Acute Chest Pain and Three Cardiac Troponin Assays
Acute Chest Pain and Three Cardiac Troponin Assays
In this prospective, observational, international, multicentre trial of 1117 unselected patients presenting with acute chest pain to the ED, we scrutinized the prognostic value of presentation values and serial measurements of three novel hs-cTn assays and compared them with a conventional cTn assay. We report five major findings.
First, hs-cTnT at presentation outperformed both hs-cTnI assays and conventional cTnT regarding the accuracy to predict 2-year mortality. This finding extends previous analyses that found similar high diagnostic accuracy for the hs-cTn assays. High-sensitivity cardiac Troponin T outperformed hs-cTnI in its prognostic accuracy both in all patients and in important subgroups such as patients with AMI at presentation, pre-existing coronary artery disease, impaired renal function, or patients older than 70 years. Optimal prognostic cut-off values in these subgroups were only slightly higher than derived from all patients. In patients with AMI at presentation only hs-cTnT and cTnT yielded statistical significance. Secondly, novel hs-cTn assays moderately predict mortality in patients with pre-existing CAD, impaired renal function, and in patients older than 70 years. Thirdly, unlike in the diagnosis of AMI, neither serial measurements of hs-cTn nor changes in the first 6 h provided important additional information alone or in combination with presentation values of hs-cTn. To the best of our knowledge, this is the first analysis to clearly show that—in contrast to diagnostic considerations—for prognostic purposes serial measurements, absolute, or relative changes of hs-cTn do not provide relevant added value regarding risk stratification of acute chest pain patients. This novel finding seems at least partly explained by the fact that most patients with acute chest pain show little change in their hs-cTn concentration. In addition, those who do show a change are predominately those with AMI, who already have the highest concentrations at presentation. It is important for clinicians to know that a patient's risk of death can reliably be estimated already with the hs-cTn value at presentation. Fourthly, the prognostic benefit of the novel hs-cTn assays was pronounced in acute chest pain patients without AMI (Figure 3D), a group of patients whose troponin concentrations often cannot be detected and quantified by conventional cTn assays. This clinically highly relevant large group of patients can now be better risk-stratified with hs-cTn assays. It is important to emphasize that in acute chest pain patients with AMI the prognostic accuracy of hs-cTn assays was only moderate (Figure 3E) and other biomarkers might be better (additional) prognosticators of mortality in this patient group. Fifthly, the accuracy of the hs-cTnI assays to predict 2-year mortality did neither differ significantly between each other nor in comparison with conventional cTnT.
The reasoning behind the varying prognostic accuracy of the three hs-cTn assays studied remains speculative: the range of hs-cTn values in general and the overlap between survivors and non-survivors was highest for the hs-cTnI (Siemens) assay, thus resulting in a lower discriminatory power. Besides, hs-cTn assays measure tiny amounts of protein: It is currently unknown whether the incidence of disturbing phenomena such as fibrin interference or interference due to antibodies varies significantly among novel hs-cTn assays. Ungerer et al. scrutinized four similar contemporary cTnT and cTnI assays and concluded that the difference in cTnT and cTnI might in parts be explained by the fact that cTnT and cTnI measure different moieties, albeit that they are supposedly released in equimolar amounts via the same pathological process. Furthermore, it is known that haemolysis of even modest extent reduces cTnT and increases cTnI with some assays. Further studies need to re-evaluate the possible impact of biological variability, assay interference such as from nonspecific antibody binding in recently introduced hs-cTn assays and detect potential further confounding factors influencing their prognostic accuracy.
Bonaca et al. showed that even small increases above the 99th percentile of a sensitive cTn assay were associated with a significantly higher risk of death in both short-term (30 days) and long-term (12 months) perspective and proposed the use of the 99th percentile also for prognostic analyses. In this study, we have been able to demonstrate that mortality differs considerably even among patients with values below the 99th percentile. Detectable hs-cTn concentrations in the 'normal range' represent subclinical cardiomyocyte injury and thus may have important clinical implications. A dichotomous outcome stratification seems to be hardly possible with any of the four troponin assays studied. As illustrated in Figure 5B any measurable amount of hs-cTnT seems to be associated with a higher risk of mortality, even at concentrations below the 99th percentile. Owing to the higher accuracy of hs-cTn assays at the 99th percentile in comparison with conventional assays, we believe that hs-cTn yields incremental prognostic benefit over the conventional cTnT assay. In particular, the large group of acute chest pain patients without AMI (and mainly low troponin concentrations) seems to benefit from the higher sensitivity of novel troponin assays and improved selection of patients at risk (Table 4; Figure 3D and E).
All new hs-cTn assays permit a quantification of cTn concentrations in significant numbers of patients without coronary obstruction, such as patients with heart failure, hypertensive crisis, patients with subclinical heart disease, and even in apparently healthy subjects.
Our data as well as findings in other cohorts suggest that levels of hs-cTn should be considered quantitative markers of cardiomyocyte damage. The higher the hs-cTn level, the more extensive is the cardiomyocyte damage. The association between hs-cTn = cardiomyocyte damage and mortality seems to be linear. When comparing hs-cTnT with the fourth generation cTnT, the ROC curves for 720-day mortality are superimposable for the measureable range of the fourth generation cTnT. This indicates that both assays provide identical sensitivity for any given specificity. In addition to these values that provide a rather high specificity (1-specificity = 0,2; therefore specificity = about 80%), hs-cTn assays provide measurements in the undetectable range with the fourth generation assay that indicate that those patients labelled undetectable with the fourth generation represent a mixed group of patients: some with actually also very low levels of hs-cTn and therefore a very low risk of dying, but also some patients detected to have mildly elevated levels of hs-cTn indicating a risk of mortality that is much higher when compared with those with really normal hs-cTn levels.
The reasoning behind the good performance of hs-cTnT in patients with cardiac, non-coronary artery disease, and patients with unknown cause of chest pain remains speculative: Yet, it seems that any quantifiable amount of cardiomyocyte damage and therefore cTn release seems to be correlated with a worse prognosis (Figure 5B). Owing to the fact that the hs-cTnT assay is highly precise around the 99th percentile (<10% coefficient of variation) and is able to quantify troponin in the majority of healthy individuals, especially patients without AMI (and low concentrations of troponin) seem to benefit from the increased prognostic accuracy of hs-cTnT (Figure 3C). Nevertheless, the prognostic value in these subgroups needs to be confirmed in further prospective studies, not least because the number of events (patients dying) in these subgroups was relatively small in this analysis.
The prognostic accuracy as to future myocardial infarction in the first 730 days did not differ significantly between the four cTn assays studied and was only modest for all four assays. Nevertheless, high NPVs could be reached by using optimal cut-off values slightly below the 99th percentile of the respective assay (Table 2).
There was an intriguing difference in correlation between hs-cTn assays in patients with and without AMI at presentation. Lippi et al. in a small study of 47 patients reached similar conclusions as in this study: assays of hs-cTnT and hs-cTnI display much higher values for correlation in patients with than without AMI. The near-perfect correlation of hs-cTnT and cTnT in patients with AMI in contrast to a low correlation of the two assays in all patients (and not significant correlation in patients without AMI) might be explainable by the fact that the same protein has been measured (with different assays) and the low sensitivity of cTnT leading to non-detectable levels of cTnT in most patients without AMI. The near-perfect correlation of the two cTnT assays in patients with AMI has also been shown by Lindahl et al. in a large study of patients with acute coronary syndrome. But, the fact that the correlation between the two cTnT assays in patients with AMI was much higher than between hs-cTnI assays and between hs-cTnT and the two hs-cTnI assays is more surprising and suggests a higher prognostic validity of hs-cTnT than hs-cTnI. This is also expressed by the AUC analyses: cTnT—although much less sensitive than hs-cTnI—was not inferior as to the prognostic accuracy than hs-cTnI assays; in patients with AMI it even outperformed the prognostic accuracy of hs-cTnI assays. Based on our data, although cTnI and cTnT reflect the same pathological process, they do not seem to be interchangeable with each other as to their prognostic accuracy. Furthermore, correlation analyses of cTn after mixing patients with and without AMI might be misleading.
First, as a prospective observational study, we cannot quantify exactly the clinical benefit associated with improved risk stratification. Secondly, we cannot comment on prognostic accuracy among patients with terminal kidney failure requiring dialysis, since such patients were excluded from our study. Thirdly, due to the size of the study and limited number of events that occurred during the follow-up we cannot fully exclude a minor additional prognostic benefit of early changes of hs-cTn. Fourthly, patients with STEMI [whose diagnosis is mainly based on the ECG, clinical presentation and not on (serial) troponin measurements] are underrepresented in this analysis since their diagnosis required prompt transfer to the catheter laboratory, often before a 1-h value of hs-cTn could be obtained. Fifthly, we cannot exclude the probability that cTn measurements and changes after the 6-h period would have improved the prognostic accuracy since no such study samples were taken. Sixthly, patients in cardiogenic shock might be underrepresented in this study, since their inclusion is hindered by proper informed consent. Seventhly, due to inter-individual difficulties in taking blood samples and varying patients' willingness to accept repetitive blood withdrawals not all serial measurements have been able to be determined in all patients. Therefore, the statistical power of our data set may have been too small to detect a small prognostic benefit of serial measurements.
High-sensitivity cardiac Troponin T predicts mortality more accurately than hs-cTnI assays in patients with suspected AMI. Unlike in the diagnosis of AMI, serial measurements and changes of hs-cTn do not seem to further improve the prognostic accuracy of presentation values of hs-cTn.
Discussion
In this prospective, observational, international, multicentre trial of 1117 unselected patients presenting with acute chest pain to the ED, we scrutinized the prognostic value of presentation values and serial measurements of three novel hs-cTn assays and compared them with a conventional cTn assay. We report five major findings.
First, hs-cTnT at presentation outperformed both hs-cTnI assays and conventional cTnT regarding the accuracy to predict 2-year mortality. This finding extends previous analyses that found similar high diagnostic accuracy for the hs-cTn assays. High-sensitivity cardiac Troponin T outperformed hs-cTnI in its prognostic accuracy both in all patients and in important subgroups such as patients with AMI at presentation, pre-existing coronary artery disease, impaired renal function, or patients older than 70 years. Optimal prognostic cut-off values in these subgroups were only slightly higher than derived from all patients. In patients with AMI at presentation only hs-cTnT and cTnT yielded statistical significance. Secondly, novel hs-cTn assays moderately predict mortality in patients with pre-existing CAD, impaired renal function, and in patients older than 70 years. Thirdly, unlike in the diagnosis of AMI, neither serial measurements of hs-cTn nor changes in the first 6 h provided important additional information alone or in combination with presentation values of hs-cTn. To the best of our knowledge, this is the first analysis to clearly show that—in contrast to diagnostic considerations—for prognostic purposes serial measurements, absolute, or relative changes of hs-cTn do not provide relevant added value regarding risk stratification of acute chest pain patients. This novel finding seems at least partly explained by the fact that most patients with acute chest pain show little change in their hs-cTn concentration. In addition, those who do show a change are predominately those with AMI, who already have the highest concentrations at presentation. It is important for clinicians to know that a patient's risk of death can reliably be estimated already with the hs-cTn value at presentation. Fourthly, the prognostic benefit of the novel hs-cTn assays was pronounced in acute chest pain patients without AMI (Figure 3D), a group of patients whose troponin concentrations often cannot be detected and quantified by conventional cTn assays. This clinically highly relevant large group of patients can now be better risk-stratified with hs-cTn assays. It is important to emphasize that in acute chest pain patients with AMI the prognostic accuracy of hs-cTn assays was only moderate (Figure 3E) and other biomarkers might be better (additional) prognosticators of mortality in this patient group. Fifthly, the accuracy of the hs-cTnI assays to predict 2-year mortality did neither differ significantly between each other nor in comparison with conventional cTnT.
The reasoning behind the varying prognostic accuracy of the three hs-cTn assays studied remains speculative: the range of hs-cTn values in general and the overlap between survivors and non-survivors was highest for the hs-cTnI (Siemens) assay, thus resulting in a lower discriminatory power. Besides, hs-cTn assays measure tiny amounts of protein: It is currently unknown whether the incidence of disturbing phenomena such as fibrin interference or interference due to antibodies varies significantly among novel hs-cTn assays. Ungerer et al. scrutinized four similar contemporary cTnT and cTnI assays and concluded that the difference in cTnT and cTnI might in parts be explained by the fact that cTnT and cTnI measure different moieties, albeit that they are supposedly released in equimolar amounts via the same pathological process. Furthermore, it is known that haemolysis of even modest extent reduces cTnT and increases cTnI with some assays. Further studies need to re-evaluate the possible impact of biological variability, assay interference such as from nonspecific antibody binding in recently introduced hs-cTn assays and detect potential further confounding factors influencing their prognostic accuracy.
Bonaca et al. showed that even small increases above the 99th percentile of a sensitive cTn assay were associated with a significantly higher risk of death in both short-term (30 days) and long-term (12 months) perspective and proposed the use of the 99th percentile also for prognostic analyses. In this study, we have been able to demonstrate that mortality differs considerably even among patients with values below the 99th percentile. Detectable hs-cTn concentrations in the 'normal range' represent subclinical cardiomyocyte injury and thus may have important clinical implications. A dichotomous outcome stratification seems to be hardly possible with any of the four troponin assays studied. As illustrated in Figure 5B any measurable amount of hs-cTnT seems to be associated with a higher risk of mortality, even at concentrations below the 99th percentile. Owing to the higher accuracy of hs-cTn assays at the 99th percentile in comparison with conventional assays, we believe that hs-cTn yields incremental prognostic benefit over the conventional cTnT assay. In particular, the large group of acute chest pain patients without AMI (and mainly low troponin concentrations) seems to benefit from the higher sensitivity of novel troponin assays and improved selection of patients at risk (Table 4; Figure 3D and E).
All new hs-cTn assays permit a quantification of cTn concentrations in significant numbers of patients without coronary obstruction, such as patients with heart failure, hypertensive crisis, patients with subclinical heart disease, and even in apparently healthy subjects.
Our data as well as findings in other cohorts suggest that levels of hs-cTn should be considered quantitative markers of cardiomyocyte damage. The higher the hs-cTn level, the more extensive is the cardiomyocyte damage. The association between hs-cTn = cardiomyocyte damage and mortality seems to be linear. When comparing hs-cTnT with the fourth generation cTnT, the ROC curves for 720-day mortality are superimposable for the measureable range of the fourth generation cTnT. This indicates that both assays provide identical sensitivity for any given specificity. In addition to these values that provide a rather high specificity (1-specificity = 0,2; therefore specificity = about 80%), hs-cTn assays provide measurements in the undetectable range with the fourth generation assay that indicate that those patients labelled undetectable with the fourth generation represent a mixed group of patients: some with actually also very low levels of hs-cTn and therefore a very low risk of dying, but also some patients detected to have mildly elevated levels of hs-cTn indicating a risk of mortality that is much higher when compared with those with really normal hs-cTn levels.
The reasoning behind the good performance of hs-cTnT in patients with cardiac, non-coronary artery disease, and patients with unknown cause of chest pain remains speculative: Yet, it seems that any quantifiable amount of cardiomyocyte damage and therefore cTn release seems to be correlated with a worse prognosis (Figure 5B). Owing to the fact that the hs-cTnT assay is highly precise around the 99th percentile (<10% coefficient of variation) and is able to quantify troponin in the majority of healthy individuals, especially patients without AMI (and low concentrations of troponin) seem to benefit from the increased prognostic accuracy of hs-cTnT (Figure 3C). Nevertheless, the prognostic value in these subgroups needs to be confirmed in further prospective studies, not least because the number of events (patients dying) in these subgroups was relatively small in this analysis.
The prognostic accuracy as to future myocardial infarction in the first 730 days did not differ significantly between the four cTn assays studied and was only modest for all four assays. Nevertheless, high NPVs could be reached by using optimal cut-off values slightly below the 99th percentile of the respective assay (Table 2).
There was an intriguing difference in correlation between hs-cTn assays in patients with and without AMI at presentation. Lippi et al. in a small study of 47 patients reached similar conclusions as in this study: assays of hs-cTnT and hs-cTnI display much higher values for correlation in patients with than without AMI. The near-perfect correlation of hs-cTnT and cTnT in patients with AMI in contrast to a low correlation of the two assays in all patients (and not significant correlation in patients without AMI) might be explainable by the fact that the same protein has been measured (with different assays) and the low sensitivity of cTnT leading to non-detectable levels of cTnT in most patients without AMI. The near-perfect correlation of the two cTnT assays in patients with AMI has also been shown by Lindahl et al. in a large study of patients with acute coronary syndrome. But, the fact that the correlation between the two cTnT assays in patients with AMI was much higher than between hs-cTnI assays and between hs-cTnT and the two hs-cTnI assays is more surprising and suggests a higher prognostic validity of hs-cTnT than hs-cTnI. This is also expressed by the AUC analyses: cTnT—although much less sensitive than hs-cTnI—was not inferior as to the prognostic accuracy than hs-cTnI assays; in patients with AMI it even outperformed the prognostic accuracy of hs-cTnI assays. Based on our data, although cTnI and cTnT reflect the same pathological process, they do not seem to be interchangeable with each other as to their prognostic accuracy. Furthermore, correlation analyses of cTn after mixing patients with and without AMI might be misleading.
Limitations
First, as a prospective observational study, we cannot quantify exactly the clinical benefit associated with improved risk stratification. Secondly, we cannot comment on prognostic accuracy among patients with terminal kidney failure requiring dialysis, since such patients were excluded from our study. Thirdly, due to the size of the study and limited number of events that occurred during the follow-up we cannot fully exclude a minor additional prognostic benefit of early changes of hs-cTn. Fourthly, patients with STEMI [whose diagnosis is mainly based on the ECG, clinical presentation and not on (serial) troponin measurements] are underrepresented in this analysis since their diagnosis required prompt transfer to the catheter laboratory, often before a 1-h value of hs-cTn could be obtained. Fifthly, we cannot exclude the probability that cTn measurements and changes after the 6-h period would have improved the prognostic accuracy since no such study samples were taken. Sixthly, patients in cardiogenic shock might be underrepresented in this study, since their inclusion is hindered by proper informed consent. Seventhly, due to inter-individual difficulties in taking blood samples and varying patients' willingness to accept repetitive blood withdrawals not all serial measurements have been able to be determined in all patients. Therefore, the statistical power of our data set may have been too small to detect a small prognostic benefit of serial measurements.
Conclusion
High-sensitivity cardiac Troponin T predicts mortality more accurately than hs-cTnI assays in patients with suspected AMI. Unlike in the diagnosis of AMI, serial measurements and changes of hs-cTn do not seem to further improve the prognostic accuracy of presentation values of hs-cTn.
Source...