Objective To examine the diagnostic accuracy of novel biomarkers of myocardial injury and troponin assays for diagnosis of myocardial infarction.
Methods 850 patients randomised to the point-of-care testing arm of the Randomised Assessment of Panel Assay of Cardiac markers (RATPAC) study in six emergency departments of low-risk patients presenting with chest pain were studied. Blood samples were obtained on admission and 90 min from admission. Myocardial infarction was defined by the universal definition of myocardial infarction. The following diagnostic strategies were compared by receiver operator characteristic curve analysis and comparison of area under the curve: individual marker values and the combination of presentation heart fatty acid binding protein (HFABP) and copeptin with troponin.
Results 68 patients had a final diagnosis of myocardial infarction. Admission samples were available from 838/1132 patients enrolled in the study. Areas under the curve were as follows (CIs in parentheses): cardiac troponin I (cTnI) Stratus CS 0.94 (0.90 to 0.98), cTnI Beckmann 0.92 (0.88 to 0.96), cTnI Siemens ultra 0.90 (0.85 to 0.95), cardiac troponin T high sensitivity 0.92 (0.88 to 0.96), HFABP 1 0.84 (0.77 to 0.90) copeptin 0.62 (0.57 to 0.68). HFABP and copeptin were diagnostically inferior to troponin. The combination of HFABP (at the 95th percentile) and troponin (at the 99th percentile) increased diagnostic sensitivity.
Conclusions High-sensitivity cardiac troponin is the best single marker. Addition of HFABP to high-sensitivity troponin increased diagnostic sensitivity. Additional measurement of copeptin is not useful in the chest pain population.
Statistics from Altmetric.com
Measurement of cardiac troponin as cardiac troponin T (cTnT) or cardiac troponin I (cTnI) is now integral to the diagnosis of acute myocardial infarction in patients presenting with acute chest pain or suspected acute coronary syndrome (ACS). The universal definition of myocardial infarction recommends that the 99th percentile is the decision limit and that this should be measurable with an assay imprecision of 10% or less. A new generation of cardiac troponin assays which meets the analytical goals has been developed. Studies of the diagnostic performance of these assays have shown that they are superior to the existing troponin assays. It has also been suggested that admission measurement alone may be sufficient.1
There are a number of limitations in the studies published to date. The populations examined have not been representative of the emergency department (ED) population where biomarkers are of most value. Patients studied have included those with ST segment elevation myocardial infarction (STEMI) and patients with high-risk changes in the ECG which suggest non-STEMI (NSTEMI). Additionally, the biochemical gold standard used has largely been the previous, less sensitive versions of the troponin assays which do not properly define a 99th percentile.
Novel biomarkers, such as heart fatty acid binding protein (HFABP) and copeptin, have also been proposed as diagnostic tests for patients presenting with chest pain. The Randomised Assessment of Panel Assay of Cardiac Markers (RATPAC) Trial was a prospective randomised controlled trial comparing point-of-care testing (POCT) with conventional management.2 The population studied was the low-risk chest pain population in whom biomarker measurement is integral to clinical decision making. Samples from this study have now been analysed to examine the diagnostic performance of the new-generation high-sensitivity troponin assays, plus the new candidate cardiac biomarkers HFABP and the neuropeptide copeptin.
Ethical permission for the study was obtained from the Leeds East Research Ethics Committee (07/Q1206/22) and the study was performed in accordance with the declaration of Helsinki.
Full details of the RATPAC trial had been published.2 The trial (ISRCTN37823923) (30 January 2008–2 June 2009) randomised low-risk patients presenting with chest pain to either diagnostic assessment by a cardiac panel measured by POCT or to diagnosis when biomarker measurement was based on central laboratory testing (CLT). Patients 18 years or older, presenting with acute chest pain to the ED of six participating hospitals were screened for eligibility. Exclusion criteria for enrolment were; ECG changes for myocardial infarction or high-risk ACS (>1 mm ST deviation or >3 mm inverted T waves), known coronary heart disease presenting with prolonged (>1 h) or recurrent episodes of cardiac-type pain, proven or suspected serious non-coronary pathology (eg, pulmonary embolus), comorbidity or social problems that require hospital admission, an obvious non-cardiac cause (eg, pneumothorax or muscular pain), more than 12 h since their most significant episode of pain, previous participants, those unable to understand the trial information and those unwilling to consent.
All those eligible for enrolment were then randomised to either the POCT or CLT arm. Patients randomised to POCT were scheduled to have a blood sample drawn on admission and at 90 min from admission for POCT measurement of the cardiac biomarkers cTnI, the MB isoenzyme of creatine kinase and myoglobin. An additional sample for subsequent biomarker measurement was drawn at the same time as the POCT sample, and the serum separated and frozen prior to −20°C prior to transfer to long-term storage at −70°C in the central laboratory. The admission and 90 min samples were subsequently analysed for cTnI (two high-sensitivity methods) cTnT (high sensitivity), HFABP and copeptin. Sample flow is summarised in figure 1 and patient characteristics in table 1.
The patient characteristics of the original data and the biomarker subset were not statistically significantly different.
Final diagnostic classification was performed by two independent clinicians with access to all the relevant information, using the 99th percentile value for the cardiac troponin value from POCT measurement, from the local laboratory and from cTnI measurements (Siemens ultra) performed in the central laboratory. All patients with a final diagnosis of AMI had at least one troponin measurement performed at >6 h from admission. All patients had POCT measurement with a cTnI method which meets current analytical goals. Four of the local laboratories used a troponin method which meets the current analytical goals for the 99th percentile, one used a cTnI method which just fails to reach these goals, and one used the current generation cTnT method. Central laboratory measurements were performed using a cTnI method that meets current analytical goals. Full details of the analytical methods are included in the online supplementary data.
Patients with a troponin rise and a final diagnosis other than ACS or AMI were reviewed to decide whether AMI was the most likely diagnosis. Disagreements were resolved by discussion, and patients classified as having AMI or no AMI. Patients were categorised as AMI (type 1 AMI, primary ischaemic cardiac injury), patients with troponin elevation not due to AMI but with a probable background of underlying coronary atheroma (type 2 MI, secondary ischaemic cardiac injury) and those with no myocardial injury. Initial, final and review diagnoses are summarised in table 2. In the final and review diagnosis, the patients initially considered to have suspected AMI were classified as AMI or not. Patients with a final diagnosis of type 2 MI were classified as no AMI for the purpose of analysis and included patients with a final diagnosis of angina (3) and other (11).
Research nurses used ED and hospital inpatient notes to identify subsequent attendances, admissions and adverse events up to 3 months. Participants were mailed a questionnaire at 1 and 3 month intervals to identify adverse events and hospital attendances. In the POCT arm there were 36 patients with events (3%) as follows: death 6 (1%) non-fatal myocardial infarction 5 (<1%); hospitalisation for ACS (without myocardial infarction) 18 (2%); life-threatening arrhythmia 6 (1%); emergency revascularisation 10 (1%). Event rates between the POCT arm and CLT arm were not statistically significantly different.
Demographics and patient characteristics were analysed by non-parametric statistics. Diagnostic test comparison was performed using AMI as the dichotomous variable. Individual markers were examined by the construction of receiver operating characteristic (ROC) curves and calculation of the area under the curve (AUC). The diagnostic strategies examined were presentation value alone, and the peak of the presentation or 90 min value to reflect the impact of repeat sampling where initial diagnosis was uncertain. Additionally, as HFABP is considered to be a sensitive early biomarker of cardiac injury, the impact of time of onset of chest pain was examined. The patient cohort was split into two groups corresponding to time from symptom onset to presentation of <3 h and <6 h.
The use of prespecified diagnostic thresholds for dichotomous classification was examined for individual markers, and combinations of markers were examined by calculation of sensitivity and specificity and by construction of 2×2 tables with comparison by Fishers exact probability test. To provide an objective independent assessment of the biomarkers, the ability of admission and peak values to predict the major adverse cardiac events of death, myocardial infarction, cardiac readmission or need for revascularisation within the 3 months follow-up period was examined.
All statistical analyses were performed using Analyse-it for Microsoft Excel (V.2.30), (http://www.analyse-it.com).
Median time of onset of chest pain to sampling was 220 min (range 0–8207, lower quartile 155, upper quartile 345 min); 284/850 (33.4%) of patients had less than 3 h from onset of chest pain to first sample, and 640/850 (75.3%) less than 6 h from onset to the first sample.
Overall, cardiac troponin (cTn) measurement (cTnT or cTnI) was diagnostically superior to HFABP and copeptin when measured on admission (figure 2 for admission values) or as the peak of the admission or 90 min sample (see online supplementary figure S1) as assessed by comparison of area under the ROC curve (see online supplementary tables S1 and S2).
Comparison of the admission and peak values showed a statistically significantly increased AUC for peak values for cTnI CS (0.98 peak vs 0.94 admission, p=0.023) and for the Siemens Ultra assay (0.94 vs 0.90, p=0.037) but not for the Beckman cTnI, or the cTnT assay. The HFABP and copeptin assays showed a reduction in diagnostic accuracy when admission value was compared with peak value (HFABP 0.84 vs 0.82, p=0.0037, copeptin 0.62 vs 0.56, p=<0.0001).
Dividing the cohort according to duration of chest pain did not alter the areas under the ROC curves but appears to improve the diagnostic performance of HFABP but not copeptin. For patients with less than 3 h duration of chest pain for samples measured on admission, differences between the AUC were no longer significant between HFABP and the Stratus CS cTnI, Siemens Ultra cTnI assay and Beckman cTnI assay, but remained so for the Roche cTnT assay (see online supplementary figure S2, and supplementary table S3). This was not true for the peak value (see online supplementary figure S3, supplementary table S4). For peak values in patients with less than 3 h duration of chest pain, all four troponins were diagnostically superior to HFABP. For patients with less than 6 h of chest pain (see online supplementary figures S4 and S5, supplementary tables S5 and S6) on admission only the Beckmann cTnI and Roche cTnT showed superior performance (although the Stratus CS cTnI just failed to reach significance), whereas for peak values, all the troponins were superior to HFABP. The differential findings may represent true differences in early detection but may also be an artefact of smaller sample numbers for patients with less than 3 and <6 h duration of chest pain.
The following diagnostic strategies were also compared. Individual marker values for cTnI and cTnT >99th percentile as recommended by the universal definition of AMI and for HFABP >95th percentile (2.5 mg/L) of a reference population. For copeptin, a decision limit of >7.4 mg/L was used. This was obtained from ROC analysis as all values obtained were below the 95th percentile). The combination of troponin plus HFABP or copeptin, were also examined. In this case, if one marker was above the diagnostic discriminant for that marker, the test was considered to be positive. The result of a combination of markers was compared with the ability of a peak of troponin value above the 99th percentile (any cTn value on admission or at 90 min from admission exceeding the 99th percentile). These results are summarised in online supplementary table S7. Measurement of HFABP and copeptin on admission were diagnostically less sensitive than cTn measurement although this was only statistically significant for copeptin. The addition of either HFABP or copeptin to admission cTn measurement improves the diagnostic sensitivity although this did not achieve statistical significance.
Comparing MACE rates at 3 months, there was no significant difference between the AUC for all troponin methods whether measured on admission or as the peak value. All the troponin methods measured on admission or as the peak value had significantly higher AUCs than copeptin (admission troponin AUCs 0.73–0.83, copeptin AUC 0.58, peak troponin AUCs 0.78–0.86, copeptin 0.55). On admission, the AUCs for the Beckmann (0.83) and Siemens (0.80) cTnI measurements were significantly better than that of HFABP (0.72). For the peak values, all the troponin I methods, but not troponin T, give significantly higher AUCs than HFABP (see online supplementary figures S6 and S7 and supplementary tables S8 and S9).
This study examined the diagnostic efficiency of the current-generation cTn assays and two novel biomarkers in patients presenting without ECG evidence of ischaemia. The measurement of cTn was the single best test, but admission measurement alone did not achieve adequate diagnostic efficiency for rule-out. Copeptin was not a useful test measured alone or in combination with cTn. Diagnostic sensitivity was improved by measurement of HFABP in addition to cTn, and should be considered for further studies.
Studies on HFABP have concentrated on its potential as a very early marker when combined with cTn. There are a number of factors which makes interpretation of the current evidence base problematic. Studies have included patients with STEMI3 ,4or the population is not completely defined.5 ,6
Many studies used a less sensitive troponin4–7 reporting sensitivities of 42–55% on admission measurement for the troponin assays, well below values reported for the current assays. HFABP did improve diagnostic sensitivity for early presentation,4 ,5 ,7 but specificity was 53–71%.
Studies using contemporary sensitive cTn assays (three of which used a sensitive HFABP assay8–10) have suggested no additional value of HFABP measurement.8–11 One study enrolled patients presenting with ST segment elevation.10 Three studies used a chest pain population. One combined cTn with HFABP for the diagnosis of non-ST segment elevation ACS and not for NSTEMI.11 The second did not analyse the diagnostic performance of a marker for the combination.8 A multicentre study of sensitive cTn and emerging biomarkers used an appropriate population and diagnostic combinations, but used a final diagnosis of ACS not NSTEMI.9 Two studies have reported an appropriate population and the ability of marker combinations. Both demonstrated that the addition of HFABP to cTn measurement did not improve diagnostic performance.12 ,13
The results of meta analysis has been mixed with early reports that hFABP does not meet the criteria for an early diagnostic test14 ,15 but more recent studies supporting a degree of added value for additional measurement of HFABP plus cTn but at the expense of a significant reduction in diagnostic specificity.16 HFABP has been shown to be a prognostic marker in patients with chest pain and suspected ACSs17–19 suggesting its role may be an ischaemia marker rather than a necrosis marker alone.
Copeptin was first systematically examined following AMI with peak values reported 24 h postevent, with greater elevation in STEMI rather than NSTEMI patients and prediction of subsequent heart failure and death but not recurrent AMI.20 Copeptin was not elevated in myocardial ischaemia.21 Reports of the role of copeptin for early rule-out of AMI have been contradictory. This may be due to the populations studied and the cTn method used as the biochemical gold standard.
In consecutive chest pain admissions, using the fourth-generation cTnT assay (AMI cut-off 40 ng/L) copeptin was only useful in patients presenting less than 10 h from onset of chest pain. By ROC curve analysis, the AUC for diagnosis of AMI was 0.75 for copeptin, 0.86 for troponin, and 0.97 for the combination and was unaffected by exclusion STEMI patients; 362/487 of the patients corresponded to the population in the RATPAC study; 23 patients had a final diagnosis of myocardial infarction; 5 (22%) had cTnT <10 ng/L and were detected solely by copeptin elevation.22 In a mixed population of patients with and without ST segment elevation on presentation with diagnosis based on a contemporary cTn method (not a sensitive assay),23 results were similar with copeptin showing higher earlier diagnostic sensitivity than the other markers studied. On exclusion of STEMI patients, the admission AUC was 0.87 for cTn and improved to 0.93 on addition of copeptin. Data for a high-sensitivity assay cTnI was also reported. The admission AUC was 0.96 which improved to 0.97 on addition of copeptin which achieved statistical significance. However, a recent study has shown that while copeptin was useful when added to a contemporary cTnT assay, it did not improve diagnostic efficiency when added to a high-sensitivity cTnT assay.24
Reports when sensitive cTn assays are used have been more mixed. In an ED population presenting with acute chest pain, copeptin levels were comparable to those found in the RATPAC study. The additive value of copeptin and high-sensitivity cTnT was not demonstrated.25 This study has been criticised as a minority of patients had myocardial infarction.26 In chest pain unit patients, the AUC of copeptin for diagnosis of AMI was 0.70 compared with 0.90 for a sensitive cTnT method. Exclusion of STEMI patients showed no benefit of measuring copeptin in combination with sensitive cTnT on ROC curve analysis. Using a prespecified cut-off of the 99th percentile for cTnT (14 ng/L) and 14 pmol/L for copeptin improved diagnostic sensitivity, but specificity was low at 0.56.27 In a population with chest pain but non-diagnostic ECG, additional measurement of copeptin did not contribute to the final diagnosis of AMI13 while there have been similar findings in other studies.28 Two recent multicentre trials have been differently interpreted. One recommends the combination of copeptin and cTnI for rule-out,29 the second does not.30 However, both demonstrate poor specificity for diagnosis, and in both studies there were patients with a final diagnosis of NSTEMI who did not show elevation of troponin or copeptin at presentation.
The major problem appears to be the specificity of copeptin and HFABP. This has a major impact in a low-risk population where it may mask true improvement in diagnostic efficiency. However, the problem with studies of chest pain patients is that while they reflect the real world, the numbers of patients with AMI are small. Large numbers, or more easily, primary data pooling studies may serve to address this problem. A limitation of the study is that because of the trial design not every patient where AMI was excluded had a >6 h from admission troponin measurement. A more general criticism of all biomarker studies is a lack of a truly independent diagnostic gold standard. Only large outcome studies may address this problem.
In conclusion, the measurement of HFABP, but not copeptin, improved the diagnostic accuracy of an initial troponin measurement but did not allow exclusion by measurement of both markers on admission measurement alone. The diagnostic accuracy of combined measurement was the same as the combination of the peak of an admission plus 90 min measurement of troponin alone.
What is already known on this subject
Sensitive troponin assays are known to allow earlier diagnosis of acute myocardial infarction (AMI) but have not been directly compared, the precise timing for blood sampling is not established, and the role of novel biomarkers has not been comprehensively compared with sensitive troponin assays.
What this study adds
This study is the first to directly compare sensitive troponin assays directly with each other, with heart fatty acid binding protein (HFABP) and copeptin in the low-risk emergency department chest pain population.
How might this impact on clinical practice
Even sensitive troponin measurements are unable on their own to diagnose AMI very early with 100% sensitivity but come close, supporting the idea of very rapid rule out is possible. Further studies combining sensitive troponin with clinical assessment or HFABP are indicated.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
Contributors PC designed the study, performed the data analysis and wrote the manuscript. DG performed the analytical work and entered all the data. SG oversaw and contributed to the clinical data review and final diagnostic classification. All the authors reviewed and contributed to the final manuscript.
Funding This project was funded by the UK NIHR HTA Programme (09/22/16: RATPAC-CBE Randomised Assessment of Treatment using Panel Assay of Cardiac Markers—Contemporary Biomarker Evaluation).
Disclaimer The views and opinions expressed therein are those of the authors and do not necessarily reflect those of the NIHR HTA.
Competing interests None.
Ethics approval Leeds East Research Ethics Committee (07/Q1206/22).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All data in the MS and data supplement is included.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.