Author Affiliation: Division of Cardiology, Department of Medicine, Duke Clinical Research Institute, Duke University Medical Center, Durham, North Carolina.
Triage of emergency department (ED) patients with possible acute myocardial infarction (MI) without ST-segment elevation remains one of the most challenging dilemmas in medical practice. The stakes are high: patients with MI inappropriately sent home have approximately 2-fold higher risk-adjusted 30-day mortality than those hospitalized.1 Conversely, it is not feasible or cost-efficient to admit all patients for MI “rule-out.” The advent of chest pain units diminished the strain on in-patient resources,2 but even these units often use serial electrocardiograms (ECGs) and cardiac marker testing over 6 to 9 hours to confidently confirm or exclude MI. With increasing ED overcrowding, more effective tools are needed to enable rapid triage of patients with possible MI. In addition, although time dependency of treatment for non–ST-segment elevation MI (non-STEMI) is uncertain, earlier diagnosis could lead to more effective use of acute therapies and more efficient, shorter hospital stays.
Cardiac troponins (cTn) are highly specific biomarkers of myocardial necrosis, are much more sensitive than creatine kinase (CK)-MB, and levels strongly correlate with subsequent mortality. These features prompted a cTn gold standard for MI diagnosis.3 However, despite nearly absolute tissue specificity and superior sensitivity, cTn is not specific for the etiology of myocardial necrosis (eg, elevated cTn levels occur in such disparate conditions as coronary ischemia, pulmonary embolism, heart failure, sepsis, and renal failure).4 Thus, clinical syndromes consistent with ischemia and a characteristic rise and/or fall in cTn levels during serial testing are critical for MI diagnosis.3
More recently, a new generation of high-sensitivity troponin (hsTn) assays has been developed. They have limits of detection approximately 10-fold lower than conventional assays, 99th percentiles in the low nanogram per liter range, and are analytically very precise (coefficients of variation of 10% at or below the 99th percentile). The ability to detect such small amounts of cTn suggests promise for diagnosing smaller MIs otherwise undetected or identifying MI earlier, when abnormal hsTn levels are below detection by conventional assays. Indeed, initial studies demonstrated that hsTn assays could detect smaller amounts of myonecrosis with greater sensitivity for MI than conventional assays at all serial time points, but highlighted challenges created by greater sensitivity and lack of disease specificity.5 That is, positive predictive value (PPV) was as low as 50%. Other studies suggested possible susceptibility of hsTn results to biological variability across age and sex (population prevalences of elevated hsTn of 1% among individuals <40 years old vs 5.2% if >65 years old, and 2.8% among men vs 1.3% among women) and demonstrated frequent elevation in asymptomatic patients with stable coronary disease (11.1%) and prior heart failure (18.9%).6,7 Combined, these factors challenge application of hsTn assays in the ED and suggest they may be better suited for population screening for subclinical disease or as markers of disease activity.
In this issue of Archives, Reichlin et al8 present evidence supporting an algorithmic approach to interpretation of hsTn results in patients with suspected MI that may offer high sensitivity and negative predictive value (NPV) for early MI rule-out combined with diagnostic specificity and PPV high enough to use in “rule-in” or even to accelerate non-STEMI treatment. Their approach extends prior work,9,10 leveraging serial testing, and expected dynamic rises in cTn levels with ischemic events vs stable levels in other conditions in which baseline hsTn elevation might be expected. Using 436 patients randomly selected from 872 with suspected ACS who had baseline and 1-hour hsTn testing and systematic, blinded MI adjudication based on hsTn levels and clinical information, they created algorithms incorporating baseline thresholds and thresholds for hsTn rise from baseline to 1 hour for both MI rule-in and rule-out.8 The final rule-in algorithm included a baseline threshold of at least 60 ng/L or at least a 15-ng/L baseline to 1-hour change threshold. It is important to note that thresholds were not sensitive to sex, ECG features or time from symptom onset. The final rule-out algorithms optimized sensitivity and NPV at 100%: baseline threshold of 12 ng/L or less and absolute change of 3 ng/L or less. Applied in a 436-patient validation cohort, a definite diagnosis was made in 72% of patients. The rule-out algorithm had 100% sensitivity and NPV, and the rule-in algorithm had 99% specificity and 91% PPV, misclassifying 5 patients as having had MIs and classifying 23 MIs into an “observational zone” in 123 patients meeting neither rule-in or rule-out criteria.
With this study, Reichlin et al8 provide an important step forward in application of hsTn as a tool for triage of ED patients with possible MI. However, much work remains to develop the evidence to bring hsTn testing and the algorithms they have developed to use in clinical practice. Most importantly, this algorithmic approach must be validated in prospective studies designed to assess not only sensitivity and NPV and specificity and PPV but also the implications for clinical outcomes and cost of widespread implementation. It is unlikely that the observed 100% sensitivity and NPV and very high specificity and PPV will hold up in general practice in which patient populations almost certainly will be less selected, the prevalence of MI may vary widely, and the prevalence of confounding comorbidities like heart failure and renal insufficiency will be higher. Such studies will also help to confirm that thresholds are satisfactory across age groups and inform whether other factors, like sex and time from symptom onset, should be considered. Finally, although touted as “simple” by the authors, the need for multicomponent algorithms that are different for rule-in and rule-out and that vary by age group or other parameters will challenge application by busy clinicians unlikely to remember or accurately process the proposed algorithm. As such, it will be imperative that hsTn algorithms, if validated, are built into clinical decision support layered onto electronic health records so that testing results are provided electronically to physicians along with the algorithmic interpretation to allow systematic application in triage and treatment. Thus, the work of Reichlin et al8 is a major advance in understanding the application of hsTn testing that with continued development could substantially improve evaluation of ED patients with suspected MI.
Correspondence: Dr Newby, Division of Cardiology, Department of Medicine, Duke Clinical Research Institute, Duke University Medical Center, PO Box 17969, Durham, NC 27715-7969 (email@example.com).
Published Online: August 13, 2012. doi:10.1001/archinternmed.2012.1808
Financial Disclosure: Dr Newby has received personal income from consulting or other services from the following companies: Amgen Inc, AstraZeneca, Biosite Inc, Bristol Myers Squibb, CV Therapeutics Inc, Daiichi Sankyo, Eli Lilly & Co, Genentech, Johnson & Johnson, Merck & Co, Mosby, Novartis Pharmaceutical Co, Proctor and Gamble, Regado Biosciences Inc, Roche Diagnostic Corp, Schering Plough Corp, Scios, and Shionogi Pharma.
Funding/Support: Dr Newby is supported by research grants or contracts from the following companies: Amylin Inc, AstraZeneca, Bristol Myers Squibb, diaDexus, Eli Lilly & Co, GlaxoSmithKline, Merck & Co, Murdock Study, National Heart, Lung, and Blood Institute, Regado Biosciences, and Roche.
Newby LK. Myocardial Infarction Rule-out in the Emergency Department: Are High-Sensitivity Troponins the Answer?Comment on “One-Hour Rule-out and Rule-in of Acute Myocardial Infarction Using High-Sensitivity Cardiac Troponin T”. Arch Intern Med. 2012;172(16):1218-1219. doi:10.1001/archinternmed.2012.1808