Strategies to overcome misdiagnosis of type 1 myocardial infarction using high sensitive cardiac troponin assays

Definition of myocardial infarction

According to the third universal definition of myocardial infarction (MI) published in 2012 [1] an MI is classified into five categories in patients with symptoms consistent with MI: type 1 MI – rise and/or fall of cardiac biomarkers and clinical evidence of an acute MI; type 2 MI has been defined if myocardial necrosis (with rise or/and fall of cardiac biomarkers) occurs not due to coronary artery disease (CAD) and plaque rupture but due to a secondary ischemic imbalance e.g. tachy- or bradyarrhytmia, arterial hypertension or coronary spasm; type 3 MI reflects suspected cardiac death before biomarkers could be obtained; type 4a MI is defined as elevation of cardiac biomarkers in the context of percutaneous coronary intervention (PCI), type 4b as stent thrombosis associated MI proven by coronary angiography or autopsy with related rise or fall in cardiac biomarkers. Coronary bypass grafting (CABG) associated MI is classified as type 5 (see also Table 1).

Cardiac troponins as biomarker of choice in diagnosis MI

Ebashi firstly described a “tropomyosin” complex in 1963 which is necessary for relaxing and contracting the muscle [3]. Three components constitute this complex named troponin I (TnI), troponin T (TnT) and troponin C (TnC) [4] and it was shown that there are cardiac specific isoforms [5]. By developing an enzyme linked immunoassays in the late eighties both for cTnI and cTnT the detection of cTn in the setting of MI was possible [6], [7]. In 2000, the Joint European Society of Cardiology (ESC)/American College of Cardiology (ACC) proposed a redefinition of MI which included cTn as the biomarkers of choice to diagnose MI paving the way for cTn into the diagnostic guidelines and the clinical routine [8].

As diagnostic threshold the assay specific upper limit of normal (ULN) was defined representing the 99th percentile concentration of the troponin assay in a healthy reference population. Furthermore, criteria of test sensitivity were defined as a coefficient of variation (CV) of 10% or less at this threshold concentration. Former cTn assays often showed a lack of sensitivity to detect cTn levels e.g. in healthy individuals. This led to an evolution of commercially available assays with improved sensitivity. These high sensitive assays are frequently able to detect low cTn levels even in apparently healthy individuals [9]. Their clinical application using the 99th percentile diagnostic threshold consistently showed an improvement in early diagnosis of MI type 1 [10], [11].

Gain of sensitivity – loss in specificity

The improved troponin assay sensitivity, facilitating a rapid diagnosis of MI type 1, could be confirmed in many clinical studies underlining the clinical benefit of sensitive troponin assays. Use of more sensitive assays allows detecting cardiac troponin levels even in healthy individuals; even a biological variation in healthy individuals is detectable without presence of an MI [12], [13], [14]. Elevated troponin levels can be found in the absence of an MI in patients with several other conditions such as in critically ill patients (e.g. acute neurological event, severe respiratory failure, severe pulmonary embolism) and troponin levels can be chronically increased in certain patient populations: this includes patients with impaired renal function or elderly patients (Figure 1). These elevations of troponin levels due to chronic or acute conditions other than MI are associated with a loss of specificity of sensitive troponin assay in the diagnosis of MI type 1.

Figure 1:

Influences on troponin levels.

Different known confounding factors and diagnoses on cardiac troponin levels. AMI, acute myocardial infarction; SAH, subarachnoid hemorrhage.

Current recommendations for diagnosing MI

While the diagnosis of an acute ST-segment-elevation myocardial infarction (STEMI) can be made primarily based on certain changes in the electrocardiogram (ECG), the diagnosis of an acute non ST-segment-elevation myocardial infarction (NSTEMI), according to the current ESC guidelines, relies on cardiac biomarkers, first and foremost on cardiac troponin measurement using a sensitive assay [15].

Besides from elevated cTn levels on admission above the ULN of the respective assay, a rise or fall in cTn levels (Delta cTn) is essential for the diagnosis of an NSTEMI in patients with suspected MI. In recent years, different approaches to evaluate Delta cTn levels have been proposed including the use of relative vs. absolute changes and the amount of change which could be considered significant [16], [17], [18], [19]. The current ESC guidelines propose different algorithms to rule-in and to rule-out NSTEMI based on baseline hs-cTn as well as the change over time including the established 0 h/3 h algorithm and recent 0 h/1 h algorithm.

Using serial troponin measurements for diagnosing MI

To account for troponin elevations especially caused by non-acute conditions the diagnostic use of change in troponin concentration within a few hours is able to improve specificity for the diagnosis of MI type I. The following two algorithms are currently recommended by the ESC guidelines.

Improving specificity by adding information of clinical risk scores

Beyond using change in troponin concentration to improve the diagnostic specificity for MI type I the use of clinical parameters has been discussed. Indeed, it could be shown that serial cTn measurements combined with risk scores such as the HEART score that include history, ECG, age, risk factors, and troponin as parameters can improve diagnosis of MI [25], [26]. Furthermore, the combination of troponin determination and the Thrombolysis In Myocardial Infarction (TIMI) score improves diagnosis of MI [27].

Improving specificity in patients with impaired renal function

Over the last 20 years, several authors described increased cTnT and cTnI levels in patients suffering from chronic kidney disease (CKD). In 1994 Hafner et al. found increased cTnI and cTnT levels in 18 patients undergoing dialysis [28]. Li et al. not only described increased cTnT levels in patients undergoing hemodialysis but also in patients with CKD [29]. Over the next years these findings were reproduced by several other authors [30], [31]. This elevation might be due to poor renal elimination, micronecrosis or even an origin other than the heart. Evidence that increased cTnT levels are seen caused by poor renal elimination were found by Keddis et al. [32]. Tsuatomoto et al. compared the transcardiac gradient in patients with and without CKD. While CKD patients showed overall higher cTnT levels, there was no difference in the transcardiac gradient, which led to the assumption that the origin of cTnT in CKD patients is the heart, and that the increased levels are based on a poor renal elimination [33]. There is data that increased cTn levels reflect underlying cardiovascular diseases such as left ventricular hypertrophy, coronary artery disease or heart failure in CKD patients [34], [35], [36] and that both, cTnT and cTnI, are markers of mortality, cardiovascular mortality or major cardiovascular events in patients with CKD [37], [38], [39], [40], [41]. This on one hand underlines the high-risk status of these patients. On the other hand, adequate diagnosis of MI type 1 with subsequent treatment initiation that is based largely on cTn is more difficult in CKD patients.

Chenevier-Gobeaux et al. analyzed the diagnostic performance of hs-cTnT in CKD patients and suspected MI upon admission. The ULN of the respective hs-cTnT assay had a high specificity in detection of MI type 1 in patients with a preserved renal function (86%); whereas in CKD patients specificity for MI decreases to 54% (Table 2) [42]. This observation is supported by data by Pfortmueller et al. showing a comparable loss in specificity of hs-TnT levels upon admission and by Twerenbold et al. who compared the diagnostic utility of baseline cTn levels of four hs-cTnI assays and one hs-cTnT assay in the evaluation of MI type 1 in CKD patients [43], [44] (Table 3). This loss in specificity poses the relevant risk for misdiagnosis in these specific patients based on the first cTn measurement. This is aggravated by the fact that CKD patients are at high risk for developing an MI [46].

The studies by Chenevier-Gobeaux et al. and Twerenbold et al. both could show that the area under the receiver operator characteristics curve (AUROC) for cTn to detect MI in CKD patients is comparable to the AUROC of patients with a preserved renal function. They state, that higher, AUROC-optimized cut-offs are in need to regain the diagnostic specificity of cTn in CKD. Chenevier-Gobeaux derived a higher optimized hs-cTnT cut-off for CKD patients that is able to detect MI with a specificity of 86% using the baseline cTn measurement, this is the same specificity provided by the ULN threshold in patients with a preserved renal function. This increase in specificity is not accompanied by a relevant loss of sensitivity (Table 2). The efficacy of higher optimized cut-offs in CKD patients to identify MI type 1 was also verified by Twerenbold et al. [44] (Table 3).

Improving specificity in elderly patients

Similar problems in the accurate diagnosis of MI type 1 as in CKD patients are seen in elderly patients. Several authors could show increased cTn levels in older patients irrespective of an acute MI. A recent study evaluated hs-cTnT levels in geriatric patients and found in over 50% of 679 hospitalized geriatric patients without MI (mean age 82.5±5.8 years) hs-cTnT levels above the ULN. Even in patients without known coronary artery disease, 46.6% had hs-cTnT levels above the ULN [47]. Saneger et al. described an association of hs-cTnT levels with age [48] and Gore et al. evaluated hs-cTnT levels in elderly patients: here hs-cTnT was analyzed in apparently cardiac healthy elderly patients from three different population based studies. In patients without structural heart disease, without impaired left ventricular function or without elevated N-terminal pro-B-type natriuretic peptide (NT-proBNP) an impact on the ULN of hs-cTnT was seen [49]. Similar findings exist for hs-cTnI. Keller et al. showed a rise in hs-cTnI levels with patients’ age [50]. This data is supported by a study of McKie et al. [51]. The authors could show that hs-cTnI levels are significantly higher in elderly patients, apparently free of cardiovascular diseases. On the basis of these findings, elevated cTn levels in elderly patients cannot only be explained by an impaired renal function that is often seen in the elderly population, as these patients were excluded. Increased cTn levels indicate a higher risk for developing chronic heart disease as they might indicate subclinical myocardial damage [52], [53].

These data suggest that elevated cTn levels in elderly patients are caused by various reasons including silent cardiovascular disease but might also be increased in patients without cardiovascular disease [49], [53]. All these studies have in common, that elevated cTnT and cTnI levels were observed in patients in the absence of MI. This increase in cTn levels therefore leads to an increase of uncertainty in evaluation of suspected MI type 1 in elderly patients.

Reiter et al. analyzed the performance of three different sensitive cTn assays to diagnose MI type I in patients older than 70 years based on baseline cTn. Use of the respective ULN as diagnostic cut-off showed a decrease in specificity in elderly patients compared to patients younger than 70 years. To regain this loss in specificity age optimized cut-off were derived based on AUROC analyses [54] (Table 4). A similar approach was carried out in the analysis of Chenevier-Gobeaux et al. [42]. The authors evaluated hs-cTnT levels in patients older than 70 years and with suspected MI measured upon admission showing a specificity with 51% if using the ULN compared to a specificity of 88% in detection of MI in patients <70 years. By applying higher, AUROC-optimized cTn cut-off levels, this loss in specificity could be regained (Table 5).

Based on the idea of increased cTn levels in elderly patients in the absence of an acute MI, a study carried out by Olivieri et al. derived age-specific cut off levels in a healthy reference population in patients older 75 years and tested their diagnostic performance in geriatric patients with suspected MI. The established 99th percentile ULN of the respective hs-cTnT assays is 14 ng/L whereas in individuals older than 75 years an ULN of 70.6 ng/L was derived. Using this higher ULN in geriatric patients with suspected MI gained a sensitivity of 94% and a specificity of 77% [55].

Also, Gore et al. defined ULN of hs-cTnT in their cohort of elderly patients apparently free of cardiovascular diseases [49]. Several ULNs for different ages, genders and races based on different population based studies were reported [49]. Mueller-Hennesen et al. recently tested the application of the reported higher ULN for individuals aged 65 years or older in a population of patients presenting with suspected MI. The authors analyzed reclassification of elderly patients with regard to the final diagnosis of acute coronary syndrome (ACS) or non-ACS. The influence on both diagnostic and prognostic reclassification was significant using an age-specific hs-cTnT ULN based cut-off level approximately two-fold of the established ULN of the assay [56].

Gender differences in troponin levels – risk of underdiagnoses in females

While chronically elevated cTn levels in patients free of an acute cardiovascular disease might make the use of optimized cTn cut-off levels beneficial, there is an ongoing debate about the need for gender specific cut-offs. It has been described, that male patients show higher cTn levels than females [49], [50]. Knowing that diagnosis of MI in female patients is often more difficult due to e.g. atypical symptoms or a lack of ECG changes [57], [58] underlines the importance of cardiac biomarkers. Given the higher troponin values observed in healthy men, use of a single ULN cut-off for both genders might lead to an underdiagnosis of MI in female patients. Several authors published gender specific ULN for several cTn assays derived in healthy reference populations stating lower ULN for females than for men:

In two different studies Apple et al. derived ULN for several hs-cTn assays in large populations of apparently healthy individuals with respect to possible gender differences with higher ULN in males compared to females [59], [60]. Giannitsis et al. described gender specific ULN in an apparently healthy population for hs-cTnT with 14.5 ng/L in males compared to 10.0 ng/L in females [61]. This gender difference in hs-cTnT ULN was also seen by others. Here, the ULN was derived in three population based cohorts and subcohorts with regard to the presence of cardiovascular diseases. In all of the three cohorts as well as in the respective subcohorts, the derived ULN for hs-cTnT were higher in men compared to women [2].

These cited publications investigated potential gender specific ULN for several cTn assays in healthy subjects, but did not investigate the impact on diagnostic performance of such gender specific ULN in patients with suspected MI and therefore their clinical relevance.

Mueller-Hennesen et al. applied the gender specific ULN cut-offs for hs-cTnT derived by Saenger et al. [48], [56] to a cohort of patients with suspected MI to evaluate a possible impact on diagnostic or prognostic reclassification. As expected, an increase of the final diagnosis of MI was seen in female patients while a decrease was observed in men. Unfortunately, this increase in MI diagnoses was accompanied by an increase of false positive diagnoses of MI in female patients. Furthermore, use of female specific cut-offs was associated with a negative impact on outcome based on the hs-cTnT level. Similar results for cTnI have been proposed by Keller et al. Gender specific cTnI ULN cut-offs were calculated in a population based cohort of apparently healthy individuals and subsequently applied to a cohort of patients with suspected MI. As anticipated more female patients would have been categorized as patients suffering from MI, but similar to the findings of Mueller-Hennesen et al. this increase of the diagnosis MI also led to an increase in the rate of false positive results [50]. In contrast Shah et al. could show an improvement in risk prediction (recurrent AMI or death) using a gender specific ULN. Women presenting with hs-cTnI levels between 16 ng/L, the gender-specific ULN for women, and 26 ng/L, the ULN without gender-specificity, showed an increased event rate compared to women presenting with a baseline hs-cTnI levels below the female specific ULN (23% vs. 4%, p<0.001) [62].

Benefit of using individual troponin cut-off levels

In several patient populations such as patients with impaired renal function or elderly patients cTn levels are chronically increased in the absence of acute cardiac diseases or MI. This complicates the diagnosis of MI type I based on cTn levels upon admission. Three strategies are recommended to improve specificity of cardiac troponins and to improve the diagnostic accuracy. First, use of serial troponin measurements to differentiate chronic elevations from increased levels due to an acute event with respective rise and later fall in marker concentration. Second, incorporation of clinical parameters such as risk scores like the HEART score.

Third, use of individually optimized cut-offs. Here, optimized cTn cut-offs need to be derived for every single cTn assay. Therefore, prospective studies are in need for different cTn assays to derive and evaluate cTn cut-offs in specific patient groups such as for elderly patients.

While for patients with CKD and for elderly patients individual cTn cut-offs seem beneficial, there is an ongoing discussion on the use of gender-specific ULN. There is ample data showing that women have lower cTn values than men leading to the risk of underdiagnosis of MI type 1 if using a unified diagnostic ULN. Otherwise, a lower female ULN might increase false positive results. So, the use of gender specific ULN has to be the aim of future prospective studies.