Analytical and diagnostic performance of troponin assays in patients suspicious for acute coronary...

Analytical and Diagnostic Performance of TroponinAssays in Patients Suspicious for Acute

Coronary Syndromes

CHRISTOPHER HEESCHEN,1,* ARIANE DEU,2 LUKAS LANGENBRINK,2 BRITTA U. GOLDMANN,2

AND CHRISTIAN W. HAMM3

1Stanford University School of Medicine, Division of Cardiovascular Medicine, Stanford, CA, USA2Hamburg University Hospital, Department of Cardiology, Hamburg, Germany

3Kerckhoff Heart Clinic, Department of Cardiology, Bad Nauheim, Germany

Background: The controversy whether there is a clinically signifi-cant difference between troponin T (cTnT) and troponin I (cTnI) inregard to predictive value and cardiac specificity is still ongoing.Methods: We evaluated enzyme-linked immunosorbent assay sys-tems for cTnI and cTnT in patients with acute coronary syndromes andmultiple control groups to define threshold values for risk stratificationand compare their predictive value.Results: In 312 patients with noncardiac chest pain, cTnI levelswere below the detection limit of 0.2 mg/L and cTnT levels were0.011 [0.010–0.013] mg/L. In patients with end-stage renal failure(n 5 26) and acute (n 5 38) or chronic (n 5 16) skeletal muscledamage, median concentrations were 0.20 [0.20–0.35], below thedetection limit, and 0.20 [0.20–0.25] for cTnI, and 0.04 [0.01–0.10],0.011 [0.005–0.025], and 0.032 [0.009–0.054] mg/L for cTnT. Inpatients with acute coronary syndromes (n 5 1130), maximizedprognostic value for 30-day outcome (death, infarction) was ob-served at a threshold level of 1.0 mg/L for cTnI (29.0% positive) andat 0.06 mg/L for cTnT (35.0% positive). Significant differences in thearea-under-the-curve values were observed between cTnI and cTnT(0.685 vs. 0.802; p 5 0.005). For both markers, the area-under-the-curve values did not increase with the second (within 24 h afterenrollment) or third (48 h) blood draw. CTnI showed a less strongassociation with 30-day outcome than cTnT. When cTnI was put ina logistic multiple-regression model first, cTnT did add significantinformation.Conclusion: By using the defined threshold values and the em-ployed test systems, single testing for cTnI and cTnT within 12 hafter symptom onset was appropriate for risk stratification. Despitethe lower cardiac specificity for cTnT, it appears to have a strongerassociation with the patients’ outcome, whereas, as previouslyshown, the ability to identify patients who benefit from treatmentwith a GP IIb/IIIa receptor antagonist is similar. Copyright © 2000The Canadian Society of Clinical Chemists

KEY WORDS: emergency room; unstable angina; acutecoronary syndromes; troponin; prognosis.

Introduction

For patients with unstable coronary artery dis-ease (i.e., unstable angina or non-Q-wave myo-

cardial infarction), the main pathophysiologicalmechanism in the form of plaque rupture or erosionis followed by exposure of thrombogenic contents,such as collagen to the circulation (1,2). The result-ing platelet activation and platelet adhesion pro-motes thrombus friable formation. Pathohistologicalstudies have disclosed focal cell necroses distal inthe myocardium supplied by the culprit artery,which have been attributed to repetitive emboliza-tion from such liable thrombi (3,4). In about onethird of such patients, the resulting minor myocar-dial damage leads to increased values of troponin I(cTnI) and troponin T (cTnT; 5–10). Although crea-tine kinase enzyme activity remains within thenormal range, for such troponin-positive patients, a5–10-fold higher incidence for mortality and nonfa-tal infarction during 30-day follow-up period hasbeen observed (5–11).

The controversy whether there is a clinically sig-nificant difference between cTnI and cTnT in regardto predictive value and cardiac specificity is stillongoing. The main purpose of the present study wasthe definition of threshold values for each of theemployed test systems—the AxSYM for cTnI andthe Elecsys 2010 for cTnT—to maximize and com-pare their prognostic value in patients with acutecoronary syndromes.

Materials and methods

STUDY PATIENTS

Patients with noncardiac chest pain

Patients had chest pain for less than 12 h beforearrival in the emergency but did not provide evi-

Correspondence: Christopher Heeschen, Stanford Uni-versity School of Medicine, Falk Cardiovascular ResearchCenter, 300 Pasteur Drive, Stanford, CA 94305-5406.E-mail: [email protected]

Received April 25, 2000; revised June 13, 2000; acceptedJune 15, 2000

Clinical Biochemistry, Vol. 33, No. 5, 359–368, 2000Copyright © 2000 The Canadian Society of Clinical Chemists

Printed in the USA. All rights reserved0009-9120/00/$–see front matter

PII S0009-9120(00)00144-2

CLINICAL BIOCHEMISTRY, VOLUME 33, JULY 2000 359

dence for coronary artery disease. These patientshad to have no ST-segment changes during serialelectrocardiographic recordings in the emergencyroom and negative treadmill testing, stress echocar-diography, or angiogram (53%). Further, during30-day follow-up, no major cardiac events (death,nonfatal infarction, or urgent revascularization)were recorded. A total of 312 patients were consec-utively enrolled between December 1997 and De-cember 1998 (85 females and 227 males, aged 60.6 611.3 years) in the emergency room of the HamburgUniversity Hospital, Germany. Blood samples werecollected within 15 min after arrival and, addition-ally, 4 h later as a supplement to the routine blooddraw.

Patients with renal failure

Patients with end-stage renal failure (10 femalesand 16 males, median age 45.8 [95% confidenceinterval (CI) 22.8–60.2] years) were enrolled at thetime of haemodialysis. Patients were asymptomaticand had no history of coronary heart disease. Bloodsamples were drawn immediately before and 1 hafter completion of haemodialysis.

Patients with skeletal muscle damage

Patients were enrolled either in the emergencyroom (acute skeletal muscle damage) or in the out-patient’s clinic of the department of cardiology(chronic skeletal muscle damage, i.e., M. Duchenne,Polymyositis). All patients underwent cardiac eval-uation, including treadmill testing and echocardiog-raphy, to exclude cardiac involvement of skeletalmuscle disease.

Patients with acute coronary syndrome

In the Platelet Receptor Inhibition in IschemicSyndrome Management (PRISM) trial, 3232 pa-tients (1034 females and 2198 males, median age61.3 years [95% CI 52.3–67.1]) were recruited from128 centers in 25 countries between March 1994 andOctober 1996 (12). Eligible patients were those whohad their most recent episode of chest pain at rest oraccelerating chest pain within 24 h before random-ization. For all patients, coronary artery disease hadto be manifested by electrocardiographic evidence ofmyocardial ischemia (ST-segment changes and/orT-wave inversions), elevation of CK enzyme activityabove twice the upper range of normal, or a positivehistory of coronary heart disease.

In this double-blinded study, patients were as-signed to treatment with either tirofiban or heparin.Regression analysis including an interaction termfor cTnI or cTnT and randomized treatment indi-cated a significant interaction between troponinlevel and benefit of treatment with tirofiban (p ,0.001). Accordingly, the tirofiban group was ex-cluded from the present analysis. Plasma samples(baseline, within 24 h and 48 h after randomization)

were available for 69% of the patients. This group,thus, concerns 1130 patients receiving standardtherapy including IV heparin and aspirin (12). Inregard to death, nonfatal infarction, and recurrentischemia all patients were followed for 30 days.Myocardial infarction was defined as typical chestpain with new ST-T changes, new pathologic Qwaves, or both, accompanied by an increase inplasma creatine kinase enzyme activity level tomore than twice the upper limit of normal.

TROPONIN MEASUREMENTS

For patients enrolled in the Hamburg UniversityHospital, plasma samples anticoagulated with Na-heparin and serum samples were collected. All sam-ples were instantly centrifuged for the off-site deter-mination of cardiac markers with a maximum delayof 30 min from venous centesis. For patients of thePRISM trial, plasma samples anticoagulated withNa-heparin were collected and centrally stored at280 °C before the analysis during winter 1998/99.Samples were only once thawed for simultaneousdetermination of all three cardiac markers. Thawingafter storage periods of as long as several years hasbeen shown not to result in degradation of thetroponins (11,13). A comparison of analytical resultsobtained from PRISM patients enrolled in 1994 vs.those enrolled during 1996 indicates no differencefor both cTnT (1994 [n 5 648]: 0.21 mg/L [95% CI0.01–8.24]; 1996 (n 5 529): 0.22 mg/L [95% CI0.01–10.51]) and cTnI (1994 (n 5 648): 2.85 mg/L[95% CI 0.3–73.8]; 1996 (n 5 529): 3.02 mg/L [95% CI0.3–81.2]). All measurements for cTnT, cTnI, andCK MB mass were simultaneously performed at thecardiac marker research laboratory of the HamburgUniversity Hospital, blinded to the patients’ histo-ries.

For quantification of cTnI, we used the AxSYMtest system (Abbott Laboratories, Abbott Park, IL,USA) and, in a limited number of control patients,the Stratus CS (Dade-Behring, Deerfield, IL, USA),a system that has been shown to provide highanalytical and diagnostic performance (8). The Ax-SYM random access analyzer utilizes microparticleenzyme immunoassay technology with a two-step,sandwich-type configuration of two antibodies—monoclonal capture anticTnI antibodies and poly-clonal anticTnI antibodies—for detection. The assayis standardized against purified cTnI and has beendeveloped for automated quantification of cTnI inhuman serum and plasma. A large multicenter trialhas established a diagnostic threshold level of 2.0mg/L for the detection of acute myocardial infarction(14). The Stratus CS is a fluorometric enzyme im-munoassay analyzer for quantitative determinationof the cardiac marker CK MB mass, myoglobin, andtroponin I only. In a recent study, a diagnosticthreshold level of 0.1 mg/L for the detection ofmyocardial infarction was established (8).

For quantification of cTnT and CK MB mass, aone-step enzyme immunoassay based on electro-

HEESCHEN ET AL.

360 CLINICAL BIOCHEMISTRY, VOLUME 33, JULY 2000

chemiluminescence technology (Elecsys 2010,Boehringer Mannheim, Indianapolis, IN, USA) wasused (15). The upper reference level for the CK MBkit is 5.0 mg/L. The established diagnostic thresholdfor the cTnT test is 0.10 mg/L (16,17).

STATISTICAL ANALYSIS

All results for continuous variables are expressedas medians with 95% CIs and comparison betweentwo subgroups was performed by the Mann–Whit-ney U-test (two-sided). Comparison of categoricalvariables was generated by the Pearson x2 test.Linear regression analysis was performed on a sin-gle test from each sample over the dynamic range ofthe assays. Logistic multiple regression analysiswas used to construct predictive models for theprimary end points, a composite of death and myo-cardial infarction, and for each component of thecomposite outcome. The following models were con-structed: one each for the cTnT and cTnI variablesalone, one in which cTnI was forced in the modelfirst followed by cTnT, and one in which the cTnTwas forced in the model first followed by cTnI.Another model was constructed to assess the rela-tive values of cTnT and cTnI and the ECG category,before and after adjustment for the other two vari-ables, in the prediction of 30-day outcome. Receiveroperating characteristics (ROC) curves including95% CIs were generated by plotting the relationshipof the true positivity (sensitivity) and false positivity(1, specificity) for the detection of high-risk patientsat various threshold levels of the tests (18). Selectedthreshold values were the values that maximizedthe sum of sensitivity and specificity. Measuredconcentrations of cardiac markers were classifiedpositive in case of results above or equal to thechosen threshold level and negative for all samplesvalues below this value. To assess the diagnosticaccuracy of troponin tests, the areas under the ROCcurves were compared. All analyses were performedwith SPSS 10.0 (SPSS Inc., Chicago, IL, USA) orwith StatXact-3 (Cytel Software Corp., Cambridge,MA, USA) software. p-Values , 0.05 were consid-ered statistically significant.

Results

ANALYTICAL PERFORMANCE

Total imprecision

Using manufacturer-provided controls, total im-precision (CV) over a 6-month period was evaluatedby performing 291 runs. For the AxSYM, CVs were7.4% at 3.1 mg/L, 6.3% at 10.2 mg/L, and 5.1% at 31.2mg/L. For the Elecsys 2010, CVs were 4.8% at 0.35mg/L, 4.3% at 1.25 mg/L, and 4.1% at 31.2 mg/L. Totalimprecision for samples close to the calculatedthreshold values (cTnI 1.0 mg/L; cTnT 0.08 mg/L)was investigated by using internal plasma controlsfrom patients with “minor myocardial damage” after

catheter ablation of ventricular arrhythmia (tro-ponins elevated, CK MB mass within normal range).Over a 30-day period, CVs were 9.5% at 1.2 mg/LcTnI with the AxSYM and 5.2% at 0.15 mg/L withthe Elecsys 2010. Total imprecision for the StratusCS (Dade Behring) was 4.5% at 0.1 mg/L cTnI.

Detection limit

Stepwise dilution tests of such plasma samplesfrom patients with minor myocardial damage re-vealed a minimal detectable troponin concentrationof 0.2 mg/L cTnI for the AxSYM and 0.01 mg/L cTnTfor the Elecsys 2010. The sample with a cTnI con-centration of 0.2 mg/L as analyzed with the AxSYMrevealed a cTnI concentration of 0.02 mg/L as ana-lyzed with the Stratus CS. The detection limit forthe latter assay was 0.01 mg/L.

Limit of quantification

The lowest troponin concentration measured withacceptable precision (CV $ 20%) was 0.02 mg/L cTnTwith the Elecsys 2010 and 0.3 mg/L cTnI with theAxSYM. The cTnI concentration as analyzed withthe Stratus CS for these samples was 0.04 mg/L. Thelimit of quantification for the Stratus CS was 0.03mg/L.

Correlation of plasma and serum samples

Increasing concentrations of cTnI ranging from0.2 to 50.0 mg/L as analyzed with the AxSYM werecompared for plasma and serum samples. Linearregression analysis (including 95% CI for slope andintercept) shows a high correlation between respec-tive samples (Figure 1):

Na-Heparin 5 1.04 (0.97–1.08) * Serum 20.25(20.31–0.13); r2 5 0.92

Results for the Elecsys 2010 and values rangingfrom 0.01 to 5.0 mg/L cTnT were:

Na-Heparin 5 0.99 (0.95–1.04) * Serum 10.18(20.13–0.24); r2 5 0.94

CLINICAL PERFORMANCE

Noncardiac chest-pain patients

Plasma and serum samples were acquired from312 patients with noncardiac chest pain (total of1386 analyzed samples). The median time fromchest-pain onset to the first blood draw was 4.6 h[95% CI 2.5–9.2]. The cTnI concentrations for theAxSYM were below the detection limit of 0.2 mg/L,both for plasma and serum samples (Figure 2).Figures for the Stratus CS were 0.013 [95% CI0.010–0.016] for plasma and 0.012 mg/L [95% CI0.010–0.15] for serum samples. The median concen-trations for cTnT were 0.011 mg/L [95% CI 0.010–0.013] for plasma and 0.012 mg/L [95% CI 0.010–0.014] for serum samples.

TROPONIN ASSAYS IN CORONARY PATIENTS


Patients with renal failure

For patients with end-stage renal failure (n 5 26;baseline creatinine 6.2 [95% CI 3.1–8.5 mg/L],plasma samples were drawn before and after com-pletion of haemodialysis. Before haemodialysis, themedian troponin concentrations were 0.2 mg/L cTnI[95% CI 0.2–0.35] for the AxSYM, 0.01 mg/L cTnI[95% CI 0.01–0.03] for the Stratus CS, and 0.04mg/L cTnI [95% CI 0.01–0.10] for the Elecsys 2010(p 5 0.012 for log-normalized data). For the AxSYM,cTnI was above 1.0 mg/L in two patients (8%) with3.2 and 1.4 mg/L, respectively. For the Stratus CS,no cTnI elevation above the threshold level of 0.1mg/L was observed. For the Elecsys 2010, cTnT wasabove 0.10 mg/L in 8 patients (31%) and above 0.06mg/L in 14 patients (54%; range 0.07–0.66 mg/L).After haemodialysis, cTnI concentrations deter-mined with the AxSYM were below the detectionlimit of 0.2 mg/L for all but one patient that had acTnI concentration of 2.6 mg/L (4%). The mediancTnI concentrations as analyzed with the StratusCS did not change with 0.01 [95% CI 0.01–0.02]mg/L. Median cTnT concentrations were slightlylower but did not decrease on a statistical significantlevel with 0.03 [95% CI 0.01–0.07] (p 5 0.21). CTnT

was above 0.10 mg/L in 6 patients (23%) and above0.06 mg/L in 12 patients (46%; range 0.06–0.49mg/L). Baseline characteristics did not differ be-tween patients with elevated and normal cTnI/cTnTlevels. None of these patients had evidence forcoronary heart disease or suffered from a cardiacevent during 30-day follow-up period.

No cTnI elevation, either with the AxSYM or theStratus CS, was found in patients with compensatedrenal insufficiency (creatinine , 2.0 mg/L)but noevidence for myocardial injury (n 5 32). For theAxSYM, all samples showed a cTnI value below thedetection limit and for the Stratus CS, cTnI concen-tration was 0.01 mg/L [95% CI 0.01–0.02]. One of 32patients had elevated cTnT levels with 0.11 mg/Land median concentration was 0.021 mg/L [95% CI0.010–0.04].

Patients with skeletal muscle damage

For patients with acute (n 5 38) and chronic (n 516) skeletal muscle damage, cTnI values as analyzed

Figure 1 — Correlation for cTnI between values obtainedfrom plasma and serum samples.

Figure 2 — Troponin concentrations in patients withoutevidence for myocardial injury (median, 75%, and 95%CIs).

HEESCHEN ET AL.


with the AxSYM were below the detection limit of0.2 mg/L in most of the patients. CTnI was elevatedin three patients with chronic skeletal muscle dam-age (19%; cTnI range 1.3–4.6 mg/L; CK MB massrange 20.1–168.7 mg/L) and in none of the patientswith acute skeletal muscle damage. CTnI concentra-tions for the Stratus CS were 0.01 mg/L [95% CI0.01–0.015] and 0.02 mg/L [95% CI 0.01–0.04], re-spectively. No cTnI elevation above the thresholdlevel 0.1 mg/L was observed. Respective cTnT valueswere 0.011 mg/L [95% CI 0.005–0.025] for patientswith acute skeletal muscle damage and 0.032 mg/L[95% CI 0.009–0.054] for patients with chronicskeletal damage. CTnT was elevated in five patientswith chronic skeletal muscle damage (31%; cTnTrange 0.19–2.13; CK MB mass range 20.1–361.5mg/L) and in two patients with acute skeletal muscledamage (5%; cTnT range 0.12–0.26; CK MB massrange 24.8–105.9 mg/L).

Patients with acute coronary syndromes

Out of the 1130 patients with acute coronarysyndromes, a total of 15 deaths and 35 nonfatalinfarctions was recorded at 7-day follow-up (totalevent rate 4.4%). At 30 days, the number of deathsand nonfatal infarctions had increased to 38 and 43,respectively (total event rate 7.1%). The event ratefor the combined endpoints death, infarction, orrecurrent ischemia was 10.2% at 7-day and 12.9% at30-day follow-up. The median time from chest-painonset to the first sample collection was 8.4 h [95% CI4.7–12.8].

The troponin concentrations of the study popula-tion ranged from 0.2 to 94.2 mg/L for cTnI and from0.01 to 10.61 mg/L for cTnT. There was only a weakcorrelation between cTnI and cTnT concentrations:

cTnT 5 0.736 (0.688 – 0.783) * cTnI 1 1.115(0.75– 0.155); r2 5 0.45; (Figure 3). For thresholdlevels of 1.0 mg/L for cTnI and 0.06 for cTnT,respectively, 929 (82.2%) had cTnT and cTnI re-sults that were in concordance (both were eitherpositive [n 5 261] or negative [n 5 668]). Of the168 patients with discordant results, 119 (10.5%)were cTnT positive and cTnI negative; the other49 patients (4.3%) showed a negative cTnT result,but were positive for cTnI. The proportion ofdiscordant patients who were cTnT positive andcTnI negative vs. the reverse differed significantly(p 5 0.005). Troponin concentrations above thedetection limit were observed for cTnI (. 0.2 mg/L)in 41.3% and for cTnT (. 0.01 mg/L) in 47.4% ofthe patients. In an analysis that divided the studypopulation with detectable troponin concentra-tions into quartiles, 57% of subjects were classi-fied concordantly and 32% of subjects varied byone quartile; 11% of the patients varied by twoquartiles.

Figure 4 illustrates that the ability of cTnI andcTnT to differentiate high-risk patients from low-risk patients. The threshold values resulting in thehighest x2 values are 1.0 mg/L (95% CI 0.8–1.2) for

cTnI and 0.06 mg/L (95% CI 0.05–0.08) for cTnT(Tables 1 and 2). Of the 1130 patients, cTnI eleva-tion above the defined threshold level of 1.0 mg/Lwas observed for 29.0% of the patients (24.4% at athreshold level of 2.0 mg/L). A cTnT elevation above0.06 mg/L was found in 35.0% of the patients (29.2%for a threshold level of 0.10 mg/L). For the combinedendpoints death or infarction, patients with elevatedcTnI values had an odds ratio of 3.86 (95% CI2.25–3.82), whereas patients with elevated cTnTvalues had an odds ratio of 5.64 (95% CI 4.12–7.65).Figures for the isolated endpoint death were 3.97(95% CI 2.11–5.61) and 7.02 (95% CI 3.31–15.28),respectively. The selected threshold values wereconfirmed by ROC curves analysis. For the com-bined endpoints death or infarction at 30-day follow-up, the area under the ROC curve for cTnI was 0.682(95% CI 0.601–0.759), a figure that was statisticalsignificant different from 0.805 (95% CI 0.726–0.882) for cTnT (p 5 0.005). Consistently, consider-ing each biochemical marker alone with logisticmultiple regression modeling (Table 1), cTnT was

Figure 3 — Correlation between cTnI and cTnT valuesfrom 1030 patients with acute coronary syndromes.



the more powerful predictor of 30-day outcome(death, infarction), as shown by the higher x2 (Ta-bles 1 and 2). After adjusting the model for cTnIfirst, cTnT did add significantly to the model’s abil-ity to predict 30-day outcome (p 5 0.02); cTnI,however, did not add significantly to a model thatfirst incorporated cTnT (p 5 0.35; Table 3). Resultswere similar for the separate endpoints death, in-farction, and recurrent ischemia.

Concerning the second (up to 24 h after enroll-ment; median 13.6 [95% CI 6.8 –21.2) h] and thirdblood draw (up to 48 h), cTnI was elevated in30.8% and 25.4% of the patients, respectively, andcTnT was above the threshold level in 35.5% and32.5% of the patients, respectively. At least onepositive test result was obtained for cTnI in 31.3%and for cTnT in 35.8% of the patients. The accu-racy of troponin measurements for risk stratifica-tion did not increase with extended measure-ments, either for cTnI or for cTnT. For cTnI,area-under-the-curve values were 0.692 (95% CI0.0.596 – 0.772) for the second blood sample and

0.657 (95% CI 0.581– 0.731) when patients with atleast one positive test result were stratified astroponin-positive. For cTnT, the respective area-under-the-curve values were 0.813 (95% CI 0.735–0.892) and 0.802 (95% CI 0.731– 0.879).

Multivariate analysis

In a model that combined the dichotomous vari-ables cTnT and cTnI and the ECG category (Table3), either cTnI or cTnT showed a greater ability topredict 30-day outcome (death, infarction) than didthe ECG (ST-segment changes, T-wave inversion,confounding factors, and negative ECG). After ad-justing for cTnI or cTnT, the ECG category, the CKMB status ($5.0 mg/L), or any other baseline char-acteristic did not significantly contribute to themodel. Either cTnT or cTnI retained a significantability to predict death or infarction during 30-dayfollow-up after adjustment of the model for the ECGcategory.

Figure 4 — Odds ratios for increasing threshold levels of troponin I and troponin T. Results are shown for the combinedendpoints death or infarction (left) and for the separate endpoint death (right) for each threshold value.

HEESCHEN ET AL.


Discussion

ANALYTICAL PERFORMANCE

In this study, the AxSYM cTnI assay providedacceptable analytical characteristics based on im-precision, detection limit, and limit of quantifica-tion. Stepwise dilution tests revealed a minimaldetectable cTnI concentration of 0.2 mg/L. By using

the Stratus CS, the same sample produced a testsignal equal to 0.02 mg/L. This figure is close to thedetermined detection limit of the Stratus CS with0.01 mg/L, a cTnI system that recently has beenshown to provide outstanding analytical and diag-nostic performance (8). The Elecsys 2010 assay forcTnT offered impressive analytical characteristicswith a CV of 5.2% at 0.15 mg/L, a detection limit of0.01 mg/L, and a limit of quantification of 0.023 mg/L.

As many emergency laboratories rely on plasmarather than serum samples to expedite analysis, wecorrelated cTnI and cTnT results for serum andplasma samples. Investigating the clinically mostimportant range from 0.2 up to 50.0 mg/L for cTnIand 0.01 to 5.0 mg/L cTnT, respectively, a highcorrelation was found. In contrast to a recent reportsuggesting that heparin binding reduces the immu-noreactivity of troponins, we did not see a significantreduction in measured troponin concentrations forplasma samples anticoagulated with Na-haparin(19). However, significantly more outliers werefound for the AxSYM when measurements wereobtained from serum samples (Figure 1). This differ-ence was observed despite particular care was takento allow complete clotting before centrifugation ofthe serum samples (20). Meanwhile, the AxSYMcTnI test has been modified to avoid such alterationsof test results by the presence of fibrin. Whether thismodified version is more refractory to incompleteseparation of serum deserves further investigation.

CLINICAL EVALUATION

The clinical evaluation of the troponin assays wasperformed in 1130 patients with acute coronarysyndromes (but no ST-segment elevation on baselineelectrocardiogram) and standard antithrombotictreatment with heparin and aspirin (12). Our inves-tigation demonstrates that a reduced threshold levelof 1.0 mg/L for cTnI and 0.08 mg/L for cTnT remark-ably increases the prognostic value for these twomarkers (Figure 4). CTnI above 1.0 mg/L was foundin 29.0% of the patients and a cTnT elevation above0.06 mg/L was found in 35.0% of the patients. For thecombined endpoints death or infarction, patientswith elevated cTnI values had an odds ratio of 3.86(95% CI 2.25–3.82), whereas patients with elevatedcTnT values had odds ratio of 5.64 (95% CI 4.12–

TABLE 1Relative Values of Troponin I in Predicting 30-Day

Outcome

ThresholdValue

Death, Infarction Death

p x2 p x2

5.0 0.64 0.3 0.54 0.564.0 0.37 0.8 0.39 0.713.0 0.064 3.9 0.065 4.152.5 0.002 11.2 0.005 8.7

2.0 0.0005 16.3 0.002 11.4

1.5 ,0.0001 19.3 0.0006 14.21.4 ,0.0001 19.8 ,0.0001 13.11.3 ,0.0001 19.1 ,0.0001 12.71.2 ,0.0001 21.2 ,0.0001 12.51.1 ,0.0001 20.9 ,0.0001 11.61.0 ,0.0001 35.1 ,0.0001 20.70.9 ,0.0001 30.5 ,0.0001 21.10.8 ,0.0001 26.2 ,0.0001 17.10.7 ,0.0001 18.8 ,0.0001 15.20.6 ,0.0001 17.7 ,0.0001 16.80.5 ,0.0001 16.1 ,0.0001 15.40.4 ,0.0001 16.3 ,0.0001 14.20.3 ,0.0001 17.6 ,0.0001 14.7

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -??????

??????

TABLE 2Relative Values of Troponin T in Predicting 30-Day

Outcome

ThresholdValue

Death, Infarction Death

p x2 p x2

0.5 0.31 1.2 0.08 3.70.4 0.21 1.8 0.07 3.50.3 0.038 5.1 0.053 4.40.2 0.005 8.6 0.016 6.70.15 ,0.0001 17.6 0.002 13.10.14 ,0.0001 16.5 ,0.0001 12.60.13 ,0.0001 18.2 ,0.0001 14.30.12 ,0.0001 27.6 ,0.0001 19.30.11 ,0.0001 30.6 ,0.0001 21

0.10 0.0001 29.2 ,0.0001 21.9

0.09 ,0.0001 29.3 0.0001 18.70.08 ,0.0001 34.4 ,0.0001 170.07 ,0.0001 39.6 ,0.0001 21.40.06 ,0.0001 56.2 ,0.0001 30.20.05 ,0.0001 52.1 ,0.0001 29.40.04 ,0.0001 48.5 ,0.0001 29.10.03 ,0.0001 49.2 ,0.0001 28.30.02 ,0.0001 32.6 ,0.0001 21.0

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -??????

??????

TABLE 3Relative Values of Troponin T in Predicting 30-Day

Outcome

ThresholdValue

Unadjusted Adjusted

Adjusted forp x2 p x2

Troponin T ,0.0001 56.2 ,0.001 23.2 Troponin I,0.001 32.9 ST changes

Troponin I ,0.0001 35.1 0.75 0.11 Troponin T,0.001 28.5 ST changes

ST changes 0.003 15.7 0.08 3.9 Troponin T0.03 6.1 Troponin I



7.65). For a threshold level of 2.0 mg/L for cTnI,however, the odds ratio was lower with 2.52 (95% CI1.65–3.91). A threshold level of 0.10 mg/L for cTnTresulted in an odds ratio of 3.55 (95% CI 2.11–5.29).

The second blood sample obtained within 24 hafter time of enrollment (median 13.6 h [95% CI6.8–21.2]), however, did not significantly increasethe predictive value of the troponins. Given the latetime point of the first blood draw with 8.4 h (95% CI4.7–12.8) after chest-pain onset, the second sampledid not provide a higher rate of positive test resultsnor did the number of patients with at least onepositive test result significantly increase. These re-sults are consistent with our previous findings indi-cating a very high predictive value for blood samplesobtained within 8 h after chest-pain onset (7). Ac-cording to these data, a first sample at admissionand a second sample collected after 6 to 8 h fromsymptom onset provides maximized prognostic in-formation.

Further, multivariate analysis indicates that thepredictive value of cTnT for patients’ 30-day out-come is superior to that of cTnI (as analyzed withthe employed test system), as shown by the higherx2 values. After adjusting the model for cTnI first,cTnT did add significantly to the model’s ability topredict 30-day outcome (p 5 0.02); cTnI, however,did not add significantly to a model that first incor-porated cTnT (p 5 0.35). Results were similar for thepredefined threshold values of 2.0 mg/L for cTnI and0.10 mg/L for cTnT, although on a lower level ofstatistical significance. Further, neither CK MBmass elevation $ 5.0 mg/L nor electrocardiographicfindings as ST-segment changes or T-wave inversionimproved the fit of the regression model when cTnTstatus was put in the model first (Table 3). Incontrast, fit of the model significantly improvedwhen cTnI status was put in the model first. Thesefindings are consistent with the results from thecTnT GUSTO-IIa substudy that also showed ahigher predictive value of the cTnT determinationcompared to cTnI determination (employed systemwas the Stratus II) (11). However, these data cannotbe extrapolated to other cTnI tests because, i.e.,antibody design and enzyme-linked immunosorbentassay technologies, important characteristics for thesystems’ analytical performance, vary. Unfortu-nately, for the PRISM study, sample volume was notsufficient to allow a cTnI determination with asecond test system that might have demonstrateddifferences in predictive value between cTnI testsystems.

At a threshold value of 1.0 mg/L for the AxSYMtest system, patients with acute skeletal muscleinjury and patients with compensated renal insuffi-ciency showed no cTnI elevation and only a minority(,10%) of patients with chronic skeletal muscleinjury and end-stage renal failure had elevated cTnIvalues. In contrast, for a threshold level of 0.06 mg/L,cTnT levels were elevated in more than 50% of thepatients with end-stage renal failure. For the pre-defined threshold value of 0.10 mg/L, still 30% of the

patients were positive for cTnT. For patients withcompensated renal insufficiency, however, cTnT el-evation was rarely found; only one of 32 patients hada cTnT elevation (0.11 mg/L). Ricchiuti et al. haverecently shown that certain cTnT isoforms are ex-pressed in the skeletal muscle of patients withchronic renal disease. However, given the epitopesrecognized by the antibodies utilized in the Elecsys2010 system, this cTnT assay should not detect suchisoforms (21). Thus, the mechanism(s) generatingthe positive signal in these patients remains un-clear. Further, also more than 30% of patients withchronic skeletal muscle disease had elevated cTnTlevels, whereas cTnT elevation in patients withacute skeletal muscle damage was a rare finding(5%). These results are consistent with a recentstudy that demonstrated cTnT mRNA expression indiseased human skeletal muscle (22). The percent-age of patients with chronic skeletal muscle andelevated CTnI levels as analyzed with the AxSYMwas lower compared to cTnT but was still consider-able (19% of the patients), whereas for the StratusCS, no cTnI elevation was observed. The findings forthe AxSYM are inconsistent with the observationthat cTnI mRNA was not expressed in diseasedhuman skeletal muscle and positive signals, thus,might be due to unspecific bindings (22).

In conclusion, the 95% CI for cTnI values in allcontrol cohorts are below the defined thresholdvalue of 1.0 mg/L. In contrast, the 95% CIs for cTnTvalues in patients with chronic skeletal muscle dam-age and renal failure are crossing the thresholdvalue of 0.06 and, to a lower extent, also the thresh-old value of 0.10 mg/L. The 95% CIs for patients withnoncardiac chest pain, acute skeletal muscle dam-age, and compensated renal insufficiency, however,were below the threshold value of 0.06 mg/L. Giventhe low incidence of patients with chronic skeletalmuscle damage and renal failure in patients sus-pected of having acute coronary syndromes and theoutstanding predictive value of the cTnT assay us-ing the lower threshold value of 0.06 mg/L (x2 values29.2 for 0.10 mg/L vs. 56.2 for 0.06 mg/L), one mightwant to accept this loss of cardiac specificity ratherthan missing high-risk patients with acute coronarysyndromes.

DIAGNOSTIC AND THERAPEUTIC IMPLICATIONS OF

TROPONIN MEASUREMENTS

A subgroup analysis of the CAPTURE study indi-cates that the excessive risk of troponin-positivepatients with refractory unstable angina can beabrogated by additional treatment with a glycopro-tein IIb/IIIa receptor antagonist (23). These resultswere recently confirmed by the troponin substudy ofthe PRISM trial, which includes a more representa-tive study population of patients with chest pain. Inthis study, both cTnT and cTnI reliably identifiedhigh-risk patients with acute coronary syndromesthat can be effectively stabilized with the glycopro-tein IIb/IIIa receptor antagonist tirofiban (24). De-

HEESCHEN ET AL.


spite the fact that the predictive value of cTnT, asdemonstrated in this study, appears to be superiorto that of cTnI as analyzed with the AxSYM system,no differences were observed in their ability topredict the benefit of tirofiban. For patients withslightly elevated troponins, cardiac risk markedlyincreased, and this was in particular observed forcTnT. This subgroup of patients with cTnT levelsbetween 0.06 and 0.10 mg/L was at increased cardiacrisk, but did not benefit from treatment with tirofi-ban. One might assume that this was a chancefinding due to the limited number of patients in thissubgroup. Similar results were, however, also re-ported for the CAPTURE troponin substudy withtreatment benefit of abciximab in patients withcTnT levels above 0.10 mg/L only (22). Further, theseretrospective data were recently confirmed in thePARAGON B trial where cTnT-positive patientswere prospectively enrolled (25). Despite the nega-tive results for the PARAGON B trial as a whole,there was a very robust benefit of treatment withlamifiban in patients with cTnT levels above 0.1mg/L (26).

In conclusion, cardiac troponins, either troponin Tor troponin I, are essential markers for the diagnos-tic and therapeutic risk stratification of patientssuspected of having acute coronary syndromes. Forthis clinical challenge, the here-evaluated troponinassays provide convincing analytical sensitivity andspecificity when the defined threshold values areused. The predefined threshold value of 2.0 mg/L forthe AxSYM cTnI test, however, does not appear to besufficient for risk stratification of patients withacute coronary syndromes.

References

1. Davies MJ, Thomas AC. Plaque fissuring—the causeof acute myocardial infarction, sudden ischemia mor-tality, and crescendo angina. Br Heart J 1985; 53:363–73.

2. Farb A, Burke AP, Tang AL, Liang Y, Mannan P,Smialek J, Virmani R. Coronary plaque erosion with-out rupture into a lipid core. Circulation 1996; 93:1354–63.

3. Davies MJ, Thomas AC, Knapman PA, HangartnerJR. Intramyocardial platelet aggregation in patientswith unstable angina suffering sudden ischemic car-diac mortality. Circulation 1986; 73: 418–27.

4. Falk E. Unstable angina with fatal outcome: dynamiccoronary thrombosis leading to infarction and/or sud-den mortality: autopsy evidence of recurrent muralthrombosis with peripheral embolization culminatingin total vascular occlusion. Circulation 1985; 711:699–708.

5. Antman EM, Tanasijevic MJ, Thompson B, et al.Cardiac-specific troponin I levels to predict the risk ofmortality in patients with acute coronary syndromes.N Engl J Med 1996; 335: 1342–9.

6. Benamer H, Steg PG, Benessiano J, et al. Elevatedcardiac troponin I predicts a high-risk angiographicanatomy of the culprit lesion in unstable angina. AmHeart J 1999; 137: 815–20.

7. Hamm CW, Goldmann BU, Heeschen C, Kreyman G,

Berger J, Meinertz T. Emergency room triage ofpatients with acute chest pain by means of rapidtesting for cardiac troponin T or troponin I. N EnglJ Med 1997; 337: 1648–53.

8. Heeschen C, Goldmann BU, Langenbrink L, Mats-chuck G, Hamm CW. Evaluation of a rapid wholeblood ELISA for quantification of Troponin I in pa-tients with acute chest pain. Clin Chem 1999; 45:1789–96.

9. Hetland O, Dickstein K. Cardiac troponins I and T inpatients with suspected acute coronary syndrome: acomparative study in a routine setting. Clin Chem1998; 44: 1430–6.

10. Luescher MS, Thygesen K, Ravkilde J, HeickendorffL, TRIM Study Group. Applicability of cardiac tropo-nin T and I for early risk stratification in unstablecoronary disease. Circulation 1997; 96: 2578–85.

11. Christenson RH, Duh S-H, Newby LK, the GUSTO-IIa Investigators, et al. Cardiac troponin T and car-diac troponin I: relative values in short-term riskstratification of patients with acute coronary syn-dromes. Clin Chem 1998; 44: 494–501.

12. The PRISM Study Investigators. A comparison ofaspirin plus tirofiban with aspirin plus heparin forunstable angina. N Engl J Med 1998; 338: 1498–505.

13. Lev EI, Osende JI, Zaman AG. Troponin T levels andabciximab therapy in unstable angina [Letter].N Engl J Med 1999; 34: 1084–5.

14. Apple FS, Maturen AJ, Mullins RE, et al. Multicenterclincial and analytical evaluation of the AySYM tro-ponin-I immunoassay to assist in the diagnosis ofmyocardial infarction. Clin Chem 1999; 45: 206–12.

15. Forest JC, Masse J, Lane A. Evaluation of the analyt-ical performance of the Boehringer Mannheim Elecsys2010 immunoanalyzer. Clin Biochem 1998; 31: 81–8.

16. Katus HA, Looser S, Hallermayer K, et al. Develop-ment and in vitro characterization of a new immuno-assay of cardiac troponin T. Clin Chem 1992; 38:386–93.

17. Katus HA, Remppis A, Neumann FJ, et al. Diagnosticefficiency of troponin T measurements in acute myo-cardial infarction. Circulation 1991; 83: 902–12.

18. Shultz EK. Multivariate receiver-operating character-istic curve analysis: prostate cancer screening as anexample [Review]. Clin Chem 1995; 41: 1248–55.

19. Gerhardt W, Nordin G, Herbert AK, et al. Troponin Tand I assays show decreased concentrations in hepa-rin plasma compared with serum: Lower recoveries inearly than in late phases of myocardial injury. ClinChem 2000;46: 817–21.

20. Nosanchuk JS. False increase of troponin I attribut-able to incomplete separation of serum [Letter]. ClinChem 1999; 45: 714.

21. Ricchiuti V, Apple FS. RNA expression of cardiactroponin T isoforms in diseased human skeletal mus-cle. Clin Chem 1999; 45: 2129–35.

22. Ricchiuti V, Voss EM, Ney A, Odland M, AndersonPAW, Apple FS. Cardiac troponin T isoforms ex-pressed in renal diseased skeletal muscle will notcause false-positive results by the second generationcardiac troponin T assay by Boehringer Mannheim.Clin Chem 1998; 44: 1919–24.

23. Hamm CW, Heeschen C, Goldmann BU, et al. Tropo-nin T predicts the benefit of Abciximab in patientswith unstable angina in the CAPTURE study. N EnglJ Med 1999; 340: 1623–9.

24. Heeschen C, Hamm CW, Goldmann B, et al. Troponinconcentrations for stratification of patients with acute



coronary syndromes in relation to therapeutic efficacyof tirofiban. Lancet 1999; 354: 1757–62.

25. Moliterno DJ. Patient-specific dosing of IIb/IIIa an-tagonists during acute coronary syndromes: rationaleand design of the PARAGON B study. The PARAGONB International Steering Committee. Am Heart J2000; 139: 563–6.

26. Harrington R. Optimal dosing of a platelet glycopro-tein IIb/IIIa antagonist, Lamifiban, using renal-basedalgorithms, in patients with acute coronary syn-dromes: results from the PARAGON B (the plateletIIb/IIIa antagonist for the reduction of acute coronarysyndrome events in a global organization network).J Am Coll Cardiol 2000; 12–15.

HEESCHEN ET AL.


Analytical and diagnostic performance of troponin assays in patients suspicious for acute coronary...

Documents

Transcript of Analytical and diagnostic performance of troponin assays in patients suspicious for acute coronary...