Ready for a Risk Stratification Robot?

To the Editor: Care patterns for emergency department (ED) patients with acute chest pain are changing rapidly. Just a few years ago patients at very low risk for acute coronary syndrome (ACS) were routinely admitted to the hospital or placed in an observation unit for serial troponins (cTn) and stress testing or computed tomography coronary angiography. These conservative practices were fueled by American College of Cardiology/American Heart Association guidelines, malpractice fears, and the TIMI and GRACE scores, which lack sensitivity and are not designed for ED use.1-4 Since that time, the derivation and validation of accelerated diagnostic protocols (ADPs) has led to low-risk patients being discharged from EDs with greater frequency. ADPs are designed to safely identify patients for early discharge from the ED. ADPs, such as ADAPT, the HEART Pathway, and EDACS, typically combine variables from the patient's history (presenting symptoms), electrocardiogram (ECG), age, and cardiovascular risk factors, with serial cTn measures at presentation and 2 or 3 hours later to risk stratify ED patients with chest pain.5-7 These tools require the healthcare provider to gather an accurate history and incorporate his or her findings and judgment into the risk stratification process. For example, providers using the HEART Pathway must determine whether the patients symptoms have highly, moderately, or slightly suspicious histories for ACS and for ADAPT, providers determine whether a patient has “severe” angina. Clinician judgment plays an important role in the determination of risk when using these ADPs. ADPs continue to evolve, and recent studies have explored integrating high-sensitivity troponin assays (hs-cTn) into ADPs. Given the improved clinical and analytical sensitivity of the newer assays many have hypothesized that hs-cTn measurement may be all that is needed for risk stratification of ED patients without a STEMI on their initial ECG. Recent studies suggest that a hs-cTn–only approach to ED chest pain risk stratification could identify more patients for early discharge than prior strategies with a high negative predictive value for acute myocardial infarction (AMI).8, 9 But, can any biomarker be so accurate that we no longer need to consider other clinical variables? Can we replace seasoned emergency care physicians with robotic hs-cTn protocols? While it is clear that hs-cTn measures above the 99th percentile upper reference limit (URL) have high sensitivity for the detection of myocardial injury, it remains to be seen whether hs-cTn measures alone or in combination with an ECG are sufficient to identify patients who can be safely discharged from the ED. Several studies suggest that hs-cTn measures used at the URL are not sufficiently sensitive. For example, in the APACE cohort, a study of 909 patients with serial hs-cTnI measures at 0 and 2 hours yielded a sensitivity of 82.7% for ACS events at 30 days. However, when hs-cTnI results were integrated with ECG data and a clinical decision aid the sensitivity for adverse events improved to >99%.10 At cut points much lower than the URL, hs-cTn have shown promise for identifying patients at very low risk for adverse cardiac events. Shah et al.8 found that patients without AMI on ED presentation and an hs-cTnI of <5 ng/L had a 99.6% negative predictive value for subsequent infarction or cardiac death at 30 days. In the TRAPID-AMI study, Mueller et al.9 evaluated an ADP based on 0- and 1-hour hs-cTnT measures and reported a 99.1% sensitivity for (index-visit) AMI. However, there are concerns about the high analytical variability (>10% coefficient of variation) that exists at the low levels of hs-cTn utilized in these studies. Furthermore, these studies did not report the sensitivity for a 30-day MACE composite outcome (major adverse cardiac events: cardiac death, AMI, and coronary revascularization). In this issue of Academic Emergency Medicine, Brody et al.11 performed a secondary analysis of the TRAPID-AMI cohort to determine if a single hs-cTnT level below the level of detection (LOD; <5 ng/L) and a nonischemic ECG could exclude AMI. Among 1,282 patients, 471 patients (37.6%) had an initial hs-cTnT level below the LOD and an ECG without acute ischemic changes. This strategy missed only two patients with an adjudicated AMI during the index encounter (a sensitivity and negative predictive value exceeding 99%). Therefore, this study adds to the growing body of literature showing improved sensitivity for the detection of index visit AMI for patients using hs-cTn cut points below the URL. However, a successful strategy for ED chest pain risk stratification must also have high sensitivity for adverse cardiac events out to 30 days. Over the 30 day follow-up period, Brody et al. found that hs-cTnT below the LOD alone (without ECG or additional clinical data) missed eight patients (1.4%, 95% confidence interval [CI] = 0.5% to 3.1%) with MACE (defined as prevalent or incident AMI, all-cause mortality, or revascularization following rehospitalization for ACS within 30 days). The addition of a nonischemic ECG to hs-cTnT decreased the number of patients with missed MACE to 6 (1.3%, 95% CI = 0.5% to 2.8%). However, an additional 11/471 (2.3%) patients with an hs-TnT below the LOD and a nonischemic ECG had coronary revascularization procedures during their index hospitalization. If those revascularization events were included in the composite MACE rate, the missed event rate was 3.6% (95% CI = 2.2% to 5.7%), well above the miss rate that most providers find acceptable.12 We applaud Dr. Brody for providing revascularization outcome data in his paper so that readers can review the adverse event rates and draw their own conclusions. Coronary revascularization is a highly controversial outcome measure and the importance of missing revascularization events remains unclear. In individual patients it is often difficult to adjudicate if revascularizations are “appropriate” and impossible to determine the relationship between the procedure and the occurrence or lack of occurrence of downstream adverse events. While revascularization procedures improve mortality and reinfarction rates among patients with an acute or prior MI, studies have questioned the role of revascularization in patients without acute cTn elevations.13 The COURAGE trial and recent systematic reviews failed to show a decreased risk of AMI or death among patients with stable coronary artery disease receiving revascularization compared to medical management.14-17 Many clinicians interpret these results as evidence that revascularization should not be included as an adverse outcome in clinical trials. However, it is important to remember that the COURAGE trial's control group was medical management not early discharge from the ED. Patients discharged from the ED based on a low-risk ADP assessment rarely receive maximal medical management. Therefore, we should not assume that use of an ADP, which identifies these patients as low risk will result in comparable outcomes. Dr. Brody's study is a well-done secondary analysis using robust cohort data. It adds to our understanding of the diagnostic performance that can be expected from an initial hs-cTnT strategy alone and with an ECG. While this study supports the widely accepted construct that hs-cTn is valuable for the early detection and exclusion of index AMI, it also casts greater doubt that an ADP that relies almost completely on a biomarker can achieve sufficient sensitivity without incorporating additional clinical variables. Clinical acumen and judgment remain necessary components of chest pain risk stratification making the emergency physician's job safe from being replaced by robots, for now …