Stress Tests:  Well Reimbursed, Rarely Helpful

 
"But you-a my doctor!" Mr. Bassimo was a young, well-dressed Italian man who spoke in what sounded to me like bursts of opera. He was visibly distressed about his atypical chest pain, and my words weren't helping. "If-a you can't tell me whether I'm gonna die, then why I come?""That's the problem Mr. Bassimo," I said. "No one can tell you that."

He sat for a moment, and his eyes widened. "How about a stress-a test?"

A few minutes later I was engaged in a similar conversation with a middle-aged Manhattanite. After a lengthy conversation about risks and options, she proposed her own solution: "What about a stress test?"

The stress test is the elemental unit of diagnostic cardiology, and patients know of stress tests all too well. Unfortunately, both doctors and laypeople have been taught and trained wrong. I, for instance, was trained to believe in the general utility of stress testing. This was wrong. I was taught to obtain a stress test history, and to be comforted by normal results. Also wrong. I was taught to believe in the power of stress tests to establish safety and identify disease in low risk chest pain patients before they leave the hospital. Wrong.

Sensitivities and specificities for stress tests are often reported as being between 70% and 90%, but these numbers are misleading. Studies of stress tests have rarely used a proper gold standard (i.e. coronary angiography), and in the one reasonably sized, high quality study to be performed rigorously, the test's sensitivity for coronary stenosis was 45%. It appears that flipping a coin would be a more sensitive mechanism for detecting CAD than relying on a stress test.

Stress tests are also terrible for predicting death or major cardiac events. In 2000, a preventive medicine group published the largest single experience of stress test screening for heart disease, with over 25,000 men (mean age 43). In the nearly ten years that followed the tests, 158 of the men suffered a cardiac death. The tests were completely normal in 40% of them.

Of course, the most important metric for any health care measure is also the simplest: does it help people? Perhaps the most definitive and most important recent study of stress test screening is the 'DIAD' trial, published this year in the Journal of the American Medical Association. The trial randomized diabetics (mean age 61) without known coronary disease, a group considered to be high risk, to either have or not to have a nuclear stress test at baseline. After 5 years an identical number of subjects in each arm of the study had suffered heart attacks or death. In other words, being screened for heart disease with a nuclear stress test (ostensibly the most accurate type of stress test) did not reduce cardiac events or deaths—the tests didn't help people.

To be fair, in the ED we aren't looking for 'stenosis', and we're not wondering whether our patients will die in the next ten years. We're looking to predict short-term risk of heart attacks or deaths. Can stress tests do this? Just weeks before his collapse due to severe three-vessel coronary disease, the venerable Tim Russert had a perfectly normal stress test. This, of course, is because stress tests aren't intended to predict short-term risk. They are designed to identify patients with ischemic, exertional EKG changes. Virtually none of the patients we classify as low risk in the emergency department are likely to fit this description. If they were, they wouldn't be low risk.

Accordingly, in the handful of studies examining outcomes of emergency department (or chest pain center) patients after stress tests, the rate of heart attacks and deaths approaches zero regardless of whether the test was positive or negative—and even regardless of whether they had a stress test.

Whether you're a frustrated Italian or Manhattanite, stress tests can rarely, if ever, help you. In the rare case where a stress test might be appropriate (e.g. a diabetic, 75-year old smoker with diaphoresis and squeezing chest pressure when he walks the third flight), the patient probably qualifies for more than just a stress test. Like many technological staples of modern medicine, it is an idea that we instituted before we ever used proper research to determine whether it would help people. It's a lot like bloodletting.

Some people are going to have a heart attack, or get cancer, or be hit by a car, and no test will efficiently identify them, not even the ubiquitous (and well reimbursed) stress test. This lack of certainty is something that both our patients and we will have to live with, and it is something that we should be telling them. No one's risk is zero, and there's one thing we know for certain: there's no use in stressing.

Dr. David Newman is the director of clinical research at Columbia University and author of the bookHippocrates' Shadow (Scribner, $15)

 
 
by Kevin Klauer, DO
{mospagebreak title=A Good Test Often Misapplied by Dr. Klauer}
The Stress Test: A Good Test Often Misapplied
by Kevin Klauer, DO
 
Are stress tests bad? Maybe. However, any test in the wrong hands can be misapplied and its value lost. Stress tests have their limitations, but applying the evidence guides us to rational use, avoiding overreliance on their results. The reality is that we have bad tests, good tests and great tests. Bad tests add little to nothing to the clinical picture, whereas great tests may clinch a diagnosis for us, avoiding false positives and false negatives. Good tests lie somewhere in between. If applied properly, their results may help us differentiate one illness or injury from another. There just aren't that many great tests at out disposal. Unfortunately, there are way too many bad tests, like a WBC count used to detect infection (e.g. appendicitis). Bad tests are usually not specific or sensitive for anything. Could you complete your career never ordering another WBC count? Sure you could. Could you make it one day without ordering a troponin? Probably not.  We need to do away with the bad, take advantage of the great tests and learn how to apply the good tests, based on the available data.

Stress tests are not great tests. They fall into the good test category. If the test is performed in the appropriate circumstances and the results applied in the proper context, the information can be valuable. They're not perfect. However, they are far from worthless. There is a long list of bad tests I would axe before I would discard stress tests.

Exercise treadmill tests? OK. I'll concede. They aren't very good tests.  50% of women and 25% of males with reversible perfusion defects detected by nuclear stress tests had a normal ETT (Hoilund- Carlsen, P.F., et al, Am J Card 95:96, January 1, 2005). Also, patients on sulfonylureas may have false negatives, as this class of drugs attenuate ST-segment change, making a standard ETT a waste of time. So, is an ETT good for anything? Actually, it is. They are fairly good predictors of cardiac events within thirty days, making low risk patients with a normal ETT very unlikely to experience an event  within a month. Such information can be helpful, with respect to risk stratification and in determining which patients should be admitted and which might be safe to discharge for an outpatient evaluation. From the patient's perspective, it might be helpful with vacation planning. If you fail your ETT, don't by tickets for the 4-week Mediterranean cruise! In June, 2008, Tim Russert, "Meet the Press" host, died of sudden cardiac death. He had a stress test two months prior. It appears that this may have been a standard ETT. If so, the test performed as predicted.

What about nuclear stress tests or myocardial perfusion imaging? Good, Bad, Ugly? It all depends on the context in which the test was performed and how the results are applied. Two Myocardial perfusion imaging studies, such as SPECT (single photon emission CT) with technetium (Tc99m)-agents, are actually relatively good for detecting acute myocardial infarction, but are less sensitive for other forms of ACS. The early data reported lower sensitivities, as many studies were done at rest only. If you add an exercise component to these studies, they perform much better.

So, when will nuclear stress tests most likely to detect a patient with ACS? When they are experiencing acute ischemia. Unfortunately, it isn't reasonable to throw a pair of Nikes on a patient with acute chest pain and put them on the treadmill or give them persantine or dobutamine. Nonetheless, if patients have non-diagnostic ECGs and negative enzymes, these tests can help guide us to the right conclusions with little risk of precipitating cardiac events. The sensitivity, for the 30-day risk of ACS, death, revascularization or life-threatening complications, of resting and exercise imaging in low risk chest pain patients was 71% and 97%, respectively (Fesmire, F.M., et al, Ann Emerg Med 38(3):207, September 2001.).  The negative predictive value of myocardial perfusion imaging is important. A large meta-analysis, J Am Coll Cardiol. 2007 Jan 16;49(2):227-37, reported the negative predictive value for AMI and cardiac death was 98.8% for the three years following the study (95% confidence interval [CI] 98.5 to 99.0).

Finally, both myocardial contrast echocardiography (MCE), investigating myocardial perfusion via micro-bubble contrast injection, and standard echocardiography to assess wall motion abnormalities (WMA), as a proxy for perfusion deficits, are both enhanced by a stress component. Contrast echocardiograpy is probably a better test than assessing for wall motion abnormalities. Regardless, both perform better when combined with stress.  The sensitivities for coronary artery disease for MCE and WMA were 84% vs. 20% at intermediate stress and 91% vs. 70% at maximal stress. (Journal of the American College of Cardiology, Volume 44, Issue 11, 7 December 2004, Pages 2185-2191).

So, what does DIAD (April 15, 2009 in JAMA), tell us about stress testing? This study was designed to assess for risk of silent AMI in asymptomatic type 2 diabetics. The same number of cardiac events were noted in those studied compared to those not studied.  However, events were less likely in those with normal MPI studies. The only thing this article proves is that, diabetic or not, asymptomatic patients don't need to be studied. It also confirms that it is pretty unrealistic to expect a test to predict rates of cardiac events up to 5 years after the test. A lot can change in that time.

Stress tests certainly aren't perfect. Very few of our tests are. The key is not to consider their results in isolation. Regardless of whether a patient presents after having a recent stress test or if such studies are obtained emergently, understanding their limitations and the applicability of the results make stress tests useful components of a very complex decision making process.