already overestimate treatment benefits and underestimate harms. This new analysis won't lead me to abandon cardiovascular risk calculators, but going forward (or until better ones are developed) I plan to acknowledge their lack of precision in discussions with patients, especially those on the lower end of the range of risk where statins are recommended.