November 25th, 2013
A 2-Hour Diagnostic Protocol for Assessing Chest Pain in the ED
CardioExchange’s John Ryan interviews Martin Than about a New Zealand study group’s randomized trial comparing a 2-hour diagnostic protocol with usual care for patients who report chest pain in the emergency department. The article is published in JAMA Internal Medicine.
At a New Zealand hospital, researchers randomized 542 adults with acute chest pain consistent with ACS (all seen by an attending physician who planned further observation and troponin testing) to one of two diagnostic pathways:
- an accelerated (experimental) protocol using TIMI score, ECG findings, and 0- and 2-hour troponin results
- a standard (usual-care control) protocol of troponin testing on arrival, prolonged observation, and repeat troponin testing 6 to 12 hours after onset of pain
The rate of discharge within 6 hours after hospital arrival was significantly higher in the experimental group than in the control group (19.3% vs. 11.0%). An additional 12.9% of patients in the experimental group were classified as low-risk but were still admitted for cardiac investigation; none of the 35 received a diagnosis of ACS after inpatient evaluation.
Ryan: Your study was powered to detect a difference in the discharge rate, but another important issue is safety. Were you able to determine whether the accelerated pathway is safe?
Than: This is a key question. It is unusual for a new diagnostic test process to have its safety tested in an implementation study before being introduced into clinical practice. That is because the number of patients required to prove noninferiority for safety, compared with an existing diagnostic strategy, is very large, particularly (as in our case) when the expected false-negative rate is about 1%. Before introduction into clinical practice, it is common for new tests to first be evaluated in large observational cohorts. The previously published ADAPT trial used contemporary troponin I (TnI) assays in 1975 patients from Australia and New Zealand. It has been confirmed using high-sensitivity TnI in 909 patients from the APACE cohort in Germany and Switzerland. In ADAPT and APACE, the negative predictive values (NPVs) of the accelerated diagnostic pathway were 99.7% and 100%, respectively. In our trial, in the experimental arm, one patient had MACE out of 94 patients classified as low-risk, yielding a similar NPV of 98.9%.
We also believe that is worth highlighting the case of the single patient who had a MACE after early hospital discharge — the MACE occurred after a clinician’s error in interpreting a diagnostic test (which could have happened in either pathway). The second TnI test for that patient was performed 8.5 hours after symptom onset; therefore, it is likely that the same outcome would have occurred if the patient had been assigned to the control group. Local procedures were modified for both pathways so that senior clinicians now interpret stress tests. We also believe that after exclusion of myocardial infarction and despite the best efforts of clinicians, a very small number of patients will still have their underlying coronary artery disease missed and later detected, leading to a subsequent hospitalization. Trying to detect 100% of such patients would burden the healthcare system considerably.
In designing this study, we considered the possibility of assessing safety, but powering the study for safety would have required a sample size of at least 7500 patients (depending on what false-negative rate was expected). We believed that such a large trial was not economically justifiable with such a low expectation of detecting a difference. After completion of our trial, the accelerated pathway was implemented at three hospitals (in New Zealand; Queensland, Australia; and Hong Kong) and has now been running for almost a year without adverse events.
Ryan: How did you get the idea for the study?
Than: This study was a logical extension of 6 years of observational research by our group, in 10 countries of the Asia-Pacific region and Australasia (see ASPECT, Lancet 2011, 377:1077-84; ADAPT, JACC 2012, 59:2091-2098; and Cullen et al. JACC 2013, in press). These studies suggest that up to 20% of patients could be safely discharged following a troponin measurement taken 2 hours after hospital arrival. These results are promising, but data on the effectiveness and the cost of the accelerated approach required testing in a randomized controlled trial conducted under real-world conditions in terms that matter to the patient and the health institution. Although we provided pathways for the intervention and control arms, the final management decision, using the TIMI score and test results, was at the attending clinician’s discretion. Clinicians do not always follow protocols or act as we would expect on the basis of test results. A pragmatic design is also the reason for recruiting a broad spectrum of patients rather than just low-risk patients with an initially normal troponin level. A focused approach may achieve an impressive discharge rate but would be much less representative of real-world care and less likely to elucidate the real benefits. An impressive increase in the discharge rate will not benefit the system if it applies only to a tiny proportion of patients.
So our trial was designed to answer these two crucial questions: (a) Would the accelerated diagnostic protocol work in practice (i.e., would it actually reduce rate and duration of hospital admissions)? (b) Would early discharge (without subsequent harm) be achieved in a significant percentage of all patients who presented with chest pain (excluding ST-segment elevation MI)? Notably, a Journal Watch Emergency Medicine reviewer of the ADAPT study wrote, “This study begs for a randomized trial in which low-risk patients are actually sent home, but such a study is unlikely to ever be conducted.”
It may seem obvious that the experimental pathway would allow more patients to be discharged earlier, but this might not actually occur in real-world practice for several reasons:
First, clinicians do not always follow protocols or act as we would expect on the basis of test results. For example, a study of 117 emergency departments found that international guidelines for investigating pulmonary embolism were not followed for 47% of patients. Our report revealed this point in the high number of patients (13% of the experimental group) who were classified as low-risk but who were still hospitalized for further investigations for ACS. This affected the difference between primary outcomes in each group, although the difference still remained significant.
Second, for the experimental pathway to be easily and reproducibly adopted, no extra staff, bed, or capital resources were provided to the experimental group. Therefore, logistical constraints might have prevented timely discharge of low-risk patients. For example, the troponin results might not be checked in a timely manner or, in a busy ED, sufficient medical staff may not be available to review and discharge patients. In the trial, this actually occurred only twice.
We believed that it was important to determine effectiveness of the experimental pathway under typical real-world clinical conditions, so we selected a pragmatic trial design (as described by the CONSORT group). Demonstrating real-world effectiveness and estimating the size of that effectiveness will help clinicians and policymakers decide whether the experimental pathway is worth adopting. In addition, although results from existing observational studies are very promising, a change in practice is more likely to occur after a randomized controlled trial is published.
Ryan: Does New Zealand have unique characteristics that make a study like this less generalizable to other countries?
Than: Thank you for this very relevant question. Although similarities exist, each country’s health system has unique features that may affect reproducibility. New Zealand has a tax-funded national health system with universal coverage for all, as well as a very strong primary health care emphasis; therefore, many patients with low risk for ACS are seen by their primary care physicians and not sent to a hospital. In this study and our previous studies, the prevalence of major adverse cardiac event and ACS excluding STEMI has been relatively high, compared with that documented in other published studies. In systems that have a lower prevalence of ACS, we think the pathway is likely to be more effective because a greater number of low-risk patients may benefit from early discharge.
It is also possible that other healthcare systems will have significantly less availability — and use — of follow-up investigations. As such, adjustments may be needed to reproduce the accelerated diagnostic protocol in other healthcare settings. However, we believe that given that all the diagnostic parameters described are easily available, other overseas health institutions may be able to adapt their processes to effectively reproduce the accelerated protocol.
Finally, even within the context of similar healthcare settings, a single-center study is never as robust as a multicenter investigation. Unfortunately, funding for multicenter studies of diagnostic protocols is extremely difficult to obtain. Government-funded studies such as this one are extremely unusual in a multinational, multicenter context.
Ryan: Are you translating the results of your study into practice in New Zealand?
Than: In Christchurch, we have been running the pathway successfully for over a year, and no adverse events have occurred. The pathway has also been running without problems for a year at Nambour Hospital in Queensland and at Queen Elizabeth II Hospital in Hong Kong. We are currently working with the Ministry of Health in New Zealand and other key clinicians to introduce the pathway to other New Zealand hospitals.
JOIN THE DISCUSSION
Share your thoughts about the diagnostic protocol investigated by Dr. Than and his colleagues in New Zealand.