- Email: [email protected]
Highly sensitive troponin T for risk stratification of acutely destabilized heart failure
Highly sensitive troponin T for risk stratification of acutely destabilized heart failure Domingo A. Pascual-Figal, PhD, a Teresa Casas, PhD, b Jordi Ordonez-LLanos, PhD, c,d Sergio Manzano-Fernández, MD, a Juan C. Bonaque, MD, a Miguel Boronat, MD, b Carmen Muñoz-Esparza, MD, a Mariano Valdés, PhD, a and James L. Januzzi, MD e Murcia, Barcelona, Spain; and Boston, MA
Background A highly sensitive assay for troponin T (hsTnT) has been recently developed, which allows for the detection of even minor myocardial necrosis with high precision. It remains unexplored whether hsTnT provides incremental prognostic accuracy beyond conventional (c)TnT in patients with acutely decompensated heart failure (ADHF). Methods A total of 202 consecutive patients admitted with ADHF and without criteria for acute myocardial infarction were studied. Troponin T was measured using the highly sensitive assay and compared with the conventional method. Patients were clinically followed up at a median of 406 days, with a primary outcome measure of all-cause mortality. Results
The high-sensitive assay detected measurable TnT in 98% of patients vs 56% for cTnT; 81% had an hsTnT above the 99th percentile for a healthy reference population, and it reclassified 60% of those with undetectable cTnT. Both TnT methods predicted the risk of death in adjusted multivariable Cox regression analyses, without a superiority of hsTnT over cTnT in the entire population (area under the curve 0.67 vs 0.71, P = .2). Among patients with a cTnT below 0.03 ng/mL (the lowest cut-point with b10% imprecision; n = 134), solely hsTnT improved the prediction of death over clinical risk factors (relative integrated discrimination improvement +36%, P = .01) and hsTnT above 20 pg/mL identified a significant higher risk of death (hazard ratio 4.7, 95% CI 1.6-13.8, P = .005).
Conclusion Among patients with ADHF, myocardial necrosis (as detected with the hsTnT assay) was nearly ubiquitous. The highly sensitive assay for TnT provides comparable prognostic information to cTnT overall, but among those in whom the cTnT method was less precise or frankly negative, the hsTnT assay provided prognostic information. (Am Heart J 2012;163:1002-10.)
Troponins are well-established tools for the diagnosis and management of acute coronary syndromes (ACS), but their role in heart failure (HF) syndromes remains in evolution. It is well known that troponin concentrations are often elevated in HF regardless of the presence of coronary disease, and when such elevation occurs, a worse outcome is predicted. 1,2 Indeed, myocardial damage—as detected by current troponin assays—has been shown to independently predict short- and longterm prognosis in both chronic 3,4 and acutely decompensated HF (ADHF). 2,5-7
From the aCardiology Department, Virgen de la Arrixaca Hospital and School of Medicine, University of Murcia, Murcia, Spain, bBiochemistry Service, Virgen de la Arrixaca Hospital, Murcia, Spain, cBiochemistry Service, Sant Pau Hospital, Barcelona, Spain, dDepartment of Biochemistry and Molecular Biology, Autonomous University of Barcelona, Barcelona, Spain, and eCardiology Division, Massachusetts General Hospital, Boston, MA. Submitted June 14, 2011; accepted March 12, 2012. Reprint requests: Domingo A. Pascual-Figal, MD, PhD, Cardiology Department, University Hospital Virgen de la Arrixaca, Ctra Madrid-Cartagena s/n, 30120 Murcia, Spain. E-mail: [email protected] 0002-8703/$ - see front matter © 2012, Mosby, Inc. All rights reserved. doi:10.1016/j.ahj.2012.03.015
The exact incidence of troponin detection in HF syndromes depends on the quantitation limit of the assay used for their measurement. Indeed, cardiac troponin concentrations in patients with HF are generally low or modestly increased, and many of the existing troponin methods are relatively inadequate for the detection of such low concentrations with acceptable precision and, thus, may be somewhat limited for risk stratification in HF when such low concentrations of troponin are present. However, newly developed “high sensitivity troponin assays” have considerably lower limits of detection and may, thus, precisely detect values of troponin in a range not possible with conventional assays. Although the value of these high-sensitive assays has been established in ACS, 8 their importance for the evaluation of patients with other cardiac pathologies remains relatively unexplored. A highly sensitive assay for troponin T (hsTnT) has been recently developed and evaluated, 9 although experience with this assay is relatively limited in HF. Using a precommercial version of the assay, Latini et al 10 reported that hsTnT was nearly universally elevated in chronic HF, typically in proportion to the severity of the syndrome; in the Latini analysis, hsTnT provided
American Heart Journal Volume 163, Number 6
incremental prognostic accuracy above and beyond natriuretic peptides. However, it remains unexplored whether hsTnT adds incremental prognostic value in ADHF. Accordingly, the present study aimed to evaluate the value of hsTnT for risk stratification in patients with ADHF, relative to other prognostic measures, including TnT evaluated by the currently existing conventional assay (cTnT).
Methods Study population and design From September 2006 to September 2008, we prospectively enrolled 202 consecutive patients admitted through the emergency department with the diagnosis of ADHF in a single center. The diagnosis of ADHF was established by experienced cardiologists on the basis of the contemporary guidelines 11 and defined as a rapid or gradual onset of signs and symptoms of HF resulting in unplanned hospitalization, including new-onset severe HF or decompensation of chronic HF. In addition, to minimize the potential confounding influence of ACS on the present analysis, patients fulfilling the criteria for acute myocardial infarction 12 were prospectively excluded. The study was approved by the local ethics committee, and informed consent was obtained from each patient at inclusion. All patients were enrolled on arrival to the emergency department, and detailed information about symptoms, clinical history, 12-lead electrocardiogram, and medication use was prospectively collected. An echocardiogram was also performed in all patients during the hospitalization, and standardized measures were obtained according to the American Society of Echocardiography recommendations. 13 Left ventricular ejection fraction (LVEF) was measured by Simpson method; preserved left ventricular systolic function was defined as an LVEF ≥50%. All patients received standard HF management as recommended by contemporary guidelines. 14 Clinical management decisions about each patient were decided by the cardiologist responsible, who was unaware of the patient's hsTnT concentrations but may have been aware of cTnT values, which were measured only for the evaluation of acute myocardial infarction at admission. All patients were clinically followed up at a median of 406 days (interquartile range [IQR] 204-742 days), and the occurrence of clinical events was registered. The primary study outcome was all-cause mortality. Rehospitalization due to HF was also recorded as a secondary study outcome. Death was ascertained from available medical records and death certificates. If hospital records were ambiguous or nonavailable, National Death Records were consulted.
Biochemical analysis Serum samples for biomarker testing (hsTnT and cTnT) were obtained at presentation of patients to the emergency department. Blood was processed immediately, and aliquots of serum were stored at −80 °C until analyzed for hsTnT analyses, whereas cTnT was processed immediately at admission. All aliquots were tested on the first freeze-thaw cycle. Concentrations of hsTnT were measured by an electrochemiluminescence immunoassay using the Elecsys Troponin T HS assay (Roche
Pascual-Figal et al 1003
Diagnostics, Basel, Switzerland) on a Elecsys 2010 analyzer (Roche Diagnostics). The hsTnT assay uses the same capture and detection antibodies as the cTnT; however, to increase analytical sensitivity and precision, an increased sample volume (from 15 to 50 AL) is required, and the signal amplification from the detection antibody is amplified by use of highly optimized signal antibody–Ruthenium conjugates, which are chimeric to avoid heterophilic interferences. The hsTnT assay is conventionally reported in picograms per milliliter and had an analytical range from 3 to 10,000 pg/mL. The 99th percentile for a reference population is 13 pg/mL, and at this value, the coefficient of variation (CV) was 9% (data provided by the manufacturer). The analytical performance of this commercial assay has been recently validated, and it complies with the Global Task Force recommendations for use in the diagnosis of myocardial necrosis. 12 Conventional TnT was measured with the fourth-generation assay from Roche Diagnostics on an Elecsys 2010 analyzer. This assay is reported in nanograms per milliliter, with a value of 0.01 ng/mL representing both the limit of detection and also the 99th percentile for a reference population; the lowest cut-point with CV b10% with this assay is 0.03 ng/mL, which was, therefore, considered the lower reference limit for this analysis. The N-terminal fragment of pro-B-type natriuretic peptide (NT-proBNP) levels were measured by the electrochemiluminescence immunoassay of Roche Diagnostics on an Elecsys 2010 analyzer, with a total CV lower than 3% and an analytical range from 5 to 35,000 pg/mL.
Statistical analysis Normally distributed data are presented as mean ± SD, and nonnormally distributed data as median (IQR). Categorical variables are expressed as percentages, and differences between groups were evaluated using the χ 2 test. All clinical and biochemical characteristics (described in Table I) were evaluated in the univariable Cox regression analysis for studying their association with outcomes, and those that were significant were included in a clinical adjusted multivariable model excluding either hsTnT or cTnT. Log transformation was performed for skewed continuous variables (cTnT, hsTnT, and NT-proBNP). This model was also adjusted by other covariates that are known to be associated with mortality. Thereafter, both hsTnT and cTnT were modeled as continuous and categorical variables and evaluated by adding them separately to the clinical adjusted multivariable model. As a continuous variable, undetectable levels of cTnT were assigned a level of 0.005 ng/mL as suggested. 15 For the categorical analyses, patients with values below the 99th percentile served as the reference group, and the remaining patients were split into approximate thirds. Logcumulative hazard plots, time-dependent covariates, and Schoenfeld residuals were used to evaluate adherence of the Cox proportional hazard assumptions. The prognostic performance of each assay was assessed with receiver operator characteristic (ROC) curves. The optimal cutpoint for each assay was identified in the ROC curve as the value that provided maximal sensitivity and specificity. To better understand the predictive value of both assays, other limits were evaluated: for hsTnT, 3 pg/mL (quantitation limit) and 13 pg/mL (99th percentile), and for cTnT, 0.01 ng/mL (quantitation limit and 99th percentile) and 0.03 ng/mL (lowest limit with CV b10%). For each cut-point, sensitivity and specificity, as well as positive and negative predictive values (all with 95% CIs) for
American Heart Journal June 2012
1004 Pascual-Figal et al
Table I. Baseline characteristics of the patients (n = 202) at admission Age (y) Male Body mass index (kg/m 2) Systolic blood pressure (mm Hg) Heart rate (beats/min) NYHA functional class III/IV Need for inotropic support LVEF (%) LVEF ≥50% Atrial fibrillation/flutter Bundle-branch block Medical history Diabetes mellitus Hypertension Hyperlipidemia Current smoking Prior HF Prior coronary artery disease Prior myocardial infarction Prior stroke Anemia Chronic obstructive pulmonary disease Laboratory Creatinine (mg/dL) Estimated GFR (mL/min per 1.73 m 2) Urea nitrogen (mg/dL) Sodium (mmol/L) Uric acid (mg/dL) Hemoglobin (g/dL) NT-proBNP (ng/L) C-reactive protein (mg/L)
Figure 1
74 (67-80) 55% 28 (25-32) 149 ± 33 95 (79-118) 86 (43%)/116 (57%) 9% 49 (32-60) 51% 54% 36% 55% 83% 47% 13% 59% 34% 28% 11% 48% 22% 1.15 (0.87-1.44) 64 ± 27 25 (18-35) 137 ± 4.5 7.6 ± 2.5 12.7 ± 2.1 3557 (1628-7112) 12 (4-31)
Histogram showing hsTnT concentrations (the dotted line indicates the 99th percentile).
Data are expressed as mean + SD or median (quartiles), and number (%). GFR, glomerular filtration rate.
Table II. Multivariate Cox regression analysis for prediction of death
death, were evaluated and compared using the McNemar test. In an effort to better understand the predictive accuracy and improvement in model performance of adding hsTnT, integrated discrimination improvement (IDI) analyses was performed as described by Pencina et al. 16 C-index and the corresponding differences and P values refer to the technique of bootstrap resampling (20,000 random samples). All P values b.05 were accepted as statistically significant. Statistical analysis was performed using PASW Statistics v. 18.0 for Windows (SPSS, Inc, Chicago, IL). This study was supported by the national network of investigation on HF “REDINSCOR”: Grant RD06/0003/0013, Instituto de Salud Carlos III, Ministry of Health, Madrid, Spain. Reagents for hsTnT were donated by Roche Diagnostics (Mannheim, Germany), which had no other role in this study. We are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the manuscript, and its final contents.
Results Study population Characteristics of the study population at presentation to the emergency department are shown in Table I.
Death χ2 Model without troponins (model NT-proBNP, per 100 pg/mL Creatinine, per mg/dL Age, per year LVEF, per 1% Model 1 including cTnT cTnT, per 0.01 ng/mL Model 1 including hsTnT hsTnT, per 13 pg/mL
1)⁎ 10.2 9.7 10.5 9.1
HR (95% CI)
1.003 2.481 1.079 0.974
P
(1.001-1.005) (1.675-3.674) (1.040-1.119) (0.957-0.991)
.001 b.001 b.001 .003
13.9
1.011 (1.004-1.017)
.001
20.8
1.015 (1.007-1.023)
b.001
⁎ The model includes those variables of Table I showing univariate statistical significance for predicting death. Also adjusted by sex, body mass index, heart rate, systolic blood pressure, NYHA class, sodium, uric acid, hemoglobin, and history of HF and myocardial infarction.
The median age was 74 years, all subjects had New York Heart Association (NYHA) functional class III or IV symptoms, half of the subjects had preserved left ventricular systolic function, and more than half had history of HF. The median length of hospitalization was 8 days (IQR 5-12 days).
American Heart Journal Volume 163, Number 6
Figure 2
Bar chart showing the 1-year mortality rate and the adjusted hazard risk ratios as a function of tertiles above the corresponding 99th percentile, for the hsTnT assay (left panel) and the cTnT assay (right panel). Pascual-Figal et al 1005
American Heart Journal June 2012
1006 Pascual-Figal et al
Table III. Comparison of prognostic performance of cardiac troponin assays for mortality AUC Sensitive assay, hsTnT Limit of detection, 3 pg/mL 99th percentile, 13 pg/mL Optimal cut-point, 20 pg/mL Standard assay, cTnT Limit of detection and 99th percentile, 0.01 ng/mL b10% CV, 0.03 ng/mL Optimal cut-point, 0.012 ng/mL
0.669 (0.600-0.734)
Sensitivity
Specificity
PPV
NPV
98 (90-100)⁎ 94 (84-99)⁎ 89 (77-96)
3 (1-7) 24 (17-32) 40 (32-48)
26 30 34
80 92 91
81 (66-90) 58 (43-71) 81 (68-90)
52 (44-60) † 75 (67-81) † 58 (50-66) †
37 44 40
89 84 90
0.706 (0.638-0.768)
⁎ P b .05 vs cTnT. † P b .001 vs hsTnT.
Outcomes Nine patients died during the index hospitalization; all other patients were clinically followed up at a median of 406 days (IQR 204-742 days). Death occurred in an additional 43 patients at a median of 87 days (IQR 35-231 days). After discharge, rehospitalization due to new ADHF occurred in 71 patients (35%), at a median of 128 days (IQR 58-273 days). Troponin T concentrations Conventional TnT was detectable (N0.01 ng/mL) in 114 (56%) patients (median 0.04 ng/mL, IQR 0.02-0.10 ng/mL); concentrations of cTnT were below 0.03 ng/mL (the lowest concentration providing acceptable precision) in 66%. A highly sensitive assay TnT was detectable, ≥3 pg/mL, in 197 (98%) patients, with a median concentration of 32 pg/mL (IQR 16-57 pg/mL); concentrations of hsTnT were below the lowest cut-point, with b10% imprecision (13 pg/mL) in only 19% (Figure 1). Among the 88 patients with undetectable cTnT concentration, a total of 53 (60%) were reclassified above the 99th percentile by the hsTnT assay. Both assays showed a good correlation in the whole population (r = 0.972). Constrained to those subjects with cTnT values ≥0.03 ng/mL, correlations were comparable (r = 0.969); in contrast, in an elimination analysis, among those with cTnT values b0.03 ng/mL, the correlation significantly worsened (r = 0.695) and remained so even when the analysis was constricted to those with detectable cTnT values (between 0.01 and 0.03 ng/mL, r = 0.501). Prognosis and TnT assays In univariable hazard Cox regression analysis, evaluated as continuous variables, both cTnT concentrations (per 0.01 ng/mL, hazard ratio [HR] 1.012, 95% CI 1.007-1.017, P b .001) and hsTnT concentrations (per 13 pg/mL, HR 1.019, 95% CI 1.012-1.026, P b .001) were associated with an incremental risk of death in the follow-up. Both
remained as independent risk factors when adding to an adjusted multivariable Cox regression model for prediction of mortality (Table II) that also included NTproBNP, creatinine, age, and LVEF. Figure 2 shows the incremental rates of 1-year mortality and adjusted hazard risks as a function of tertiles of concentrations above their corresponding reference groups. Using the cTnT assay, 88 patients were below the 99th percentile and had a mortality of 10%; in contrast, only 39 patients had hsTnT below the corresponding 99th percentile and exhibited a lower mortality of 5%. We also studied separately the association with HF rehospitalization; however, TnT concentrations were not associated with a higher risk for this outcome as measured using either the conventional (per 0.01 ng/mL, HR 1.001, 95% CI 0.988-1.014, P = .90) or the highsensitive assay (per 13 pg/mL, HR 1.008, 95% CI 0.9931.024, P = .30).
Comparison of prognostic performance of assays Table III shows the predictive ability of different limits of both assays for the prediction of death. In the ROC analysis, the optimal cutoff point for the cTnT assay was close to the 99th percentile and detection limit; this means that the optimal cTnT cutoff is acquired with nonacceptable imprecision. The optimal hsTnT cutoff was 20 pg/mL, which is measured with an imprecision largely of only 4% (Jordi Ordonez-LLanos, personal communication). As shown in Table III, the area under the curve (AUC) did not differ (P = .20) and sensitivity did not differ between assays at the optimal cutoff values, but specificity was higher for cTnT than for hsTnT at all values. In Cox proportional hazards, cTnT N0.012 ng/mL (P b .001, HR 4.6, 95% CI 2.3-9.2) and hsTnT N20 pg/mL (P = .001, HR 4.4, 95% CI 1.9-10.3) predicted death. Figure 3 details Kaplan-Meier time-to-event analyses for each assay at their optimal cut-point. The IDI from the addition of hsTnT to cTnT in all subjects was nonsignificant (IDI 0.013, P = .12).
American Heart Journal Volume 163, Number 6
Figure 3
Pascual-Figal et al 1007
Kaplan-Meier survival curves for the hsTnT assay (left panel) and the cTnT assay (right panel) at their corresponding optimal cutoff points.
1008 Pascual-Figal et al
Figure 4
Kaplan-Meier survival curves as a function of elevated hsTnT in those subjects with low concentrations of cTnT, below the value of unacceptable imprecision (N10%) of the assay (b0.03 ng/mL).
Patients with low concentrations of cTnT In an effort to better understand the value of hsTnT relative to cTnT, we then turned our attention to those with very low measured cTnT values within the range of unacceptable precision (N10%) of the assay (b0.03 ng/ mL, n = 134). In this group, a total of 22 (16%) died, and the AUC for the ability of hsTnT to predict death was 0.63, which identified 20 pg/mL as the optimal cutoff point; in Cox proportional hazards, this cutoff predicted death (HR 4.70, 95% CI 1.6-13.8, P = .005). The KaplanMeier survival curves detailing the time-to-event analyses as a function of hsTnT in those subjects with cTnT b0.03 ng/mL are shown in Figure 4. As a supplemental analysis (Table IV), we also evaluated the improvement in model performance of adding cTnT and hsTnT over the final clinical model (creatinine, LVEF, age, and NT-proBNP). In this analysis, both assays added similar predictive information for the entire population, but solely, the high-sensitive assay improved the prediction of death over clinical risk factors for those with cTnT b0.03 ng/mL.
Discussion When using conventional assays, measurable to frankly elevated troponin blood levels have been reported in several cohorts of patients with HF, and the magnitude of such troponin elevation has typically been correlated with the severity of the disease; thus, in general,
American Heart Journal June 2012
troponins above the 99th percentile of a healthy population have been more commonly observed in patients with ADHF and in those with more advanced chronic disease. 10 For example, 10.4% of patients with chronic HF in the Valsartan Heart Failure Trial (Val-HeFT) had elevated cTnT values, whereas 56% of subjects in the ADHF population evaluated in the present study had elevated cTnT concentrations, similar to the report of Latini et al 10 and Sakhuja et al. 17 On the other hand, when newly developed “highly sensitive assays” (ie, assays with improved analytical precision at the detection and quantification limits) are used to evaluate patients with HF, a different picture emerges: although concentrations of hsTnT are still very prognostically meaningful, they are detectable in almost all patients with HF. In our analysis, we found a similar percentage of subjects with detectable hsTnT (98%) as was found by Latini et al 10 in the Val-HEFT ambulatory population (91%). Thus, in context with the results from Val-HeFT, our findings suggest that the hsTnT assay detects myocardial necrosis in almost all patients with HF, regardless of compensation status. However, as expected in a more severe syndrome, concentrations were higher in our ADHF population (median 32 pg/mL) than the Val-HEFT population where more than half of the subjects were below 12 pg/mL. 10 Based on these findings, when using a highly sensitive assay, clinicians should be aware of the extraordinarily high frequency of abnormal troponin values in patients with ADHF, above and beyond the presence of ACS in this population. Although acute myocardial ischemia is an important cause of ADHF and should always be suspected in this setting, when using highly sensitive assays in patients with ADHF, clinicians should not assume that an elevated value is necessarily indicative of an ACS and integrate clinical judgment when encountering such a situation. In the present study, cardiac troponin elevation was an independent and powerfully predictor of death and added prognostic information to other risk factors, consistent with other works. 1-5,18 As expected, we observed that TnT concentrations, measured with the conventional or the highly sensitive assay, had a graded association with death, were predictive of mortality, and retained an independent prognostic value in the multivariable model. In the group as a whole, the 2 assays correlated well, and we did not find any difference between the assays for the prediction of death. This equivalence speaks to the fact that the cTnT and hsTnT assays are based on the same antibodies, and when TnT is measurable, one would not envision a significant difference between the 2 tests. Notably, cTnT showed a good predictive performance for the entire population, whereas the highly sensitive assay showed better performance for those patients with low concentrations of cTnT, both in a range when cTnT was either undetectable or imprecise. Indeed,
American Heart Journal Volume 163, Number 6
Pascual-Figal et al 1009
Table IV. Measures of improvement in model performance of adding cTnT and hsTnT over the clinical risk model for predicting mortality
All patients (n = 202) Clinical risk factors⁎ +cTnT +hsTnT cTnT b0.03 ng/mL (n = 134) Clinical risk factors⁎ +cTnT +hsTnT
C-index (95% CI)
ΔC-index
P
IDI (relative, %)
P
0.754 (0.679-0.830) 0.797 (0.729-0.865) 0.789 (0.720-0.857)
– +0.042 ± 0.023 +0.034 ± 0.019
– .069 .095
– +33 +23
– .001 .018
0.690 (0.570-0.810) 0.732 (0.622-0.843) 0.752 (0.651-0.853)
– +0.043 ± 0.041 +0.052 ± 0.043
– .296 .074
– +13 +36
– .094 .01
C-index and the corresponding differences and P values refer to the technique of bootstrap resampling. ⁎ Clinical risk factors as detailed in Table II.
differences between the cTnT and hsTnT assays emerge when one examines the 2 as a function of concentration. Conventional TnT values between the detection limit value (0.01 ng/mL, 99th percentile) and the lowest cut-point with b10% imprecision (0.03 ng/mL) are prognostically important, however, accompanied by a high degree of imprecision. In fact, when evaluated in the population with detectable values below the 10% CV, correlation between the 2 assays was considerably lower (r = 0.50). Nevertheless, hsTnT was still prognostically meaningful, adding predictive information when the contemporary assay was undetectable or detectable with unacceptable imprecision. This is relevant because such patients could benefit from an assay providing more reliable risk stratification at these low concentrations. We found that an elevated hsTnT is nearly universal in ADHF, where 81% of cases were above the 99th percentile. Mechanistically, outside HF after ACS, the release of troponin in HF is thought to represent a largely noncoronary mechanism due to necrosis, apoptosis, or autophagy of myocardium in response to increased transmural pressure, subendocardial ischemia, and tissue-level oxygen supply-demand inequity. That most analyses find an independent association with risk beyond that associated with ACS supports these mechanisms, as does the extraordinarily frequent observation of elevated troponin in patients with documented nonischemic HF syndromes. Although the mechanism of troponin release remains speculative, the ramifications of this phenomenon are clear: when elevated in ADHF, above and beyond the presence of ACS, hsTnT identifies a clear risk for mortality. Whether this risk is related to a heightened vulnerability for pump failure or arrhythmic death remains unclear and a relevant goal to understand in future analyses. Because retrospective analyses have suggested that the risk associated with an elevated troponin may be heightened through the use of inotropic drugs, 5 whereas βadrenergic blockers may reduce it, 17 the use of troponins to assist in the selection of therapy for
ADHF management is a worthy consideration for future trials of therapeutics in this area. Limitations of our cohort include the fact that it is small and derived from a single tertiary care medical center. Nonetheless, the clinical characteristics of the population are typical of a usual HF population, and the subjects are well characterized. That hsTnT was most noticeably advantageous in patients with cTnT values is not unexpected; however, given our limited power, the prognostic importance of hsTnT in those with undetectable cTnT was marginally nonsignificant. Another limitation is that we analyzed hsTnT and cTnT using 99th percentile values generated from a healthy population; it is likely that the “true” 99th percentile for a hospitalized population of patients such as ours is significantly higher. The use of a higher cut-point might render any difference between hsTnT and cTnT, even smaller is likely. Nonetheless, the cut-points used are recommended by consensus 12 and consistent with the assays currently in clinical use. Indeed, our results provide useful information to the clinicians already using the hsTnT assay. In conclusion, TnT assessed with a highly sensitive assay was prognostically elevated in almost all patients presenting with ADHF at the emergency department. Although comparable with cTnT for prognostication, in those subjects in whom the cTnT was undetectable or detectable with unacceptable imprecision, the hsTnT assay remained prognostic. Further studies are necessary for defining the role of these new assays in stratifying risk in this cardiovascular population.
Disclosures Dr Pascual-Figal, Dr Ordonez-Llanos, and Dr Januzzi report grant support from Roche Diagnostics.
References 1. Kociol RD, Pang PS, Gheorghiade M, et al. Troponin elevation in heart failure prevalence, mechanisms, and clinical implications. J Am Coll Cardiol 2010;56:1071-8.
1010 Pascual-Figal et al
2. Mueller C. Risk stratification in acute decompensated heart failure: the role of cardiac troponin. Nat Clin Pract Cardiovasc Med 2008;5: 680-1. 3. Pascual-Figal DA, Manzano-Fernandez S, Pastor F, et al. Troponin-T monitoring in outpatients with nonischemic heart failure. Rev Esp Cardiol 2008;61:678-86. 4. Setsuta K, Seino Y, Takahashi N, et al. Clinical significance of elevated levels of cardiac troponin T in patients with chronic heart failure. Am J Cardiol 1999;84:608-11. 5. Peacock WF, De MT, Fonarow GC, et al. Cardiac troponin and outcome in acute heart failure. N Engl J Med 2008;358:2117-26. 6. Ishii J, Nomura M, Nakamura Y, et al. Risk stratification using a combination of cardiac troponin T and brain natriuretic peptide in patients hospitalized for worsening chronic heart failure. Am J Cardiol 2002;89:691-5. 7. Perna ER, Macin SM, Cimbaro Canella JP, et al. Minor myocardial damage detected by troponin T is a powerful predictor of long-term prognosis in patients with acute decompensated heart failure. Int J Cardiol 2005;99:253-61. 8. Reichlin T, Hochholzer W, Bassetti S, et al. Early diagnosis of myocardial infarction with sensitive cardiac troponin assays. N Engl J Med 2009;361:858-67. 9. Giannitsis E, Kurz K, Hallermayer K, et al. Analytical validation of a high-sensitivity cardiac troponin T assay. Clin Chem 2010;56: 254-61. 10. Latini R, Masson S, Anand IS, et al. Prognostic value of very low plasma concentrations of troponin T in patients with stable chronic heart failure. Circulation 2007;116:1242-9. 11. Nieminen MS, Bohm M, Cowie MR, et al. Executive summary of the guidelines on the diagnosis and treatment of acute heart failure. Rev Esp Cardiol 2005;58:389-429.
American Heart Journal June 2012
12. Thygesen K, Alpert JS, White HD. Universal definition of myocardial infarction. J Am Coll Cardiol 2007;50:2173-95. 13. Lang RM, Bierig M, Devereux RB, et al. Recommendations for chamber quantification: a report from the American Society of Echocardiography's Guidelines and Standards Committee and the Chamber Quantification Writing Group, developed in conjunction with the European Association of Echocardiography, a branch of the European Society of Cardiology. J Am Soc Echocardiogr 2005;18: 1440-63. 14. Hunt SA, Abraham WT, Chin MH, et al. Yancy CW. 2009 Focused update incorporated into the ACC/AHA 2005 guidelines for the diagnosis and management of heart failure in adults: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines developed in collaboration with the International Society for Heart and Lung Transplantation. J Am Coll Cardiol 2009;53:e1-e90. 15. D'Angelo G, Weissfeld L. An index approach for the Cox model with left censored covariates. Stat Med 2008;27:4502-14. 16. Pencina MJ, D'Agostino Sr RB, D'Agostino Jr RB, et al. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008;27: 157-72. 17. Sakhuja R, Green S, Oestreicher EM, et al. Amino-terminal pro–brain natriuretic peptide, brain natriuretic peptide, and troponin T for prediction of mortality in acute heart failure. Clin Chem 2007;53: 412-20. 18. Bertinchant JP, Combes N, Polge A, et al. Prognostic value of cardiac troponin T in patients with both acute and chronic stable congestive heart failure: comparison with atrial natriuretic peptide, brain natriuretic peptide and plasma norepinephrine. Clin Chim Acta 2005;352:143-53.
Background A highly sensitive assay for troponin T (hsTnT) has been recently developed, which allows for the detection of even minor myocardial necrosis with high precision. It remains unexplored whether hsTnT provides incremental prognostic accuracy beyond conventional (c)TnT in patients with acutely decompensated heart failure (ADHF). Methods A total of 202 consecutive patients admitted with ADHF and without criteria for acute myocardial infarction were studied. Troponin T was measured using the highly sensitive assay and compared with the conventional method. Patients were clinically followed up at a median of 406 days, with a primary outcome measure of all-cause mortality. Results
The high-sensitive assay detected measurable TnT in 98% of patients vs 56% for cTnT; 81% had an hsTnT above the 99th percentile for a healthy reference population, and it reclassified 60% of those with undetectable cTnT. Both TnT methods predicted the risk of death in adjusted multivariable Cox regression analyses, without a superiority of hsTnT over cTnT in the entire population (area under the curve 0.67 vs 0.71, P = .2). Among patients with a cTnT below 0.03 ng/mL (the lowest cut-point with b10% imprecision; n = 134), solely hsTnT improved the prediction of death over clinical risk factors (relative integrated discrimination improvement +36%, P = .01) and hsTnT above 20 pg/mL identified a significant higher risk of death (hazard ratio 4.7, 95% CI 1.6-13.8, P = .005).
Conclusion Among patients with ADHF, myocardial necrosis (as detected with the hsTnT assay) was nearly ubiquitous. The highly sensitive assay for TnT provides comparable prognostic information to cTnT overall, but among those in whom the cTnT method was less precise or frankly negative, the hsTnT assay provided prognostic information. (Am Heart J 2012;163:1002-10.)
Troponins are well-established tools for the diagnosis and management of acute coronary syndromes (ACS), but their role in heart failure (HF) syndromes remains in evolution. It is well known that troponin concentrations are often elevated in HF regardless of the presence of coronary disease, and when such elevation occurs, a worse outcome is predicted. 1,2 Indeed, myocardial damage—as detected by current troponin assays—has been shown to independently predict short- and longterm prognosis in both chronic 3,4 and acutely decompensated HF (ADHF). 2,5-7
From the aCardiology Department, Virgen de la Arrixaca Hospital and School of Medicine, University of Murcia, Murcia, Spain, bBiochemistry Service, Virgen de la Arrixaca Hospital, Murcia, Spain, cBiochemistry Service, Sant Pau Hospital, Barcelona, Spain, dDepartment of Biochemistry and Molecular Biology, Autonomous University of Barcelona, Barcelona, Spain, and eCardiology Division, Massachusetts General Hospital, Boston, MA. Submitted June 14, 2011; accepted March 12, 2012. Reprint requests: Domingo A. Pascual-Figal, MD, PhD, Cardiology Department, University Hospital Virgen de la Arrixaca, Ctra Madrid-Cartagena s/n, 30120 Murcia, Spain. E-mail: [email protected] 0002-8703/$ - see front matter © 2012, Mosby, Inc. All rights reserved. doi:10.1016/j.ahj.2012.03.015
The exact incidence of troponin detection in HF syndromes depends on the quantitation limit of the assay used for their measurement. Indeed, cardiac troponin concentrations in patients with HF are generally low or modestly increased, and many of the existing troponin methods are relatively inadequate for the detection of such low concentrations with acceptable precision and, thus, may be somewhat limited for risk stratification in HF when such low concentrations of troponin are present. However, newly developed “high sensitivity troponin assays” have considerably lower limits of detection and may, thus, precisely detect values of troponin in a range not possible with conventional assays. Although the value of these high-sensitive assays has been established in ACS, 8 their importance for the evaluation of patients with other cardiac pathologies remains relatively unexplored. A highly sensitive assay for troponin T (hsTnT) has been recently developed and evaluated, 9 although experience with this assay is relatively limited in HF. Using a precommercial version of the assay, Latini et al 10 reported that hsTnT was nearly universally elevated in chronic HF, typically in proportion to the severity of the syndrome; in the Latini analysis, hsTnT provided
American Heart Journal Volume 163, Number 6
incremental prognostic accuracy above and beyond natriuretic peptides. However, it remains unexplored whether hsTnT adds incremental prognostic value in ADHF. Accordingly, the present study aimed to evaluate the value of hsTnT for risk stratification in patients with ADHF, relative to other prognostic measures, including TnT evaluated by the currently existing conventional assay (cTnT).
Methods Study population and design From September 2006 to September 2008, we prospectively enrolled 202 consecutive patients admitted through the emergency department with the diagnosis of ADHF in a single center. The diagnosis of ADHF was established by experienced cardiologists on the basis of the contemporary guidelines 11 and defined as a rapid or gradual onset of signs and symptoms of HF resulting in unplanned hospitalization, including new-onset severe HF or decompensation of chronic HF. In addition, to minimize the potential confounding influence of ACS on the present analysis, patients fulfilling the criteria for acute myocardial infarction 12 were prospectively excluded. The study was approved by the local ethics committee, and informed consent was obtained from each patient at inclusion. All patients were enrolled on arrival to the emergency department, and detailed information about symptoms, clinical history, 12-lead electrocardiogram, and medication use was prospectively collected. An echocardiogram was also performed in all patients during the hospitalization, and standardized measures were obtained according to the American Society of Echocardiography recommendations. 13 Left ventricular ejection fraction (LVEF) was measured by Simpson method; preserved left ventricular systolic function was defined as an LVEF ≥50%. All patients received standard HF management as recommended by contemporary guidelines. 14 Clinical management decisions about each patient were decided by the cardiologist responsible, who was unaware of the patient's hsTnT concentrations but may have been aware of cTnT values, which were measured only for the evaluation of acute myocardial infarction at admission. All patients were clinically followed up at a median of 406 days (interquartile range [IQR] 204-742 days), and the occurrence of clinical events was registered. The primary study outcome was all-cause mortality. Rehospitalization due to HF was also recorded as a secondary study outcome. Death was ascertained from available medical records and death certificates. If hospital records were ambiguous or nonavailable, National Death Records were consulted.
Biochemical analysis Serum samples for biomarker testing (hsTnT and cTnT) were obtained at presentation of patients to the emergency department. Blood was processed immediately, and aliquots of serum were stored at −80 °C until analyzed for hsTnT analyses, whereas cTnT was processed immediately at admission. All aliquots were tested on the first freeze-thaw cycle. Concentrations of hsTnT were measured by an electrochemiluminescence immunoassay using the Elecsys Troponin T HS assay (Roche
Pascual-Figal et al 1003
Diagnostics, Basel, Switzerland) on a Elecsys 2010 analyzer (Roche Diagnostics). The hsTnT assay uses the same capture and detection antibodies as the cTnT; however, to increase analytical sensitivity and precision, an increased sample volume (from 15 to 50 AL) is required, and the signal amplification from the detection antibody is amplified by use of highly optimized signal antibody–Ruthenium conjugates, which are chimeric to avoid heterophilic interferences. The hsTnT assay is conventionally reported in picograms per milliliter and had an analytical range from 3 to 10,000 pg/mL. The 99th percentile for a reference population is 13 pg/mL, and at this value, the coefficient of variation (CV) was 9% (data provided by the manufacturer). The analytical performance of this commercial assay has been recently validated, and it complies with the Global Task Force recommendations for use in the diagnosis of myocardial necrosis. 12 Conventional TnT was measured with the fourth-generation assay from Roche Diagnostics on an Elecsys 2010 analyzer. This assay is reported in nanograms per milliliter, with a value of 0.01 ng/mL representing both the limit of detection and also the 99th percentile for a reference population; the lowest cut-point with CV b10% with this assay is 0.03 ng/mL, which was, therefore, considered the lower reference limit for this analysis. The N-terminal fragment of pro-B-type natriuretic peptide (NT-proBNP) levels were measured by the electrochemiluminescence immunoassay of Roche Diagnostics on an Elecsys 2010 analyzer, with a total CV lower than 3% and an analytical range from 5 to 35,000 pg/mL.
Statistical analysis Normally distributed data are presented as mean ± SD, and nonnormally distributed data as median (IQR). Categorical variables are expressed as percentages, and differences between groups were evaluated using the χ 2 test. All clinical and biochemical characteristics (described in Table I) were evaluated in the univariable Cox regression analysis for studying their association with outcomes, and those that were significant were included in a clinical adjusted multivariable model excluding either hsTnT or cTnT. Log transformation was performed for skewed continuous variables (cTnT, hsTnT, and NT-proBNP). This model was also adjusted by other covariates that are known to be associated with mortality. Thereafter, both hsTnT and cTnT were modeled as continuous and categorical variables and evaluated by adding them separately to the clinical adjusted multivariable model. As a continuous variable, undetectable levels of cTnT were assigned a level of 0.005 ng/mL as suggested. 15 For the categorical analyses, patients with values below the 99th percentile served as the reference group, and the remaining patients were split into approximate thirds. Logcumulative hazard plots, time-dependent covariates, and Schoenfeld residuals were used to evaluate adherence of the Cox proportional hazard assumptions. The prognostic performance of each assay was assessed with receiver operator characteristic (ROC) curves. The optimal cutpoint for each assay was identified in the ROC curve as the value that provided maximal sensitivity and specificity. To better understand the predictive value of both assays, other limits were evaluated: for hsTnT, 3 pg/mL (quantitation limit) and 13 pg/mL (99th percentile), and for cTnT, 0.01 ng/mL (quantitation limit and 99th percentile) and 0.03 ng/mL (lowest limit with CV b10%). For each cut-point, sensitivity and specificity, as well as positive and negative predictive values (all with 95% CIs) for
American Heart Journal June 2012
1004 Pascual-Figal et al
Table I. Baseline characteristics of the patients (n = 202) at admission Age (y) Male Body mass index (kg/m 2) Systolic blood pressure (mm Hg) Heart rate (beats/min) NYHA functional class III/IV Need for inotropic support LVEF (%) LVEF ≥50% Atrial fibrillation/flutter Bundle-branch block Medical history Diabetes mellitus Hypertension Hyperlipidemia Current smoking Prior HF Prior coronary artery disease Prior myocardial infarction Prior stroke Anemia Chronic obstructive pulmonary disease Laboratory Creatinine (mg/dL) Estimated GFR (mL/min per 1.73 m 2) Urea nitrogen (mg/dL) Sodium (mmol/L) Uric acid (mg/dL) Hemoglobin (g/dL) NT-proBNP (ng/L) C-reactive protein (mg/L)
Figure 1
74 (67-80) 55% 28 (25-32) 149 ± 33 95 (79-118) 86 (43%)/116 (57%) 9% 49 (32-60) 51% 54% 36% 55% 83% 47% 13% 59% 34% 28% 11% 48% 22% 1.15 (0.87-1.44) 64 ± 27 25 (18-35) 137 ± 4.5 7.6 ± 2.5 12.7 ± 2.1 3557 (1628-7112) 12 (4-31)
Histogram showing hsTnT concentrations (the dotted line indicates the 99th percentile).
Data are expressed as mean + SD or median (quartiles), and number (%). GFR, glomerular filtration rate.
Table II. Multivariate Cox regression analysis for prediction of death
death, were evaluated and compared using the McNemar test. In an effort to better understand the predictive accuracy and improvement in model performance of adding hsTnT, integrated discrimination improvement (IDI) analyses was performed as described by Pencina et al. 16 C-index and the corresponding differences and P values refer to the technique of bootstrap resampling (20,000 random samples). All P values b.05 were accepted as statistically significant. Statistical analysis was performed using PASW Statistics v. 18.0 for Windows (SPSS, Inc, Chicago, IL). This study was supported by the national network of investigation on HF “REDINSCOR”: Grant RD06/0003/0013, Instituto de Salud Carlos III, Ministry of Health, Madrid, Spain. Reagents for hsTnT were donated by Roche Diagnostics (Mannheim, Germany), which had no other role in this study. We are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the manuscript, and its final contents.
Results Study population Characteristics of the study population at presentation to the emergency department are shown in Table I.
Death χ2 Model without troponins (model NT-proBNP, per 100 pg/mL Creatinine, per mg/dL Age, per year LVEF, per 1% Model 1 including cTnT cTnT, per 0.01 ng/mL Model 1 including hsTnT hsTnT, per 13 pg/mL
1)⁎ 10.2 9.7 10.5 9.1
HR (95% CI)
1.003 2.481 1.079 0.974
P
(1.001-1.005) (1.675-3.674) (1.040-1.119) (0.957-0.991)
.001 b.001 b.001 .003
13.9
1.011 (1.004-1.017)
.001
20.8
1.015 (1.007-1.023)
b.001
⁎ The model includes those variables of Table I showing univariate statistical significance for predicting death. Also adjusted by sex, body mass index, heart rate, systolic blood pressure, NYHA class, sodium, uric acid, hemoglobin, and history of HF and myocardial infarction.
The median age was 74 years, all subjects had New York Heart Association (NYHA) functional class III or IV symptoms, half of the subjects had preserved left ventricular systolic function, and more than half had history of HF. The median length of hospitalization was 8 days (IQR 5-12 days).
American Heart Journal Volume 163, Number 6
Figure 2
Bar chart showing the 1-year mortality rate and the adjusted hazard risk ratios as a function of tertiles above the corresponding 99th percentile, for the hsTnT assay (left panel) and the cTnT assay (right panel). Pascual-Figal et al 1005
American Heart Journal June 2012
1006 Pascual-Figal et al
Table III. Comparison of prognostic performance of cardiac troponin assays for mortality AUC Sensitive assay, hsTnT Limit of detection, 3 pg/mL 99th percentile, 13 pg/mL Optimal cut-point, 20 pg/mL Standard assay, cTnT Limit of detection and 99th percentile, 0.01 ng/mL b10% CV, 0.03 ng/mL Optimal cut-point, 0.012 ng/mL
0.669 (0.600-0.734)
Sensitivity
Specificity
PPV
NPV
98 (90-100)⁎ 94 (84-99)⁎ 89 (77-96)
3 (1-7) 24 (17-32) 40 (32-48)
26 30 34
80 92 91
81 (66-90) 58 (43-71) 81 (68-90)
52 (44-60) † 75 (67-81) † 58 (50-66) †
37 44 40
89 84 90
0.706 (0.638-0.768)
⁎ P b .05 vs cTnT. † P b .001 vs hsTnT.
Outcomes Nine patients died during the index hospitalization; all other patients were clinically followed up at a median of 406 days (IQR 204-742 days). Death occurred in an additional 43 patients at a median of 87 days (IQR 35-231 days). After discharge, rehospitalization due to new ADHF occurred in 71 patients (35%), at a median of 128 days (IQR 58-273 days). Troponin T concentrations Conventional TnT was detectable (N0.01 ng/mL) in 114 (56%) patients (median 0.04 ng/mL, IQR 0.02-0.10 ng/mL); concentrations of cTnT were below 0.03 ng/mL (the lowest concentration providing acceptable precision) in 66%. A highly sensitive assay TnT was detectable, ≥3 pg/mL, in 197 (98%) patients, with a median concentration of 32 pg/mL (IQR 16-57 pg/mL); concentrations of hsTnT were below the lowest cut-point, with b10% imprecision (13 pg/mL) in only 19% (Figure 1). Among the 88 patients with undetectable cTnT concentration, a total of 53 (60%) were reclassified above the 99th percentile by the hsTnT assay. Both assays showed a good correlation in the whole population (r = 0.972). Constrained to those subjects with cTnT values ≥0.03 ng/mL, correlations were comparable (r = 0.969); in contrast, in an elimination analysis, among those with cTnT values b0.03 ng/mL, the correlation significantly worsened (r = 0.695) and remained so even when the analysis was constricted to those with detectable cTnT values (between 0.01 and 0.03 ng/mL, r = 0.501). Prognosis and TnT assays In univariable hazard Cox regression analysis, evaluated as continuous variables, both cTnT concentrations (per 0.01 ng/mL, hazard ratio [HR] 1.012, 95% CI 1.007-1.017, P b .001) and hsTnT concentrations (per 13 pg/mL, HR 1.019, 95% CI 1.012-1.026, P b .001) were associated with an incremental risk of death in the follow-up. Both
remained as independent risk factors when adding to an adjusted multivariable Cox regression model for prediction of mortality (Table II) that also included NTproBNP, creatinine, age, and LVEF. Figure 2 shows the incremental rates of 1-year mortality and adjusted hazard risks as a function of tertiles of concentrations above their corresponding reference groups. Using the cTnT assay, 88 patients were below the 99th percentile and had a mortality of 10%; in contrast, only 39 patients had hsTnT below the corresponding 99th percentile and exhibited a lower mortality of 5%. We also studied separately the association with HF rehospitalization; however, TnT concentrations were not associated with a higher risk for this outcome as measured using either the conventional (per 0.01 ng/mL, HR 1.001, 95% CI 0.988-1.014, P = .90) or the highsensitive assay (per 13 pg/mL, HR 1.008, 95% CI 0.9931.024, P = .30).
Comparison of prognostic performance of assays Table III shows the predictive ability of different limits of both assays for the prediction of death. In the ROC analysis, the optimal cutoff point for the cTnT assay was close to the 99th percentile and detection limit; this means that the optimal cTnT cutoff is acquired with nonacceptable imprecision. The optimal hsTnT cutoff was 20 pg/mL, which is measured with an imprecision largely of only 4% (Jordi Ordonez-LLanos, personal communication). As shown in Table III, the area under the curve (AUC) did not differ (P = .20) and sensitivity did not differ between assays at the optimal cutoff values, but specificity was higher for cTnT than for hsTnT at all values. In Cox proportional hazards, cTnT N0.012 ng/mL (P b .001, HR 4.6, 95% CI 2.3-9.2) and hsTnT N20 pg/mL (P = .001, HR 4.4, 95% CI 1.9-10.3) predicted death. Figure 3 details Kaplan-Meier time-to-event analyses for each assay at their optimal cut-point. The IDI from the addition of hsTnT to cTnT in all subjects was nonsignificant (IDI 0.013, P = .12).
American Heart Journal Volume 163, Number 6
Figure 3
Pascual-Figal et al 1007
Kaplan-Meier survival curves for the hsTnT assay (left panel) and the cTnT assay (right panel) at their corresponding optimal cutoff points.
1008 Pascual-Figal et al
Figure 4
Kaplan-Meier survival curves as a function of elevated hsTnT in those subjects with low concentrations of cTnT, below the value of unacceptable imprecision (N10%) of the assay (b0.03 ng/mL).
Patients with low concentrations of cTnT In an effort to better understand the value of hsTnT relative to cTnT, we then turned our attention to those with very low measured cTnT values within the range of unacceptable precision (N10%) of the assay (b0.03 ng/ mL, n = 134). In this group, a total of 22 (16%) died, and the AUC for the ability of hsTnT to predict death was 0.63, which identified 20 pg/mL as the optimal cutoff point; in Cox proportional hazards, this cutoff predicted death (HR 4.70, 95% CI 1.6-13.8, P = .005). The KaplanMeier survival curves detailing the time-to-event analyses as a function of hsTnT in those subjects with cTnT b0.03 ng/mL are shown in Figure 4. As a supplemental analysis (Table IV), we also evaluated the improvement in model performance of adding cTnT and hsTnT over the final clinical model (creatinine, LVEF, age, and NT-proBNP). In this analysis, both assays added similar predictive information for the entire population, but solely, the high-sensitive assay improved the prediction of death over clinical risk factors for those with cTnT b0.03 ng/mL.
Discussion When using conventional assays, measurable to frankly elevated troponin blood levels have been reported in several cohorts of patients with HF, and the magnitude of such troponin elevation has typically been correlated with the severity of the disease; thus, in general,
American Heart Journal June 2012
troponins above the 99th percentile of a healthy population have been more commonly observed in patients with ADHF and in those with more advanced chronic disease. 10 For example, 10.4% of patients with chronic HF in the Valsartan Heart Failure Trial (Val-HeFT) had elevated cTnT values, whereas 56% of subjects in the ADHF population evaluated in the present study had elevated cTnT concentrations, similar to the report of Latini et al 10 and Sakhuja et al. 17 On the other hand, when newly developed “highly sensitive assays” (ie, assays with improved analytical precision at the detection and quantification limits) are used to evaluate patients with HF, a different picture emerges: although concentrations of hsTnT are still very prognostically meaningful, they are detectable in almost all patients with HF. In our analysis, we found a similar percentage of subjects with detectable hsTnT (98%) as was found by Latini et al 10 in the Val-HEFT ambulatory population (91%). Thus, in context with the results from Val-HeFT, our findings suggest that the hsTnT assay detects myocardial necrosis in almost all patients with HF, regardless of compensation status. However, as expected in a more severe syndrome, concentrations were higher in our ADHF population (median 32 pg/mL) than the Val-HEFT population where more than half of the subjects were below 12 pg/mL. 10 Based on these findings, when using a highly sensitive assay, clinicians should be aware of the extraordinarily high frequency of abnormal troponin values in patients with ADHF, above and beyond the presence of ACS in this population. Although acute myocardial ischemia is an important cause of ADHF and should always be suspected in this setting, when using highly sensitive assays in patients with ADHF, clinicians should not assume that an elevated value is necessarily indicative of an ACS and integrate clinical judgment when encountering such a situation. In the present study, cardiac troponin elevation was an independent and powerfully predictor of death and added prognostic information to other risk factors, consistent with other works. 1-5,18 As expected, we observed that TnT concentrations, measured with the conventional or the highly sensitive assay, had a graded association with death, were predictive of mortality, and retained an independent prognostic value in the multivariable model. In the group as a whole, the 2 assays correlated well, and we did not find any difference between the assays for the prediction of death. This equivalence speaks to the fact that the cTnT and hsTnT assays are based on the same antibodies, and when TnT is measurable, one would not envision a significant difference between the 2 tests. Notably, cTnT showed a good predictive performance for the entire population, whereas the highly sensitive assay showed better performance for those patients with low concentrations of cTnT, both in a range when cTnT was either undetectable or imprecise. Indeed,
American Heart Journal Volume 163, Number 6
Pascual-Figal et al 1009
Table IV. Measures of improvement in model performance of adding cTnT and hsTnT over the clinical risk model for predicting mortality
All patients (n = 202) Clinical risk factors⁎ +cTnT +hsTnT cTnT b0.03 ng/mL (n = 134) Clinical risk factors⁎ +cTnT +hsTnT
C-index (95% CI)
ΔC-index
P
IDI (relative, %)
P
0.754 (0.679-0.830) 0.797 (0.729-0.865) 0.789 (0.720-0.857)
– +0.042 ± 0.023 +0.034 ± 0.019
– .069 .095
– +33 +23
– .001 .018
0.690 (0.570-0.810) 0.732 (0.622-0.843) 0.752 (0.651-0.853)
– +0.043 ± 0.041 +0.052 ± 0.043
– .296 .074
– +13 +36
– .094 .01
C-index and the corresponding differences and P values refer to the technique of bootstrap resampling. ⁎ Clinical risk factors as detailed in Table II.
differences between the cTnT and hsTnT assays emerge when one examines the 2 as a function of concentration. Conventional TnT values between the detection limit value (0.01 ng/mL, 99th percentile) and the lowest cut-point with b10% imprecision (0.03 ng/mL) are prognostically important, however, accompanied by a high degree of imprecision. In fact, when evaluated in the population with detectable values below the 10% CV, correlation between the 2 assays was considerably lower (r = 0.50). Nevertheless, hsTnT was still prognostically meaningful, adding predictive information when the contemporary assay was undetectable or detectable with unacceptable imprecision. This is relevant because such patients could benefit from an assay providing more reliable risk stratification at these low concentrations. We found that an elevated hsTnT is nearly universal in ADHF, where 81% of cases were above the 99th percentile. Mechanistically, outside HF after ACS, the release of troponin in HF is thought to represent a largely noncoronary mechanism due to necrosis, apoptosis, or autophagy of myocardium in response to increased transmural pressure, subendocardial ischemia, and tissue-level oxygen supply-demand inequity. That most analyses find an independent association with risk beyond that associated with ACS supports these mechanisms, as does the extraordinarily frequent observation of elevated troponin in patients with documented nonischemic HF syndromes. Although the mechanism of troponin release remains speculative, the ramifications of this phenomenon are clear: when elevated in ADHF, above and beyond the presence of ACS, hsTnT identifies a clear risk for mortality. Whether this risk is related to a heightened vulnerability for pump failure or arrhythmic death remains unclear and a relevant goal to understand in future analyses. Because retrospective analyses have suggested that the risk associated with an elevated troponin may be heightened through the use of inotropic drugs, 5 whereas βadrenergic blockers may reduce it, 17 the use of troponins to assist in the selection of therapy for
ADHF management is a worthy consideration for future trials of therapeutics in this area. Limitations of our cohort include the fact that it is small and derived from a single tertiary care medical center. Nonetheless, the clinical characteristics of the population are typical of a usual HF population, and the subjects are well characterized. That hsTnT was most noticeably advantageous in patients with cTnT values is not unexpected; however, given our limited power, the prognostic importance of hsTnT in those with undetectable cTnT was marginally nonsignificant. Another limitation is that we analyzed hsTnT and cTnT using 99th percentile values generated from a healthy population; it is likely that the “true” 99th percentile for a hospitalized population of patients such as ours is significantly higher. The use of a higher cut-point might render any difference between hsTnT and cTnT, even smaller is likely. Nonetheless, the cut-points used are recommended by consensus 12 and consistent with the assays currently in clinical use. Indeed, our results provide useful information to the clinicians already using the hsTnT assay. In conclusion, TnT assessed with a highly sensitive assay was prognostically elevated in almost all patients presenting with ADHF at the emergency department. Although comparable with cTnT for prognostication, in those subjects in whom the cTnT was undetectable or detectable with unacceptable imprecision, the hsTnT assay remained prognostic. Further studies are necessary for defining the role of these new assays in stratifying risk in this cardiovascular population.
Disclosures Dr Pascual-Figal, Dr Ordonez-Llanos, and Dr Januzzi report grant support from Roche Diagnostics.
References 1. Kociol RD, Pang PS, Gheorghiade M, et al. Troponin elevation in heart failure prevalence, mechanisms, and clinical implications. J Am Coll Cardiol 2010;56:1071-8.
1010 Pascual-Figal et al
2. Mueller C. Risk stratification in acute decompensated heart failure: the role of cardiac troponin. Nat Clin Pract Cardiovasc Med 2008;5: 680-1. 3. Pascual-Figal DA, Manzano-Fernandez S, Pastor F, et al. Troponin-T monitoring in outpatients with nonischemic heart failure. Rev Esp Cardiol 2008;61:678-86. 4. Setsuta K, Seino Y, Takahashi N, et al. Clinical significance of elevated levels of cardiac troponin T in patients with chronic heart failure. Am J Cardiol 1999;84:608-11. 5. Peacock WF, De MT, Fonarow GC, et al. Cardiac troponin and outcome in acute heart failure. N Engl J Med 2008;358:2117-26. 6. Ishii J, Nomura M, Nakamura Y, et al. Risk stratification using a combination of cardiac troponin T and brain natriuretic peptide in patients hospitalized for worsening chronic heart failure. Am J Cardiol 2002;89:691-5. 7. Perna ER, Macin SM, Cimbaro Canella JP, et al. Minor myocardial damage detected by troponin T is a powerful predictor of long-term prognosis in patients with acute decompensated heart failure. Int J Cardiol 2005;99:253-61. 8. Reichlin T, Hochholzer W, Bassetti S, et al. Early diagnosis of myocardial infarction with sensitive cardiac troponin assays. N Engl J Med 2009;361:858-67. 9. Giannitsis E, Kurz K, Hallermayer K, et al. Analytical validation of a high-sensitivity cardiac troponin T assay. Clin Chem 2010;56: 254-61. 10. Latini R, Masson S, Anand IS, et al. Prognostic value of very low plasma concentrations of troponin T in patients with stable chronic heart failure. Circulation 2007;116:1242-9. 11. Nieminen MS, Bohm M, Cowie MR, et al. Executive summary of the guidelines on the diagnosis and treatment of acute heart failure. Rev Esp Cardiol 2005;58:389-429.
American Heart Journal June 2012
12. Thygesen K, Alpert JS, White HD. Universal definition of myocardial infarction. J Am Coll Cardiol 2007;50:2173-95. 13. Lang RM, Bierig M, Devereux RB, et al. Recommendations for chamber quantification: a report from the American Society of Echocardiography's Guidelines and Standards Committee and the Chamber Quantification Writing Group, developed in conjunction with the European Association of Echocardiography, a branch of the European Society of Cardiology. J Am Soc Echocardiogr 2005;18: 1440-63. 14. Hunt SA, Abraham WT, Chin MH, et al. Yancy CW. 2009 Focused update incorporated into the ACC/AHA 2005 guidelines for the diagnosis and management of heart failure in adults: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines developed in collaboration with the International Society for Heart and Lung Transplantation. J Am Coll Cardiol 2009;53:e1-e90. 15. D'Angelo G, Weissfeld L. An index approach for the Cox model with left censored covariates. Stat Med 2008;27:4502-14. 16. Pencina MJ, D'Agostino Sr RB, D'Agostino Jr RB, et al. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008;27: 157-72. 17. Sakhuja R, Green S, Oestreicher EM, et al. Amino-terminal pro–brain natriuretic peptide, brain natriuretic peptide, and troponin T for prediction of mortality in acute heart failure. Clin Chem 2007;53: 412-20. 18. Bertinchant JP, Combes N, Polge A, et al. Prognostic value of cardiac troponin T in patients with both acute and chronic stable congestive heart failure: comparison with atrial natriuretic peptide, brain natriuretic peptide and plasma norepinephrine. Clin Chim Acta 2005;352:143-53.