Although it has been a bit of a convoluted path, hopefully these 4 posts have illustrated why I don’t order stress tests for my patients. We make a lot of assumptions when we are ordering stress tests, but those assumptions appear to be faulty. We assume that patients with negative chest pain work-ups are high enough risk to warrant further investigation, when in reality our miss rate is 1-3 in 1,000. We assume that stress tests help us identify patients at short term risk of MI, when in reality almost every short term MI was missed by stress testing in these studies. If we aren’t worried about short term MI, we assume that stress testing is accurate enough to rule in or rule out stable coronary artery disease, when in reality it is only moderately accurate at best. Finally, we assume that we are helping patients by identifying those at higher risk, when in reality invasive treatment doesn’t help these patients, but does expose them to unnecessary harm.
The only RCTs of stress testing in emergency department patients demonstrate no benefit. (Lim 2013; Frisoli 2017) RCTs in other patient populations have also demonstrated no benefit, but an increase in invasive procedures. (Young 2009, Poldermans 2006)
Ultimately, this literature makes it pretty clear that stress testing results in net harm for our patients.
- This post is part of a series exploring the practice of stress testing.
The stress test math
Let’s consider how the stress testing numbers play out. Exercise stress test is still the most commonly used, and in part 3 we discovered that it has a sensitivity of 45% and a specificity of 85% for coronary artery disease, leading to a negative likelihood ratio of 0.65 and a positive likelihood ratio of 3. As we discovered in part 1, we rarely miss ACS. After a negative ED work-up with multiple ECGs and troponins, the risk that a patient will have a short term MI is around 0.2%. So what happens when we apply a test with a 45% sensitivity and 85% specificity to a population with a pretest probability of disease of 0.2%?
If the test is negative, the patient has a 1 in 1000 chance of having an MI. (Essentially unchanged from the 2 in 1000 chance prior to the test.) If the test is positive, the patient has about a 1% chance of having a short term MI. For every 1,000 patients we put expose to stress testing, there are 2 at risk for MI and the stress test will likely pick up 1 of them. However, for that 1 true positive test, there will be 150 false positives.
Even if you think stress testing is more accurate than the studies say, a test with an 85% sensitivity and specificity would still have almost 100 false positive tests for every true positive test. The only difference is instead of missing 1 of the 2 true positives, the test is somewhat more likely to catch both. It still isn’t clear how you distinguish the 2 true positives from the 100 false positives.
Bottom line: The math just doesn’t work. Stress testing cannot help up in such a low risk population.
What about coronary artery disease and longer term risk?
The above math assumes that you are trying to find patients at risk of short term ACS. I think this is a fair assumption, because this is what emergency physicians tell me they are looking for with a stress test. However, if instead of looking for patients at risk of ACS, you are just hoping to find stable coronary artery disease, the math changes because coronary artery disease is so common.
Let’s imagine a relatively high risk chest pain patient. He is 65, has type 2 diabetes, mild hypercholesterolemia, hypertension, but doesn’t smoke. Using the Framingham score, this man has between a 15% chance of MI or death in the next 10 years. So what happens when we apply a test with a 45% sensitivity and 85% specificity to a population with a pretest probability of disease of 15%?
If the test is negative, the patient has a 10% chance of death or MI at 10 years. (Probably not a significant change from the 15% chance before the test.) If the test is positive, the patient has about a 35% chance death or MI. For every 1,000 patients we put expose to stress testing, 150 are at risk of death or MI, and the stress test will only pick up 68 (82 false negatives). On the other hand, there will be 128 false positives. So even with the higher pretest probability, it isn’t clear that the stress testing math works out.
Furthermore, in order for a test to be helpful, it needs to lead to a change in management. As we discussed in part 4, revascularization does not help these patients (with negative ECGs and troponins). In fact, it is probably harmful. So the question remains: what are you going to do with the results of the test?
Stress testing to decide medical management
Stress testing does not predict short term outcomes. It is not accurate enough to rule in or rule out coronary artery disease, and it cannot tell us if this episode of pain was cardiac in nature. It cannot be used to direct invasive management, because none of these patients should have invasive management. However, it is possible that the stress test could be used to help guide medical management.
In the studies looking at revascularization, invasive management was not helpful, but there is an important caveat: all patients in the control group received maximal medical therapy for their coronary artery disease.
All patients with chest pain warrant medical management. They should have risk factor modification, with discussions about diet, exercise, and smoking cessation. That should be followed with appropriate management of cholesterol, diabetes, and blood pressure. However, this is all part of good primary care, and should be done regardless of any stress test result.
On the other hand, not everyone should be prescribed life long anti-platelet medications. The decision to prescribe anti-platelet medications depends on whether you think the chest pain was truly cardiac (primary vs secondary prevention). Even so, pretest probability matters. Some people are so low risk (a 25 year old with no coronary risk factors) that ASA shouldn’t be prescribed, no matter what a stress test says. Other patients will be high enough risk to warrant ASA based on risk factors alone, and so also don’t need a stress test. However, there may be patients in the middle, in whom a stress test could theoretically help.
Aside from the fact that this decision is probably more appropriate for primary care than emergency medicine, I still think there are reasons to doubt the value of the stress test. The accuracy of the stress test is awful. You are going to have a lot of false positives and a lot of false negatives. More importantly, even if a stress test could tell you about the presence of coronary artery disease, it doesn’t tell you if this episode of chest pain was actually cardiac. Furthermore, all of the RCT evidence we have indicates that stress testing doesn’t help. Therefore, I think we are better of making this decision using clinical judgement, rather than trying to rely on an inaccurate test. Typical chest pain should be treated. Atypical chest pain should have risk factor modification, but no further intervention.
We need to act like doctors and make difficult decisions. We don’t need more tests. (Especially not tests as awful as the stress test.)
Of course, at this point we are firmly in the realm of primary care. These decisions require lengthy discussion with patients. They require a nuanced understanding of risk factor modification and the risks and benefits of lifelong daily medications. I don’t think this can be done adequately in a single emergency department visit, and if you aren’t going to be be making those decisions, why are you ordering test?
Stress tests are harmful
Before closing, it is important to recognize the harms that result from stress tests. There is pretty clear evidence of overdiagnosis. Lucas (2006) looked at stress testing use among Medicare beneficiaries between 1993 and 2001. Over that time period, stress testing use increased by almost 3-fold (from 19 to 82 per 1000). This was associated with a doubling in the rate of angiography and almost a 7 fold increase in the rate of stents (since they started in 1995). However, the rate of MI didn’t change at all (9 per 1000 per year). Therefore, patients were undergoing a lot of extra testing and invasive intervention, without any observable benefit. Obviously, not all the extra invasive procedures can be blamed on stress testing, but one thing that was clear from the stress testing literature was that stress testing leads to more invasive procedures, and all the harms they entail.
Although the data here is pretty complex, I actually think the conclusion is pretty easy. I haven’t been ordering stress tests for years.
If you are interested in catching patients at risk of short term MI or death – it’s pretty clear stress test can’t do that. Even out to 1 year, stress testing doesn’t seem to predict important outcomes. If you are interested in trying to figure out which patients need invasive management – well stress testing can’t do that, because these patients don’t need invasive management. If you are interested in finding stable coronary artery disease – well, stress testing isn’t very good at that either, and you’d have to ask yourself, what am I going to do with the results?
People worry about stopping the workup with no further testing, but we have to remember, tests can hurt patients. In emergency medicine we get so focused on not missing anything, that we frequently overlook the harms of over-testing. And this data clearly shows harms. Stress testing leads to patients getting invasive procedures that they don’t need. Those invasive procedures cause harm. So for me it is clear, stress testing has no role in emergency medicine.
Amsterdam EA, Wenger NK, Brindis RG, et al. 2014 AHA/ACC Guideline for the Management of Patients With Non–ST-Elevation Acute Coronary Syndromes Journal of the American College of Cardiology. 2014; 64(24):e139-e228.
Frisoli TM, Nowak R, Evans KL, et al. Henry Ford HEART Score Randomized Trial. Circ Cardiovasc Qual Outcomes. 2017; 10(10).
Lim SH, Anantharaman V, Sundram F, et al. Stress myocardial perfusion imaging for the evaluation and triage of chest pain in the emergency department: A randomized controlled trial. J. Nucl. Cardiol.. 2013; 20(6):1002-1012.
Lucas FL, DeLorenzo MA, Siewers AE, Wennberg DE. Temporal trends in the utilization of diagnostic testing and treatments for cardiovascular disease in the United States, 1993-2001. Circulation. 2006; 113(3):374-9. PMID: 16432068
Poldermans D, Bax JJ, Schouten O, et al. Should major vascular surgery be delayed because of preoperative cardiac testing in intermediate-risk patients receiving beta-blocker therapy with tight heart rate control? Journal of the American College of Cardiology. 2006; 48(5):964-9. PMID: 16949487
Young LH, Wackers FJ, Chyun DA, et al. Cardiac outcomes after screening for asymptomatic coronary artery disease in patients with type 2 diabetes: the DIAD study: a randomized controlled trial. JAMA. 2009;301:(15)1547-55. PMID: 19366774 [free full text]
Morgenstern, J. Stress Tests Conclusion: Putting it all together, First10EM, March 15, 2019. Available at: