High-Sensitivity Troponin and the Selection of Patients for Cardiac Imaging in the Outpatient Clinic

Suspected stable angina is a common presenting complaint, and the process of establishing the etiology of these symptoms consumes substantial diagnostic resources. Perhaps the greatest challenge relates to individuals with low to intermediate pretest probability for obstructive coronary artery disease. In these patients, additional noninvasive testing—either functional or anatomic—is frequently required to provide patients and clinicians alike with sufficient reassurance that invasive coronary angiography can be deferred. Demand for further investigation continues to grow despite falling baseline risk, and as a consequence, there has been a marked reduction in the prevalence of functionally significant coronary artery disease in patients undergoing stress imaging in recent years (1). Clearly then, there is a need to improve the selection of patients for whom further testing is necessary, and cardiac troponin appears a suitable candidate for such an indication. High-sensitivity assays allow accurate quantification of plasma or serum troponin concentrations in the majority of adults, and numerous studies have demonstrated the prognostic importance of troponin in primary prevention populations (2, 3) and among patients with a variety of stable (4, 5) and acute (6) cardiac disorders. Furthermore, when compared to the reference standard of obstructive disease identified on coronary computed tomography angiography, high-sensitivity troponin I assays have been shown to improve the accuracy of existing risk models for the estimation of pretest probability (7, 8). Within this context, Mueller et al. report in this issue of Clinical Chemistry on a large, well-conducted investigation that reinforces the potential diagnostic value of troponin in the setting of suspected inducible myocardial ischemia (9). An important strength is the use of the most widely available high-sensitivity troponin (hs-cTn) I and T assays with blinded classification of functionally relevant coronary artery disease (fCAD)4, determined from myocardial perfusion imaging with single-photon emission computed tomography as the reference standard (9). Although not a novel finding, the comparatively modest correlation (r = 0.62) between hs-cTnI and hs-cTnT remains an intriguing one and implies each assay may be offering different insights into cardiovascular pathology. Nevertheless, both assays demonstrated similar overall discrimination, with a c statistic of 0.68 (hs-cTnI) and 0.66 (hs-cTnT). Similarly, the authors confirmed that both assays provide a small but statistically significant increase in diagnostic accuracy when added to the existing coronary artery disease consortium model (CADC). This is the scoring system currently recommended by the European Society of Cardiology guidelines for the estimation of risk among patients presenting with suspected stable angina. The CADC risk model is used to identify individuals with a pretest probability of <15% who may have further diagnostic testing safely deferred (10). Mueller et al. then attempt to bring troponin testing in this setting closer to clinical application by deriving possible rule-in and rule-out thresholds with sufficient diagnostic sensitivity and specificity to satisfy the European Society of Cardiology guideline risk threshold. Interestingly, as with the universal definition of myocardial infarction, they suggest that sex-specific troponin cut points may be preferable with lower rule-out limits for women and higher rule-in limits for men. Finally, they conclude by describing the now well-established association between cardiac troponin concentrations and future cardiovascular events. The authors should be commended on undertaking a large prospective study and a robust comparison of the 2 most widely used high-sensitivity assays in clinical practice. However, in considering how we might apply these findings to current clinical practice several caveats are worth noting. First, this was not an “all-comers” population. For inclusion, patients must have been referred for evaluation of fCAD by rest/stress myocardial perfusion imaging with single-photon emission computed tomography at a single tertiary-level hospital. Furthermore, 45% of patients consented for the study were excluded from the primary analysis, predominantly owing to the requirement for pharmacological stress testing or the presence of established structural heart disease, which may have increased troponin concentrations in a manner unrelated to myocardial ischemia. These exclusions do not apply to the CADC risk model, and therefore, the findings reported here may not be directly applicable to a substantial proportion of patients seen in an outpatient practice. The combination of a selected cohort and the lack of external validation detracts from the generalizability of the study conclusions and the authors' suggestion that the results of this study make troponin testing an “attractive option in the clinical setting of a general practitioner” (9). Second, the present study cohort comprises a high-risk cohort. Almost one-half of the analysis population had an established diagnosis of coronary artery disease, with nearly a third having experienced a prior myocardial infarction. The composition of the study cohort matters because the baseline prevalence of coronary disease has important consequences for the diagnostic performance of the troponin assays. The ability of a biomarker to discriminate between cases and noncases depends on the spread of risk among the population investigated. Likewise, the prevalence of disease among a population directly influences the sensitivity and specificity of a diagnostic test (11), while even with a given sensitivity and specificity any decrease in pretest probability will result in a lower positive predictive value offset by a higher negative predictive value at each troponin cutpoint. Third, the troponin concentrations identified as suitable to rule out the need for further noninvasive testing are at the limits of assay performance. The limit of detection for a biochemical assay describes the concentration of an analyte that can be reliably distinguished from a blank sample—i.e., a concentration that is not zero. Even at these very low limits, the hs-cTnT assay achieved a diagnostic sensitivity of only 82%, corresponding to a negative predictive value of 82%, insufficient to meet current European (10) or North American (12) guideline standards for deferring additional noninvasive testing. Biological and analytical variability is substantial relative to these very low concentrations. The same research group has previously established that troponin concentrations increase in response to exercise regardless of the presence or absence of inducible ischemia (13), and such variation creates the very real risk of misclassification when a single measurement for clinical decision-making is used. Fourth, having established the complementarity of troponin with the CADC model when performing the receiver operating characteristics analysis, the authors discard this prior knowledge and report the diagnostic performance of troponin at several concentration thresholds without regard to the information provided by this well-established, albeit imperfect, risk score. It is likely that troponin concentrations will have a differential diagnostic significance according to an individual's baseline risk factor profile and, although not reported, it is reasonable to assume that at least some of the 10%–15% of patients with fCAD ruled out by merit of very low troponin concentrations would have been appropriately identified at low risk on the basis of the CADC score alone. The interaction between baseline risk factors and troponin is nicely demonstrated within the article itself, in which a lower rule-out threshold is required to achieve comparable diagnostic sensitivity between women and men, while even then the higher prevalence of fCAD among men results in a lower negative predictive value. In truth, physicians are unlikely to consider a biomarker in isolation when determining whom to investigate further; unfortunately, the authors have missed an opportunity to describe how we might combine clinical assessment with troponin testing to optimize diagnostic testing strategies. Finally, by attempting to define optimal diagnostic troponin thresholds, the authors fall into the oft lamented trap of dichotomania, the arbitrary categorization of continuous variables (14). There is no troponin concentration threshold of risk in the setting of suspected acute coronary syndrome (6), and it seems similarly improbable that one should exist in a stable outpatient population. Although conceptually simple, dichotomizing may lead to misleading results as it implies diagnostic certainty that does not exist (15). If the authors' intention is to encourage clinical application of their findings (accepting our earlier cautions regarding lack of external validation), clinicians should be given access to a risk calculator that incorporates the established CADC model along with the troponin concentration and returns the estimated probability of fCAD along with confidence intervals. In the age of smartphone ubiquity, such an approach seems eminently feasible and wholly preferable to oversimplified binary decision-making thresholds. In summary then, it is increasingly clear that hs-cTn provides important information regarding pathophysiological processes within the cardiovascular system. Furthermore, although troponin concentrations correlate with established cardiovascular risk factors such as age and sex, the diagnostic and prognostic information they provide is independent of these characteristics. However, although Mueller et al. have added to the observational evidence base, the appropriate clinical application of this information in the setting of suspected stable angina remains unclear. What is now required is a randomized diagnostic intervention trial in which such information is applied in a prospective manner with clinically relevant endpoints. Healthcare funders may want to consider supporting such an initiative since given the ever-expanding demand for noninvasive testing in this setting, the potential for cheap and widely available tests to reduce unnecessary testing could prove highly cost-effective. functionally relevant coronary artery disease high-sensitivity cardiac troponin coronary artery disease consortium.