Evaluation of High-Sensitivity Cardiac Troponin I Levels in Patients With Suspected Acute Coronary Syndrome

Importance  Low concentrations of high-sensitivity cardiac troponin I determined on presentation to the emergency department (ED) have been shown to have an excellent negative predictive value (NPV) for the identification of acute myocardial infarction. The sensitivity, and therefore clinical applicability, of such testing strategies is unknown.

Objective  To determine the diagnostic performance of low concentrations of high-sensitivity cardiac troponin I in patients with suspected cardiac chest pain and an electrocardiogram showing no ischemia as an indicator of acute myocardial infarction.

Design, Setting, and Participants  A pooled analysis of 5 international (Australia, New Zealand, and England) prospective, observational cohort studies with blinded outcome assessment and 30-day follow-up was conducted. A total of 3155 patients presenting with symptoms suggestive of cardiac ischemia were included in the analysis. Eligible patients had a nonischemic electrocardiogram determined and high-sensitivity troponin I measured at presentation. The lower limit of detection (1.2 ng/L) as well as cutoff concentrations rounded to the nearest integer for a high-sensitivity troponin I assay were used in the analysis. Recruitment was undertaken from November 1, 2007, to August 10, 2013.

Main Outcomes and Measures  The primary outcome was fatal or nonfatal acute myocardial infarction occurring within 30 days of ED presentation, adjudicated with serial troponin testing. The secondary outcome was the proportion of patients potentially suitable for early discharge at each cutoff concentration.

Results  Of the 3155 eligible patients, 1771 were male (56.1%), and mean (SD) age was 57.4 (13.3) years. Acute myocardial infarction developed in 291 individuals (9.2%). The 1.2-ng/L limit of detection gave a sensitivity of 99.0% (95% CI, 96.8%-99.7%) and an NPV of 99.5% (95% CI, 98.4%-99.9%). This cutoff level would allow for early discharge of 594 patients (18.8%). All higher rounded cutoff values had sensitivities less than 98.0%. Diagnostic performance of the limit of detection was maintained when patients were stratified by age, sex, risk factors, presence of coronary artery disease, and early presentation.

Conclusions and Relevance  High-sensitivity troponin I concentrations determined at presentation to the ED that were below the limit of detection identified 18.8% of patients potentially suitable for discharge, with a high sensitivity for acute myocardial infarction. Rounded cutoff values above the limit of detection may not have the required sensitivity for clinical implementation.

Patients with suspected cardiac chest pain account for more than 6 million emergency department (ED) visits annually across the United States.¹ Current American Heart Association guidelines² recommend serial measurements of contemporary cardiac troponin at presentation and 3 to 6 hours after symptom onset. As a result, most patients require prolonged assessment prior to safe discharge. This diagnostic approach leads to a large number of costly, potentially avoidable hospital admissions.^1,3 Strategies that could safely identify a large proportion of patients suitable for discharge after a single sample of blood is taken on arrival in the ED would have major benefits to health care systems.

Low concentrations of high-sensitivity cardiac troponin (hs-cTn) assays determined at presentation to the ED in combination with nonischemic electrocardiogram (ECG) findings demonstrate excellent prognostic ability in patients with suspected acute coronary syndromes.^4,5 It has therefore been suggested⁵ that the clinical application of strategies that use these low cutoff values could transform the assessment of chest pain by identifying a large proportion of patients at low risk for myocardial infarction (MI). These patients may be suitable for immediate discharge after a single sample of blood is taken on arrival. However, large-scale cohort studies^4,5 that reported excellent negative predictive values (NPVs) (>99%) using very low or undetectable cutoff strategies have not openly reported the more important measure of diagnostic performance: sensitivity. As practitioners may intuitively interpret low cutoff hs-cTn strategies as diagnostic—rather than prognostic—tools, the provision of sensitivity values for informed decision making in exclusion of acute coronary syndromes is critical. In a meta-analysis⁶ of the diagnostic performance of undetectable hs-cTnT levels in which sensitivity was provided, the point estimate was 97.4% (95% CI, 94.9%-98.7%). This finding is below the minimum clinically acceptable threshold of 99.0%.⁷

We aimed to evaluate the diagnostic performance of low concentrations of hs-cTnI in patients with suspected cardiac chest pain and a nonischemic ECG from 5 international ED cohorts of prospectively recruited patients. We specifically intended to report sensitivity for the diagnosis of fatal or nonfatal acute MI (AMI).

Box Section Ref ID

The study population consisted of eligible patients recruited into 5 prospective ED cohort studies. The first 2 cohorts were enrolled at 2 tertiary teaching hospitals (Royal Brisbane and Women’s Hospital, Brisbane, Australia, and Christchurch Hospital, Christchurch, New Zealand) in the ADAPT (2-Hour Accelerated Diagnostic Protocol to Assess Patients With Chest Pain Symptoms Using Contemporary Troponins as the Only Biomarker) study.⁸ The third cohort was enrolled at Poole Hospital National Health Service Trust, Dorset, England, in the TRUST (Triage Rule-Out Using High-Sensitivity Troponin) study.⁹ The fourth cohort was enrolled at Central Manchester University Hospitals National Health Service Trust, Manchester, England, in the Novel Biomarkers of Acute Coronary Syndromes With Metabolomics Study.¹⁰ The final cohort was enrolled at Stockport National Health Service Trust, Stockport, England, in the Validation of the Manchester Acute Coronary Syndromes decision rule study.¹¹

Written informed consent was obtained from all patients, and the study protocol was approved by the respective local ethics committees (Human Research Ethics Committee, Royal Brisbane and Women's Hospital [Australia], Upper South A Regional Ethics Committee [New Zealand], Frenchay Research Ethics Committee [Poole], and North West Cheshire Research Ethics Committee [Manchester and Stockport]). The present analysis, with waiver of informed consent and use of deidentified data, was approved by those same ethics committees. Recruitment was undertaken from November 1, 2007, to August 10, 2013.

All patients recruited in the trials were eligible for enrollment if they presented with symptoms suggestive of cardiac ischemia (acute chest, epigastric, neck, jaw, or arm pain, or discomfort or pressure without an apparent noncardiac source). For all cohorts, patients were excluded if any of the following were present: ST-segment elevation MI or new left bundle-branch block, new-onset ECG changes diagnostic of ischemia (ST-segment depression ≥1 mm or T-wave inversion consistent with ischemia), significant arrhythmias (sustained supraventricular tachycardia, second-degree or complete heart block, or sustained or recurrent ventricular arrhythmias), age younger than 18 years, a clear cause of the symptoms other than acute coronary syndromes, pregnancy, inappropriate recruitment (eg, terminal illness), unwillingness to consent, and if follow-up was considered impossible. Specific to the present analysis, participants in whom hs-cTnI assay results determined on presentation to the ED were not available were excluded. We selected patients in whom new-onset ECG changes diagnostic of ischemia were absent to maximize clinical applicability of the results and reflect clinical practice in which patients with ECG changes are immediately defined as high risk and therefore not suitable for discharge. All ECG findings were adjudicated by blinded research staff (E.C., S.C., C.H., W.P., and K.G.).

All participants had laboratory troponin concentrations measured at presentation and at least 6 hours after presentation (Brisbane, Christchurch, and Poole) or at least 12 hours after the development of peak symptoms (Manchester and Stockport) as part of clinical care. Electrocardiograms were recorded on presentation. Treatment was managed according to local protocols. All clinical management, including the decision to perform stress testing or coronary angiography, was at the discretion of the attending physician.

Patient data were recorded according to standardized data collection forms using a published data dictionary.¹² Follow-up events were monitored by dedicated research staff through a combination of telephone contact, corroboration by review of hospital online patient management systems, and query to the national death registries at least 12 months after index presentation.

Research samples obtained on presentation were centrifuged, and serum was stored frozen at −70°C or below for later analysis in a blinded fashion using an hs-cTnI assay (Architect Stat; Abbott Diagnostics) with a lower limit of detection of 1.2 ng/L (to convert to nanograms per milliliter, divide by 1000; micrograms per liter, multiply by 10⁻⁶) (range, 1.2-1.9 ng/L) and a 10% coefficient of variation of 4.7 ng/L. Samples were thawed, mixed, and centrifuged for 30 minutes at 3000g and 4°C for serum samples or twice for 10 minutes at 3000g for plasma samples before analysis and according to the manufacturer’s instructions. Long-term stability of hs-cTnI has been demonstrated.¹³

We determined a priori that primary cutoff concentrations for analysis were the lower limit of detection of the hs-cTnI assay (1.2 ng/L) and rounded cutoff concentrations below 5 ng/L, the threshold previously identified⁵ as allowing the maximal number of patients for potential discharge with an NPV of greater than 99.5%. The analysis of rounded concentrations was done because it is common practice for clinical laboratories to round hs-cTnI results before reporting them. Rounding occurred up or down to the nearest whole number (eg, from 2.2 to 2 ng/L and from 2.6 to 3 ng/L).

The primary outcome was the presence of fatal or nonfatal AMI occurring within 30 days of hospital attendance (including the index ED visit). The secondary outcome was the proportion of patients potentially suitable for early discharge at each cutoff concentration.

The presence of AMI was defined according to the third universal definition of MI,¹⁴ which states that a rise and/or fall in troponin, with at least 1 value above the 99th centile value in the context of a patient with ischemic symptoms or signs (ECG changes, such as new significant ST-segment T-wave changes and pathologic Q waves, or imaging evidence, such as new regional wall motion abnormality or intracoronary thrombus by angiography) would satisfy the diagnosis. The diagnosis of AMI was adjudicated using presentation and late troponin results, according to assays in use at each institution at the time of recruitment (eTable 1 in the Supplement). Blood samples obtained 6 to 12 hours after patient presentation for clinical management were not available for hs-cTnI; therefore, it was not possible to evaluate the outcome separately using this assay. The end point was adjudicated by researchers (E.C., R.B., S.C., C.H., W.P., and K.G.) blinded to the presentation hs-cTnI results but with the knowledge of local troponin assay results.

Baseline characteristics of the study population were analyzed with conventional group descriptive statistics according to study cohort. Results were pooled for calculation of sensitivity, specificity, NPV, and positive predictive value. To illustrate the validity of each cutoff value as an exclusion tool, a minimum clinically acceptable sensitivity threshold of 99% was chosen, below which the missed-event rate becomes unacceptable to practitioners.⁷ Subgroup analysis for age, sex, cardiac risk factor burden, history of coronary artery disease, and time from symptom onset to ED presentation was undertaken for the lower limit of detection cutoff concentration (1.2 ng/L). Data were analyzed using Stata, version 12 (StataCorp).

Data were available for 3155 patients across 5 cohorts, including 1164 (36.9%) from Australia, 810 (25.7%) from New Zealand, and 1181 (37.4%) from the United Kingdom (867 from Poole, 134 from Manchester, and 180 from Stockport). Of these, 1771 (56.1%) individuals were male and the mean (SD) age was 57.4 (13.3) years. From the original cohorts, 626 patients were excluded owing to the unavailability of hs-cTnI results; these patients were not included in the 3155 individuals in the present analysis.

Baseline characteristics of the study population, classified according to recruiting center, are summarized in the Table. Of the 3155 patients included in the analysis, 291 developed a fatal or nonfatal AMI within 30 days of the index presentation, resulting in a prevalence of 9.2%, ranging from 3.4% (Australia) to 18.1% (New Zealand). Patients from New Zealand were significantly older, had a higher cardiac risk factor burden (with the exception of smoking), and were more likely to have a history of cardiac disease compared with the cohort with the lowest prevalence of AMI (Australia) (P < .05 for all). Of the 291 patients adjudicated as having an AMI, 277 (95.2%) individuals received the diagnosis at the index presentation and 14 (4.8%) were identified during the following 30 days. Calculation of diagnostic performance using 2 × 2 tables is available in eTable 2 in the Supplement.

Troponin concentrations were below the lower limit of detection (1.2 ng/L) in 594 (18.8%) of 3155 patients. This cutoff, together with a nonischemic ECG, would allow up to 18.8% of patients to be discharged with a sensitivity of 99.0% (95% CI, 96.8%-99.7%) and an NPV of 99.5% (95% CI, 98.4%-99.9%) (Figure 1A and B and eTable 3 in the Supplement). None of the 3 (0.5%) patients with false-negative results died within 30 days of presentation. eTable 4 in the Supplement summarizes the clinical characteristics of patients with a presentation hs-cTnI below the lower limit of detection and nonischemic ECG with a diagnosis of fatal or nonfatal AMI. All 3 patients were men and from a cohort (Poole) in which outcomes were adjudicated using an alternative high-sensitivity troponin assay (Elecsys hs-cTnT; Roche Diagnostics). At the upper limit of detection for the hs-cTnI assay (<2 ng/L), 807 (25.6%) patients would have been eligible for early discharge but with a sensitivity of 97.9% (95% CI, 95.4%-99.2%) and NPV of 99.3% (95% CI, 98.3%-99.7%).

The sensitivity decreased with increasing cutoff concentrations (Figure 1A and eTable 3 in the Supplement). All rounded cutoff values above 1.2 ng/L, up to and including 5 ng/L, had sensitivities less than 98% (albeit with NPVs >99%). The cutoff concentration of 5 ng/L had an excellent NPV (99.2% [95% CI, 98.8%-99.5%]) but poor sensitivity (94.5% [95% CI, 91.1%-96.7%]). The proportion of patients with an hs-cTnI value below each cutoff concentration and the cumulative missed-event rate are shown in Figure 2.

The sensitivity of an hs-cTnI concentration less than 1.2 ng/L for AMI was similar in men and women and when stratified by age, cardiac risk factor burden, and history of coronary artery disease (Figure 3 and eTable 5 in the Supplement). No patient aged 80 years or older (187 [5.9%]) had a presentation hs-cTnI concentration below 1.2 ng/L. In a cohort with documented time from symptom onset to ED presentation, 1047 of 3124 patients (33.5%) were classified as early presenters (time from symptom onset to ED presentation ≤2 hours). In these patients, test sensitivity for AMI at 1.2 ng/L was maintained at 98.6% (95% CI, 91.8%-99.9%) (Figure 3 and eTable 5 in the Supplement). The reduction in diagnostic performance of hs-cTnI cutoff concentrations above 1.2 ng/L to exclude AMI cutoff was more marked in early presenters (eFigure in the Supplement).

In more than 3000 patients with symptoms suggestive of cardiac ischemia obtained from 5 diverse international cohorts, we demonstrate that hs-cTnI obtained at presentation to the ED can, at a cutoff level of 1.2 ng/L in combination with a nonischemic ECG, identify 18.8% of patients as being potentially suitable for immediate discharge, with a high diagnostic performance in excluding AMI. To place these results in the context of absolute numbers of presenting patients, a number-needed-to-diagnose approach shows that, for the 1.2-ng/L cutoff level, for every 10 630 patients assessed, 1990 would be correctly reassured that they are not having an AMI, 10 would be falsely reassured, and 8630 would undergo further investigation, of whom 990 would ultimately receive a diagnosis of AMI. We also demonstrate that cutoff values above the lower limit of detection may not have the required diagnostic performance for clinical implementation. In particular, we identified sensitivity of less than 95% for the cutoff value of 5 ng/L, which has recently been recommended⁵ for implementation on the basis of an NPV greater than 99.5%. Although this high NPV is maintained in our analysis, the possibility of missing 5 or more AMIs for every 100 AMIs presenting to the ED is unacceptable for most physicians.⁷ Regarding sensitivity, we were unable to determine whether our findings differed significantly from those reported by Shah et al⁵ since sufficient information to calculate a missed AMI rate in a comparable cohort was not provided in that study.

Our analysis complements the work of studies and meta analyses^4,6,15-17 demonstrating that the limit of detection of hs-cTn assays (both I and T) show promise for excluding AMI early in ED patients with suspected cardiac chest pain especially when used in combination with ECG findings. However, to our knowledge, this analysis is the first to report full diagnostic performance statistics for a range of low concentrations of hs-cTnI across a range of international sites with varying degrees of disease prevalence.

Although the proportion of patients potentially suitable for discharge using the lower limit of detection is smaller than the proportion determined using rounded cutoff values, such as 5 ng/L as proposed by Shah and colleagues,⁵ the low sensitivity at concentrations above 1.2 ng/L brings into question the safety of these cutoff values as an exclusion strategy within a clinical environment. Furthermore, it has been demonstrated⁵ that the diagnostic performance of higher rounded cutoff values, such as 5 ng/L, may be reduced in patients presenting within 2 hours of chest pain onset. Our data suggest that diagnostic performance is retained in early presenters when the lower limit of detection cutoff value is used.

We demonstrate the importance of presenting both sensitivity and NPV to evaluate the diagnostic accuracy of early strategies to exclude AMI. The NPV is directly related to the prevalence of the target disease in the specific population under consideration and represents the posttest probability of a negative test result. When clinical implementation of an exclusion strategy is considered, it is therefore important to establish the NPV for each hospital so that an attending physician can better interpret a negative test result. The NPV should not be used to recommend generalization of a test across populations with varying disease prevalence. However, sensitivity is not affected by the disease prevalence, represents the true positive rate, and is therefore more useful to physicians in establishing the validity of a diagnostic test at an individual patient level.

We chose a sensitivity threshold of 99%, below which the miss-rate of AMI becomes unacceptable.⁷ When point estimates are considered, the only hs-cTnI cutoff value that reached this sensitivity threshold was the lower limit of detection (1.2 ng/L). However, the lower limits of the 95% CI at these cutoff points fall below this 99% sensitivity threshold. Therefore, we suggest that implementation of limit of detection exclusion strategies should undergo further evaluation in randomized clinical trials that test both clinical effect and cost-effectiveness in routine practice.

A key strength of our study is that all patients had outcomes adjudicated using serial troponin testing. This consistent evaluation enabled the identification of any patients who may subsequently have had an AMI and ensured accurate reporting of event rates. This method is in contrast to previous work^4,5 using registry or administrative data. More than one-third of the patients in these studies were not assessed using serial troponin test results; therefore, the number of missed events may have been higher than reported. Our data demonstrate that this issue may be more relevant in the era of hs-cTn assays. All 3 patients with an unidentified AMI according to the lower limit of detection of hs-cTnI were evaluated using serial testing with an hs-cTnT assay, and no events were missed when outcomes were evaluated with contemporary assays. When a high-sensitivity assay is used for outcome adjudication, the false-negative rate for AMI determined with a single baseline test may have increased when compared with contemporary assays. The diagnostic sensitivity of the index test may therefore be overestimated when an hs-cTn assay is used for outcome adjudication.

Our findings can be applied only to the assay tested (hs-cTnI) and cannot be translated to other assays, even high-sensitivity ones. The hs-cTnI assay is recommended for clinical use as a high-sensitivity assay by consensus guidelines.¹⁸ However, physicians should be aware that, at low concentrations, test results are less reliable than at the 99th percentile¹⁹; therefore, any clinical implementation should be a multidisciplinary decision between ED physicians, cardiologists, and laboratory staff. No hs-cTn assay is currently approved for use by the US Food and Drug Administration, and whether the use of low concentrations of hs-cTn in clinical practice will be approved remains unknown.

Our analysis has several limitations. First, outcomes were adjudicated using troponin assays, both sensitive and high-sensitivity, in clinical use at the time of patient recruitment rather than the hs-cTnI assay under evaluation. This approach may lead to misclassification bias. Second, many patients were excluded from the original cohorts as a result of the unavailability of hs-cTnI results. This exclusion may result in selection bias. Third, we selected patients in whom new-onset ECG changes diagnostic of ischemia were absent to reflect clinical practice. This method may serve to reduce the outcome prevalence in the cohort as a whole and therefore raise the NPV of the index test. Furthermore, the population selected for this analysis was likely to be of lower risk compared with those of prior studies.^4,5,15-17 Finally, no patient was discharged according to presentation hs-cTnI assay results and the proportion of patients undergoing further cardiac testing was high across all cohorts (Table). Therefore, whether the strategies tested can be successfully implemented into clinical practice and what further testing is required remains unknown. The choice of outcome measure for this analysis (fatal or nonfatal AMI occurring within 30 days) fails to incorporate the full spectrum of acute coronary syndromes (eg, patients who may require urgent revascularization). Therefore, a key concept in ensuring the safety of this biomarker strategy is the availability of timely outpatient testing to ensure early detection of such patients. Further studies are required (ideally a randomized clinical trial) to determine the clinical effectiveness of low cutoff concentrations of high-sensitivity troponin.

High-sensitivity troponin I concentrations determined at presentation to the ED that were below the limit of detection identified 18.8% of patients potentially suitable for discharge with a high sensitivity for AMI. Rounded cutoff values above the limit of detection may not have the required sensitivity for clinical implementation.

Corresponding Author: Edward Carlton, PhD, Emergency Department, Southmead Hospital, North Bristol National Health Service Trust, Southmead Road, Bristol, Avon BS10 5NB, England (eddcarlton@gmail.com).

Accepted for Publication: April 8, 2016.

Published Online: June 1, 2016. doi:10.1001/jamacardio.2016.1309.

Open Access: This article is published under the JAMA Cardiology open access model and is free to read on the day of publication.

Author Contributions: Dr Carlton had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Carlton, Cullen, Body, Aldous, Lord, Greaves.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Carlton, Than, Pickering, Carley, Greaves.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Carlton, Greenslade, Body, Carley, Lord.

Obtained funding: Carlton, Greenslade, Cullen, Than, Greaves.

Administrative, technical, or material support: Kendall, Greaves.

Study supervision: Cullen, Than, Aldous, Greaves.

Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Dr Carlton has undertaken research under collaborative agreements with Abbott and Randox Laboratories. Dr Cullen has received funding from Abbott, Roche, Alere, Siemens, and Radiometer Pacific for clinical trials and from Alere, Boehringer Ingelheim, Pfizer, AstraZeneca, Abbott, Novartis, and Radiometer Pacific for speaking and education. Dr Body has undertaken research under collaborative agreements with Roche, Siemens Diagnostics, Alere, and Randox Laboratories and has accepted travel and accommodation for conferences from Roche Diagnostics and Randox Laboratories. Dr Than has received funding from Alere, Abbott, Beckman, and Roche for speaking and support for other research. Dr Aldous received funding from the National Heart Foundation (New Zealand) for cardiac research. Dr Kendall has received funding from Boehringer Ingelheim and Novartis for speaking, education, and support for other research. Dr Parsonage received funding from the Queensland Emergency Medicine Research Foundation, Abbott Diagnostics, Roche, Alere, and Beckmann Coulter for research. He also has received honoraria, travel expenses, and consultancy fees from Abbott, AstraZeneca, Hospira, and Aventis. Dr Greaves has received funding from AstraZeneca for related research. No other disclosures were reported.

Funding/Support: Australia/New Zealand cohorts: Funding for the ADAPT study was predominantly provided by the Christchurch Heart Institute and the Queensland Emergency Medicine Research Foundation with a small (20%) contribution from industry (Abbott and Alere). Poole, England, cohort: Funding was provided by the Royal College of Emergency Medicine of the United Kingdom, Bournemouth University. Manchester, England, cohort: Funding was received from the Royal College of Emergency Medicine of the United Kingdom, as well as fellowship funding from the United Kingdom National Institute for Health Research (NIHR) and by the NIHR Clinical Research Network (UK CRN 8376). Stockport, England, cohort: Funding was received from the Royal College of Emergency Medicine of the United Kingdom as well as fellowship funding from the NIHR and by the NIHR Clinical Research Network (UK CRN 8376).

Role of the Funder/Sponsor: For all cohorts, no commercial organization or sponsor was involved in the design and conduct of the study; collection, management, analysis and interpretation of the data; and preparation, review or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We are indebted to the patients who participated in the study. We thank the research staff, emergency department staff, and laboratory technicians of all participating facilities for their valuable efforts.