Accuracy of biomarkers to diagnose acute cardiac ischemia in the emergency department: A meta-analysis

Abstract

Study Objective: We sought to evaluate quantitatively the evidence on the diagnostic performance of presentation and serial biochemical markers for emergency department diagnosis of acute cardiac ischemia (ACI), including acute myocardial infarction (AMI) and unstable angina. Methods: We conducted a systematic review and meta-analysis of the English-language literature published between 1966 and December 1998. We examined the diagnostic performance of creatine kinase, creatine kinase-MB, myoglobin, and troponin I and T testing. Diagnostic performance was assessed by using estimates of test sensitivity and specificity and was summarized by summary receiver-operating characteristic curves. Results: Only 4 studies were found that evaluated all patients with ACI; 73 were found that focused only on a diagnosis of AMI. To diagnose ACI, presentation biomarker tests had sensitivities of 16% to 19% and specificities of 96% to 100%; serial biomarker tests had sensitivities of 31% to 45% and specificities of 95% to 98%. Considering only the diagnosis of AMI, presentation biomarker tests had summary sensitivities of 37% to 49% and summary specificities of 87% to 97%; serial biomarker tests had summary sensitivities of 79% to 93% and summary specificities of 85% to 96%. Variation of test sensitivity was best explained by test timing. Longer symptom duration or time between serial tests yielded higher sensitivity. Conclusion: The limited evidence available to evaluate the diagnostic accuracy of biomarkers for ACI suggests that biomarkers have very low sensitivity to diagnose ACI. Thus, biomarkers alone will greatly underdiagnose ACI and will be inadequate to make triage decisions. For AMI diagnosis alone, multiple testing of individual biomarkers over time substantially improves sensitivity, while retaining high specificity, at the expense of additional time. Further high-quality studies are needed on the clinical effect of using biomarkers for patients with ACI in the ED and on optimal timing of serial testing and in combination with other tests. [Balk EM, Ioannidis JPA, Salem D, Chew PW, Lau J. Accuracy of biomarkers to diagnose acute cardiac ischemia in the emergency department: a meta-analysis. Ann Emerg Med. May 2001;37:478-494.]

Introduction

The original 1997 report of the National Heart Attack Alert Program Working Group summarized data from studies on creatine kinase (CK) and CK-MB; however, there were few studies on other biomarkers.¹ Many studies on CK, CK-MB, myoglobin, and troponins T and I have been published since.

As noted in the original report, biomarkers were developed for the diagnosis of acute myocardial infarction (AMI) and were not intended for the overall diagnosis of acute cardiac ischemia (ACI), which includes unstable angina. Biomarkers detect evidence of necrotic myocardium but not noninfarction ACI (unstable angina), which is an important reason for appropriate hospitalization and further diagnostic evaluation. For this reason, very few biomarker studies report test performance data for ACI.

The present systematic review and meta-analyses aimed to characterize the test performance of these biomarkers to diagnose ACI, both AMI and unstable angina, in the emergency department setting. However, the vast majority of evidence found focused on AMI alone. Because there were few studies that addressed ACI in the ED and to provide a more complete picture of the diagnostic performance of biomarker tests, we also included studies that assessed AMI diagnosis only. In addition, we included studies that carried out testing in the ED; many of these, however, analyzed only hospitalized patients. When possible, we restricted analysis to studies that included all ED patients (including those discharged from the ED).

We included relevant articles found in MEDLINE through December 1998, with additional articles known to our domain experts. The methods for the systematic review, meta-analysis, general study selection criteria, and characterization of population category and study quality assessments are presented in the accompanying synopsis of the evidence report on ACI.²

We investigated the use of both presentation and serial testing (single and multiple serum samples, respectively) for each of the biochemical tests reviewed (CK, CK-MB, myoglobin, and troponin I and T tests). For each test, we describe the excluded studies and the reasons for their exclusion. We summarize the eligible articles, including study eligibility criteria, ACI prevalence, outcome (AMI, unstable angina, or both) definition, and thresholds used to define positive test results. When meta-analysis was feasible, we report combined estimates of the test performance for each of the tests. In our primary analysis, we included all studies in which laboratory testing was performed in the ED setting; however, when possible, we also analyzed only those studies that included all eligible ED patients (ie, those in which patients discharged from the ED were also included).

Table 1 presents the study characteristics and test performance for studies of ACI. Tables 2 through 13 present the characteristics and test performance for studies that evaluated each of the biomarkers for the diagnosis of AMI only.

Tables 14 and 15 present summaries of AMI-only studies; they include the total number of studies analyzed, the total number of patients included, and the diagnostic performance (test sensitivity and specificity) of each test. Results of analyses are presented for all included articles and, when feasible, for the subgroup of studies that included all ED patients. Further detailed analyses, including diagnostic odds ratios (which allow direct comparisons between diagnostic tests), are reported in the 2000 updated National Heart Attack Alert Program report.³Figures 1 through 4 present summary receiver-operating characteristic curves for specific tests.

Among studies in each of the biomarker groups, there was a large degree of heterogeneity in eligibility criteria. As discussed in greater detail in the accompanying article,² this resulted in a wide range of both ACI and AMI prevalence. Across all studies, ACI prevalence ranged from 20% to 75%, and AMI prevalence ranged from 1% to 78%. In addition, individual studies used many different criteria to define test positivity and timing of serial testing (Table 1, Table 15). We examined all these factors to explain variability in study results.