The guidelines of the Global Task Force for the Universal Definition of Myocardial Infarction (MI)1 and the complementary laboratory medicine practice guidelines of the National Academy of Clinical Biochemistry (NACB), both published in 2007, established cardiac troponin (cTn) as the standard biomarker for the diagnosis of MI. Ongoing improvements in technology over the past several years have advanced the analytical detection of low concentrations of cTn (I and T) in blood, serum, and plasma. “Contemporary” assays demonstrate excellent diagnostic accuracy for detection of MI by using the recommended 99th percentile concentration as the medical decision cutoff value in specimens obtained at presentation. Optimal accuracy is seen with these assays in samples obtained 6 or more hours after presentation. However, there are 2 important biomarker issues mentioned in these guidelines that will be critical in defining the next and possibly final generation of high-sensitivity (hs) cTn assays. (1) In both guidelines, total assay imprecision is recommended but not required to be ≤10% at the 99th percentile. As previously shown and discussed in the literature, cTn assay imprecision ≤20% CV at the 99th percentile value does not lead to substantial increases in false-positive or -negative misclassifications. (2) Whereas study of a normal reference population is advocated in both guidelines for calculation of the 99th percentile before assay implementation in clinical practice, the specifications of that population (age, sex, ethnicity, race) and the number of individuals that should be included in the “normal” group have not been clearly stipulated.
An opinion paper recently published in this journal suggested that hs-cTn assays need to have a total imprecision ≤10% at the 99th percentile and that >95% of apparently healthy subjects need to have measurable concentrations for the assay to be “guideline acceptable.” An assay with an imprecision between >10% and ≤20% was deemed “clinically usable.” It appears, however, that no real consensus exists on how to define a “high-sensitivity assay,” in analytical or clinical terms. With the introduction of the hs-cTnT assay in Europe and the likelihood that this same assay and possibly 1 or more hs-cTnI assays will be cleared by the US Food and Drug Administration (FDA) for use in the US, consistent criteria are needed for the assessment of hs-cTn assays to better define what constitutes a high-sensitivity assay. In addition, the impact of improving diagnostic sensitivity by lowering the concentration threshold needs to be carefully evaluated to address any analytical concerns that may potentially give false-positive results. With improvements to diagnostic sensitivity, the diagnostic clinical specificity will decrease, likely as low as 65% to 75%. As a result of the increased diagnostic sensitivity of the new hs-cTn assays, clinicians as well as laboratorians will need to be educated about the non–acute coronary syndrome (non-ACS) etiologies that can cause increases in cTn. Below, we discuss these critical issues with 4 experts in the field.
Can you provide what minimal criteria are required to determine the cTn 99th percentile reference value, addressing how many healthy people need to be studied and their age, sex, and ethnicity?
Alan Wu2 : I suggest that the guidelines established by the Clinical Laboratory Standards Institute (CLSI) for establishing reference limits for any clinical laboratory test be followed. At this time, there is no consensus on the need for establishing different 99th percentile values for troponin based on age, sex, or ethnicity. There is debate about whether apparently healthy subjects recruited for this determination should be young, old, or a mixture of both. In my opinion, the age range should be broad.
Paul Collinson3 : The minimum sample size will be determined by the ability to generate a 99th percentile limit with sufficiently small confidence intervals. With respect to defining age, sex, and ethnicity requirements, several statistical techniques would be used. Comparison of 2 populations of around 200 in each subgroup is likely to be sufficient to detect any significant differences in sex and ethnicity. Age effects would be determined by examining the strength of the association between age and the troponin values obtained. If a strong association were identified, then a larger sample size partitioned by age ranges would be required with direct comparison of the distributions. As an added complication, if the assay being evaluated produces the majority of its values as below the limit of detection of the assay or below the 10% CV concentration, the power to generate a reliable 99th percentile and to compare population subgroups will be substantially reduced. Finally, patient selection will be paramount. Two recent evaluations of the hs-cTnT assay have shown sex differences. Our own experiences show that, as in the case of cTnI, sex differences are abolished by patient selection approaches that exclude any possibility of underlying cardiac disease.
Allan Jaffe4 : Substantial effort should be devoted to rigorously defining reference values for the new “high-sensitivity” assays, also providing sex, age, and ethnically diverse ranges as needed. If age-adjusted ranges are needed, the total number of individuals studied will depend on how many age groups are defined. If decile age ranges of 30–39, 40–49, 50–59, 60–69, 70–79, and ≥80 years were defined for each sex, 120 subjects per decile might be sufficient, but, to reduce the size of the confidence interval around the 99th percentiles for each decile, some have argued for 320 individuals per decile. The real challenge of performing such a study is how to define and select healthy individuals. Usually, a simple history form is filled out. For the selection to be definitive, I would suggest a history, physical examination, electrocardiogram, and chest x-ray. I would exclude anyone with any cardiac history and anyone on any cardiac-related medication. Then, ideally, I would suggest cardiac magnetic resonance (cMR) imaging to eliminate those with otherwise undiagnosed cardiac disease. An echocardiogram would be second best. Such an approach, however, would be very costly. Therefore, assuming a good normal physical examination, a normal natriuretic peptide value corrected for age and sex might be sufficient. This appears daunting, but if a cooperative study were performed across countries, companies, and research funding sources, this would need to be done only once. Samples from such a study need to be banked and provided, in exchange for funding at the front end, to those who had financially sponsored the activity. Given the large number of facilities doing screening on normal subjects in the US, if one could leverage such resources, this approach might be even more cost effective.
David Morrow5 : I think that the characterization of the reference population is one of the most important issues that we face in going forward with newer generations of assays for troponin. For scientific purposes, there is a strong need to conduct reference testing in a very thoroughly characterized set of younger individuals of mixed sex and ethnicity that are completely free of any identifiable structural heart disease by using advanced imaging such as magnetic resonance imaging (MRI) and computed tomography (CT) angiography to examine both the myocardium and vessels—the “normal, healthy population.” At the same time, there is a growing perspective in the clinical community that we are in need of more complete data regarding the distribution of troponin concentrations among clinically stable individuals who may have cardiovascular risk factors and may be of advanced age but who have no acute cardiovascular disease. It is this population that will help us better understand what changes may constitute an acute alteration in patients presenting for emergency care.
Can you define the criteria you would propose to characterize a cardiac troponin assay as a high-sensitivity assay?
Alan Wu: Some scientific criteria are warranted for defining a troponin assay as high sensitivity. I would suggest that a high-sensitivity designation be given to commercial troponin assays that are able to detect the majority of healthy subjects as being above the limit of detection of the assay, when tested by following the CLSI C28-A3 protocol. The percentage of the values that are detectable can be debated. Fred Apple recently suggested such values for 3 generations of high-sensitivity assays, with percent of apparently healthy subjects detected being between 50% and 95%. Whatever approach is used, a discussion with the FDA is essential.
Paul Collinson: A true high-sensitivity assay would define the complete reference distribution. Such an assay would be able to fully define a reference population and detect a significant change within the reference distribution on serial sampling. Pragmatically, definition of 90% or better of a well-characterized reference population would suffice. The current assays detect <40% to 50% of a reference population and are therefore usable but do not meet the true criteria of a high-sensitivity assay. The inability to measure low-level values impairs the statistical power of test evaluations.
Allan Jaffe: Initially, I was impressed by the idea of defining this based on the number of normal subjects with a measured value, and I still think that idea has merit. However, I have been concerned that in some studies, such as those recently published in the New England Journal of Medicine, such assays did not appear to fare better clinically. This being the case, I would still use the number of healthy individuals with measurable concentrations of troponin as an index but would also want to see clinical studies in chest pain or ACS patients (which are the most frequent and thus easiest to study) showing increased sensitivity to detecting changing patterns and abnormal clinical status with the use of these assays.
David Morrow: First, from a standpoint of educating clinicians, I think it is of greater importance that clinicians gain a deeper interest and understanding of the performance characteristics of the assays [limit of detection (LOD), 99% percentile, lowest concentration at a 10% CV] in clinical use in their institution—“know your assay”—rather than applying a term of “high-sensitivity” to the assay. However, the current absence of consensus about the term presents challenges for the discussion. Therefore, I advocate a relatively simple approach to categorization, such as the one proposed by Fred Apple, that incorporates a 10% CV at or below the 99th percentile and a metric confirming that the assay is able to measure troponin at a low range, by use of either a reference population as suggested by Apple or an NIST standard reference material (SRM 2129). It may be that measurable concentrations in >95% of “normal subjects” is a higher bar than might be relevant to clinical practice as high sensitivity. If you ask the average clinician, most contemporary assays have already surpassed a bar of high sensitivity from a clinical perspective.
As hs-cTn assays are evaluated, what minimal analytical studies should be conducted to confidently alleviate the concern for potential false-positive or false-negative results?
Alan Wu: There are 2 types of evaluations for sensitivity and specificity. Diagnostic sensitivity and specificity assess the ability of the marker to detect a clinical entity. If the objective of troponin is to detect MI, the diagnostic sensitivity is 100% because it is part of the definition of MI. The diagnostic specificity will be less, as there are many other etiologies of myocardial injury, particularly minor ones. If the objective of measuring troponin is to detect myocardial injury, both the diagnostic sensitivity and specificity will be very high. The literature is full of clinical studies on the use of troponin for MI diagnosis and risk stratification, and additional studies may be unnecessary. Analytical sensitivity is the lowest concentration at which an assay can detect the analyte with some statistical degree of confidence. Analytical sensitivity can be determined by use of spiking studies in which the analyte of interest is added to an analyte-free matrix or by serially diluting a positive sample until the analyte is no longer detected. Total precision at low troponin concentrations is also important. Analytical specificity studies involve testing the assay for cross-reactivity against other proteins and cardiac markers. Although early assays suffered from nonspecific binding, this has largely been resolved. Analytic false positives exist only with the presence of abnormal antibodies (heterophile and human antianimal antibodies). Analytic false-negative results are possible with the presence of autoantibodies to troponin, but too few studies have been performed so far to adequately characterize autoantibody interference in commercial assays.
Paul Collinson: The current protocols to detect interference from cross-reaction, heterophilic antibodies, and other interference should be applied. The problem comes in the nature of the reference standard. On the basis of current history, it is more likely that the assays themselves will be more sensitive than techniques they are being compared with. It is difficult to define a false positive unless there is an independent outcome assessment. The best that can be achieved will be to compare against a panel of other tests supported by good-quality clinical assessment. The closest analogy is serological testing for hepatitis, where comparison is against a validated panel of positive and negative sera.
Allan Jaffe: Because minor analytical problems near the detection threshold could cause the determined concentrations to cross cutoff values of importance, the detection of even small amounts of analytical confounding in cTn assays is critically important. For example, minor differences demonstrated between serum, EDTA, and heparin plasma for the novel high-sensitivity Beckman assay may be of importance. Not only does this require us to lower our threshold for considering these effects important, but it would also include making sure that manufacturing is capable of producing consistent reagents free of large lot-to-lot variation. Such attention to minor assay variations is important for both those effects that might cause false positives (hemolysis for cTnI assays and heterophilic antibodies) and those that can cause false negatives, like antibodies to cTnI and the most recently reported antibodies to cTnT. An additional consideration is that if these assays have very high precision near the cutoff values, this may allow smaller (and more sensitive) thresholds for change to be used. Finally, the specificity studies done previously for cTnI and cTnT in cardiac disease should be redone with these increasingly sensitive probes.
David Morrow: The extent of nonspecific binding to other serum or plasma constituents, as well as the influence of matrix selection, must be investigated. However, there will be limits to what can be established analytically, as there are inherent challenges in reinventing the gold standard. Clinical investigations demonstrating prognostic relevance will be most convincing of clinical relevance to the practitioner.
Based on the preliminary clinical data in the literature using both hs-cTnT and hs-cTnI assays, what clinical studies should be conducted, at a minimum, to provide sufficient evidence-based support to confidently introduce these assays into clinical practice?
Alan Wu: It is necessary that a clinical study be conducted (and published ideally) for each commercial hs-cTnT and hs-cTnI assay that is used as a diagnostic marker and as a marker of risk stratification. Separate clinical studies may be necessary to address these separate questions. Owing to the lack of assay standardization, these studies are necessary to establish the 99th percentile cutoff and the diagnostic and prognostic performance of each assay at this particular cutoff.
Paul Collinson: The current published studies are very encouraging. Future evaluation requires similar studies to those that have already been published by using, as far as is possible, independent diagnostic criteria. Outcome studies comparing existing troponin assays and the high-sensitivity troponin assays are also required to see if there is incremental improvement in risk stratification, not just in the ACS population but rather in the general ischemic (chest pain) population where these tests will be used. In other words, does the detection of an increased troponin substantially improve discrimination? Such studies must address absolute, not relative, risk. The ability to distinguish a relative risk increase of 25% that translates into an absolute risk change of 0.25% is not clinically useful. The analogy is with statin treatment for primary prevention. One important question that needs to be answered is whether the detection of low-level troponin values should be used to guide intervention. This question can probably be answered by analysis of sample data banks taken as part of intervention studies or possibly revisiting existing trial sample data, but the same caveats apply as above—absolute vs relative risk modification. One key question that remains to be answered is the comparison of the high-sensitivity troponin assays with the other putative markers that have been suggested as alternatives to the original generation of troponin assays.
Allan Jaffe: The assays presently on the market have been validated particularly in patients with ACS. We know that patients with increased cTn values benefit from aggressive anticoagulation, aggressive antiplatelet therapies, and an early invasive strategy [usually percutaneous coronary intervention (PCI)]. Given that some patients with putatively stable coronary artery disease and those with other comorbidities, e.g., heart failure, can have increases in cTn even with contemporary assays, it is unclear whether patients with increased cTn by these new assays will manifest similar benefit from the above interventions. Studies to answer this question will be essential. In addition, both analytical and clinical studies will be necessary to define what constitutes a significant change in values, since it is likely that with these new assays, a changing pattern of values will be even more important than with contemporary assays, and how to define the minimum change that is clinically important is unclear. Finally, although studies often focus on diagnosing patients earlier, there is a lack of information about how many patients, if any, may manifest troponin increases only after many hours. Answering this question will require a separate study.
David Morrow: (1) Evaluation in a reference population. (2) Rigorous assessment of diagnostic performance in an wide-spectrum population with nontraumatic chest pain, in which the diagnosis of MI was adjudicated by clinical experts with the use of criteria that incorporate both the clinical presentation and biomarker results and for which other non-ACS etiologies of myocardial injury are recognized and captured. (3) Prognostic assessment in large, well-characterized cohorts of patients both with nontraumatic chest pain and separately those with a clinical diagnosis of ACS, with rigorous individual patient follow-up to at least 30 days, ideally 6–12 months, and endpoints adjudicated by a clinical events committee. Endpoints should include cardiovascular death, new or recurrent MI, and possibly recurrent unstable angina, and should not include revascularization procedures that may be driven by the initial troponin result. Completion of the first 2 sets of studies may be sufficient for introduction as a diagnostic test. However, all 3 lines of investigation will be needed to support clinical acceptance.
What advice would you provide to the FDA to assist in the 510K clearance process of hs-cTnT and hs-cTnI assays?
Alan Wu: Because the 99th percentile is used as the cutoff, it is critically important that manufacturers provide data on their reference value study and how that study was conducted. What criteria were used to determine health? What was the age, sex, and ethnicity distribution of the population used? This will particularly be important if an older population was used, given the prevalence of cardiovascular disease in that population. The FDA should also establish a protocol for the evaluation of analytic interferences, heterophile, human antimouse, and autoantibodies, so that each assay is evaluated consistently. There has been discussion of the creation of a universal healthy subject sample bank that contains aliquots and/or serum or plasma that manufacturers can use to establish their reference values. The addition of samples with abnormal antibodies to this bank would be helpful for interference studies. Such an approach would allow assays to be directly compared with each other.
Paul Collinson: This will require 2 types of study. Direct comparison with existing assays to show equivalence across comparable measuring ranges is required. The second type of studies is similar to those that have already been published that look at the prognostic increment of increased sensitivity in a well-defined ACS population (but see comments above).
Allan Jaffe: 510K clearance must critically evaluate assay performance around the important cutoff values. A close association overall is expected across assays, but it is assay performance in the vicinity of the cutoffs that distinguishes sensitive from insensitive assays. The Venge paper some years ago describing the differences between the Beckman, Roche, and Abbott assays is a prototype for the sort of clinical study that should be mandatory to allow a 510K clearance.
David Morrow: I am certain that my colleagues will comment on the requisites outlined in responses to questions 1–4, so I will focus on the major clinical struggle of selecting cutpoints for clinical practice. Professional society guidelines and regulatory labeling may sometimes differ for valid reasons, because their objectives are not completely aligned. However, discordance between product labeling and current professional guidelines has contributed to the present confusion by practitioners over appropriate clinical decision limits. The outdated use of MI cutpoints based on comparison with creatine kinase-MB has confused the field. An approach that incorporates complete reporting of assay performance [limit of blank (LOB), LOD, limit of quantitation (LOQ), 10% CV, 99th percentile) along with clinical data described above will best serve the practitioner.
Author Contributions: All authors confirmed they have contributed to the intellectual content of this paper and have met the following 3 requirements: (a) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (b) drafting or revising the article for intellectual content; and (c) final approval of the published article.
Authors’ Disclosures of Potential Conflicts of Interest: Upon manuscript submission, all authors completed the Disclosures of Potential Conflict of Interest form. Potential conflicts of interest:
Employment or Leadership: None declared.
Consultant or Advisory Role: F.S. Apple, Ortho Clinical Diagnostics, Abbott Laboratories, Beckman Coulter, and Sensera; A. Jaffe, Beckman Coulter, Siemens, Ortho Clinical Diagnostics, Inverness, Critical Diagnostics, Singulex, Nanosphere, Novartis, GSK, and Merck; D. Morrow, Beckman Coulter, Ortho Clinical Diagnostics, Roche, and Siemens.
Stock Ownership: None declared.
Honoraria: F.S. Apple, Ortho Clinical Diagnostics, Beckman Coulter, Biosite Inverness, and Roche.
Research Funding: F.S. Apple, Radiometer, Roche, Response, Siemens, Biosite, Nanosphere, and Abbott; A. Jaffe, Beckman Coulter and Siemens; D. Morrow, Beckman Coulter, Ortho Clinical Diagnostics, Roche, Siemens, Singulex, and Nanosphere.
Expert Testimony: None declared.
Role of Sponsor: The funding organizations played no role in the design of study, choice of enrolled patients, review and interpretation of data, or preparation or approval of manuscript.
↵1 Nonstandard abbreviations: MI, myocardial infarction; NACB, National Academy of Clinical Biochemistry; cTn, cardiac troponin; hs, high-sensitivity; FDA, US Food and Drug Administration; ACS, acute coronary syndrome; CLSI, Clinical Laboratory Standards Institute; cMR, cardiac magnetic resonance; MRI, magnetic resonance imaging; CT, computed tomography; LOD, limit of detection; SRM, standard reference material; PCI, percutaneous coronary intervention; LOB, limit of blank; LOQ, limit of quantitation.
↵2 Alan Wu, Department of Laboratory Medicine, University of California, and San Francisco General Hospital, San Francisco, CA.
↵3 Paul Collinson, Consultant Chemical Pathologist, St. George’s Hospital, London, UK.
↵4 Allan Jaffe, CV Division, Department of Medicine, and CCLS Division, Mayo Clinic, Rochester, MN.
↵5 David Morrow, Cardiovascular Division, Department of Medicine, Brigham and Women’s Hospital, Boston, MA.
- © 2010 The American Association for Clinical Chemistry