CLB-08846; No. of pages: 6; 4C: Clinical Biochemistry xxx (2014) xxx–xxx
Contents lists available at ScienceDirect
Clinical Biochemistry journal homepage: www.elsevier.com/locate/clinbiochem
Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial☆ Alan B. Storrow a,⁎, Robert H. Christenson b, Richard M. Nowak c, Deborah B. Diercks d, Adam J. Singer e, Alan H.B. Wu f, Erik Kulstad g, Frank LoVecchio h, Christian Fromm i, Gary Headden j, Tracie Potis k, Christopher J. Hogan l, Jon W. Schrock m, Daniel P. Zelinski n, Marna R. Greenberg o, James C. Ritchie p, Janna S. Chamberlin q, Kurtis R. Bray q, Daniel W. Rhodes q, Deirdre Trainor q, Dawn Holmes q, Paula C. Southwick q a
Dept. of Emergency Medicine, Vanderbilt University Medical Center, Nashville, TN, USA Dept. of Pathology, University of Maryland Medical Center, Baltimore, MD, USA Dept. of Emergency Medicine, Henry Ford Health System, Detroit, MI, USA d Dept. of Emergency Medicine, University of California Davis, Sacramento, CA, USA e Dept. of Emergency Medicine, State University of New York at Stony Brook, Stony Brook, NY, USA f Clinical Chemistry Laboratory, San Francisco General Hospital, University of California, San Francisco, CA, USA g Dept. of Emergency Medicine, Advocate Christ Medical Center, Oak Lawn, IL, USA h Dept. of Emergency Medicine, Maricopa Medical Center, Phoenix, AZ, USA i Dept. of Emergency Medicine, Maimonides Medical Center, Brooklyn, NY, USA j Chest Pain Center, Medical University of South Carolina, Charleston, SC, USA k Dept. of Emergency Medicine, Covenant Medical Center, Saginaw, MI, USA l Dept. of Emergency Medicine, Virginia Commonwealth University Medical Center, Richmond, VA, USA m Dept. of Emergency Medicine, Case Western Reserve University School of Medicine, MetroHealth Medical Center, Cleveland, OH, USA n Dept. of Emergency Medicine, The Ohio State University Medical Center, Columbus, OH, USA o Dept. of Emergency Medicine, Lehigh Valley Health Network, Allentown, PA, USA p Dept. of Pathology and Laboratory Medicine, Emory University, Atlanta, GA, USA q Dept. of Clinical Research, Beckman Coulter, Inc., Carlsbad, CA, USA b c
a r t i c l e
i n f o
Article history: Received 25 June 2014 Received in revised form 24 August 2014 Accepted 25 August 2014 Available online xxxx Keywords: Myocardial infarction Acute coronary syndromes Troponin Biomarkers Diagnosis Chest pain Emergency medicine
a b s t r a c t Objectives: To compare emergency department TnI serial sampling intervals, determine optimal diagnostic thresholds, and report representative diagnostic performance characteristics for early rule-in and rule-out of MI. Methods: We prospectively measured TnI (AccuTnI+3™, Beckman Coulter) at serial time intervals in 1929 subjects with chest pain or equivalent ischemic symptoms suggestive of acute coronary syndromes at 14 medical centers. Diagnosis was adjudicated by an independent central committee. Results: TnI ≥ 0.03 ng/mL provided 96.0% sensitivity and 89.4% specificity at 1–3 h after admission, and 94.9% sensitivity and 86.7% specificity at 3–6 h. NPV (rule-out, non-MI) was 99.5% at 1–3 h, and 99.0% at 3–6 h when TnI is b0.03 ng/mL. NPV was 99.1% when TnI is b 0.03 ng/mL and time of symptom onset is ≥ 8 h. Approximately 50–58% (PPV) of patients with TnI ≥ 0.03 ng/mL were diagnosed with MI, depending upon time from onset or admission; PPVs emphasize the importance of serial samples and delta TnI (rising or falling pattern) when low cutoffs are used. Nevertheless, even a single elevated TnI value increased the risk of MI. As TnI values rose, the probability of MI increased. Values ≥ 0.20 ng/mL were associated with nearly 90% probability of MI. Conclusions: We report a large multicenter prospective adjudicated trial assessing troponin for early rule-in and rule-out using the Universal Definition of MI and conducted in primary care hospital-associated
Abbreviations: ACS, acute coronary syndromes; AUC, Area Under the Curve; IQR, interquartile range; MI, myocardial infarction; NPV, Negative Predictive Value; PPV, Positive Predictive Value; ROC curve, Receiver Operating Characteristic curve; TnI, Troponin I; URL, Upper Reference Limit. ☆ Financial support: This work was supported by research grants from Beckman Coulter, Inc. (Brea, California, USA). ⁎ Corresponding author at: Department of Emergency Medicine, Vanderbilt University, 703 Oxford House, 1313 21st Ave South, Nashville, TN 37232, USA. Fax: +1 615 936 4490. E-mail address: [email protected] (A.B. Storrow).
http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018 0009-9120/© 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Please cite this article as: Storrow AB, et al, Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial, Clin Biochem (2014), http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018
2
A.B. Storrow et al. / Clinical Biochemistry xxx (2014) xxx–xxx
emergency departments. Our study demonstrates high diagnostic accuracy at early observation times, and reinforces consensus recommendations for sampling on admission and 3 h later, repeated at 6 h when clinical suspicion remains high. © 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Introduction As more sensitive troponin assays are introduced and lower diagnostic thresholds recommended [1–3], an increasing number of patients are identified with chronic cardiovascular comorbidities or acute nonischemic conditions, making the diagnosis of acute myocardial infarction (MI) more clinically challenging. Sensitivity for detection of MI is significantly improved, but diagnostic specificity is decreased. To improve diagnostic accuracy, the Third Universal Definition of MI [1] requires serial samples demonstrating a rise and/or fall in troponin concentration, with at least one value above the decision level for the diagnosis of MI. The optimal diagnostic threshold has become increasingly important, yet few studies exist reporting performance characteristics in large representative populations using the current standards of care for diagnosis. The objectives of this multicenter, prospective, adjudicated study using the Universal Definition of MI were to investigate and compare pre-specified serial sampling intervals (admission, 1–3 h, 3–6 h, 6–9 h), determine the optimal diagnostic threshold for TnI (AccuTnI+3™, Beckman Coulter, Inc.) in patients presenting to the emergency department (ED) with chest pain or equivalent ischemic symptoms suggestive of acute coronary syndromes (ACS), and report representative diagnostic performance characteristics (sensitivity, specificity, Positive Predictive Value, Negative Predictive Value) that would be observed in clinical practice for early rule-in and rule-out of MI. Methods Patient population This prospective trial enrolled 1929 subjects at 14 primary care hospital-associated emergency departments. Study entry criteria included presentation to the ED with chest pain or equivalent ischemic symptoms suggestive of ACS, and a 12-lead ECG taken at time of presentation. The trial was approved by the relevant institutional review boards at all participating centers. Written informed consent was obtained from all patients.
The AccuTnI+3 assay coefficient of variation (CV) as reported by the manufacturer is 10% at 0.04 ng/mL, 15% at 0.03 ng/mL, and 20% at 0.02 ng/mL. Actual total CVs obtained during the study for the controls at the low end of the range analyzed in duplicate twice per day were: 2.8% at 0.96 ng/mL, 3.0% at 0.63 ng/mL, 5.1% at 0.04 ng/mL and 15.8% at 0.02 ng/mL (Bio-Rad Liquichek Cardiac Markers Plus Control LT, and MORE Diagnostics Cardiac Markers Control samples). The assay Limit of Detection is b 0.01 ng/mL. The reportable measuring range of the assay is defined as the range from 20% CV to the S5 calibrator (~100 ng/mL). The independent external laboratories also tested all stored study samples using the Siemens ADVIA Centaur® TnI-Ultra™ assay, for later comparison to the AccuTnI+3 diagnostic accuracy results. Statistical methods The minimum sample size required for a 95% two-tailed confidence interval about a proportion is 73 to 139 MI patients assuming diagnostic sensitivity of 90% to 95%. Incidence of MI in the US in patients presenting with chest pain is approximately 7% to 15%. Therefore, nearly 2000 enrolled patients presenting with chest pain are required to yield 139 MI patients. To evaluate the clinical performance of the AccuTnI+3 assay, Receiver Operating Characteristic (ROC) curves, Areas Under the Curve (AUC), diagnostic sensitivity and specificity, Positive Predictive Values (PPVs) and Negative Predictive Values (NPVs) were determined for pre-specified serial time intervals after symptom onset (b8 h, ≥8 h) [5] and after admission to the hospital (baseline, 1–3 h, 3–6 h, 6–9 h). The following potential cutoffs were evaluated: the 99th percentile Upper Reference Limit (URL) (0.02 ng/mL), and the cutoff derived from ROC curve analysis of a feasibility trial of 363 ED patients presenting with chest pain (0.03 ng/mL, diagnostic sensitivity 95.5%). Other potential sequential cutoffs ranging from 0.02 to 0.20 were also assessed by ROC analysis. For continuous variables, the most comprehensive description of the spectrum of diagnostic sensitivity and specificity is provided by ROC analysis, which examines all possible decision thresholds and all corresponding combinations of sensitivity and specificity [6,7]. Results
Adjudication of clinical endpoint Baseline characteristics The primary clinical endpoint for this study was the diagnosis of myocardial infarction (MI). The diagnosis was based on the Universal Definition of myocardial infarction [4]. The troponin assay used for the diagnosis was the standard of care assay at each institution (Siemens ADVIA Centaur® TnI-Ultra™ assay). The clinical endpoint was adjudicated by an independent central endpoints committee, consisting of four cardiologists, blinded to AccuTnI+3 assay results and attending physicians' diagnosis. A rise and/or fall in TnI values with at least one value above the TnI-Ultra 99th percentile URL (0.04 ng/mL) was required to establish a diagnosis of MI [4].
Baseline characteristics of the study population (N = 1929) are listed in Table 1. Patients diagnosed with MI (N = 253, 13% [11.6–14.7]) were older, more frequently male, and were more likely to have known cardiovascular risk factors and comorbidities including hypertension, past smoker, coronary artery disease (CAD), prior MI, heart failure, and chronic kidney disease. The number of samples in each time interval was: baseline (1929), ≥ 1–3 h (1138), ≥ 3–6 h (1098), and ≥ 6–9 h (289). If patients are subcategorized according to time from symptom onset, 47.7% were b 8 h and 52.3% were ≥8 h after symptom onset.
Biomarker testing Distribution of troponin values by diagnosis Blood samples were collected and processed at each patient enrollment site. Plasma was separated from cells, dispensed into cryo-tubes, stored at −20 °C, and shipped to four independent external laboratories for testing. Troponin I was measured (Access AccuTnI+3™ on the Access 2® Immunoassay System, Beckman Coulter, Inc.) by personnel blinded to the standard of care assay results at each institution and patient diagnosis.
The median TnI value of all sample draws for MI patients was 0.11 ng/mL (interquartile range 0.04–0.47; 5th–95th 0.02–4.99) versus a median value for non-MI patients of 0.00 ng/mL (interquartile range 0.00–0.01; 5th–95th 0.00–0.04). MI values were significantly higher than non-MI values (p b 0.0001).
Please cite this article as: Storrow AB, et al, Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial, Clin Biochem (2014), http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018
A.B. Storrow et al. / Clinical Biochemistry xxx (2014) xxx–xxx
3
Table 1 Baseline characteristics by diagnosis. Data are expressed as median (interquartile range), or as n/N (%). p-Values are shown for each characteristic; p-values for sex, race, and TIMI risk score represent the relationship between diagnosis and all levels of the characteristic. Total (n = 1929) Age, median (IQR), years Weight, median (IQR), lbs
MI (n = 253)
56.0 (49.0–66.0) 190.0 (160.0–222.0)
61.0 (51.0–71.0) 185.0 (160.0–214.0)
Non-MI (n = 1676) 56.0 (49.0–65.0) 190.0 (160.0–224.0)
p-Value b0.001 0.085
Sex Male Female
1066/1929 (55.3) 863/1929 (44.7)
180/253 (71.1) 73/253 (28.9)
886/1676 (52.9) 790/1676 (47.1)
b0.001
Race White Black Other
1126/1839 (61.2) 622/1839 (33.8) 99/1839 (5.4)
147/233 (63.1) 72/233 (30.9) 16/233 (6.9)
979/1606 (61.0) 550/1606 (34.2) 83/1606 (5.2)
0.386
Risk factors Hypertension (HTN) Dyslipidemia or hypercholesterolemia Diabetes mellitus (DM) Current smoker Past smoker Known coronary artery disease (CAD), documented stenosis N50% Prior myocardial infarction (MI) Prior coronary artery stent or angioplasty Prior coronary artery bypass graft (CABG) Heart failure (HF) chronic kidney disease or chronic renal insufficiency
1363/1923 (70.9) 1023/1895 (54.0) 570/1923 (29.6) 529/1909 (27.7) 637/1863 (34.2) 592/1863 (31.8) 459/1907 (24.1) 451/1915 (23.6) 216/1926 (11.2) 308/1917 (16.1) 173/1922 (9.0)
193/252 (76.6) 142/248 (57.3) 86/253 (34.0) 76/251 (30.3) 99/242 (40.9) 112/242 (46.3) 86/249 (34.5) 84/250 (33.6) 37/253 (14.6) 60/251 (23.9) 33/253 (13.0)
1170/1671 (70.0) 881/1647 (53.5) 484/1670 (29.0) 453/1658 (27.3) 538/1621 (33.2) 480/1621 (29.6) 373/1658 (22.5) 367/1665 (22.0) 179/1673 (10.7) 248/1666 (14.9) 140/1669 (8.4)
0.032 0.267 0.104 0.329 0.018 b0.001 b0.001 b0.001 0.065 b0.001 0.016
TIMI risk score: 0–2 (low risk) 3–4 (intermediate risk) 5–7 (high risk)
1238/1929 (64.2) 627/1929 (32.5) 64/1929 (3.3)
122/253 (48.2) 115/253 (45.5) 16/253 (6.3)
1116/1676 (66.6) 512/1676 (30.5) 48/1676 (2.9)
b0.001
Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC)
Fig. 2 presents the probability of MI (PPV) at baseline, stratified by baseline TnI value.
ROC curves are shown in Fig. 1A for time from symptom onset. AUCs were identical for b 8 h (AUC 0.96 [0.95–0.97]) and ≥8 h after symptom onset (AUC 0.96 [0.94–0.98]). ROC curves are shown in Fig. 1B for time from admission: baseline (AUC 0.94 [0.92–0.95]), 1–3 h (AUC 0.97 [0.96–0.98]), 3–6 h (AUC 0.96 [0.95–0.98]), and 6–9 h (AUC 0.94 [0.89–0.99]). Fig. 1C shows ROC curves for males (AUC 0.96 [0.94– 0.98]) and females (AUC 0.96 [0.93–0.98]) ≥8 h after symptom onset. Results were not significantly different (p = 0.76). Results were also not significantly different when analyzed by race: White AUC 0.96 (0.94–0.99), Black AUC 0.95 (0.91–0.98).
Comparison of AccuTnI+3 and TnI-Ultra assays
Diagnostic sensitivity and specificity Table 2A shows diagnostic sensitivity and specificity for sequential TnI cutoffs by time after symptom onset. The pre-specified hypothesized ROC-curve-derived cutoff of ≥0.03 ng/mL yielded 91.5% sensitivity and 89.1% specificity b 8 h after symptom onset, and 94.3% sensitivity and 86.8% specificity ≥8 h after symptom onset. The URL (0.02 ng/mL) yielded 96.9% sensitivity and 81.7% specificity ≥ 8 h after symptom onset. Specificity was significantly higher using a diagnostic threshold for MI of 0.03 ng/mL (86.8%) versus 0.02 ng/mL (81.7%) (p b 0.001). Table 2B shows diagnostic sensitivity and specificity for sequential TnI cutoffs by time after admission. Sensitivity at a cutoff ≥0.03 was as follows: baseline (87.4%), 1–3 h (96.0%), 3–6 h (94.9%), and 6–9 h (90.7%). Of the 253 MI patients, 12.6% did not show an elevation in TnI (b0.03 ng/mL) upon admission. Of the 1676 cases for which MI was ruled out, 10.7% had values above this TnI cutoff at admission. Positive Predictive Value (PPV) and Negative Predictive Value (NPV) Table 3A shows PPV and NPV for the URL (0.02 ng/mL) and the AccuTnI+3 diagnostic cutoff for MI (0.03 ng/mL) by time after symptom onset. Table 3B shows PPV and NPV by time after admission.
Troponin measurements were obtained for all study samples using both the AccuTnI+3 and TnI-Ultra assays in order to allow comparison of diagnostic performance results on the identical cohort of patients and samples. Table 4 shows diagnostic sensitivity and specificity (A), and PPV and NPV (B), for assay URLs and ROC-derived cutoffs, ≥8 h after symptom onset. Sensitivity, specificity, PPV and NPV values were similar for both assays at each sequential cutoff. The ROC-derived cutoff (0.03) optimized clinical performance for both assays (~95% sensitivity and ~86% specificity). The AccuTnI+3 URL (0.02 [0.01–0.05], number of healthy individuals = 527) yielded significantly lower specificity (81.7%) than the ROC-derived cutoff (0.03) (86.8%) (p b 0.001). The TnI-Ultra URL (0.04 [0.02–0.06], number of healthy individuals = 648) yielded significantly lower sensitivity (88.7%) than the ROC-derived cutoff (0.03) (95.0%) (p b 0.05). PPV values were low at 0.02 (~ 40%). Values were higher at 0.04 (~60%), but at the cost of lower sensitivity (~88%). Discussion We report results for a large multicenter prospective adjudicated trial assessing troponin for early rule-in and rule-out using the Universal Definition of MI and conducted in primary care hospital-associated emergency departments. Incidence of MI in our study population was 13%, representative of the ED populations in which troponin would be used. We found that AccuTnI+3 troponin concentrations above a diagnostic threshold of 0.03 ng/mL provided 96.0% sensitivity and 89.4% specificity 1–3 h after admission, and 94.9% sensitivity and 86.7% specificity 3–6 h after admission, in patients presenting with chest discomfort or equivalent ischemic symptoms suggestive of acute coronary syndromes (ACS).
Please cite this article as: Storrow AB, et al, Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial, Clin Biochem (2014), http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018
4
A.B. Storrow et al. / Clinical Biochemistry xxx (2014) xxx–xxx
B 1.0
1.0
0.9
0.9
0.8
0.8
0.7
0.7
Sensitivity
Sensitivity
A
0.6 0.5 0.4
0.6 0.5 0.4
0.3
0.3
0.2
0.2
0.1
0.1
Contents lists available at ScienceDirect
Clinical Biochemistry journal homepage: www.elsevier.com/locate/clinbiochem
Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial☆ Alan B. Storrow a,⁎, Robert H. Christenson b, Richard M. Nowak c, Deborah B. Diercks d, Adam J. Singer e, Alan H.B. Wu f, Erik Kulstad g, Frank LoVecchio h, Christian Fromm i, Gary Headden j, Tracie Potis k, Christopher J. Hogan l, Jon W. Schrock m, Daniel P. Zelinski n, Marna R. Greenberg o, James C. Ritchie p, Janna S. Chamberlin q, Kurtis R. Bray q, Daniel W. Rhodes q, Deirdre Trainor q, Dawn Holmes q, Paula C. Southwick q a
Dept. of Emergency Medicine, Vanderbilt University Medical Center, Nashville, TN, USA Dept. of Pathology, University of Maryland Medical Center, Baltimore, MD, USA Dept. of Emergency Medicine, Henry Ford Health System, Detroit, MI, USA d Dept. of Emergency Medicine, University of California Davis, Sacramento, CA, USA e Dept. of Emergency Medicine, State University of New York at Stony Brook, Stony Brook, NY, USA f Clinical Chemistry Laboratory, San Francisco General Hospital, University of California, San Francisco, CA, USA g Dept. of Emergency Medicine, Advocate Christ Medical Center, Oak Lawn, IL, USA h Dept. of Emergency Medicine, Maricopa Medical Center, Phoenix, AZ, USA i Dept. of Emergency Medicine, Maimonides Medical Center, Brooklyn, NY, USA j Chest Pain Center, Medical University of South Carolina, Charleston, SC, USA k Dept. of Emergency Medicine, Covenant Medical Center, Saginaw, MI, USA l Dept. of Emergency Medicine, Virginia Commonwealth University Medical Center, Richmond, VA, USA m Dept. of Emergency Medicine, Case Western Reserve University School of Medicine, MetroHealth Medical Center, Cleveland, OH, USA n Dept. of Emergency Medicine, The Ohio State University Medical Center, Columbus, OH, USA o Dept. of Emergency Medicine, Lehigh Valley Health Network, Allentown, PA, USA p Dept. of Pathology and Laboratory Medicine, Emory University, Atlanta, GA, USA q Dept. of Clinical Research, Beckman Coulter, Inc., Carlsbad, CA, USA b c
a r t i c l e
i n f o
Article history: Received 25 June 2014 Received in revised form 24 August 2014 Accepted 25 August 2014 Available online xxxx Keywords: Myocardial infarction Acute coronary syndromes Troponin Biomarkers Diagnosis Chest pain Emergency medicine
a b s t r a c t Objectives: To compare emergency department TnI serial sampling intervals, determine optimal diagnostic thresholds, and report representative diagnostic performance characteristics for early rule-in and rule-out of MI. Methods: We prospectively measured TnI (AccuTnI+3™, Beckman Coulter) at serial time intervals in 1929 subjects with chest pain or equivalent ischemic symptoms suggestive of acute coronary syndromes at 14 medical centers. Diagnosis was adjudicated by an independent central committee. Results: TnI ≥ 0.03 ng/mL provided 96.0% sensitivity and 89.4% specificity at 1–3 h after admission, and 94.9% sensitivity and 86.7% specificity at 3–6 h. NPV (rule-out, non-MI) was 99.5% at 1–3 h, and 99.0% at 3–6 h when TnI is b0.03 ng/mL. NPV was 99.1% when TnI is b 0.03 ng/mL and time of symptom onset is ≥ 8 h. Approximately 50–58% (PPV) of patients with TnI ≥ 0.03 ng/mL were diagnosed with MI, depending upon time from onset or admission; PPVs emphasize the importance of serial samples and delta TnI (rising or falling pattern) when low cutoffs are used. Nevertheless, even a single elevated TnI value increased the risk of MI. As TnI values rose, the probability of MI increased. Values ≥ 0.20 ng/mL were associated with nearly 90% probability of MI. Conclusions: We report a large multicenter prospective adjudicated trial assessing troponin for early rule-in and rule-out using the Universal Definition of MI and conducted in primary care hospital-associated
Abbreviations: ACS, acute coronary syndromes; AUC, Area Under the Curve; IQR, interquartile range; MI, myocardial infarction; NPV, Negative Predictive Value; PPV, Positive Predictive Value; ROC curve, Receiver Operating Characteristic curve; TnI, Troponin I; URL, Upper Reference Limit. ☆ Financial support: This work was supported by research grants from Beckman Coulter, Inc. (Brea, California, USA). ⁎ Corresponding author at: Department of Emergency Medicine, Vanderbilt University, 703 Oxford House, 1313 21st Ave South, Nashville, TN 37232, USA. Fax: +1 615 936 4490. E-mail address: [email protected] (A.B. Storrow).
http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018 0009-9120/© 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Please cite this article as: Storrow AB, et al, Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial, Clin Biochem (2014), http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018
2
A.B. Storrow et al. / Clinical Biochemistry xxx (2014) xxx–xxx
emergency departments. Our study demonstrates high diagnostic accuracy at early observation times, and reinforces consensus recommendations for sampling on admission and 3 h later, repeated at 6 h when clinical suspicion remains high. © 2014 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.
Introduction As more sensitive troponin assays are introduced and lower diagnostic thresholds recommended [1–3], an increasing number of patients are identified with chronic cardiovascular comorbidities or acute nonischemic conditions, making the diagnosis of acute myocardial infarction (MI) more clinically challenging. Sensitivity for detection of MI is significantly improved, but diagnostic specificity is decreased. To improve diagnostic accuracy, the Third Universal Definition of MI [1] requires serial samples demonstrating a rise and/or fall in troponin concentration, with at least one value above the decision level for the diagnosis of MI. The optimal diagnostic threshold has become increasingly important, yet few studies exist reporting performance characteristics in large representative populations using the current standards of care for diagnosis. The objectives of this multicenter, prospective, adjudicated study using the Universal Definition of MI were to investigate and compare pre-specified serial sampling intervals (admission, 1–3 h, 3–6 h, 6–9 h), determine the optimal diagnostic threshold for TnI (AccuTnI+3™, Beckman Coulter, Inc.) in patients presenting to the emergency department (ED) with chest pain or equivalent ischemic symptoms suggestive of acute coronary syndromes (ACS), and report representative diagnostic performance characteristics (sensitivity, specificity, Positive Predictive Value, Negative Predictive Value) that would be observed in clinical practice for early rule-in and rule-out of MI. Methods Patient population This prospective trial enrolled 1929 subjects at 14 primary care hospital-associated emergency departments. Study entry criteria included presentation to the ED with chest pain or equivalent ischemic symptoms suggestive of ACS, and a 12-lead ECG taken at time of presentation. The trial was approved by the relevant institutional review boards at all participating centers. Written informed consent was obtained from all patients.
The AccuTnI+3 assay coefficient of variation (CV) as reported by the manufacturer is 10% at 0.04 ng/mL, 15% at 0.03 ng/mL, and 20% at 0.02 ng/mL. Actual total CVs obtained during the study for the controls at the low end of the range analyzed in duplicate twice per day were: 2.8% at 0.96 ng/mL, 3.0% at 0.63 ng/mL, 5.1% at 0.04 ng/mL and 15.8% at 0.02 ng/mL (Bio-Rad Liquichek Cardiac Markers Plus Control LT, and MORE Diagnostics Cardiac Markers Control samples). The assay Limit of Detection is b 0.01 ng/mL. The reportable measuring range of the assay is defined as the range from 20% CV to the S5 calibrator (~100 ng/mL). The independent external laboratories also tested all stored study samples using the Siemens ADVIA Centaur® TnI-Ultra™ assay, for later comparison to the AccuTnI+3 diagnostic accuracy results. Statistical methods The minimum sample size required for a 95% two-tailed confidence interval about a proportion is 73 to 139 MI patients assuming diagnostic sensitivity of 90% to 95%. Incidence of MI in the US in patients presenting with chest pain is approximately 7% to 15%. Therefore, nearly 2000 enrolled patients presenting with chest pain are required to yield 139 MI patients. To evaluate the clinical performance of the AccuTnI+3 assay, Receiver Operating Characteristic (ROC) curves, Areas Under the Curve (AUC), diagnostic sensitivity and specificity, Positive Predictive Values (PPVs) and Negative Predictive Values (NPVs) were determined for pre-specified serial time intervals after symptom onset (b8 h, ≥8 h) [5] and after admission to the hospital (baseline, 1–3 h, 3–6 h, 6–9 h). The following potential cutoffs were evaluated: the 99th percentile Upper Reference Limit (URL) (0.02 ng/mL), and the cutoff derived from ROC curve analysis of a feasibility trial of 363 ED patients presenting with chest pain (0.03 ng/mL, diagnostic sensitivity 95.5%). Other potential sequential cutoffs ranging from 0.02 to 0.20 were also assessed by ROC analysis. For continuous variables, the most comprehensive description of the spectrum of diagnostic sensitivity and specificity is provided by ROC analysis, which examines all possible decision thresholds and all corresponding combinations of sensitivity and specificity [6,7]. Results
Adjudication of clinical endpoint Baseline characteristics The primary clinical endpoint for this study was the diagnosis of myocardial infarction (MI). The diagnosis was based on the Universal Definition of myocardial infarction [4]. The troponin assay used for the diagnosis was the standard of care assay at each institution (Siemens ADVIA Centaur® TnI-Ultra™ assay). The clinical endpoint was adjudicated by an independent central endpoints committee, consisting of four cardiologists, blinded to AccuTnI+3 assay results and attending physicians' diagnosis. A rise and/or fall in TnI values with at least one value above the TnI-Ultra 99th percentile URL (0.04 ng/mL) was required to establish a diagnosis of MI [4].
Baseline characteristics of the study population (N = 1929) are listed in Table 1. Patients diagnosed with MI (N = 253, 13% [11.6–14.7]) were older, more frequently male, and were more likely to have known cardiovascular risk factors and comorbidities including hypertension, past smoker, coronary artery disease (CAD), prior MI, heart failure, and chronic kidney disease. The number of samples in each time interval was: baseline (1929), ≥ 1–3 h (1138), ≥ 3–6 h (1098), and ≥ 6–9 h (289). If patients are subcategorized according to time from symptom onset, 47.7% were b 8 h and 52.3% were ≥8 h after symptom onset.
Biomarker testing Distribution of troponin values by diagnosis Blood samples were collected and processed at each patient enrollment site. Plasma was separated from cells, dispensed into cryo-tubes, stored at −20 °C, and shipped to four independent external laboratories for testing. Troponin I was measured (Access AccuTnI+3™ on the Access 2® Immunoassay System, Beckman Coulter, Inc.) by personnel blinded to the standard of care assay results at each institution and patient diagnosis.
The median TnI value of all sample draws for MI patients was 0.11 ng/mL (interquartile range 0.04–0.47; 5th–95th 0.02–4.99) versus a median value for non-MI patients of 0.00 ng/mL (interquartile range 0.00–0.01; 5th–95th 0.00–0.04). MI values were significantly higher than non-MI values (p b 0.0001).
Please cite this article as: Storrow AB, et al, Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial, Clin Biochem (2014), http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018
A.B. Storrow et al. / Clinical Biochemistry xxx (2014) xxx–xxx
3
Table 1 Baseline characteristics by diagnosis. Data are expressed as median (interquartile range), or as n/N (%). p-Values are shown for each characteristic; p-values for sex, race, and TIMI risk score represent the relationship between diagnosis and all levels of the characteristic. Total (n = 1929) Age, median (IQR), years Weight, median (IQR), lbs
MI (n = 253)
56.0 (49.0–66.0) 190.0 (160.0–222.0)
61.0 (51.0–71.0) 185.0 (160.0–214.0)
Non-MI (n = 1676) 56.0 (49.0–65.0) 190.0 (160.0–224.0)
p-Value b0.001 0.085
Sex Male Female
1066/1929 (55.3) 863/1929 (44.7)
180/253 (71.1) 73/253 (28.9)
886/1676 (52.9) 790/1676 (47.1)
b0.001
Race White Black Other
1126/1839 (61.2) 622/1839 (33.8) 99/1839 (5.4)
147/233 (63.1) 72/233 (30.9) 16/233 (6.9)
979/1606 (61.0) 550/1606 (34.2) 83/1606 (5.2)
0.386
Risk factors Hypertension (HTN) Dyslipidemia or hypercholesterolemia Diabetes mellitus (DM) Current smoker Past smoker Known coronary artery disease (CAD), documented stenosis N50% Prior myocardial infarction (MI) Prior coronary artery stent or angioplasty Prior coronary artery bypass graft (CABG) Heart failure (HF) chronic kidney disease or chronic renal insufficiency
1363/1923 (70.9) 1023/1895 (54.0) 570/1923 (29.6) 529/1909 (27.7) 637/1863 (34.2) 592/1863 (31.8) 459/1907 (24.1) 451/1915 (23.6) 216/1926 (11.2) 308/1917 (16.1) 173/1922 (9.0)
193/252 (76.6) 142/248 (57.3) 86/253 (34.0) 76/251 (30.3) 99/242 (40.9) 112/242 (46.3) 86/249 (34.5) 84/250 (33.6) 37/253 (14.6) 60/251 (23.9) 33/253 (13.0)
1170/1671 (70.0) 881/1647 (53.5) 484/1670 (29.0) 453/1658 (27.3) 538/1621 (33.2) 480/1621 (29.6) 373/1658 (22.5) 367/1665 (22.0) 179/1673 (10.7) 248/1666 (14.9) 140/1669 (8.4)
0.032 0.267 0.104 0.329 0.018 b0.001 b0.001 b0.001 0.065 b0.001 0.016
TIMI risk score: 0–2 (low risk) 3–4 (intermediate risk) 5–7 (high risk)
1238/1929 (64.2) 627/1929 (32.5) 64/1929 (3.3)
122/253 (48.2) 115/253 (45.5) 16/253 (6.3)
1116/1676 (66.6) 512/1676 (30.5) 48/1676 (2.9)
b0.001
Receiver Operating Characteristic (ROC) curves and Area Under the Curve (AUC)
Fig. 2 presents the probability of MI (PPV) at baseline, stratified by baseline TnI value.
ROC curves are shown in Fig. 1A for time from symptom onset. AUCs were identical for b 8 h (AUC 0.96 [0.95–0.97]) and ≥8 h after symptom onset (AUC 0.96 [0.94–0.98]). ROC curves are shown in Fig. 1B for time from admission: baseline (AUC 0.94 [0.92–0.95]), 1–3 h (AUC 0.97 [0.96–0.98]), 3–6 h (AUC 0.96 [0.95–0.98]), and 6–9 h (AUC 0.94 [0.89–0.99]). Fig. 1C shows ROC curves for males (AUC 0.96 [0.94– 0.98]) and females (AUC 0.96 [0.93–0.98]) ≥8 h after symptom onset. Results were not significantly different (p = 0.76). Results were also not significantly different when analyzed by race: White AUC 0.96 (0.94–0.99), Black AUC 0.95 (0.91–0.98).
Comparison of AccuTnI+3 and TnI-Ultra assays
Diagnostic sensitivity and specificity Table 2A shows diagnostic sensitivity and specificity for sequential TnI cutoffs by time after symptom onset. The pre-specified hypothesized ROC-curve-derived cutoff of ≥0.03 ng/mL yielded 91.5% sensitivity and 89.1% specificity b 8 h after symptom onset, and 94.3% sensitivity and 86.8% specificity ≥8 h after symptom onset. The URL (0.02 ng/mL) yielded 96.9% sensitivity and 81.7% specificity ≥ 8 h after symptom onset. Specificity was significantly higher using a diagnostic threshold for MI of 0.03 ng/mL (86.8%) versus 0.02 ng/mL (81.7%) (p b 0.001). Table 2B shows diagnostic sensitivity and specificity for sequential TnI cutoffs by time after admission. Sensitivity at a cutoff ≥0.03 was as follows: baseline (87.4%), 1–3 h (96.0%), 3–6 h (94.9%), and 6–9 h (90.7%). Of the 253 MI patients, 12.6% did not show an elevation in TnI (b0.03 ng/mL) upon admission. Of the 1676 cases for which MI was ruled out, 10.7% had values above this TnI cutoff at admission. Positive Predictive Value (PPV) and Negative Predictive Value (NPV) Table 3A shows PPV and NPV for the URL (0.02 ng/mL) and the AccuTnI+3 diagnostic cutoff for MI (0.03 ng/mL) by time after symptom onset. Table 3B shows PPV and NPV by time after admission.
Troponin measurements were obtained for all study samples using both the AccuTnI+3 and TnI-Ultra assays in order to allow comparison of diagnostic performance results on the identical cohort of patients and samples. Table 4 shows diagnostic sensitivity and specificity (A), and PPV and NPV (B), for assay URLs and ROC-derived cutoffs, ≥8 h after symptom onset. Sensitivity, specificity, PPV and NPV values were similar for both assays at each sequential cutoff. The ROC-derived cutoff (0.03) optimized clinical performance for both assays (~95% sensitivity and ~86% specificity). The AccuTnI+3 URL (0.02 [0.01–0.05], number of healthy individuals = 527) yielded significantly lower specificity (81.7%) than the ROC-derived cutoff (0.03) (86.8%) (p b 0.001). The TnI-Ultra URL (0.04 [0.02–0.06], number of healthy individuals = 648) yielded significantly lower sensitivity (88.7%) than the ROC-derived cutoff (0.03) (95.0%) (p b 0.05). PPV values were low at 0.02 (~ 40%). Values were higher at 0.04 (~60%), but at the cost of lower sensitivity (~88%). Discussion We report results for a large multicenter prospective adjudicated trial assessing troponin for early rule-in and rule-out using the Universal Definition of MI and conducted in primary care hospital-associated emergency departments. Incidence of MI in our study population was 13%, representative of the ED populations in which troponin would be used. We found that AccuTnI+3 troponin concentrations above a diagnostic threshold of 0.03 ng/mL provided 96.0% sensitivity and 89.4% specificity 1–3 h after admission, and 94.9% sensitivity and 86.7% specificity 3–6 h after admission, in patients presenting with chest discomfort or equivalent ischemic symptoms suggestive of acute coronary syndromes (ACS).
Please cite this article as: Storrow AB, et al, Diagnostic performance of cardiac Troponin I for early rule-in and rule-out of acute myocardial infarction: Results of a prospective multicenter trial, Clin Biochem (2014), http://dx.doi.org/10.1016/j.clinbiochem.2014.08.018
4
A.B. Storrow et al. / Clinical Biochemistry xxx (2014) xxx–xxx
B 1.0
1.0
0.9
0.9
0.8
0.8
0.7
0.7
Sensitivity
Sensitivity
A
0.6 0.5 0.4
0.6 0.5 0.4
0.3
0.3
0.2
0.2
0.1
0.1
