Author + information
- Received October 20, 1999
- Revision received March 15, 2000
- Accepted June 28, 2000
- Published online November 15, 2000.
- Michael C Kontos, MD, FACCa,b,* (, )
- F.Philip Anderson, PhDc,
- Ramin Alimard, MDa,
- Joseph P Ornato, MD, FACCb,
- James L Tatum, MDd and
- Robert L Jesse, MD, PhD, FACCa
- ↵*Reprint requests and correspondence:
Dr. Michael C. Kontos, Room 7-074, Heart Station, 1300 E. Marshall Street, P.O. Box 980051, Richmond, Virginia 23298-0051
We sought to determine the predictive ability of troponin I (TnI) in a heterogeneous group of patients with chest pain admitted from the emergency department (ED) for exclusion of myocardial infarction (MI).
Previous studies in high-risk patients demonstrated that troponin elevations are associated with increased cardiac events. Little information is available on its predictive ability in more heterogeneous, lower risk patients.
Consecutive patients admitted from the ED for possible MI underwent serial myocardial marker sampling of TnI and creatine kinase, CK-MB over an 8-h period. Patients with ST segment elevation were excluded. End points included MI, death, significant complications (e.g., cardiac or respiratory arrest, intra-aortic balloon pump, pulmonary artery catheter or pacemaker placement, revascularization or inotropic infusion) and significant disease.
Events occurred in 513 (27%) of the 1,929 patients evaluated: MI in 175 (9.1%) and death in 34 (1.8%); an additional 248 patients (13%) without MI had complications, and 323 (17%) without MI had significant disease. Sensitivity of TnI for MI was high (96%). Patients without MI who were TnI-positive were more likely to have complications (43% vs. 12%) or significant disease (41% vs. 17%) as compared with those who were TnI-negative; however, the sensitivity of TnI for these two end points was low (14% and 21%, respectively). Predictive values were unchanged after excluding patients with ischemic electrocardiograms.
Troponin I had a high sensitivity for MI when used as part of a rapid rule-in protocol; however, the sensitivity for other end points was low. Use of TnI alone failed to identify the majority of patients who had either significant disease or complications.
Patients presenting with chest pain or other symptoms suggestive of myocardial ischemia represent a spectrum of diagnoses from musculoskeletal pain to acute myocardial infarction (MI). In patients who have ST segment elevation consistent with myocardial injury, MI is highly likely (1,2). However, these patients constitute only a minority of those presenting with chest pain. Accurate early identification of MI or myocardial ischemia in the remaining patients is difficult, resulting in the hospital admission of many patients who are ultimately found to have nonischemic causes for their symptoms (1).
Cardiac troponin T (TnT) and I (TnI) can identify patients who have minimal amounts of myocardial necrosis (3,4). Patients who have troponin elevations are at increased risk for adverse cardiac events, even in the absence of MI identified by traditional criteria (3,5). Most previous investigations examining the predictive ability of troponin were limited by inclusion of only high risk patients (3,6–8), the majority of whom had MI (7,8), used only the initial sample (7,8) or prolonged sampling for up to 24 to 48 h (3,9,10), thus limiting our knowledge of these markers’ performance when used in rapid rule-out protocols. Therefore, we examined the relation between cardiac events and TnI in a large, heterogeneous nonselected patient group undergoing a rapid MI diagnostic pathway.
The chest pain protocol used at our institution has been described in detail previously (11). All patients who present to the Medical College of Virginia Hospitals’ Emergency Department (ED) with symptoms suggestive of myocardial ischemia undergo prompt clinical evaluation by ED house staff and attending physicians. After the initial evaluation, patients thought to be at high risk (ischemic electrocardiographic [ECG] changes or typical symptoms in patients with known coronary disease) are admitted directly to the Coronary Care Unit (CCU), whereas those at low to moderate risk for acute coronary syndromes undergo further risk stratification using early rest myocardial perfusion imaging (11). Moderate risk patients are admitted, whereas low risk patients undergo perfusion imaging from the ED and are discharged and scheduled for outpatient stress testing if the images are negative. All monitoring and data collection were done as part of an ongoing quality-improvement process and were not considered to need approval from the Committee on the Conduct of Human Research. From June 1996 through May 11, 1998, 2,194 consecutive patients were admitted to the CCU from the ED for suspected MI. Patients with ST segment elevation meeting criteria for fibrinolytic therapy were excluded (n = 125), as MI was highly likely. An additional 124 patients did not have an 8-h TnI sample, and 16 patients left against medical advice before marker sampling was completed, leaving 1,929 patients. Only the initial hospital admission was included for patients admitted more than once. All patients underwent serial testing for creatine kinase, CK-MB by mass assay and total CK at 0, 3, 6 and 8 h after presentation. Troponin I was sampled at the time of admission and at 8 h. Sampling was continued at 6- to 8-h intervals in patients diagnosed with MI or those who had recurrent or continuing symptoms until a diagnosis was made. Decisions regarding further diagnostic evaluation were made by the attending cardiologist in the CCU.
The CK-MB assay (Behring Diagnostics, Boston, Massachusetts) used in this study has been described in detail previously (12). Total serum CK activity (Vitros, Johnson & Johnson, Rochester, New York) was determined using N-acetylcysteine activation without pretreatment. Troponin I was measured using the Opus Magnum Analyzer (Behring Diagnostics). Centrifuged plasma was filtered before analysis using 0.2-μm sterile filters fitted to 3-ml Becton Dickinson (Franklin Lakes, New Jersey) syringes. The lower limit for detectability for this assay is 0.5 ng/ml, and the manufacturers’ suggested diagnostic value for MI is 2.0 ng/ml.
Diagnosis of MI required symptoms consistent with myocardial ischemia with a CK-MB level ≥8.0 ng/ml and a relative index ≥4 ([CK-MB × 100]/total CK). The initial ECG from the ED was interpreted by a cardiologist who had no knowledge of the clinical variables, TnI results and outcomes. The ECGs showing transient ST segment elevation, ST segment depression ≥1 mm or T-wave inversion ≥2 mm were considered consistent with ischemia. All other ECGs were considered nonischemic. Patients undergoing coronary angiography had it performed using the Judkin’s technique, with views of the coronary arteries obtained in multiple projections. Significant coronary artery disease was defined as ≥50% stenosis of the left main coronary artery or ≥70% stenosis in a major coronary artery, its branches or a bypass graft. Cardiac death was defined as death secondary to MI or arrhythmia, or an unexpected death of unknown cause.
Cardiac end points included 1) MI within one week; 2) cardiac death within five weeks; 3) significant disease demonstrated on coronary angiography within five weeks; 4) significant complications, defined as revascularization (coronary artery bypass graft surgery [CABG] or percutaneous transluminal coronary angioplasty [PTCA]) within five weeks; or 5) any of the following which occurred during hospital admission: cardiac or respiratory arrest, intubation, insertion of an intra-aortic balloon pump, second- or third-degree heart block requiring placement of a transvenous pacemaker, ventricular tachycardia requiring treatment with an intravenous antiarrhythmic drug or cardioversion, ventricular fibrillation or the need for inotropic support. For patients who had more than one event, only the initial one was used for analysis. Predictive values for the non-MI end points were calculated after exclusion of the patients with MI.
Relative risks (13) were calculated as described previously. The results were compared using the Student t test for continuous variables and the chi-square test for dichotomous variables. A p value ≤0.05 was considered significant. Multivariate analysis was performed using stepwise logistic regression (SAS, version 6.11). Candidate variables included clinical variables (Table 1), as well as an ischemic ECG and a TnI level ≥1.0 ng/ml. Variables with a p value <0.10 were then entered into the logistic regression model.
The demographic data of the 1,929 patients who formed the study cohort are shown in Table 1. End points occurred in 513 patients and included death in 34 (1.8%) and MI in 175 (9.1%). Coronary angiography was performed in 534 patients without MI, 323 of whom had significant coronary disease. A total of 248 patients without MI had at least one significant complication, which included death in 20, revascularization in 193 (PTCA in 143 and CABG in 50), intravenous pressor support in 23, intubation in 19, pulmonary artery catheterization in 10, transvenous pacemaker placement in 6, intra-aortic balloon pump in 4 and ventricular tachycardia or fibrillation in 6.
A receiver-operating characteristic curve, derived using the end point of MI, demonstrated that the optimal cutoff value for MI was 1.0 ng/ml (Fig. 1), which was used for all later analyses.
A total of 288 patients (15%) had at least one positive TnI. Troponin I was positive in 175 (91%) of the 195 patients who had a MI or who died, and in 168 (96%) of the 175 patients with MI alone. The initial TnI was positive in 62 of the patients (35%) who had MI. Troponin I was positive in 52 patients (21%) who had significant complications and in 45 patients (14%) who had significant coronary disease. Predictive values and relative risks for TnI are shown in Table 2. Patients with a positive TnI were significantly more likely to have a MI, significant complications or significant disease (Fig. 2).
Troponin I was negative in seven patients diagnosed with MI. Four of the patients had a peak CK-MB level <10 ng/ml. A fifth patient had a peak CK-MB level of 30 ng/ml, but repeated TnI samples were all <0.5 ng/ml. Subsequent coronary angiography demonstrated no significant coronary disease. There was a minimal change from the first to the fourth CK-MB measurement in these five patients, suggesting that the elevations resulted from a noncardiac source. In the sixth patient, CK-MB was not elevated until the third sample; TnI 12 h after presentation was 3.0 ng/ml. In the seventh patient, CK-MB was not elevated until the fourth sample; TnI 10 h after presentation was 4.6 ng/ml. Troponin I was also negative in nine patients who had cardiac death without MI (mean duration from hospital admission to death 18 ± 12 days). Troponin I was positive in 59 patients who did not have one of the end points. The mean value was 3.4 ± 4.9 ng/ml (median 1.8 ng/ml [range 1.0 to 29]).
Ischemic ECG changes were present in 213 patients (11%). Of these, 51 patients died or had a MI, 41 had significant complications and 53 had significant coronary disease. The predictive values for an ischemic ECG are shown in Table 3. Patients who had an ischemic ECG were significantly more likely to have adverse events (Fig. 3).
The sensitivity of TnI was significantly higher than that of an ischemic ECG for most end points (Fig. 4). However, this was primarily a result of the high sensitivity of TnI for MI; sensitivities were not significantly different in patients who had complications or who had significant disease.
The predictive ability of TnI was reexamined after excluding the 213 patients with ischemic ECGs. Myocardial infarction or death occurred in 144 patients (8.4%), significant complications in 208 (12%) and significant disease in 269 (16%). The predictive ability of TnI was unchanged (Table 2).
Multivariate analysis demonstrated that a positive TnI value was the most important predictor of the combination of MI, death and significant coronary disease (Table 4). A positive TnI value was also the most important multivariate predictor of significant complications (Table 5)and of significant coronary disease (Table 6).
In the largest study to date to examine the predictive value of either TnT or TnI, we found that TnI had a high sensitivity for identifying patients who had MI, when it was used as part of an 8-h rapid diagnostic protocol. The sensitivities of TnI for the different end points were unchanged after excluding patients with an ischemic ECG. However, the sensitivity for identifying patients who had significant complications or significant coronary disease was low and was similar to the presence of ischemic ECG changes.
Troponin I in patients with MI
The sensitivity of TnI was high for identifying patients with MI, when it was used in this rapid 8-h diagnostic pathway. Although one study reported a relatively low sensitivity (9), most studies using prolonged troponin sampling have reported sensitivities similar to the current one (7,8,10). Reasons for variations include differences in underlying risk and frequency of MI, timing of presentation, time course of sampling and use of different assays and cutoff values. Inclusion of patients who have ST segment elevation can overestimate early sensitivity because of the larger infarct size and more rapid marker rise (14). Sensitivity is also affected by the choice of cutoff value. In contrast to studies that used the manufacturers’ suggested cutoff values, we used receiver-operator characteristic curve analysis to determine the optimal cutoff value.
Although sensitivity was high, some patients with MI did not have TnI elevations. Because peak troponin values in patients with MI are proportional to infarct size, they will be lower in patients with small MIs (15). Sensitivity is affected by the choice of the reference standard, which, if not perfect, results in some patients without necrosis being incorrectly diagnosed as having MI (16–18). This likely occurred in some of the patients who had minor CK-MB elevations without TnI elevations.
Finally, if sampling occurs early after the onset of necrosis, troponin values may be undetectable, as was seen in two patients with MI having late TnI elevations. It has been proposed that troponin sampling could be used as the only marker for identifying patients with myocardial necrosis (19–21). Our data indicate that if a rapid diagnostic rule-out pathway is used, inclusion of another marker that is released earlier after the onset of necrosis, such as CK-MB (22), is necessary to ensure that patients with MI are not misidentified.
Troponin I in patients without MI
The prevalence of patients without MI who had a positive TnI value was only 9.6%, which is lower than some (3,6–8) but not all (9,19,23) prevalences in previous studies. Initial studies were conducted in patients enrolled in clinical protocols (3,6–8) that included only high risk patients, with an associated higher incidence of troponin positivity. Studies performed in more heterogeneous groups of patients with chest pain have found a proportion of patients with troponin elevations similar to that of the current one (9,10,19,20).
Early identification of patients with acute coronary syndromes allows initiation of appropriate therapy, which can improve outcomes. Patients without MI who had TnI elevations were a high risk group, as the majority had complications or significant coronary disease. Importantly, excluding patients with ischemic ECG changes did not alter the predictive accuracy of TnI. This is significant, as this patient group presents the greatest diagnostic challenge for excluding myocardial ischemia.
Although TnI was an independent predictor of the non-MI cardiac events, the sensitivity was low, indicating that most patients with these end points did not have TnI elevations. Therefore, the absence of TnI elevations identifies a lower risk, but not necessarily a low risk, patient group. Previous studies have also found limitations of troponin for identifying high risk patients. These include little to no difference in the incidence of recurrent ischemia (6), revascularization (6,24–26) or the extent of coronary disease (8,27,28) in patients with and without troponin elevations. Although it has been suggested that patients with negative troponin values within 6 h of symptom onset could be safely discharged home (19,20), the low sensitivity and positive predictive value of TnI found in this study, as well the results of others (9,10), indicate that this strategy would misidentify many patients who are at risk of complications.
Our study differs from previous investigations in several ways. This is the largest study reported to examine the predictive ability of troponin in a heterogeneous group of patients with chest pain; therefore, it better defines the test’s diagnostic accuracy. We excluded patients with ST segment elevation, because this is a high risk group in whom the initial decision for reperfusion therapy is based only on clinical variables. Although they provide prognostic information, biochemical markers are not indicated for initial triage and treatment decisions in these patients. We analyzed the predictive ability of TnI after excluding patients with ischemic ECGs, because these are the patients in whom a diagnosis is in doubt and in whom markers play a critical role.
The TnI results were available to the clinicians caring for the patient and may have affected decisions for further care, such as the decision to perform coronary angiography. However, this would not affect the sensitivity for MI or for other complications. We used revascularization as an end point. Although there is a potential selection bias, all patients had rest angina in combination with significant coronary disease on the coronary angiogram. Despite its potential limitations, revascularization has been used as an end point in other studies of unstable angina (9,24,27,29). The incidence of revascularization in the current study was not excessive and was similar to (30) or lower than (3,6,29) the revascularization rates in previous studies. Use of other complications as end points is limited by the low overall risk in heterogeneous groups of patients (9,10). We did not include patients who were discharged from the ED after the initial evaluation. However, because our protocol requires rest perfusion imaging to be negative in low risk patients before they are discharged, the risk of serious complications is very low (<1%) (11).
We found that TnI had a high sensitivity for identifying patients who had a MI, when it was used in a rapid diagnostic protocol. Excluding patients with an ischemic ECG did not change the sensitivities of TnI for the different end points. However, the sensitivity for identifying patients who had significant complications or significant coronary disease was low and failed to identify the majority of non-MI end points.
- coronary artery bypass graft surgery
- coronary care unit
- creatine kinase
- electrocardiogram or electrocardiographic
- emergency department
- myocardial infarction
- percutaneous transluminal coronary angioplasty
- troponin I
- troponin T
- Received October 20, 1999.
- Revision received March 15, 2000.
- Accepted June 28, 2000.
- American College of Cardiology
- Yusuf S,
- Pearson M,
- Sterry H,
- et al.
- FRISC Study Group,
- Lindahl B,
- Venge P,
- Wallentin L
- Adams J.E III.,
- Bodor G.S,
- Davila-Roman V.G,
- et al.
- TRIM Study Group,
- Luscher M.S,
- Thygesen K,
- Ravkilde J,
- Heickendorff L
- Polanczyk C.A,
- Lee T.H,
- Cook E.F,
- et al.
- Sackett D.L,
- Haynes R.B,
- Guyatt G.H,
- Tugwell P
- de Winter R.J,
- Koster R.W,
- Sturk A,
- Sanders G.T
- Katus H.A,
- Remppis A,
- Neumann F.J,
- et al.
- Adams J.E III.,
- Schechtman K.B,
- Landt Y,
- Ladenson J.H,
- Jaffe A.S
- Wu A.H.B,
- Apple F.S,
- Warshaw M.M,
- Valdes R Jr.,
- Jesse R.L,
- Gibler W.B
- Galvani M,
- Ottani F,
- Ferrini D,
- et al.
- Stubbs P,
- Collinson P,
- Moseley D,
- Greenwood T,
- Noble M
- Mockel M,
- Stork T,
- Heller G.J,
- et al.
- Anderson H.V,
- Cannon C.P,
- Stone P.H,
- et al.