A new study published in the Annals of Internal Medicine[1] confirmed something that ought to be obvious: predicting the future is hard—especially when it comes to cardiovascular events.
We know cardiovascular disease is the number-one killer of humans; we know its first manifestation is often heart attack, stroke, or death; and we know all medical therapy comes with trade-offs. Medical treatment of healthy people in the name of preventing something that may or may not happen in the future is dicey. Think do no harm.
That is where risk prediction comes in. You have to know the odds of something (or nothing) happening without treatment. The gamble of statins and aspirin, for instance, looks most favorable in patients who are most likely to have an event.
But where to draw that line, at what future risk is it worth taking a chemical, is the issue at hand. The extreme cases are easy. Most everyone agrees that statins and aspirin provide enough benefit in patients who have suffered a cardiovascular event. For secondary prevention, future risk is high, so benefits outweigh harms. It's the opposite in very low-risk patients. The middle ground is not so easy.
Here is where we have to consider the tools—calculators—to predict future risk. We know certain conditions, such as age, gender, blood pressure, diabetes, smoking, biomarkers, family history, and coronary calcium, contribute to future risk. Numerous expert panels, including the American College of Cardiology and American Heart Association, have compiled different calculators to predict the future. The ACC/AHA risk calculator for atherosclerotic CVD (ASCVD) has sparked debate because many experts believe it overestimates risk.[2]
The Annals Study
Dr Andrew DeFillippis (University of Louisville, KY) and a team of Multi-Ethnic Study of Atherosclerosis (MESA) coinvestigators used this community-based, sex-balanced, multiethnic cohort to compare the calibration and discrimination of the new ASCVD risk score with alternative risk scores.
They compared the observed and expected events for the ASCVD score with three Framingham-based scores and the Reynolds risk score in 4227 MESA subjects aged 50 to 74 years over a 10-year follow-up.
Using this real-world population, they found four of the five risk scores overestimated risk. Calibration was worse in men: overestimates ranged from 37% to 154%. In women, three of four scores overestimated risk by 46% to 67%, and the Reynolds Risk score underestimated risk by 21%.
It's worth saying this another way: when the ACC/AHA ASCVD score predicted event rates of 7.5 to 10%—a range deemed above the statin-benefit cutoff—the actual events were just 3%.
Speaking by phone (we live in the same city), lead author Dr DeFillippis explained to me the important business of looking only at untreated patients. He described their sensitivity analysis, which excluded all patients who received aspirin or any lipid-lowering or antihypertensive drug. To lessen the chance of bias, they analyzed this drug-free group of 790 patients separately and found the same overprediction.
The authors concluded that if these findings are validated, overestimation of ASCVD risk may have substantial implications for individual patients and the healthcare system.
Comments
In our conversation, Dr DeFillippis emphasized the strengths and limits of this study.
The strengths include the fact that the MESA cohort is America. The multiethnic group of patients derived from six cities are like those seen by most physicians. "America is an ethnic melting pot," said DeFillippis.
MESA is also a modern-day cohort study. Dr DeFillippis noted that the risk scores they studied and the ones recommended for use today were validated in another era of medicine. It was not surprising, then, that they found overestimation of risk.
On that modern theme, Dr DeFillippis made an interesting point to me about the overall best-performing Reynolds Risk score. He noted the Reynolds score uses genetics (family history) and CRP (inflammation) levels to predict the future. Bookmark that for the future—genetics and inflammation, that is.
Dr DeFillippis was cautious and balanced—he allowed for the fact that one of the possible explanations of overestimation was that they missed actual events. He said they made every effort to find all cardiac events, but it is possible that some were missed. He also tempered his conclusions by saying these findings should be validated by other work with modern cohorts.
I won't be so measured.
These findings have major implications. Drugs are not free. Aspirin and statins come with side effects and dollar costs. The patient who takes these drugs in hopes of preventing future events makes the gamble that the costs are worth the benefit. Policy makers who recommend these drugs expose millions of people to a therapy that turns on delicate balance between future benefit and harm.
We recently learned that statin-proponent Dr Rory Collins (Oxford University, UK) will now examine patient-level safety data to assess for statin side effects. According to UK newspapers, he has not viewed the actual data, which seems unusual, given his status. This study suggests that published data and patient-level data can differ.[3]
The final point to make is that the use of statins and other drugs for the prevention of future events is not a doctor's or professional society's decision. The human being who swallows a drug must ultimately decide whether the gamble is favorable. Consider, just for argument's sake, the possibility of over-the-counter statin drugs. Here, the consumer would have to make a decision on value. To do that, she would balance the benefits, harms, and costs, then make the decision that fits best with her goals.
These data inform the decision. As caregivers, our job is to assist patients in that difficult decision—because the matter of treating asymptomatic human beings is far from clear.
JMM
© 2015 WebMD, LLC
Cite this: Statins in Primary Prevention: Welcome to the Gray Zone - Medscape - Feb 20, 2015.
Comments