OUP user menu

Early diagnosis of acute myocardial infarction in patients with pre-existing coronary artery disease using more sensitive cardiac troponin assays

Miriam Reiter, Raphael Twerenbold, Tobias Reichlin, Benjamin Benz, Philip Haaf, Julia Meissner, Willibald Hochholzer, Claudia Stelzig, Michael Freese, Corinna Heinisch, Cathrin Balmelli, Beatrice Drexler, Heike Freidank, Katrin Winkler, Isabel Campodarve, Joaquim Gea, Christian Mueller
DOI: http://dx.doi.org/10.1093/eurheartj/ehr376 988-997 First published online: 1 November 2011

Abstract

Aims We sought to examine the diagnostic and prognostic utility of sensitive cardiac troponin (cTn) assays in patients with pre-existing coronary artery disease (CAD).

Methods and results We conducted a multicentre study to examine the diagnostic accuracy of one high-sensitive and two sensitive cTn assays in 1098 consecutive patients presenting with symptoms suggestive of acute myocardial infarction (AMI), of whom 401 (37%) had pre-existing CAD. Measurements of Roche high-sensitive cTnT (hs-cTnT), Siemens cTnI-Ultra, Abbott-Architect cTnI and the standard assay (Roche cTnT) were performed in a blinded fashion. The final diagnosis was adjudicated by two independent cardiologists. Acute myocardial infarction was the final diagnosis in 19% of CAD patients. Among patients with diagnoses other than AMI, baseline cTn levels were elevated above the 99th percentile with Roche hs-cTnT in 40%, with Siemens TnI-Ultra in 15%, and Abbott-Architect cTnI in 13% of them. In patients with pre-existing CAD, the diagnostic accuracy at presentation, quantified by the area under the receiver operator characteristic curve (AUC), was significantly greater for the sensitive cTn assays compared with the standard assay (AUC for Roche hs-cTnT, 0.92; Siemens cTnI-Ultra, 0.94; and Abbott-Architect cTnI, 0.93 vs. AUC for the standard assay, 0.87; P < 0.01 for all comparisons). Elevated levels of cTn measured with the sensitive assays predicted mortality irrespective of pre-existing CAD, age, sex, and cardiovascular risk factors.

Conclusion Sensitive cTn assays have high-diagnostic accuracy also in CAD patients. Mild elevations are common in non-AMI patients and test-specific optimal cut-off levels tend to be higher in CAD patients than in patients without history of CAD. Sensitive cTn assays also retain prognostic value. (ClinicalTrials.gov number, NCT00470587).

  • Acute myocardial infarction
  • Coronary artery disease
  • Troponin
  • Diagnosis
  • Prognosis

See page 944 for the editorial comment on this article (doi:10.1093/eurheartj/ehs009)

Introduction

Acute myocardial infarction (AMI) is a major cause of death and disability. Its rapid and accurate diagnosis is critical for effective evidence-based medical management and treatment,1,2 but still an unmet clinical need. Delayed ‘rule-in’ increases morbidity and mortality, particularly in patients with pre-existing coronary artery disease (CAD).3,4 Delayed ‘rule-out’ prolongs the time spent in the emergency department (ED), increasing patients' anxiety, and causes enormous costs for the health-care system.5

More sensitive cardiac troponin (cTn) assays with a limit of detection (LoD) below the 99th percentile of a reference population and improved precision have recently become available in clinical practice.68 These assays improved the early diagnosis of AMI in unselected patients with acute chest pain.9,10 However, their diagnostic accuracy in patients with pre-existing CAD is uncertain, as recently elevated cTn levels were found in >10% of patients with stable CAD.11,12

Also for several other reasons, patients with pre-existing CAD merit particular attention. First, they are at increased risk for both AMI as well as anxiety related to non-cardiac causes of chest pain. Secondly, interpretation of a 12-lead electrocardiography (ECG) is challenging in these patients: pre-existing ST-segment and T-wave alterations are frequent, and new ST-segment elevation is less common in patients with pre-existing CAD.13 Thirdly, the utility of CT angiography is considerably reduced in such patients.14,15 Fourthly, the impact of myocardial loss is particularly devastating when the ventricles have already suffered previous assaults, and delayed diagnosis of AMI yields especially severe consequences.3,4 We therefore examined the diagnostic performance of more sensitive cTn assays for the early diagnosis of AMI in patients with pre-existing CAD, presenting with acute chest pain to the ED.

Methods

Study design and population

The Advantageous Predictors of Acute Coronary Syndrome Evaluation (APACE) Study is an ongoing prospective multicentre study designed, coordinated by the University Hospital Basel. From April 2006 to June 2009, a total of 1247 consecutive patients presenting to the ED with chest pain suggestive of AMI with onset or peak within the last 12h were recruited. Patients with end-stage renal failure requiring dialysis were excluded. Pre-existing CAD was defined as history of previous AMI, previous coronary revascularization for obstructive CAD, or known coronary artery stenosis exceeding 50%. For analysis, patients were included if baseline values of all four cTn assays were available.

The study was carried out according to the principles of the Declaration of Helsinki and approved by the local ethics committees. Written informed consent was obtained from all patients. The authors designed the study, gathered and analysed the data, vouch for the data and analysis, wrote the paper, and made the decision to submit it for publication. The assays were donated by the manufacturers, who had no role in the design of the study, data analysis, manuscript, or decision to submit for publication.

Routine clinical assessment

All patients underwent an initial clinical assessment that included history-taking, a physical examination, 12-lead ECG, continuous ECG monitoring, pulse oximetry, standard blood tests, and chest radiography. Cardiac troponin I or cTnT, CK-MB, and myoglobin were measured at presentation and 6–9h after, or as long as clinically indicated. The precise timing of clinical post-baseline measurements and the treatment of patients were left to the discretion of the attending physician.

Adjudicated final diagnosis

To determine the final diagnosis for each patient, two independent cardiologists reviewed all available medical records from the time of the patient's arrival in the ED to the end of the 90-day follow-up period. When there was disagreement about the diagnosis, cases were reviewed and adjudicated in conjunction with a third cardiologist.

An AMI, ST-elevation or Non-ST-elevation myocardial infarction, was defined in accordance with current guidelines.16 In brief, an AMI was diagnosed when there was evidence of myocardial necrosis in association with clinical signs of myocardial ischaemia and/or ECG findings suggestive of myocardial ischaemia. Necrosis was diagnosed by a 30% rising and/or falling pattern of the local cTn level, with at least one value above the 99th percentile, at a level of imprecision of <10% (for detailed information see Supplementary material online, Appendix).6,17 The following cTn assays were used for the adjudication of the final diagnosis at participating hospitals: Abbott-Axsym cTnI ADV, Beckmann Coulter Accu cTnI, and Roche cTnT. All three are well-validated current cTn assays with comparable performance in the diagnosis of AMI.6,17 Unstable angina (UA) was diagnosed when a patient had normal cTn levels and typical angina at rest, a deterioration of a previously stable angina, in cases of positive cardiac exercise testing or cardiac catheterization showing coronary arteries with stenosis of 70% or more of the vessel diameter, or when the diagnosis was uncertain but follow-up information showed that the patient had an AMI or a sudden cardiac death within 60 days after presentation. Further predefined diagnostic categories included cardiac but not coronary symptoms (e.g. tachyarrhythmias), non-cardiac causes, and symptoms of unknown origin. If AMI was ruled out in the ED but no sufficient diagnostic procedures were performed to establish a conclusive diagnosis, symptoms were classified as being of unknown origin.

Cardiac troponin analysis

Blood samples for determination of cTn levels with four cTn assays one high-sensitive cTnT (hs-cTnT) assay: Roche high-sensitive-cTnT;18 two sensitive cTnI assays: Siemens cTnI-Ultra,7,8 Abbott-Architect cTnI;18 and one standard cTnT assay: Roche cTnT,18,19 were collected into tubes containing potassium EDTA or serum within the first hour of the patient's presentation to the ED. Additional samples were collected at 1, 2, 3, and 6h. Serial sampling was discontinued when the diagnosis of AMI was certain and treatment required transferring the patient to the catheter laboratory or coronary care unit. After centrifugation, samples were frozen at −80°C until they were assayed in a blinded fashion in two batches in a dedicated core laboratory. In contrast to the standard assay, the more sensitive cTn assays have a LoD below the 99th percentile of a normal reference population.7,8,18

All Roche assays were performed with the use of the Elecsys 2010 system (Roche Diagnostics): cTnT (fourth generation) with a LoD of 0.01 ng/mL, a 99th percentile cut-off point of <0.01 ng/mL, and a coefficient of variation of <10% at 0.035 ng/mL; and high-sensitive-cTnT with a LoD of 0.003 ng/mL (3 ng/L), a 99th percentile cut-off point of 0.014 ng/mL (14 ng/L), and a coefficient of variation of <10% at 0.013 ng/mL (13 ng/L).20 The Siemens cTnI-Ultra assay was performed with the use of the ADVIA Centaur immunoassay system (Siemens), with a LoD of 0.006 ng/mL (6 ng/L), a 99th percentile cut-off point of 0.04 ng/mL (40 ng/L), and a coefficient of variation of <10% at 0.03 ng/mL (30 ng/L), as specified by the manufacturer.79 The Abbott-Architect cTnI assay was performed with the use of the Architect system (Abbott Diagnostics), with a LoD of 0.01 ng/mL (10 ng/L), a 99th percentile cut-off point of 0.028 ng/mL (28 ng/L), and a coefficient of variation of <10% at 0.032 ng/mL (32 ng/L), as specified by the manufacturer.

Statistical analysis

Continuous variables are presented as means (±SD) or medians (with the inter-quartile range), and categorical variables as numbers and percentages. Continuous variables were compared with the use of the Mann–Whitney test and categorical variables with the use of the Pearson-χ2-square test. Receiver operating characteristic (ROC) curves were constructed to assess the sensitivity and specificity of cTn measurements obtained at specific times with the four assays and to compare their ability to diagnose AMI. Logistic regression was used to combine cTn levels at presentation with early changes in cTn levels. The comparison of areas under the ROC curves (AUC) was performed as recommended by DeLong et al.21 The optimal cut-off values were determined by the minimal distance of the ROC-curve to the point (0;1) of the graph. We used the relevant cross table at this cut-off point to calculate sensitivity and its 95% confidence interval (95% CI), and determined the troponin values around this cut-off, that corresponded to the 95% CI.22 Sensitivities and specificities were compared with a Mc Nemar χ2 test in the case of paired binary outcomes.23 In the case of independent binary outcomes, we used the χ2 test to compare sensitivity, specificity, and positive, and negative predictive values. For the analysis of the prognostic value of the sensitive cTn assays, we did Kaplan–Meier analysis and presented cumulative survival rates at 1 year, subgrouping for pre-existing CAD, diagnosed AMI and elevated sensitive cTn levels above the 99th percentile. We estimated 95% CIs estimated by the standard error. Furthermore, we performed a separate Cox regression analysis for each assay including the cTn elevation above the 99th percentile, pre-existing CAD, age, sex, and cardiovascular risk factors that represented independent predictors for death (arterial hypertension and diabetes) and for AMI during follow-up (arterial hypertension and hypercholesterolaemia) in univariate regression models. All hypothesis testing was two-tailed, and P-values of <0.05 were considered to indicate statistical significance. All statistical analyses were performed with the use of SPSS for Windows, version 15.0 (SPSS), MedCalc software, version 10.3.0 (MedCalc), and the R statistical package (online at http://www.R-project.org).

Results

Characteristics of the patients

Of the 1247 consecutively enrolled patients, measurement of all four cTn assays was obtained at presentation from 1098 patients, of whom 401 (37%) had pre-existing CAD. Patients with pre-existing CAD differed in several baseline characteristics from those without pre-existing CAD (Table 1).

View this table:
Table 1

Baseline characteristics of the patients

 All patients (n = 1098)Patients with a history of CADa (n = 401)Patients without  a history of CAD (n = 697)P-value*Patients with a history of CAD
Acute myocardial infarctionP-value
Yes (n = 77)No (n = 324)
Male gender, no (%)756 (67)307 (77)424 (61)<0.00158 (75)249 (77)0.78
Age, year
 Median647259<0.0017570<0.001
 Inter-quartile range51–7559–7947–7268–8457–78
Risk factors, no (%)
 Hypertension693 (63)320 (82)361 (53)<0.00163 (82)267 (82)0.90
 Hypercholesterolaemia492 (45)282 (71)210 (30)<0.00149 (64)233 (72)0.15
 Diabetes217 (20)126 (31)91 (13)<0.00126 (34)100 (31)0.59
 Current smoking265 (24)74 (19)191 (27)0.00118 (23)56 (17)0.21
 History of smoking391 (36)194 (48)197 (28)<0.00131 (40)163 (50)0.11
History, no (%)
 Previous myocardial infarction271 (25)271 (68)0<0.00154 (70)217 (67)0.60
 Previous revascularization296 (27)295 (74)1 (0)b<0.00149 (64)245 (76)0.03
  Previous PCI254 (23)253 (63)1 (0)b<0.00137 (48)216 (67)0.002
  Previous CABG111 (10)111 (28)0<0.00124 (31)87 (27)0.45
 Peripheral artery disease76 (7)57 (14)19 (3)<0.00115 (20)42 (13)0.14
 Impaired kidney function115 (11)88 (22)27 (4)<0.00123 (30)65 (20)0.06
 Previous stroke64 (6)32 (8)32 (5)0.0211 (14)21 (7)0.02
Vital Status, median (IQR)
 Heart rate, b.p.m.75 (66–89)71 (62–82)78 (68–92)<0.00180 (65–97)69 (61–79)<0.001
 Systolic blood pressure, mmHg142 (127–160)138 (124–157)144 (129–161)<0.001140 (118–162)138 (124–155)0.87
 Diastolic blood pressure, mmHg84 (74–93)80 (70–89)86 (77–95)<0.00178 (67–88)80 (71–89)0.28
 Body mass index26 (24–30)27 (24–30)26 (24–29)0.06226 (24–29)27 (24–30)0.05
Medication
 ACE inhibitors/AT-II blockers536 (49)270 (67)266 (38)<0.00153 (69)203 (63)0.33
 ASA420 (38)297 (74)123 (18)<0.00156 (73)241 (75)0.77
 Beta-blockers411 (37)288 (72)123 (18)<0.00147 (61)241 (74)0.02
 Calcium antagonists184 (17)120 (30)64 (9)<0.00121 (5)99 (30)0.57
 Diuretics299 (27)170 (42)129 (19)<0.00145 (58)125 (38)0.002
 Lipid-lowering drugs393 (36)293 (73)100 (14)<0.00148 (62)245 (76)0.02
 Nitrates/molsidomin132 (12)109 (27)23 (3)<0.00129 (38)80 (25)0.02
ECG
 Potential ischaemic ECG changes247 (23)109 (27)138 (20)0.00441 (53)68 (21)<0.001
  ST-segment elevation58 (5)19 (5)39 (6)0.55612 (16)7 (2)<0.001
  ST-segment depression131 (12)57 (14)74 (11)0.07129 (38)28 (8)<0.001
  Abnormal Q-wave112 (10)75 (19)37 (5)<0.00114 (18)61 (19)0.90
  Left bundle branch block42 (4)25 (6)17 (2)0.00113 (17)12 (4)<0.001
 T-wave inversion147 (13)71 (18)76 (11)0.00124 (31)47 (14)0.001
  • aCAD, coronary artery disease.

  • bPatient with aortal dissection and consecutive coronary dissection but without relevant coronary artery disease.

  • 2 test for comparison of proportions of patients with a history of coronary artery disease and patients without coronary artery disease.

Acute myocardial infarction was the adjudicated final diagnosis in 19% of patients with pre-existing CAD when compared with 14% in patients without pre-existing CAD (P < 0.01). In patients with pre-existing CAD, other adjudicated diagnoses included UA in 27%, cardiac symptoms from causes other than CAD in 10%, non-cardiac causes in 34%, and symptoms of unknown origin in 10% (Table 2).

View this table:
Table 2

Final diagnoses of the patients

All patients (n = 1098)History of CADa (n = 401)No history of CAD (n = 697)P-value*
Acute myocardial infarction173 (16)77 (19)96 (14)0.02
 ST-segment elevation41 (4)15 (4)26 (4)0.99
 Non-ST-segment elevation132 (12)62 (16)70 (10)0.01
UA152 (14)108 (27)44 (6)<0.001
Cardiac cause, but not CAD147 (13)39 (10)108 (16)0.01
Non-cardiac cause528 (48)136 (34)392 (56)<0.001
Unknown98 (9)41 (10)57 (8)0.25
  • aCAD, coronary artery disease.

  • 2 test for comparison of proportions of patients with a history of coronary artery disease and patients without coronary artery disease.

Cardiac troponin levels at presentation

Among the patients, whose final diagnosis was not an AMI, patients with pre-existing CAD had significantly higher baseline levels of all three more sensitive cTn compared with patients without a history of CAD: median levels in CAD patients were 0.014 mg/dL (IQR: 0.009–0.024), with hs-cTnT; 0.01 mg/dL (IQR: 0.004–0.025), with cTnI-Ultra; and 0.003 mg/dL (IQR: 0–0.011), with Abbott-Architect cTnI; compared with 0.005 mg/dL (IQR: 0.003–0.009), with hs-cTnT; 0.004 mg/dL (IQR: 0.001–0.011), with cTnI-Ultra; and 0.000 mg/dL (IQR: 0.000–0.002), with Abbott-Architect cTnI in patients without a history of CAD (P < 0.001 for all comparisons).

Forty per cent of the CAD patients, with a final diagnosis other than AMI, had elevated baseline levels above the 99th percentile with the hs-cTnT, 15% had elevated baseline levels above the 99th percentile with the Siemens cTnI-Ultra, and 13% had elevated baseline levels above the 99th percentile with the Abbott-Architect cTnI assay. Among patients without a history of CAD the percentages were significantly smaller (18%, 9 and 7%; P < 0.001, P = 0.002, and P = 0.004, respectively; see Figure 1). Among all patients with elevated cTn levels above the 99th percentile measured with the hs-cTnT, the Siemens cTnI-Ultra, and the Abbott-Architect cTnI assay 24%, 10 and 9%, had UA while 34%, 9 and 7% had non-cardiac chest pain, respectively (see Table 3).

View this table:
Table 3

Final diagnoses of patients with cardiac troponin levels above the 99th percentile

High-sensitive troponin T (n = 402)Siemens troponin I ultra (n = 257)Abbott-Architect troponin I (n = 229)
Acute myocardial infarction162 (40)155 (60)146 (64)
 ST-segment elevation125 (31)119 (46)114 (50)
 Non-ST-segment elevation37 (9)36 (14)32 (14)
UA58 (14)26 (10)21 (9)
Cardiac cause, but not CAD71 (18)44 (17)39 (17)
Non-cardiac cause81 (20)23 (9)16 (7)
Unknown30 (8)9 (4)7 (3)
  • CAD, coronary artery disease.

Figure 1

Baseline levels of sensitive troponin assays at presentation. Cardiac troponin levels at presentation displayed as multiples of the 99th percentile. Boxes represent inter-quartile ranges, while whiskers display ranges (without outliers further than 1.5 inter-quartile ranges). CAD denotes coronary artery disease. Left side: in patients with final diagnosis of acute myocardial infarction, troponin levels compared within assays were similar in patients with pre-existing coronary artery disease compared with patients without a history of coronary artery disease (all P > 0.05). Right side: in patients with final diagnosis other than acute myocardial infarction, troponin levels compared within assays were significantly higher in patients with pre-existing coronary artery disease (all P< 0.001).

Diagnostic accuracy of cardiac troponin in the early diagnosis of acute myocardial infarction

In patients with pre-existing CAD, the diagnostic accuracy for AMI, quantified by the AUC, was significantly higher with the sensitive cTn assays than that with the standard assay (AUC for Roche hs-cTnT, 0.92; 95% CI: 0.89–0.95; for Siemens cTnI-Ultra, 0.94; 95% CI: 0.91–0.96; and for Abbott-Architect cTnI, 0.93; 95% CI: 0.90–0.95; vs. AUC for the standard assay, 0.87; 95% CI: 0.83–0.90; P = 0.01, P = 0.003, P = 0.007, respectively, for comparisons; Table 4, Supplementary material online, Table S4B and Figure 2A). Overall, the diagnostic accuracy was similar among the three sensitive assays (P > 0.05).

View this table:
Table 4

Diagnostic performance of sensitive troponin assays at the 99th percentile; at 10% coefficient of variation for the standard assay (95% CI)

History of CADNo history of CADP-value*
Sensitive troponin assays
 Roche high-sensitive troponin T 99th percentile [0.014 ng/mL (14 ng/L)]Sensitivity94 (85–98)94 (87–98)0.998
Specificity59 (54–65)81 (78–84)<0.001
Negative predictive value97 (94–99)99 (97–100)0.452
Positive predictive value35 (29–41)45 (39–52)0.137
 Siemens troponin I ultra 99th percentile [0.040 ng/mL (40 ng/L)]Sensitivity91 (82–96)89 (81–94)0.880
Specificity85 (80–88)91 (88–93)0.007
Negative predictive value98 (95–99)98 (97–99)0.883
Positive predictive value58 (49–67)62 (53–69)0.832
 Abbott-Architect troponin I 99th percentile [0.028 ng/mL (28 ng/L)]Sensitivity83 (73–91)85 (77–92)0.918
Specificity87 (83–91)93 (91–95)0.016
Negative predictive value96 (93–98)98 (96–99)0.292
Positive predictive value61 (51–70)66 (57–74)0.719
Standard troponin assay
 Roche troponin T 4th generation 99th percentile (unknown) 10% CV (0.035 ng/mL)Sensitivity69 (57–79)83 (57–77)0.988
Specificity97 (94–99)95 (96–99)0.788
Negative predictive value93 (90–95)95 (93–96)0.417
Positive predictive value84 (73–92)83 (73–90)0.958
  • CAD, coronary artery disease; CV, coefficient of variation.

  • 2 test for comparison of proportions of patients with a history of coronary artery disease and patients without coronary artery disease.

Figure 2

Diagnostic performance of cardiac troponin at presentation. Receiver operating characteristic curves describing the diagnostic performance of different cardiac troponin assays at presentation for the diagnosis of acute myocardial infarction in patients with a history of coronary artery disease (top), and in those without a history of coronary artery disease (bottom). Left side: presenting within 12h from chest pain onset/maximum. Right side: presenting within 3h. (A) and (C) The calculated optimal cut-off values for patients with and without a history of coronary artery disease are defined by the point farthest from the point (0,1) of the graph.

Optimal cut-off for cardiac troponin in the early diagnosis of acute myocardial infarction determined by receiver operating characteristic curve

The optimal cut-off value to separate AMI from non-AMI determined by ROC analysis in CAD patients was more than twice the 99th percentile for hs-cTnT [0.030 ng/mL (30 ng/L)], and close to the 99th percentile for both sensitive cTnI assays [0.034 ng/mL (34 ng/L) for Abbott-Architect cTnI and 0.046 ng/mL (46 ng/L) for Siemens cTnI-Ultra; see Table 5 and Figure 2A]. The optimal cut-off value to separate AMI from non-AMI in patients without a history of CAD was close to the 99th percentile for hs-cTnT [0.020 ng/mL (20 ng/L)] and half the 99th percentile for Abbott-Architect cTnI [0.015 ng/mL (15 ng/L); see Table 5 and Figure 2C].

View this table:
Table 5

Diagnostic performance of sensitive troponin assays in patients with pre-existing coronary artery disease at the optimal cut-off determined by the receiver operating characteristic curve (95% CI)

Optimal cut-off ROCCorresponding cTn valuesaOptimal cut-off ROC99th percentileP-value*
Sensitive troponin assays
Patients with pre-existing coronary artery disease
 Roche high-sensitive troponin T 99th percentile [0.014 ng/mL (14 ng/L)]0.030 ng/mL (30 ng/L)0.025–0.035Sensitivity83 (73–91)94 (85–98)0.008
Specificity86 (82–89)60 (54–65)<0.001
 Siemens troponin I ultra 99th percentile [0.040 ng/mL (40 ng/L)]0.046 ng/mL (46 ng/L)0.034–0.083Sensitivity88 (79–95)91 (82–96)0.500
Specificity89 (85–92)85 (80–88)<0.001
 Abbott-Architect troponin I 99th percentile [0.028 ng/mL (28 ng/L)]0.034 ng/mL (34 ng/L)0.020–0.061Sensitivity83 (73–91)83 (73–91)1
Specificity90 (86–93)87 (83–91)0.008
Sensitive troponin assays
Patients without pre-existing coronary artery disease
 Roche high-sensitive troponin T 99th percentile [0.014 ng/mL (14 ng/L)]0.020 ng/mL (20 ng/L)0.016–0.025Sensitivity91 (83–96)94 (87–98)0.250
Specificity88 (86–91)82 (79–85)<0.001
 Siemens troponin I ultra 99th percentile [0.040 ng/mL (40 ng/L)]0.045 ng/mL (45 ng/L)0.029–0.066Sensitivity89 (80–94)89 (80–94)1
Specificity93 (90–95)91 (89–93)0.016
 Abbott-Architect troponin I 99th percentile [0.028 ng/mL (28 ng/L)]0.015 ng/mL (15 ng/L)0.010–0.024Sensitivity90 (82–95)85 (77–92)0.125
Specificity90 (87–92)93 (91–95)<0.001
  • *Comparisons among patients by McNemar χ2 test.

  • acTn values that correspond to the 95% confidence interval of the sensitivity at the ROC-derived optimal cut-off.

Diagnostic performance in the early diagnosis of acute myocardial infarction at the 99th percentile

Overall, at the 99th percentile, all cTn assays showed lower specificity in patients with pre-existing CAD when compared with patients without a history of CAD (Table 4). The decrease in specificity was particularly pronounced with hs-cTnT (59% in CAD patients vs. 81% in patients without a history of CAD, P < 0.001).

Diagnostic accuracy of cardiac troponin in the diagnosis of ACS (acute myocardial infarction or unstable angina)

The diagnostic accuracy for acute coronary syndroms (ACS), quantified by the AUC, was similarly low with the three sensitive cTn assays in patients with pre-existing CAD, (AUC for Roche hs-cTnT, 0.66; 95% CI: 0.61–0.70; for Siemens cTnI-Ultra, 0.67; 95% CI: 0.62–0.72; and for Abbott-Architect cTnI, 0.67; 95% CI: 0.63–0.72), but moderate to high in patients without a history of CAD (AUC for Roche hs-cTnT 0.89; 95% CI: 0.86–0.91; AUC for Siemens cTnI-Ultra,0.86, 95% CI: 0.83–0.89; AUC for Abbott-Architect cTnI, 0.86, 95% CI: 0.83–0.88; P < 0.001 for all comparisons of AUC in patients with vs. without a history of CAD). For the diagnosis of acute coronary syndromes (AMI or UA), the negative predictive value of a measured cTn-value below the 99th percentile was 64% (57–71%) for the Roche hs-cTnT, 65% (60–71%) for the Siemens cTnI-Ultra, and 64% (58–69%), for the Abbott-Architect cTnI assay. In patients without a history of CAD, the negative predictive value was 93% (91–95%), 92% (89–94%), and 91% (88–93%), respectively.

Cardiac troponin levels in patients with recent onset of chest pain

In patients with pre-existing CAD the superiority of the sensitive cTn assays in the diagnosis of AMI was most pronounced among patients with recent onset of chest pain (Figure 2B, Figure 3, and Supplementary material online, Table S4A).

Figure 3

Diagnostic accuracy at presentation according to chest pain onset. Area under the receiver operating characteristic curves and 95% confidence intervals for the different cardiac troponin assays at presentation in the diagnosis of acute myocardial infarction according to the time from the onset of chest pain. Left: patients with a history of coronary artery disease. Right: patients without a history of coronary artery disease.

Among CAD patients who presented within 3h after the onset of chest pain (n = 167), the AUCs for the four assays were as follows: Roche hs-cTnT, 0.86 (95% CI: 0.80–0.91); Siemens cTnI-Ultra, 0.96 (95% CI: 0.92–0.98); Abbott-Architect cTnI, 0.93 (95% CI: 0.89–0.97); and the standard assay, 0.76 (95% CI: 0.64–0.88) (P < 0.01, P < 0.001, P < 0.001, respectively, for the comparisons of the sensitive assays with the standard assay). The AUCs of the sensitive cTnI assays, Siemens cTnI-Ultra and Abbott-Architect cTnI, were higher than the AUC of the Roche hs-cTnT assay (P = 0.012 and P = 0.051, respectively; Figure 2B and Figure 3).

For patients without a history of CAD, who presented within 3h to the ED, the three sensitive cTn assays had comparable accuracy (AUC for Roche hs-cTnT, 0.93; 95% CI: 0.90–0.96; Siemens cTnI-Ultra, 0.93; 95% CI: 0.90–0.96; Abbott-ArchitectcTnI, 0.91; 95% CI: 0.88–0.94; Figure 2D and Figure 3).

Serial cardiac troponin levels

During serial sampling the AUC for all cTn assays increased (Supplementary material online, Table S4B). Absolute values of changes in high-sensitive cTn levels from presentation to 1 and 2h alone had similar diagnostic accuracy as the baseline high-sensitive cTn levels. The combination of baseline levels plus early changes improved the performance of the baseline level for all cTn assays. With the hs-cTnT assay and the Siemens cTnI-Ultra assay this increase in accuracy was statistically significant for the combination of the baseline level and the change already within the first hour after presentation (P = 0.032 and P = 0.039, respectively); with the Abbott-Architect cTnI assay this increase was only significant for the combination of the baseline level and the change within 2h after presentation (P = 0.02; Supplementary material online, Table S4C).

With the standard assay the diagnostic performance of the combination of baseline levels plus early change at 2h was superior to that of the combination of baseline levels and early change at 1h. The diagnostic performance of the combinations was higher than that of early changes alone, and superior to the single measurement at presentation (all P-values < 0.05; see Supplementary material online, Table S4C).

Prognostic value of sensitive cardiac troponin assays

Median follow-up was 379 days (IQR: 107–721) days. Among the whole cohort, 58 patients died and 53 patients sustained an AMI during follow-up. Cumulative survival rates for patients with pre-existing CAD were 0.88 at 1 year vs. 0.98 in patients without a history of CAD (log-rank test: P < 0.001). In patients with AMI, survival rates at 1 year were 0.82 vs. 0.97 in patients with other diagnoses than AMI (log-rank test: P < 0.001). In patients with elevated levels of Roche hs-cTnT, Siemens cTnI-Ultra, and Abbott-Architect cTnI above the 99th percentile, survival rates were 0.87, 0.83, 0.85 vs. 0.99, 0.98, 0.97 in patients with cTn levels below the 99th percentile (all comparisons by log-rank test <0.001; for details see Supplementary material online, Table S5A and B in the Supplementary material online, Appendix).

Elevated levels of Roche hs-cTnT as well as Siemens cTnI-Ultra above the 99th percentile strongly predicted mortality independent of the presence of pre-existing CAD, age, sex, arterial hypertension, and diabetes in all patients (HR: 2.3, 95% CI: 1.1–5.1, P = 0.034; HR: 2.3, 95% CI: 1.2–4.4, P = 0.009; respectively).

In Cox regression analyses, adjusting for pre-existing CAD, age, sex, arterial hypertension, and hypercholesterolaemia, none of the sensitive cTn assays predicted AMI during follow-up (all P-values not significant).

Discussion

In this prospective multicentre study, we address important issues related to the clinical application of sensitive cTn assays and examined the impact of pre-existing CAD on their diagnostic and prognostic accuracy. We provide seven important findings with impact on their best possible use in the early diagnosis of AMI:.

First, the prevalence of elevated sensitive cTnI and hs-cTnT levels above the 99th percentile in CAD patients with a final diagnosis other than AMI was high and differed largely among the three novel cTn assays, ranging from 13 to 40%. Clinically, the high incidence of elevated cTn levels in CAD patients challenges the application of the 99th percentile for the decision limit for the diagnosis of AMI, as suggested in current guidelines. Careful clinical assessment and thoughtful differential diagnosis are required to separate AMI from a variety of acute and chronic disorders also associated with low-level myocardial necrosis.24 In addition, the difference regarding the incidence of cTn levels above the 99th percentile in CAD patients without AMI might indicate the presence of important differences in the release of cTnI and cTnT in these non-AMI settings, which is further supported by recent data.11,12 An alternative explanation for the difference regarding the incidence of elevated cTn levels is the fact that the 99th percentiles for the three sensitive assays were not determined in the same reference population.

Second, for all three sensitive cTn assays, the diagnostic accuracy at presentation was nevertheless significantly higher than with the standard assay in CAD patients as well as in patients without a history of CAD.

Third, the ROC-derived optimal cut-off levels for CAD patients tended to be higher than in patients without a history of CAD, although the AUC of sensitive cTn assays did not differ significantly comparing CAD and non-CAD patients. All cTn assays showed higher sensitivity but lower specificity in CAD patients when compared with patients without a history of CAD, reflecting the higher baseline levels in CAD patients without AMI. These findings highlight the clinical need to develop test-specific algorithms that fine tune the application of these novel tests in patients with acute chest pain.12

Fourthly, the superiority of the sensitive cTn assays was most pronounced among CAD patients with a recent onset of chest pain, offering the opportunity to minimize myocardial damage by extending early treatment options to AMI patients without ST-segment elevation.1,2

Fifthly, the sensitive cTnI assays seemed to outperform the hs-cTnT assay in early presenters.

Sixthly, the accuracy of sensitive cTn assays to diagnose ACS (AMI or UA) was significantly lower in patients with pre-existing CAD when compared with patients without pre-existing CAD. Further research is necessary to identify biomarkers that reliably detect myocardial ischaemia irrespective of necrosis, particularly in patients with pre-existing CAD.25

Seventhly, the Roche hs-cTnT and the sensitive Siemens cTnI-Ultra assay predict mortality independent of age, sex, pre-existing CAD, and cardiovascular risk factors. Our findings extend the results of previous studies, investigating the mortality of apparently healthy subjects with elevated levels measured with sensitive cTnI assays.11,26,27

The following limitations of the current study merit consideration. First, we evaluated three sensitive cTn assays. We hypothesize that our findings can be generalized to other cTn assays with similar sensitivity and precision. However, additional studies need to confirm this hypothesis. Secondly, in this ongoing prospective study, the subgroup analysis of patients with pre-existing CAD was not predefined at the time of the initial protocol written in 2005. It was added while we were still blinded to the results in 2009, with regard to recent investigations, challenging the diagnostic accuracy of sensitive cTn assays in CAD patients.11,12 Therefore, e.g. our analysis of the assay-specific ROC-derived optimal cut-off values to differentiate AMI from other causes of acute chest pain should be considered exploratory and requires confirmation in additional studies. Third, this observational study cannot quantify exactly the clinical benefit associated with the increase in early diagnostic accuracy. To add this important information, intervention studies seem warranted. Fourth, the first 800 blood samples for the hs-cTnT assay were collected into tubes containing serum, while all other blood samples were collected into tubes containing potassium EDTA, which might lead to slightly different concentration values. Fifthly, some of the patients with positive sensitive cTn values classified as non-AMIs might have had small AMIs below the decision value of conventional cTn. Presumably, this contributed to the reduced specificity of the sensitive assays.

In conclusion, sensitive cTn assays introduce diagnostic improvements as well as challenges. The excellent diagnostic performance of sensitive cTn assays in the early diagnosis of AMI can be extended to patients with pre-existing CAD. However, elevated cTn levels at presentation are common also in CAD patients with diagnoses other than AMI, challenging differential diagnosis. Accordingly, the accuracy to diagnose ACS was lower in patients with pre-existing CAD when compared with patients without pre-existing CAD and optimal cut-off levels tend to be higher. Sensitive cTn assays have prognostic value in patients with a final diagnosis other than AMI.

Funding

The study was supported by research grants from the Swiss National Science Foundation (PP00B-102853), the Swiss Heart Foundation, Abbott, Roche, Siemens, and the Department of Internal Medicine, University Hospital Basel.

Conflict of interest: We disclose that C.M. has received research support from the Swiss National Science Foundation (PP00B-102853), the Swiss Heart Foundation, the Novartis Foundation, the Krokus Foundation, Abbott, Astra Zeneca, Biosite, Brahms, Nanosphere, Roche, Siemens, and the Department of Internal Medicine, University Hospital Basel, as well as speaker honoraria from Abbott, Biosite, Brahms, Roche, and Siemens.

Acknowledgements

We are indebted to the patients who participated in the study and to the ED staff as well as the laboratory technicians for their most valuable efforts, and we thank Kris Denhaerynck and Dr Schindler for expert statistical advice.

Footnotes

  • Both authors have contributed equally and should be considered first author.

References

View Abstract