Diagnostic and Prognostic Value of Cardiac Troponin I Assays in Patients Admitted With Symptoms Suggestive of Acute Coronary Syndrome

Recent consensus guidelines by the European Society of Cardiology (ESC), the American College of Cardiology (ACC), and the American Heart Association (AHA) have redefined the criteria for patients with acute myocardial infarction (MI) as well as the management of patients with chest pain and acute coronary syndrome (ACS), the latter 2 of which are predicated on monitoring cardiac troponin I or T (cTnI or cTnT) in the clinical setting of ischemic symptoms.1–4 Increased serum and plasma cardiac troponins in patients with chest pain now aid in the diagnosis of MI and in the differentiation and guidance of the management of patients with ischemia having unstable angina (UA; normal cardiac troponin) and non–ST-elevation MI (NSTEMI; increased cardiac troponin).1–5 Studies have also shown that increases of either serum or plasma cTnI and cTnT are powerful predictors of both 30-day (short-term) and 6-month to 3-year (long-term) adverse outcomes in patients with ACS, including those with NSTEMI and UA, and in patients with ST-segment–elevation MI.6–8 Further, studies have now demonstrated that increases in both cTnI and cTnT identify a subgroup of patients with ACS who benefit from treatment with dalteparin, enoxaparin, and glycoprotein IIb/IIIa inhibitors.1,2,5,9 

One of the important recommendations that evolved from the biochemistry working group from the Joint ESC/ACC Committee for the redefinition of MI was that each cardiac troponin assay be subjected to the rigors of a clinical and analytic trial to validate its clinical and analytic performance.3,4 The purpose of this study was to compare the analytic and clinical characteristics of the Ortho-Clinical Diagnostics (Rochester, NY) Vitros Troponin I assay (Ortho cTnI) and the Beckman-Coulter (Brea, Calif) Access AccuTnI (Beckman cTnI) assay for the diagnosis of MI and to aid in the risk stratification for patients presenting to a community hospital with myocardial ischemia symptoms suggestive of ACS.

We prospectively studied 200 patients with myocardial ischemia symptoms suggestive of ACS (designated “ACS patients”) consecutively admitted to Hennepin County Medical Center (Minneapolis, Minn) during a 3-month period to rule in or rule out MI. Appropriate institutional review board approval was obtained. Included were patients older than 18 years of any race and of either sex who presented within 12 hours of experiencing symptoms of ischemia (eg, chest pain, chest discomfort), with or without electrocardiographic (ECG) evidence of myocardial ischemia (ST-segment depression, ST-segment elevation, and T-wave inversion). Patients were considered to have UA or NSTEMI on the basis of serial ECG determinations and determinations of cTnI (the Dade-Behring Dimension cTnI assay used in clinical practice at Hennepin).10 An NSTEMI was considered present when cTnI was greater than 0.6 μg/L in the presence of ischemic symptoms either at entry to the study or any time up to 12 to 24 hours thereafter. The severity of MI clinical findings was not factored into either diagnostic or risk assessment decisions.

Plasma (lithium heparin) specimens were obtained in the patients with chest pain at the time of presentation and serially every 3 to 4 hours to 12 hours after enrollment and per physician request up to 24 hours after enrollment. Although 200 subjects were enrolled, because of low specimen volumes, not all patients were analyzed by both assays (Ortho, n = 186; Beckman, n = 188; 180 samples had results for both assays). Biochemical analyses were performed on fresh or refrigerated (4°C) specimens within 48 hours of the drawing of blood by individuals unaware of the clinical status of the patients. The maximum concentration was defined as the highest concentration from any serial specimen per patient. cTnI was measured by the first-generation Ortho cTnI assay, which has a detection limit of 0.038 μg/L and a total imprecision of 10% at 0.12 μg/L.11–14 The 99th percentile of a reference population is 0.08 μg/L. cTnI was also measured by the second-generation Beckman cTnI assay, which has a detection limit of 0.01 μg/L and a total 10% coefficient of variation (CV) at 0.06 μg/L.14 The 99th percentile of a reference population is 0.04 μg/L. Each assay uses a different set of antibodies that recognize different epitopes of the varying cTnI forms known to circulate in the blood following myocardial injury. Both the 99th percentile cutoff value as recommended by the ESC/ACC consensus document for the redefinition of MI3 and the lowest concentration giving a total imprecision of 10% (10% CV)13,14 were determined according to NCCLS protocols when investigations are conducted on duplicate samples that are analyzed for 20 days across 3 calibration lots of reagent. These are the values that were used in this study as cutoff concentrations for risk stratification classification. Only baseline samples were used for risk assessment analysis. Note that the 10% CV concentration is not an evidence-based medical decision type of cutoff but a published value used to address the lowest troponin concentration to allow a 10% CV, because the 99th percentile value does not meet the 10% imprecision suggestion per ESC/ACC guidelines.14 Regression correlations were performed for all 659 specimens analyzed from the 200 patients with chest pain as well as for selected concentration ranges (<10 and <1 μg/L). The receiver operating characteristic (ROC) curves were determined using only the maximum cTnI concentrations for both assays. Areas under the curve were calculated and statistically compared by means of the Mann-Whitney version of the nonparametric 2-sample statistical test.15 

For risk stratification, the clinical endpoint was all-cause death (death) as determined by a chart review during the 180 days following enrollment. The time to event was calculated from the date of admission (only the admission sample was used) to the date of event, date of last contact, or through 180 days, whichever came first. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated. ORs are reported because of the small number of events. Survival curves were generated using the Kaplan-Meier method, and comparisons between risk stratification groups were made using the log-rank statistic. In calculating an OR, a death at day 1 is treated the same as a death at day 180, and an exposure removed or censored at day 1 is treated the same as an exposure removed or censored at day 180. In the current study, the event rates from the Kaplan-Meier curves are higher because the statistic has taken into account that a number of patients have been censored before 180 days. When there are not many patients censored and the event rate is not high, the OR and Kaplan-Meier rates come out about the same. We have added the “number of patients at risk” at 60-day intervals to the Kaplan-Meier survival curves to show the effect of patient removal or censoring. A power analysis was performed to validate the differences detected for all-cause death between a positive and negative cardiac troponin in 200 patients, and we determined that our data were sufficiently powered to achieve greater than 80% power, using a P = .05 significance level. All statistical tests were 2-sided, and statistical significance was accepted at the P = .05 level. Analyses were performed by SPSS (Mac v10) software (SPSS Inc, Chicago, Ill).

The 200 ACS patients had a mean age of 58 years (range, 19–96 years), and 58% were men. Forty-five percent of the patients were white, 35% were black, 13% were American Indian, 3% were Hispanic, 3% were Asian, and 2% were other/unknown. A history of coronary disease was present in 34%, a history of diabetes was present in 30%, and a history of chronic renal disease was present in 5%. Correlation plots between the Beckman (x-axis) and Ortho (y-axis) cTnI assays across different concentration ranges are shown in Figure 1 (all samples [n = 659]—Ortho cTnI = 0.82 Beckman cTnI + 0.03, r = 0.98; cTnI < 10 μg/L [n = 648]—Ortho cTnI = 0.73 Beckman cTnI + 0.2, r = 0.96; cTnI < 1.0 [n = 631]—Ortho cTnI = 0.73 Beckman cTnI + 0.00, r = 0.85). From the 200 ACS patients studied, 10 MIs (5%) were diagnosed. The areas under the ROC curves that were constructed with only the maximum cTnI concentrations for both assays are shown in Figure 2. No significant differences were found (Ortho, .991; Beckman, .995). At the 99th percentile cutoff, sensitivity and specificity calculations were as follows: Ortho, 100% and 80% and Beckman, 100% and 78%, respectively. At the 10% CV cutoff, sensitivity and specificity calculations were as follows: Ortho, 100% and 82% and Beckman, 100% and 82%, respectively.

Kaplan-Meier cumulative survival curves were constructed for both Ortho and Beckman cTnI assays using both the 99th percentile cutoffs (0.08 and 0.04 μg/L, respectively) and the 10% CV cutoffs (0.12 and 0.06 μg/L, respectively) as shown in Figure 3, A and B. For the Ortho cTnI assay, significant differences were found at 60 days follow-up and through 180 days of follow-up for death when comparisons were made between a normal and an increased cTnI at the 99th percentile (Figure 3, A; P = .03) and at the 10% CV cutoffs (Figure 3, B; P < .001). The 10% CV cutoff demonstrated a greater percentage of patients with poor outcomes (15.8% vs 10.3%). Similar findings were obtained for the Beckman assay at the 99th percentile (Figure 3, A; P < .001) and the 10% CV cutoffs (Figure 3, B; P < .001), with a greater percentage of patients with poor outcomes when using the 10% CV (26%) versus the 99th percentile (16%) cutoff. Mortality rates were significantly greater using either cutoff for an increased versus a normal cardiac troponin, as shown in the Table. The ORs of death with increased cTnI (P ≤ .04) were as follows: for Ortho, the 99th percentile cutoff was 5.9 (95% CI, 1.1–30.9), and the 10% CV cutoff was 10.3 (95% CI, 1.9–55); for Beckman, the 99th percentile cutoff was 31.4 (95% CI, 3.5–280), and the 10% CV cutoff was 15.3 (95% CI, 2.6–88).

For the Ortho and Beckman cTnI assays, we provide evidence of similar findings in ACS patients for MI diagnostics and risk stratifications with regard to death when using the 99th percentile as well as the 10% CV cutoff concentrations. Both the lower 99th percentile and the higher 10% CV cutoffs showed a significant risk of death during the 180-day follow-up in ACS patients when comparisons were made between a normal and an increased troponin value. These data support the previous studies using both similar and different cTnI assays that have demonstrated that increased baseline troponin concentrations show strong prognostic value for risk stratification in patients with chest pain or ACS.6–9,16–18 The current findings substantiate that increased events occur even at troponin concentrations above the 99th percentile reference limit.16–19 Patients in the current study with an increased plasma cTnI had a 3- to 31-fold higher risk of death than patients with a normal value. Thus, both assays provide the same potential benefits of risk assessment in patients with chest pain that will assist in identifying a subset of patients for aggressive medical therapy.1,2,5 Trials have shown that ACS patients with an increased cardiac troponin benefit more from low-molecular-weight heparin and platelet glycoprotein IIb/IIIa receptor inhibitor therapies than do patients with a normal cardiac troponin.1,2,5,9 In the current study, ACS patients with normal cardiac troponin values still have an overall mortality of 2%. These findings point to the importance of using the clinical history and ECG of a patient, as well as the cardiac troponin findings, because this type of information will assist in identifying even low-risk patients.20 Regression analysis of the 2 cTnI assays, Figure 1, demonstrated only minor assay concentration differences across different concentration ranges. Any differences would likely reflect differences in antibody epitope recognition patterns as well as differences in calibration that are known to exist between different cTnI assays.14 

The recent ESC/ACC consensus document for the redefinition of MI recommends the use of a 99th percentile cutoff value for cTnI and cTnT assays as the discriminating value for the detection of myocardial injury.3,4 Using the 99th percentile cutoffs, both assays demonstrated 100% sensitivity and similar specificities for ruling in and ruling out MI. However, because no manufacturer of cardiac troponin assays is able to achieve a 10% CV at the 99th percentile as recommended by the ESC/ACC document, it has been proposed that the lowest concentration permitting a 10% CV (total imprecision) be used as an alternative cutoff until manufacturers improve low-end assay imprecisions.13,14 This provides our rationale for presenting data for both the 99th percentile and the 10% CV cutoffs for risk stratification in this study. Our findings are supported by recent substudies of the FRISC II trial and the TACTICS TIMI trial as strong predictors of adverse cardiac events. First, Morrow et al9 (TACTICS) demonstrated that minor elevations for cTnI equal to 0.1 to 0.4 μg/L or greater (using the Bayer ACS 180 assay) offer important prognostic information in ACS patients. Second, Lindahl et al19 (FRISC II) showed that any detectable elevation of cTnT (>0.01 μg/L; Roche Elecsys assay) was associated with an increased risk of reinfarction and death. Third, Venge et al18 (FRISC II) demonstrated that ACS patients with nonelevated concentrations had a significantly better prognosis, as measured by the Abbott AxSYM cTnI (>0.6 μg/L), the Beckman Access (second generation, >0.02 μg/L), and the Roche cTnT (>0.01 μg/L), than did patients with increased concentrations. Further, they identified a 10% to 12% cohort of patients with a poor prognosis that was found only at low cTnI concentrations (between 0.03 and 0.06 μg/L) as measured by the Access cTnI assay.

Several limitations of this study need to be noted regarding our evidence-based findings. First, the limited number of ACS patients and endpoints in our study did not allow a stratification or an adjustment for other known risk factors, such as ECG findings, renal function, and diabetes.21 Further, because the numbers of subjects and endpoints in the intermediate group between the 99th percentile and the 10% CV values were small, we were unable to perform statistical analyses to determine differences between assays. Second, our study did not address the possible serum/heparin plasma biases that may exist for troponin assays.22 However, as heparinized plasma is widely used internationally and is the recommended specimen for both the Ortho and Beckman assays, our findings substantiate risk stratification claims for heparin plasma for both methods.6–8 Third, although we realize that this study was performed on a small sample size, our findings parallel observations made in studies with larger cohorts.6,7 As noted in “Materials and Methods,” the Kaplan-Meier survival analysis takes into account the removal or censoring of patients who were lost to follow-up before the entire 180-day follow-up period. Therefore, the inverse of the survival rate shown in the Kaplan-Meier curves would be higher than the simple mortality rate calculated in the chi-square calculation in the OR values in the Table. However, the Kaplan-Meier analysis showed significant differences at the 60-day follow-up as well as at 180 days, when only about half of the initial subjects at risk were still available for follow-up. Again, our findings parallel observations made in studies with larger cohorts.6,7 However, we cannot exclude the possibility of patient selection bias at the 180-day follow-up. Fourth, instead of using maximum concentrations, only baseline admission specimens were used for all patients (n = 200) for both the ROC curve and the risk outcome analyses, which is typical in most study designs.6,7 Previous findings have shown that added statistical significance for predicting outcomes can be gained by monitoring maximum troponin concentrations obtained during the 24 hours following presentation.23 Analysis of diagnostic sensitivities and specificities performed on specimens that were available for cardiac troponin analyses from the same subjects (n = 180) for both assays showed no differences compared to the ROC curves shown in Figure 2 (data not shown).

Overall, our findings demonstrate the importance of the independent validation of individual cardiac troponin assays prior to use by clinicians as diagnostic risk stratifications and as medical and therapeutic management tools. The decision thresholds we have used and recommend for clinical practice are consistent with the proposed ESC/ACC redefinition of the MI consensus document, the ACC/AHA guideline updates for the management of patients with UA and NSTEMI, and the suggested use of the lowest cardiac troponin concentration, which shows a 10% imprecision when not met at the 99th percentile reference limit for the detection of myocardial injury. Our findings support and add to the previous data that cardiac troponin increases even below the level of a 10% CV but increased above the 99th percentile reference limit may offer important risk outcome assessment in ACS patients. For clinical application, we support the use of the 10% CV concentration cutoff value for both diagnostic and risk assessment use to avoid the potential of false-positive results above the 99th percentiles due to imprecision issues.

This study was supported in part by Ortho-Clinical Diagnostics, Rochester, NY. The design and data analyses of the study were solely conducted by the authors.