Whether you are aware of it or not, you make three major assumptions every time you employ any diagnostic strategy in the emergency department—that the diagnostic test will identify patients at risk of a poor outcome because of an undiagnosed process, that an effective intervention will avert that outcome, and that discovering this process early will make the intervention more effective than if we had waited for the disease to manifest clinically obvious characteristics.
These assumptions are frequently based purely on physiological reasoning and good intentions, but these are poor surrogates for patient-oriented outcomes and often fail to survive the test of scientific inquiry. Such is the case for noninvasive cardiac testing. Multiple studies suggest its ineffectiveness, but the stress test has maintained its lofty position for managing patients presenting to the ED with chest pain.
A recent study by Sandhu, et al., utilized a large insurance claims database to identify patients who presented to the ED with chest pain and compared outcomes in patients who did or did not undergo noninvasive testing. (JAMA Intern Med 2017;177:1175.) The authors used a fairly novel and elegant approach to control for the many imbalances one would expect from such a large, heterogeneous, nonrandomized cohort.
Using what is called an instrumental-variables approach, the authors exploited the fact that care is not delivered consistently across all seven days of the week. Their premise, according to previous data, was that patients seen in the ED on the weekend (Friday-Sunday) were less likely to undergo stress testing than those who presented Monday-Thursday, based not on differences in patient-level characteristics, but rather the universal distaste of working on the weekend.
The researchers excluded patients with diagnoses suggestive of acute ischemia and those whose chest pain could be explained with an alternate diagnosis, and they identified 926,633 unique adult ED visits from 2011 to 2012. Unsurprisingly, patients who received testing were older, with more risk factors than patients who did not undergo testing. Conversely, patients who presented on the weekend appeared to be fairly similar at baseline when compared with those who presented during the week.
As the authors predicted, patients evaluated on the weekend underwent less stress testing when compared with those who presented during the week (18.18% vs 12.30%). They also observed more early angiography (2.10% vs 1.30%) and downstream testing, defined as any invasive or noninvasive testing done over the next 30 days (26.10% vs 21.35%). Even after adjusting for possible bias, not controlled for with their instrumental approach, the authors noted an increase in the rates of invasive angiography in patients who presented on the weekday when compared with those who presented on the weekend. They also noted that this increase in invasive or noninvasive testing did not lead to an observed decrease in the rate of myocardial infarction.
Despite the elegance with which these authors manipulated this large unwieldy dataset, its innate structure creates the potential for multiple sources of bias that cannot be controlled by any statistical manipulations. That said, their results are fairly consistent with the majority of the previous literature examining noninvasive stress testing. More importantly, the stress test has failed to meet the three initial assumptions required when examining any diagnostic testing strategy.
The concept that noninvasive stress tests identify a population at higher risk for a myocardial infarction is based on poor data. Amsterdam, et al., using a prospective dataset, claimed that patients with positive stress tests were at significantly higher risk of adverse events than those with negative tests (17% vs 0.16%). (J Am Coll Cardiol 2002;40:251.)
More Harm than Good
The reality is somewhat different. These authors enrolled 1000 patients presenting to the ED with chest pain who had a negative ECG and initial troponin, and performed exercise stress tests. A total of 640 had a negative stress test and were discharged home. Only one had an MI during follow-up. The vast majority of adverse events in the 125 patients with a positive stress test were revascularization procedures (J Am Coll Cardiol 2002;40:251), and only four had an MI. A closer examination showed that all four MIs were identified on the second troponin in the ED.
This study actually demonstrated that patients presenting to the ED with a negative ECG and two negative troponins were already at fairly low risk for a myocardial infarction, and the only thing the stress test added was to increase downstream invasive procedures. These results are not limited to exercise stress tests. Even more anatomically accurate forms of tests, such as CTCA, have failed to demonstrate a decrease in the rates of death or MI when used in an ED population. (J Am Coll Cardiol 2013;61:880).
This is likely due partly to the fact that we have stratified these patients to such a low risk using ECG and cardiac biomarkers that any further risk stratification is likely to result in more harm than good. (JAMA Intern Med 2013;173:1128; 2015;175:428.)
But let's say for the sake of argument that stress tests do identify a subset of patients who are at higher risk for myocardial events. To what end? No discernible evidence demonstrates that patients identified by positive noninvasive testing will benefit from invasive catheterization. The majority of the data in patients not actively experiencing a myocardial infarction favors medical management alone. (Arch Intern Med 2012;172:312; JAMA Intern Med 2014;174:232.) Without an effective way to prevent the prophesized myocardial consequences, stress testing only causes unneeded testing, interventions, and harm.
I am sure our hesitancy to discard this needless diagnostic strategy is in part because no one has suggested a viable replacement. The right answer may be simply to do nothing, but such self-restraint is rarely palatable to the modern physician. A bad solution is not better than no solution, especially when it has the potential to hurt patients. Surely the ever-growing literature discrediting noninvasive stress testing on ED patients means the real question to ask is, just who exactly are we treating when ordering these potentially harmful and costly tests, us or the patient?
Share this article on Twitter and Facebook.
Access the links in EMN by reading this on our website or in our free iPad app, both available at www.EM-News.com.
Comments? Write to us at firstname.lastname@example.org.Copyright © 2018 Wolters Kluwer Health, Inc. All rights reserved.