Do we need to consider age and gender for accurate diagnosis of myocardial infarction?

Introduction

In patients presenting with acute symptoms to an emergency department (ED), assessment and interpretation of cardiac troponin (cTn) is paramount for correct diagnosis of an acute myocardial infarction (AMI) [1], [2], [3]. Thus, it is extremely important for initiation of further diagnostic workup and guideline conform therapy, that appropriate cTn decision cut-offs for rule-in and rule-out of AMI are used [3]. Following the introduction of more sensitive and high-sensitivity (hs) cTn assays with lower limits of detection, it has become possible to measure cTn in the majority of healthy subjects [4], [5]. As a consequence, age- and gender-dependent differences in hs-cTn concentrations can now be detected, not only in the initial assay validation cohorts, but also in patients presenting to EDs with symptoms suggestive of AMI.

Consequently, there is an increasing number of studies which have focused on age- and gender-dependent differences in cTn. These differences may impact diagnostic decisions, but also all related algorithms incorporating cTn for therapy and risk stratification. Thus, there is an immediate need to assess the potential effects for different cut-off values and diagnostic reclassification in cardiovascular (CV) risk stratification. The risk for under-diagnosis of AMI by using general cut-offs that are higher than the age- or gender-specific cut-offs may severely disadvantage the respective individuals concerning receipt of guideline-recommended treatment leading to a potential negative impact on outcome. Contrary, over-diagnosis of AMI using a general cut-off that is lower than the age- or gender-specific cut-off can result in prolongation of ED stays and unnecessary hospitalizations that are cost-intensive, as well as initiation of potentially harmful therapies including antithrombotic treatment and invasive coronary angiography.

To date, there is still uncertainty about the pros and cons of age- or gender-specific cut-offs and no clear recommendation is given how to implement these different cut-offs into clinical practice. At present, almost all recommendations concerning the currently available cTn assays support the use of a uniform 99th percentile cut-off derived from healthy reference populations [2], [6]. However, the third version of the universal MI definition states that sex-specific cut-offs for cTn may be recommended, without giving any further comment on the way it should be handled in the clinical setting [2]. Moreover, the current ESC non-ST-elevation (NSTE)-acute coronary syndrome (ACS) guidelines have added a section concerning gender and age differences in NSTE-ACS and support to implement diagnostic and prognostic strategies that take these subgroups into consideration [3]. Albeit, statements on impact of age and gender on cTn algorithms are completely lacking as well as any specific recommendation. Thus, clinicians may be confused by the variety of study results on the one hand and by the relatively absence of clear recommendations on implementation in guidelines on the other hand.

cTn elevations in the elderly and prognostic implications

cTn elevations above the diagnostic cut-offs have been found to be relatively common in the elderly, even in presumably healthy populations [7], [8]. In the Dallas Heart Study (DHS), a multi-ethnic, community study of healthy subjects aged 30 to 65 years, one fourth exhibited elevations above the 99th percentile cut-off for the Roche hs-cTnT assay, whereas only 0.7% had values above the diagnostic cut-off for conventional TnT [9]. In addition, the prevalence of hs-cTnT above the limit of blank (≥3 ng/L) in individuals younger than 40 years was found to be 14% and increased to 57.6% in those aged 60 years and older [9]. Similarly, elevated levels of cTnI with a more sensitive assay were found in 21.7% of individuals aged 70 years or older living in the community [8]. Emerging data indicate that higher age represents a potential reason for elevated cTn values in community-based populations [9], [10] as well as in patients with ACS and non-ACS conditions [11], [12], [13]. Moreover, cTnT elevations above the diagnostic cut-off were related to structural heart disease, reduced kidney function and higher mortality risk [9], [14].

In the DHS, elevations of hs-cTnT above the 99th percentile carried a significantly higher risk for all-cause death compared to patients with normal hs-cTnT levels [9]. In the Cardiovascular Health Study (CHS) and the Atherosclerosis Risk in Communities (ARIC) study, similar findings concerning improved risk stratification using hs-cTnT could be observed for elderly adults aged over 65 years without a history of heart failure [10] or individuals who were 54 to 74 years of age [15], which indicates that hs-cTnT is able to identify patients at increased risk of adverse events even in normal populations of asymptomatic older individuals. In a sub-analysis of the ActiFE study, a longitudinal cohort study on 1506 asymptomatic volunteers from the community aged 65 years or older, hs-cTnT and hs-cTnI were independent predictors for adverse outcomes in the elderly [16]. However, differences in outcome prediction were found between the genders with higher adjusted hazard ratios for all-cause death for increasing cTn in women, which draws further attention to variations in the prognostic value of cTn in asymptomatic elderly males and females [16]. Higher prevalence of CV risk factors, reduced cardiac function, increasing left ventricular (LV) mass, subclinical micro-embolizations as well as attenuated cardiac integrity in the absence of myocardial injury have been discussed as possible explanations of elevated cTn values in the general population and in asymptomatic elderly adults [8], [17].

Age-dependent 99th percentile cTn cut-offs

Given the apparent differences in cTn depending on age and the concomitant increased mortality and morbidity risk, the accurate assessment of reference cut-offs in healthy individuals based on age as potential decision thresholds is even more important. However, studies reporting age-dependent 99th percentiles in healthy populations are sparse due to under-representation of elderly patients in the majority of reference studies and, thus, are mainly derived from population studies after exclusion of patients with overt or subclinical CVD. In a previous study using a Beckman-Coulter hs-cTnI prototype assay, Venge et al. [18] observed 99th percentiles of 10 ng/L for healthy subjects aged <60 years and 19 ng/L for those aged ≥60 years. In a sub-analysis of the Prospective Investigation of the Vasculature in Uppsala Seniors (PIVUS) study in individuals aged 70 years from the community, the 99th percentiles for the Abbott ARCHITECT STAT hs-cTnI assay were found to be 55.2 ng/L for the entire population, 69.3 ng/L for male and 26.3 ng/L for female participants [19]. In individuals without CVD at baseline and after 5 years of follow-up, the 99th percentiles were 31.6 ng/L at age 70 and 51.3 ng/L at age 75 years [19]. These observed cut-offs were markedly higher compared to the 99th percentile from younger reference populations for the hs-cTnI assay [7], [20]. Together with the observation of a relative rise of 62% for the 99th percentiles in individuals 70 vs. 75 years, the data suggests a strong association of age with cTn levels even in patients without CVD [19]. However, it should be noted that absence from CVD did not include echocardiographic data, which were also available in the study. Thus, it cannot be excluded that the cut-offs would be lower with consideration of imaging data, especially considering the higher association of subclinical structural heart disease in the elderly [8]. Conversely, the 99th percentile was twice as high at baseline in participants with CVD as assessed by history, ECG and the surrogate marker NT-proBNP, underscoring the need to establish universal criteria for normal reference populations and the association of cTn to cardiac abnormalities [21].

Similar for hs-cTnT, Hammarsten et al. more recently evaluated different populations in which hs-cTnT determined by the Roche Elecsys assay was elevated and found that hs-cTnT percentiles varied depending on age leading to increased 99th percentile value in elderly patients [14]. While the 99th percentile in a random population sample of individuals 39–59 years of age was 12.7 ng/L, inclusion of subjects ≥60 years of age resulted in an increase to around 25 ng/L [14]. In an ED population without any signs of ACS or any other apparent condition associated with cTnT increases, the 99th percentile in patients <65 years was 12.0 ng/L, whereas in patients >65 years, a value of 81.9 ng/L could be detected [14]. In one of the largest studies to date to evaluate age-dependent 99th percentile cut-offs, Gore et al. determined 99th percentile cut-off values for hs-cTnT from 12,618 patients of three large population-based studies and found that a considerable amount of 10% of male patients aged between 65 and 74 years of age without CVD had hs-cTnT elevations beyond the currently proposed hs-cTnT cut-off of 14 ng/L. Moreover, within each cohort, higher 99th percentile cut-off values were observed in elderly and male patients. In the sub-cohort of the ARIC study on patients >65 years of age, 99th percentile values were found to be 36 ng/L for patients, in which overt CVD, renal dysfunction and recent hospitalization was excluded, and 28 ng/L when patients with subclinical structural heart disease were further excluded.

Impact of age-dependent cut-offs on diagnosis and prognosis

Similar to study data on 99th percentiles, there is only sparse information regarding the age-dependent influence on the diagnostic and prognostic performance of hs-cTn assays. In the APACE trial on 1098 consecutive patients, Reiter et al. analyzed three more sensitive Tn assays in patients >70 years of age presenting with acute chest pain in order to evaluate the diagnostic performance of cTn in the early detection of AMI [11]. They found that 51% of non-AMI patients had elevated baseline cTn values with hs-cTnT (Roche), 17% with TnI-Ultra (Siemens), and 13% with hs-cTnI (Abbott ARCHITECT) [11]. As a consequence, ROC-optimized cut-offs in elderly patients were substantially higher than in younger patients [11]. For hs-cTnT, the cut-off value (54ng/L) determined by ROC-analysis was markedly higher than the currently proposed 99th percentile of 14 ng/L. However, the diagnostic performance of all three more sensitive Tn assays in the elderly was significantly superior compared to the standard assay for the early AMI detection [11].

In accordance, recently published findings of our group showed that patients older than 75 years more frequently exhibited hs-cTnT elevations above the 99th percentile. In addition, ROC analysis demonstrated improved diagnostic performance using a more than threefold higher cut-off in patients ≥75 years [13]. By the use of ROC-optimized cut-offs in the elderly, sensitivities decreased and specificities increased, bearing a potential risk to underdiagnose AMI in these patients. However, elevated hs-cTnT in older patients was associated with a higher prevalence of non-AMI conditions, thus impairing diagnostic performance for discrimination of AMI. In comparison to the elderly, a significantly better diagnostic performance of hs-cTnT for rule-out of AMI using the 99th percentile cut-off value was observed in younger patients [13].

In order to evaluate the clinical impact using age-dependent compared to general 99th percentile cut-offs for hs-cTnT, we recently analyzed 1282 patients presenting to the ED with suspected AMI as part of the TRAPID-AMI study [22]. We found that application of a higher age-specific 99th percentile cut-off value of 28 ng/L resulted in a reduction of patients fulfilling AMI criteria from 29.8% to 18.3% in all unselected chest pain patients aged 65 years or older, corresponding to a reduction from 54.7% to 40.9% in a cohort of adjudicated elderly ACS patients [22]. This diagnostic reclassification also translated into improved prognostication for elderly patients in unselected chest pain patients as well as in the adjudicated ACS cohort as assessed by net reclassification improvement, especially in short-term follow-up [22].

Gender-related differences

Gender disparities have been observed for patients with ACS concerning clinical presentation, diagnostic or therapeutic procedures and outcomes. Female ACS patients are less frequently referred for coronary angiography [23] and less frequently undergo diagnostic and invasive procedures [24], which might lead to a delay in AMI diagnosis and treatment [25]. Even women with diagnosed AMI were found to be treated differently from men with less invasive strategies [26], and lower utilization of antithrombotic therapy [18]. In coronary angiography, women had less extensive CAD [27]. Taken together, all these findings might lead to the perception of clinicians that women are at lower risk [3]. However, there are also a number of studies which did not observe significant gender differences for utilization of subsequent procedures and therapies [23], [27], [28], [29]. Concerning outcomes, some studies found that women are at higher risk for in-hospital mortality [30] and for 1-year MACE after AMI [28], while others observed similar long-term outcomes in female AMI with PCI [31] or attenuated mortality risks after age-adjustment [32]. An explanation for these differences may be attributed to the fact, that women more frequently present with atypical clinical symptoms and less frequently with chest pain [30].

Given these ambiguous findings, precise assessment of cut-offs for cardiac troponins depending on sex may be even more important to provide guidance for correct diagnosis and decide on initiation of further diagnostic testing and therapeutic consequences. Cardiac biomarkers including cTn are more frequently within the normal range as defined by the general 99th percentile cut-off in women compared to men [3]. Moreover, gender-dependent differences in cTn levels may be attributed to intrinsic cTn differences in release kinetics or diverse ACS pathophysiology with more frequent non-occlusive causes in females [33], [34].

Gender-specific 99th percentile cTn cut-offs

In the initial hs-cTnT validation study, males and females showed significantly different 99th percentile cut-off values. While the cut-off for females was found to be 9.0 ng/L, males exhibited a cut-off value of 15.5 ng/L that was approximately 1.7 times higher [35]. More recently, Apple et al. tested the 99th percentile value of 19 cTn assays in a population of 272 presumably healthy males and 252 females [20]. In comparison to females, male patients showed a 1.2 to 2.4-fold higher 99th percentile cut-off value for the hs-cTnI assays (i.e. Abbott ARCHITECT STAT, Beckman Access 2, Siemens Dimension Vista and the Singulex Erenna). For the Roche hs-cTnT assay, cut-off values of 13 ng/L were found in women and 20 ng/L in men [20].

Diagnostic reclassification and prognostic impact of gender-specific cTn

Bohula May et al. evaluated the risk of CV death or MI at 30 days stratified by sex-specific 99th percentile cut-offs for the Abbott ARCHITECT hs-cTnI assay based on data from EARLY-ACS and SEPIA-ACS1-TIMI 42 trial on 4695 patients [36]. Use of gender-specific cut-offs (16 ng/L for females and 34 ng/L for males) compared to the general 99th percentile (26 ng/L) reclassified six patients to a lower risk-group as assessed by values below the 99th percentile, with three of these patients experiencing a negative outcome event [36]. Due to the small number of reclassified patients despite the large size of the study cohort and the misclassification of half of these reclassified patients from a higher to a lower risk group, the authors argued against the use of sex-specific cut-offs for prognostication, especially considering the negative effects of confusing clinicians with the use of manifold decision cut-offs for one assay [36].

Similarly, we recently compared the 99th percentile cut-offs of 14 ng/L for the Roche hs-cTnT assay to sex-specific cut-offs (9 ng/L in females and 15.5 ng/L in males) in a population of 1282 patients with recent onset of chest pain suggestive of AMI [22]. We found that gender-specific cut-offs resulted in reclassification of only 3.5% of all chest pain patients. While AMI rates in females were marginally increased from 16.6% to 22.6% in the entire chest pain cohort, rates decreased in male patients from 23.1% to 21.1% [22]. In parallel, we observed an increase of 5.3% of female patients with an adjudicated final diagnosis of non-AMI now fulfilling AMI biomarker criteria indicated by an elevated value above the sex-specific 99th percentile in combination with a relevant kinetic change. This misclassification of non-AMI patients as AMI may lead to potentially harmful initiation of AMI guideline therapies including antithrombotic and early invasive treatment in female patients [22]. In addition, we did not observe any relevant benefit on outcomes by reclassification according to gender-specific cut-offs for the combination of male and female patients [22]. More intriguing, the prognostic reclassification even led to a negative impact on outcomes as assessed with net reclassification improvement in female patients. Thus, in accordance to the findings of Bohula May et al. for the hs-cTnI assay, our results did not support the use of gender-specific thresholds for detection of AMI and risk stratification.

In a large registry-based study on nearly 50,000 patients admitted to cardiac units in Sweden, the prediction of negative outcomes within 1 year using hs-cTnT 99th percentiles adapted according to gender was evaluated [37]. A diagnostic reclassification was observed for 3.0% of male patients and 8.4% of female patients when using gender-specific baseline hs-cTnT cut-offs instead of general cut-offs. Whether this translates into improved risk stratification was consequently investigated using multivariate Cox-regression analysis. However, in accordance with our results, the authors did not observe an improved prognostication by gender-dependent differentiation of hs-cTnT cut-offs for hs-cTnT [37].

Conversely, Shah et al. found that use of sex-specific 99th percentiles for the Abbott Architect hs-TnI assay doubled AMI diagnosis from 11% to 22% in female patients in comparison to a contemporary cTnI assay with a single cut-off of 50 ng/L [38]. Nevertheless, results were similar to our findings concerning diagnostic reclassification when the gender-specific 99th percentile for hs-cTnI was compared to the general hs-cTnI 99th percentile cut-off [22], [38]. In this setting, AMI rates in females increased only slightly from 16% to 22% [38], indicating that a large amount of the suggested under-diagnosis of AMI in women in this study was not due to gender-related differences but resulted from analytical differences between the high-sensitivity and the contemporary assay. Concerning outcomes, the combined endpoint of all-cause death and recurrent MI at 12 months was observed six-fold more frequently in females with hs-cTnI levels between the single (26 ng/L) and the gender-specific cut-off (16 ng/L) than in women with hs-cTnI levels below the gender-specific cut-off suggesting that more female patients at risk will be identified [38]. Similarly, a recent observational multicenter study showed improved identification of women at 1-year-risk for MACE (MI, emergency coronary revascularization or death) when sex-specific cut-offs are used, while no reclassification improvement was observed in the overall cohort for patients with or without MACE [39], [40]. Thus, the authors only recommend the use of gender-specific thresholds for prognostication in female patients [39], [40]. However, both latter studies did not report short-term outcome data, which might be more relevant to assess the immediate impact of diagnostic and therapeutic treatment of the index event on risks for negative outcomes.

Specific cut-off based on age or gender necessary for AMI diagnosis?

The apparent age- and gender-dependent differences in cTn levels in different populations including asymptomatic, healthy individuals but also symptomatic patients at risk for AMI question the appropriateness of solitary cTn cut-offs for medical decisions. However, despite this clear evidence of sex- and age-dependent differences in cTn levels, current studies investigating the diagnostic and prognostic impact of age- and gender-specific cut-offs observed very mixed results. Some studies concluded that age or gender may significantly influence diagnostic reclassification as well as risk assessment, thus resulting in potential AMI under-diagnosis in women and over-diagnosis in the elderly. Importantly, in both of these patient groups presentation to EDs is often ambiguous and differentiation of AMI from non-AMI causes even more challenging. On the other hand, results by other groups only observed very low to moderate reclassification rates when translating the findings to ACS and non-ACS patients presenting with acute symptoms to EDs. In addition, the influence on prognostication in these studies was only poor.

One of the major findings, which all studies investigating the diagnostic and prognostic impact have in common, is the fact that larger differences in cTn between specific cut-offs led to a more pronounced influence of either age or gender on diagnosis and outcomes. For example, in the study of Shah et al. using the Abbott ARCHITECT hs-cTnI assay, the difference between the gender-cut offs was 18 ng/L, resulting in a marked impact on prognostication [38], while in the gender differences for the Roche hs-cTnT assay no significant impact on risk stratification was 6.5 ng/L [35] was observed. In comparison, the difference for age-dependent hs-cTnT 99th percentile cut-offs, which in recent studies showed a relevant prognostic reclassification of outcome events, was slightly higher at 14 ng/L (14 ng/L for age<65 vs. 28 ng/L for age ≥65 years) [22], [41]. This illustrates that diagnostic reclassification and prognostic impact is highly influenced by the difference of the specific 99th percentile cut-off for the respective cTn assays [39]. As a consequence, the results cannot be generalized to all cTn assays, but should be assessed individually for each cTn assay [39]. Beside the fact that the implementation of gender and/or age-specific cut-offs (i.e. two to four new cut-offs) into the clinical routine of EDs may be impractical, further consideration of these specific-cut-offs for each cTn assay may be even more confusing to ED physicians [22], [39]. Moreover, the majority of studies which have described age- and gender-difference in 99th percentile values were conducted in populations which were biased by the lack of uniform criteria to define healthy populations, poor clinical phenotyping and differences in the relative age of the study participants compared to typical ACS risk populations. Thus, it is even unclear whether the observed 99th percentile cut-offs are appropriate as reference values.

Summary

In summary we conclude that there is currently no clear evidence for the use of age- and/or gender dependent 99th percentiles. Moreover, the use of different cut-offs in diverse populations such as age or gender groups is complex and acceptance by clinicians is anticipated to be low. In addition, before starting implementation of such complex recommendations, further prospective studies on the clinical significance need to address this important issue.