Validation of an accelerated high-sensitivity troponin T assay protocol in an Australian cohort with chest pain

Elevation of cardiac troponin (cTn) levels is central to the definition of acute myocardial infarction (AMI).1 Increasingly sensitive cTn assays offer the potential for AMI to be diagnosed earlier than the time points recommended in established clinical guidelines.2 Earlier diagnosis allows for more timely initiation of therapy for those with AMI and earlier rule-out of AMI for others. This is relevant to overcrowded hospital emergency departments, where chest pain remains one of the most common reasons for presentation.3 In Australia, time-based political imperatives such as the National Emergency Access Target have further increased the need for efficient yet safe means of assessing patients with chest pain.4

No universally agreed nomenclature of cTn assays exists. It is generally held that a highly sensitive assay should have a coefficient of variation (CV) of < 10% at the level of the 99th percentile of a healthy population and should be able to measure cTn in > 50% of such a population. Recent guidance includes preliminary advice regarding the appropriate use of highly sensitive assays.2 One such assay has been in clinical use in Australia since 2011 and although its uptake has been brisk, a corresponding change in the diagnostic approach according to this guidance has been variable. The reasons for this are unknown but may include the reference to a diagnostic cut-off that is at odds with the universal definition of myocardial infarction,5,6 reluctance to change long established practice, or limited local clinical outcome data using this assay.

Our primary aim was to examine the diagnostic accuracy of the high-sensitivity cardiac troponin T (hs-cTnT) assay for the diagnosis of AMI, using serial measurements at 0 h and 2 h after presentation, in an unselected, well characterised cohort of patients with chest pain. Our secondary aim was to examine the accuracy of predicting 30-day major adverse cardiac events (MACE) using a model incorporating 0 h and 2 h hs-cTnT assay measurements and risk stratification according to the guidelines of the National Heart Foundation of Australia and Cardiac Society of Australia and New Zealand (NHFA/CSANZ).7

This was a prespecified substudy of a prospective, observational cohort study conducted between November 2008 and February 2011. Consecutive eligible patients presenting during office hours to a single, large, metropolitan tertiary hospital emergency department with symptoms suggestive of cardiac chest pain were recruited. Inclusion criteria were age ≥ 18 years, and at least 5 minutes of possible cardiac symptoms where the attending physician intended to perform serial cTn tests. We used American Heart Association definitions for possible cardiac symptoms (ie, acute chest, epigastric, neck, jaw or arm pain; or discomfort or pressure without an apparent non-cardiac source).8 Patients were excluded for any of the following: a clear cause of symptoms other than acute coronary syndrome (ACS); inability or unwillingness to provide consent or be contacted after discharge; recruitment considered inappropriate by staff (eg, palliative treatment); interhospital transfer; pregnancy; and previous enrolment. Perceived high risk of ACS was not an exclusion criterion. Patients with missing blood samples were also excluded. The study was carried out according to the principles of the Declaration of Helsinki, approved by the Royal Brisbane and Women’s Hospital Human Research Ethics Committee (2008/101), and registered with the Australian New Zealand Clinical Trials Registry (ACTRN12610000053022). Written informed consent was obtained.

There were 1495 patients who met the initial inclusion criteria. Of these, 517 were subsequently excluded and 214 did not have serial blood samples stored. Characteristics of the remaining 764 patients are shown in Box 1.

There were 141 major cardiac events in 83 patients (10.9% of the cohort) within 30 days of presentation (Box 2).

Box 3 describes the accuracy of the hs-cTnT assay at 0 h, 2 h and 0 h or 2 h in diagnosing AMI within 24 h of presentation (there were no cardiac deaths within 24 h), compared with the adjudicated end points. Compared with the reference standard using cTnI measurement at 6 h, the hs-cTnT assay resulted in two false negative cases (Box 4). Both these patients had an adjudicated diagnosis of NSTEMI in the context of a prior history of coronary artery disease.

NHFA/CSANZ risk stratification incorporating the hs-cTnT assay at 0 h and 2 h performed similarly to the conventional strategy incorporating cTnI testing at 0 h and 6 h (Box 5). One patient with an adjudicated diagnosis of NSTEMI was classified as intermediate risk using the hs-cTnT assay but high risk using the 6 h cTnI assay; and one patient with UAP was classified as high risk using the hs-cTnT assay but intermediate risk using the 6 h cTnI assay (Box 6). No patients were incorrectly classified as low risk using the 2 h strategy when compared with the standard 6 h strategy.

Most of the 123 patients who had elevated hs-cTnT assay results and an adjudicated diagnosis other than AMI had some other significant cardiac diagnosis (Box 7).

Exclusion of a diagnosis of AMI based on clinical grounds alone has been shown to be unsatisfactory, and discharge with a missed AMI carries a poor prognosis.10 Therefore, a period of clinical observation and serial testing has been recommended for those at intermediate risk of ACS.7 Critically, we have shown that the negative predictive value of serial measurements with the hs-cTnT assay at or below the 99th percentile in the 2 h after presentation excluded a diagnosis of AMI with a very high level of certainty (over 99%).

Two patients with an adjudicated end point of NSTEMI had negative hs-cTnT assays at 0 h and 2 h, yielding an overall sensitivity of 96%. One patient had early objective investigation and was found to have reversible ischaemia on non-invasive testing. The other patient would have been considered at high risk of ACS due to a prior history of coronary artery bypass surgery. In keeping with current guidelines, further investigation would have been recommended for both patients. These cases clearly illustrate that cardiac biomarkers alone cannot be relied on for diagnosis of patients presenting with chest pain. The results of any biomarker strategy must be interpreted in the context of a recognised systematic approach that incorporates other clinical factors.

We also found that the strategy of early serial testing with the hs-cTnT assay performed well when incorporated into the model of risk stratification for MACE recommended by the NHFA/CSANZ,7 compared with conventional serial testing over a longer period. Although more patients were considered to be at high risk using the hs-cTnT assay (277 v 255), the 30-day MACE rates in high-risk patients were comparable using either assay (20% with hs-cTnT v 22% with cTnI). Only two patients with 30-day MACE were reclassified when the hs-cTnT assay was incorporated into risk stratification. Importantly, no patients classified as low risk using either assay had a major cardiac event during the 30-day follow-up period.

As expected, an increase in sensitivity of the troponin assay led to some corresponding fall in specificity. Nevertheless, most patients with an elevated hs-cTnT assay result had a significant cardiac diagnosis even in the absence of AMI. Thus, even when using a highly sensitive assay, a level above the 99th percentile is rarely insignificant.

Our study has some limitations. First, only patients presenting with chest pain were recruited, so our results cannot be extrapolated to other modes of presentation of ACS, such as dyspnoea or syncope in the absence of chest pain. Also, consecutive recruitment was performed only within office hours, but this is unlikely to have had any significant effect on our findings.

Second, an emerging concept is that the diagnostic performance of high-sensitivity troponin assays can be optimised by a more sophisticated approach to their interpretation that incorporates delta values as well as absolute cut-offs. One recent study using early biomarker testing supports such an approach to enhance the accuracy of the test.11 Although we did not address this hypothesis in our analysis, it is included in a larger multicentre study that is underway.

Finally, we did not examine the performance of the hs-cTnT assay in line with guidance regarding the appropriate timing of measurement and magnitude of delta troponin required to diagnose AMI, as suggested in the 2011 addendum to the NHFA/CSANZ guidelines.2 Our study commenced before this publication appeared, and the decision to sample cTn at 2 h after presentation was based on data suggesting a high degree of accuracy of sensitive cTn assays12 and our previous studies of accelerated diagnostic protocols for chest pain using other biomarker strategies at this time point.13,14 Additionally, the NHFA/CSANZ recommendations regarding interpretation of changing cTn levels are based largely on studies of assay performance and biological variability in healthy populations, and they await confirmation in studies based on clinical outcomes.15 Nevertheless, we cannot exclude the possibility that later blood sampling would have altered the performance of the assay in our cohort.

In summary, we have shown that, for patients presenting with chest pain, the hs-cTnT assay measured at presentation and 2 h later performed well in the initial diagnosis of AMI. We also found it is effective when incorporated into a comprehensive risk stratification strategy that includes clinical evaluation and early non-invasive testing for ischaemia, where appropriate. The time saved by this approach is considerable and, given the burden of patients presenting with chest pain, could have a major impact on health service delivery. Translation into improved patient outcomes or significant cost savings remains an important area for further research. However, our findings provide validation in an Australian cohort of the hs-cTnT assay that is already in widespread clinical use.