Comparison of Usefulness of Heart-Type Fatty Acid Binding Protein Versus Cardiac Troponin T for Diagnosis of Acute Myocardial Infarction
Article Outline
We aimed to assess the additive diagnostic value of measuring the serum levels of soluble human heart-type fatty acid binding protein (H-FABP) in the early diagnosis of acute myocardial infarction (AMI) in unselected patients with chest pain. A total of 97 consecutive patients with acute ischemic-type chest pain were prospectively enrolled and classified according to the American Heart Association/American College of Cardiology guidelines. The test characteristics of H-FABP and cardiac troponin T serum levels at admission revealed a greater sensitivity of H-FABP in the first 4 hours of symptoms (86% vs 42%, p <0.05). Combining H-FABP and cardiac troponin T also improved the sensitivity in the detection of AMI (97% vs 71%, p <0.05) but demonstrated a greater misclassification rate (25% vs 9%, p <0.05). The specificity of H-FABP was poor (65%, 95% confidence interval 58% to 71%). Receiver operating characteristics revealed a poor performance of H-FABP in patients with non–ST-elevation myocardial infarction. Classification tree analysis demonstrated that an H-FABP–related improvement in the early definite rule-out of AMI (reduction of false-negative rate from 11% to 3%) was at the expense of an increase in the false-positive rate to 5%. In conclusion, measurement of H-FABP, in addition to cardiac troponin T, serum levels within the first 4 hours of symptoms improves the sensitivity and negative predictive value for the detection of AMI at the cost of test accuracy and precision, especially in patients with non–ST-elevation myocardial infarction.
In patients presenting with ischemic-type chest pain, the early establishment of a definite final diagnosis of acute myocardial infarction (AMI) is crucial for treatment in the emergency department (“timely diagnosis”). For a rapid and accurate diagnosis, the indicators of early coronary pathophysiologic events (i.e., the time of plaque rupture before complete coronary occlusion) are crucial (“early diagnosis”). The ideal biomarker for myocardial ischemia would therefore be indicative early in the cascade of events and allow a prompt diagnosis, aiding therapeutic decisions in the clinical setting.1 According to the American Heart Association/American College of Cardiology ST-elevation myocardial infarction (STEMI) and non-STEMI (non-STEMI) guidelines of myocardial infarction, a positive serum level of cardiac troponin is a constituent part of the final diagnosis.2 However, because of its large molecular size, cardiac troponin T (cTnT) does not peak until approximately 6 to 12 hours after the onset of symptoms.3 In addition, the electrocardiogram has only 50% sensitivity in the diagnosis of AMI.4, 5 Heart-type fatty acid binding protein (H-FABP), a small cytoplasmatic molecule, has been suggested as a timelier biomarker for the initial diagnosis of myocardial infarction because of its high sensitivity for AMI in the first few hours after symptom onset.6 Several studies in various clinical settings have reported favorable diagnostic results.7, 8, 9, 10, 11, 12 However, little is known about the diagnostic value of serum H-FABP levels in the complex setting of diagnostic decision making in a cohort of unselected patients presenting with chest pain. Therefore, we evaluated the diagnostic value of serum concentrations of H-FABP in an acute setting on consecutive unselected patients with ischemic-type chest pain.
Methods
A total of 97 unselected consecutive patients with ischemic-type chest pain presenting at the emergency department of the HELIOS Heart Center Wuppertal (Wuppertal, Germany) were enrolled in this prospective study. The exclusion criteria were the inability or unwillingness to give informed consent, age <18 years, and interhospital transfer. Three patients were excluded from the study because of improper sample collection or timing that made it impossible to establish or rule out the final diagnosis of AMI; thus, 94 patients were included in the analysis. The clinical parameters assessed included history of cardiac events, Goldman risk score (reflecting clinical risk of myocardial infarction),13, 14 type and duration of symptoms, arteriosclerotic risk factors, and presence of renal impairment with an estimated glomerular filtration rate of <60 ml/min. At admission to the hospital, the cTnT and H-FABP serum levels were measured from the initial blood samples, and a second cTnT measurement was performed ≥12 hours after admission. All samples were centrifuged and frozen at −70°C; H-FABP is known to show no significant changes in the serum concentration after 12 months of storage. Serum creatinine was also measured in the initial sample to assess chronic renal failure as a source of variation in the biomarker levels.15, 16 The serum cTnT levels were measured using the Elecsys troponin T immunoassay (Roche Diagnostics, Mannheim, Germany). The upper reference limit (ninety-ninth percentile) is 0.01 μg/L and the lowest concentration with a coefficient of variation of ≤10% was 0.03 μg/L. H-FABP was measured by quantitative enzyme-linked immunosorbent assay (Hycult Bio/Technology Human H-FABP enzyme-linked immunosorbent assay test kit HK402, HyCult Biotechnology BV, Eindhoven, The Netherlands). The detection limit of the test was 250 pg/ml, and the detection range was 100 to 25,000 pg/ml (manufacturer's data).
To determine the diagnostic test characteristics of H-FABP, we preclinically tested serum drawn from 51 patients with AMI at admission (STEMI, age 60 ± 13 years, 70% men) retrospectively. These patients with anterior (n = 18, maximum creatinine kinase serum level 960 ± 1,037 U/L), inferior (n = 24, maximum creatinine kinase 615 ± 391 U/L), and posterolateral (n = 7, maximum creatinine kinase 541 ± 226 U/L) STEMI were admitted 291 ± 207 minutes after the onset of symptoms. Blood samples from 80 patients without angiographic evidence of coronary artery disease and 80 patients with chronic stable coronary artery disease but no AMI (elective patients admitted for diagnostic coronary angiography) served as the controls. The median H-FABP serum concentration of normal patients without coronary artery disease was 5.4 ng/ml (95% CI 5.1 to 5.6), with an upper reference limit of 7.1 ng/ml (90% CI 6.8 to 7.4). The median H-FABP concentrations did not differ significantly from those for the patients with coronary artery disease (5.6 ng/ml, 95% CI 5.5 to 5.8). Receiver operating characteristics analysis found the optimal test accuracy for differentiating between patients with myocardial infarction (STEMI) and normal patients to be a cutoff value of 7.3 ng/ml (sensitivity 83%, specificity 97%). The upper limit of normal for H-FABP was therefore defined as 7.3 ng/ml for the clinical trial. Receiver operating characteristics analysis also revealed that the test characteristics of H-FABP (area under the receiver operating characteristics curve [AUC]) were not inferior to cTnT in all patients with STEMI and significantly better than cTnT in patients with STEMI presenting <4 hours after symptom onset (AUC 0.98 vs 0.86). On the basis of the retrospective control group analysis (STEMI vs controls without AMI, AUC for cTnT and H-FABP), the sample size estimation revealed a required sample size of 93 patients for the comparison of the 2 AUCs (derived from the same cases). To show that the discriminating power of the H-FABP and cTnT assays, performed on the same cases, was significantly different, we aimed at a sample size of 100 (assumptions: α level = 0.05, β level = 0.10, rank correlation coefficient = 0.4).
In the clinical study, we blindly assessed the initial 12-lead electrocardiograms of all patients. ST elevations ≥1 and ≥2 mm (at the J point) were noted. ST depression ≥0.5 mm at 80 ms after the J point and T inversion of ≥1.0 mm at the nadir were classified. Q waves were recorded if ≥0.03 s and ≥25% of the following R. The presence of left bundle branch block was noted separately. AMI was diagnosed when either the cTnT serum levels at admission or at 12 hours were >0.03 ng/ml, irrespective of the presence of ischemic features on the electrocardiogram in the absence of any other cause for the chest pain.17 In the absence of cTnT levels at 12 hours, a typical increase and decrease in creatinine kinase levels of more than twice the upper level of normal at 24 hours also confirmed the final diagnosis of AMI (type 1 or 2).2 When this definition was met, we then further classified cases as STEMI and non-STEMI according to the electrocardiographic features. STEMI was diagnosed when ST elevation was found in 2 contiguous leads of >1 mV in leads I to III, aVL, aVF, V5 to V6, and ≥2 mV in V1 to V3. Classification of non-STEMI was by exclusion of STEMI. Unstable angina pectoris was diagnosed when the history and/or electrocardiographic changes were consistent with an acute coronary syndrome but cTnT negative at 12 hours and/or no typical increase and decrease in creatinine kinase levels at 24 hours. The history parameters included previous myocardial infarction, percutaneous coronary intervention, or coronary artery bypass grafting. The electrocardiographic parameters included significant ischemic changes on the admission electrocardiogram (ST depression ≥0.5 mV, T inversion ≥1 mV) or evidence of coronary artery disease during the index hospital stay (positive coronary angiographic findings, positive stress test). Uninterpretable electrocardiographic tracings (pacemaker, left bundle branch block) did not contribute to the diagnosis of unstable angina but required a positive history of coronary artery disease and/or evidence of coronary artery disease with additional testing. Nonischemic chest pain was diagnosed by the exclusion of AMI and unstable angina.
The baseline patient characteristics were assessed using the Mann-Whitney U test for continuous variables and using the chi-square test for categorical variables. Two-tailed p values <0.05 were taken as significant. We compared the biomarker levels between independent groups using the Mann-Whitney U test. The median values and 95% confidence intervals (95% CI) of median were given for continuous variables that were not distributed normally. Receiver operating characteristics analysis was performed to analyze the performance of H-FABP as an early indicator of AMI. The AUC (AUC ± SE) was calculated as a quantitative measurement for test performance, taking into account sensitivity and specificity. Significant differences from random (AUC = 0.50) and differences between the AUCs of H-FABP and cTnT (repeated measurements) and between independent samples (STEMI and non-STEMI) were tested. The sensitivities and specificities were compared using McNemar's test. Data analysis was performed using a standard statistical software package (MedCalc, version 10.0, MedCalc Software, Mariakerke, Belgium). We evaluated decision making in the context of acute chest pain using binary recursive partitioning. Decision tree analysis was performed using commercial decision tree software (DTREG, version 4.5, evaluation, Phillip H. Sherrod; available from: www.dtreg.com). Tree building was performed by splitting each node into 2 child nodes aimed at minimizing node impurity and misclassification. Splitting was evaluated using the entropy algorithm by finding the best fit for each possible split for each predictor variable. The tree building stopping criterion was controlled by setting maximum splitting levels to 10 and a minimum node size for a split to 10. Tree pruning was performed using V-fold cross validation to find the optimal tree size. We used 10 V-fold cross-validation trees to minimize classification error size (cross-validated error, cross validation cost).
The present study complied with the Declaration of Helsinki, and the local ethics committee approved it. All patients gave informed consent at study enrollment.
Results
Patients with a final diagnosis of AMI did not differ significantly from those without AMI with respect to mean age, gender, median symptom duration, or presence of cardiac history (Table 1). A total of 49 patients (52%) presented within 4 hours of symptom onset. They did not differ in terms of age, gender, history, clinical risk score, or electrocardiographic features from those patients presenting >4 hours after symptom onset. However, early presenters were less likely to have a positive initial cTnT test, with a greater rate of normal findings on the initial electrocardiograms and a lower rate of ST depression/T inversion found on the electrocardiograms than were late presenters (Table 2). Acute management characteristics are listed in Table 3. Figure 1 compares the median H-FABP serum concentrations in different patient subgroups, along with controls. In particular, patients with STEMI displayed significantly greater median H-FABP values than did patients without AMI, in controls and prospective patients alike. However, we found no significant difference in the H-FABP concentrations between patients with non-STEMI and those with unstable angina pectoris. The H-FABP serum levels in patients presenting at the emergency department with acute chest pain but no AMI were still significantly greater than those in the controls without AMI.
Table 1. Clinical, electrocardiographic, and biomarker characteristics at initial presentation
| Variable | All Patients (n =94) | AMI (n=31) | No AMI (n=63) | p Value |
|---|---|---|---|---|
| Age (mean years±SD) | 67±14 | 69±14 | 66±14 | 0.322 |
| Men | 55(59%) | 18(58%) | 37(59%) | 0.951 |
| Pain-to-admission (hours)(median, 95%confidence interval of median) | 4(3–6) | 6(3–9) | 4(3–7) | 0.626 |
| Previous AMI/PCI/CABG | 39(41%) | 11(35%) | 28(44%) | 0.407 |
| Renal failure | 9(10%) | 5(16%) | 4(6%) | 0.130 |
| Clinical risk of AMI(Goldman score) | 0.016 | |||
| Very low/low | 47(50%) | 10(32%) | 37(59%) | 0.016 |
| Moderate/high | 47(50%) | 21(67%) | 26(41%) | 0.016 |
| Initial electrocardiographic findings | <0.001 | |||
| ST elevation | 11(12%) | 11(35%) | 0(0%) | <0.001 |
| ST depression/T inversion | 34(36%) | 11(35%) | 23(37%) | 0.923 |
| Left bundle branch block | 14(15%) | 4(13%) | 10(16%) | 0.704 |
| Normal | 35(37%) | 5(16%) | 30(48%) | 0.003 |
| Tachyarrhythmia | 9(10%) | 2(7%) | 7(11%) | 0.470 |
| Initial biomarkers | ||||
| Cardiac troponin T(ng/ml)(median, 95%CI) | 0.03(0.03–0.03) | 0.10(0.04–0.26) | 0.03(0.03–0.03) | <0.001 |
| Cardiac troponin T positive | 23(24%) | 23(74%) | 0(0%) | <0.001 |
| Heart type fatty acid binding protein(ng/ml)(median, 95%CI) | 7.14(6.92–7.65) | 8.10(7.33–11.62) | 6.95(6.58–7.27) | 0.001 |
| Heart type fatty acid binding protein positive | 44(47%) | 22(71%) | 22(35%) | 0.006 |
Table 2. Patient characteristics stratified for symptom duration
| Characteristic | Symptom-to-Admission Time | p Value | |
|---|---|---|---|
| ≤4 Hours (n=49; 52%) | >4 Hours (n=45; 48%) | ||
| Age (years) (mean ± SD) | 69±12 | 65±15 | 0.346 |
| Men | 27(55%) | 28(62%) | 0.484 |
| Pain-to-admission (hours) (median, 95% CI) | 2(1.5–3) | 12(8–39) | <0.001 |
| Previous AMI/PCI/CABG | 23(47%) | 16(36%) | 0.263 |
| Renal failure | 6(12%) | 3(7%) | 0.490 |
| Initial electrocardiographic findings | 0.153 | ||
| ST elevation | 6(12%) | 5(11%) | 0.864 |
| ST depression/T inversion | 13(27%) | 21(47%) | 0.042 |
| Left bundle branch block | 7(14%) | 7(16%) | 0.863 |
| Normal | 23(47%) | 12(27%) | 0.042 |
| Tachyarrhythmia | 3(6%) | 6(13%) | 0.303 |
| Initial biomarkers | |||
| cTnT (ng/ml) (median, 95% CI) | 0.03(0.03–0.03) | 0.03(0.03–0.11) | 0.005 |
| cTnT positive | 6(12%) | 17(38%) | 0.004 |
| H-FABP (ng/ml) (median, 95% CI) | 7.28(6.87–7.83) | 7.03(6.58–7.69) | 0.425 |
| H-FABP positive | 24(49%) | 20(44%) | 0.440 |
| Final diagnosis | 0.221 | ||
| AMI | 14(29%) | 17(38%) | 0.343 |
| STEMI | 6(12%) | 5(11%) | 0.864 |
| Non-STEMI | 8(16%) | 12(21%) | 0.221 |
| Unstable angina pectoris | 26(53%) | 15(33%) | 0.054 |
| Nonischemic chest pain | 9(18%) | 13(23%) | 0.229 |
Table 3. Management and final treatment of patients with acute chest pain
| All Patients (n=94; 100%) | AMI (n=31; 33%) | No AMI (n=63; 67%) | p Value | |
|---|---|---|---|---|
| Acute management | ||||
| In-patient | 80(85%) | 31(100%) | 49(78%) | 0.004 |
| CCU/catheter laboratory | 28(30%) | 17(55%) | 11(17%) | <0.001 |
| ICU | 13(14%) | 9(29%) | 4(6%) | 0.003 |
| Coronary angiography | 62(66%) | 24(77%) | 38(60%) | 0.100 |
| Interval to catheterization (hours) (median, 95%confidence interval of median) | 24.0(15.4–58.3) | 2.9(1.3–43.8) | 52.7(19.9–78.8) | 0.002 |
| Coronary artery disease positive (>50%lumen diameter reduction) | 51(82%) | 23(96%) | 28(74%) | 0.026 |
| 1-Vessel | 16(26%) | 6(25%) | 10(26%) | 0.908 |
| 2-Vessel | 15(24%) | 7(29%) | 8(21%) | 0.467 |
| 3-Vessel | 20(32%) | 10(42%) | 10(26%) | 0.208 |
| Revascularization/treatment | <0.001 | |||
| PCI | 28(30%) | 18(58%) | 10(16%) | <0.001 |
| CABG | 11(12%) | 3(10%) | 8(13%) | 0.668 |
| Conservative | 55(59%) | 10(32%) | 45(71%) | <0.001 |
Figure 1.
Median serum concentrations of H-FABP in patients with chest pain and controls. Boxes represent 25% and 75% percentiles, whiskers represent 10% and 90% percentiles. *p <0.05 (Mann-Whitney U test). CAD = coronary artery disease; MI = myocardial infarction; NICP = nonischemic chest pain; NSTEMI = non–ST-elevation MI; UA = unstable angina pectoris.
The cutoff values established from the preclinical analysis were used to dichotomize the biomarker concentrations. The test characteristics of H-FABP and initial cTnT serum levels are listed in Table 4. The sensitivity of cTnT at presentation was lowest when the symptom duration was <4 hours and increased to 100% when the symptom duration was >4 hours. The sensitivity of H-FABP for AMI was superior to cTnT for patients admitted within 4 hours of symptom onset. The sensitivity of H-FABP was greatest at 2 to 4 hours after symptom onset and decreased after >8 hours. In contrast, cTnT maintained a sensitivity of 100% at later points (Figure 2). Combining cTnT and H-FABP (either elevated) provided a significant improvement in sensitivity for patients presenting within the first 4 hours after symptom onset. This improvement in sensitivity for AMI could be maintained at later points. The specificity of H-FABP, however, was significantly lower than cTnT in all subgroups and for all periods. Taking into account the sensitivities and specificities, the AUC was calculated for the initial cTnT and H-FABP level as a quantitative parameter of test performance (Table 5). In the total patient population, the AUC was significantly higher for initial cTnT than for H-FABP. However, when testing for patients presenting within 4 hours after symptom onset, no significant difference was found in test performance between the 2 biomarkers. When testing for acute STEMI, the AUC was slightly higher for H-FABP than for to the initial cTnT, without statistical significance. However, in those patients with STEMI who presented early (<4 hours), the AUC for H-FABP was significantly higher than for the initial cTnT. The same relation between the biomarker test performance in the first 4 hours held true for controls (STEMI). In contrast, when testing for acute non-STEMI, the AUC was markedly lower for H-FABP than for cTnT and did not prove significantly different from random (p = NS compared with an AUC of 0.50; Figure 3).
Table 4. Test characteristics of initial cTnT, H-FABP or either for diagnosis of acute myocardial infarction (AMI) stratified by symptom duration
| Variable | Initial cTnT Positive | H-FABP Positive | Either cTnT or H-FABP Positive |
|---|---|---|---|
| All patients (n=94)⁎ | |||
| Sensitivity | 74%(66–74%) | 71%(57–83%) | 97%(86–99%)†‡§ |
| Specificity | 100%(96–100%) | 65%(58–71%)†¶ | 65%(60–66%)†§ |
| Negative predictive value | 89%(85–89%) | 82%(73–89%) | 98%(90–100%) |
| Positive predictive value | 100%(88–100%) | 50%(40–58%) | 58%(51–59%) |
| Misclassification rate | 9%(9–14%) | 33%(25–42%) | 25%(23–32%) |
| False-negative rate | 26%(26–34%) | 29%(17–43%) | 3%(1–14%) |
| False-positive rate | 0%(0–4%) | 35%(11–42%) | 35%(34–40%) |
| Number needed to diagnose | 1.4(1.4–1.6) | 2.8(1.9–6.8) | 1.6(1.5–2.2) |
| Symptom to admission <4 hours(n=49)∥ | |||
| Sensitivity | 42%(28–43%) | 86%(64–96%)†¶ | 93%(73–99%)†§ |
| Specificity | 100%(94–100%) | 66%(57–70%)†¶ | 66%(58–68%)†§ |
| Negative predictive value | 81%(77–81%) | 92%(80–98%) | 96%(84–99%) |
| Positive predictive value | 100%(65–100%) | 50%(38–56%) | 52%(41–55%) |
| Misclassification rate | 16%(16–25%) | 29%(23–41%) | 27%(23–38%) |
| False-negative rate | 58%(57–72%) | 14%(4–36%) | 7%(1–27%) |
| False-positive rate | 0%(0–6%) | 34%(30–43%) | 34%(32–42%) |
| Number needed to diagnose | 2.3(2.3–4.6) | 1.9(1.5–4.7) | 1.7(1.5–3.3) |
| Symptom to admission >4 hours(n=45)# | |||
| Sensitivity | 100%(90–100%) | 59%(40–75%)†¶ | 100%(85–100%)†‡ |
| Specificity | 100%(94–100%) | 64%(53–74%)†¶ | 64%(55–64%)†§ |
| Negative predictive value | 100%(94–100%) | 72%(59–83%) | 100%(86–100%) |
| Positive predictive value | 100%(90–100%) | 50%(34–64%) | 63%(54–63%) |
| Misclassification rate | 0%(0–8%) | 38%(26–52%) | 22%(22–33%) |
| False-negative rate | 0%(0–10%) | 41%(25–60%) | 0%(0–15%) |
| False-positive rate | 0%(0–6%) | 36%(26–47%) | 36%(36–45%) |
| Number needed to diagnose | 1.0(1.0–1.2) | 4.3(2.0–15.0) | 1.6(1.6–2.5) |
⁎Prevalence of AMI: 33% (95% |
†p <0.05 McNemar's test. |
‡Either versus H-FABP. |
§Either versus cTnT. |
¶H-FABP versus cTnT. |
∥Prevalence of AMI: 29% (17–43%). |
#Prevalence of AMI: 38% (24–58%). |
Figure 2.
Sensitivities of initial cTnT compared to H-FABP and a combined approach (either positive) with varying symptom durations. *p <0.05 (McNemar's test).
Table 5. Test performance of quantitative cardiac troponin T (cTnT) and heart-type fatty acid binding protein (H-FABP) in subgroups of acute myocardial infarction (AMI) and controls
| Variable | AUC | |||||
|---|---|---|---|---|---|---|
| All Patients | Early Presentation⁎ | |||||
| AUC±SE | 95% CI | p Value (area=0.5) | AUC±SE | 95% CI | p Value (area=0.5) | |
| AMI | n=31/94(33%) | n=14/49(29%) | ||||
| cTnT | 0.87±0.04 | 0.79–0.93 | <0.001 | 0.71±0.09 | 0.57–0.83 | <0.05 |
| H-FABP | 0.71±0.06 | 0.61–0.80 | <0.001 | 0.76±0.08 | 0.62–0.87 | <0.01 |
| p(ΔAUC) | <0.05 | NS | ||||
| STEMI | n=11/94(12%) | n=6/49(12%) | ||||
| cTnT | 0.69±0.09 | 0.58–0.78 | <0.05 | 0.48±0.13 | 0.34–0.63 | NS |
| H-FABP | 0.79±0.08 | 0.70–0.87 | <0.001 | 0.81±0.11 | 0.68–0.91 | <0.001 |
| p(ΔAUC) | NS | <0.05 | ||||
| Non-STEMI | n=20/94(21%) | n=8/49(16%) | ||||
| cTnT | 0.88±0.05 | 0.79–0.94 | <0.001 | 0.81±0.10 | 0.67–0.90 | <0.01 |
| H-FABP | 0.60±0.07 | 0.49–0.70 | NS | 0.64±0.11 | 0.49–0.78 | NS |
| p(ΔAUC) | <0.01 | NS | ||||
| Controls (STEMI) | n=51/131(39%) | n=23/131(18%) | ||||
| cTnT | 0.87±0.04 | 0.79–0.92 | <0.001 | 0.83±0.06 | 0.74–0.90 | <0.001 |
| H-FABP | 0.95±0.02 | 0.89–0.98 | <0.001 | 0.97±0.02 | 0.92–1.00 | <0.001 |
| p(ΔAUC) | NS | <0.05 | ||||
⁎Symptom to admission <4 hours. |
Figure 3.
Receiver operating characteristics for H-FABP. True-positive rate (sensitivity) plotted as a function of false-positive rate (100 − specificity) for different H-FABP serum concentrations. Each point on receiver operating characteristics plot represents sensitivity/specificity pair corresponding to particular decision threshold. The closer the receiver operating characteristics plot to the upper left corner, the greater the overall accuracy of the test. Discriminative value of H-FABP was particularly poor in patients with non-STEMI with the curve close to the diagonal (AUC = 0.5). (For statistics, see Table 5.)
Adding electrocardiographic information, we evaluated decision making using binary recursive partitioning and classification trees. Tree building yielded diagnostic algorithms (Figure 4). The classification of patients using traditional decision making for AMI (Figure 4) predicted the presence of AMI with a sensitivity 90% (95% CI 74% to 98%), specificity 100% (95% CI 94% to 100%), AUC of 0.95 ± 0.03 (95% CI 0.89 to 0.99; p <0.001). An alternative decision tree (Figure 4) adding the H-FABP test results and symptom durations to abnormal electrocardiographic findings and cTnT test results predicted the presence of AMI with a sensitivity of 97% (95% CI 83% to 99%), specificity 95% (95% CI 87% to 100%), AUC of 0.96 ± 0.03 (95% CI 0.89 to 0.98; p <0.001). Comparing both classification trees according to the receiver operating characteristics curves for the predicted and true presence of AMI between the 2 decision models, we found no significant difference, implicating that the harm of not including H-FABP testing for AMI was not significant (AUC, p = 0.8, NS).
Figure 4.
Decision trees found by binary recursive partitioning. (A) With ST elevation and initial cTnT as relevant (defining) predictors of AMI, a guideline-conform decision tree was built with a final misclassification rate of 3% (false-negative rate 11%, false-positive rate 0%. (B) Confining electrocardiographic interpretation to discriminating normal from abnormal initial electrocardiographic tracings, H-FABP proved a usable predictor of AMI in cTnT-negative patients with abnormal electrocardiographic findings and early presentation within first 4 hours after symptoms began. Misclassification rate was 4% (false-negative rate 3%, false-positive rate 5%). mc, absolute number of misclassified cases per node for classification as “no-MI” or classification as “AMI.”
Discussion
The use of biomarkers aids in the process of diagnosing myocardial infarction in the emergency department and helps in risk stratification of patients, allowing appropriate treatment to be given, and has proved superior to electrocardiographic guidance alone.1 Cardiac troponins have revolutionized decision making and led to a redefinition of myocardial infarction in 200017 that included elevated troponins as an obligatory diagnostic criterion. With respect to revascularization, the period for optimal treatment is within the first 4 hours after the onset of plaque rupture. However, a number of studies have reported a low sensitivity of cardiac troponins in the first 6 to 12 hours.3, 12, 18, 19 The data of the present study have confirmed the early sensitivity issue of cTnT.
In a number of studies, H-FABP has been reported to be particularly sensitive within the first few hours after the onset of coronary occlusion and symptoms.9, 10, 11, 12, 20, 21 The reason for this high sensitivity has been explained by its small molecular weight (15 kDa) and its cytoplasmatic unbound abundance, resulting in rapid release from damaged myocardial cells.6 A high sensitivity early in myocardial infarction has been noted in several, mainly selected, study population groups, demonstrating a promising potential role of H-FABP as a decision tool in the early diagnosis of myocardial infarction.7, 8, 9, 11 The release characteristics of H-FABP after occlusive AMI showed elevated serum concentrations 30 to 90 minutes after the onset of symptoms, with peak levels reached at 4 to 6 hours, and normalization of serum levels to baseline levels owing to rapid renal clearance within 20 hours. The data of the present study support the idea that H-FABP is a sensitive marker in the first 4 hours after symptom onset. Given the relatively high prevalence of AMI in the study population (constantly between 25% and 35% in this tertiary cardiology service at the HELIOS Heart Centre chest pain unit), high sensitivity translated into high negative predictive values for H-FABP, suggesting a possible use of this marker, especially for the early rule out of AMI.10
The low H-FABP specificity of 65% was in the range of what other researchers found in similar settings (49% to 86%).9, 10, 20, 21 The reasons for the poor specificity of H-FABP for the final diagnosis of AMI have been attributed to numerous factors. First, renal insufficiency is known to significantly increase the serum level of renally excreted H-FABP.15, 16 However, in the present study, the frequency of renal impairment (estimated glomerular filtration rate <60 ml) between the groups with or without AMI or between the subgroups was equally distributed. The prevalence of renal impairment was 10% in the total study population. The specificity did not change in our data when patients with renal impairment were excluded from the analysis. Second, H-FABP can be released from ischemic myocardium, as well as from infarcted myocardium.6 When studying median H-FABP serum concentrations in the non-AMI subgroups of the study population (Figure 1), we found the median H-FABP concentrations were not significantly elevated in patients diagnosed with unstable angina pectoris compared to patients diagnosed with nonischemic chest pain. Remarkably, however, the H-FABP serum levels in both non-AMI subgroups were significantly greater than in the non-AMI controls. Because the control patients had been electively recruited in an outpatient setting without acute cardiac distress symptoms, this striking finding could point to a possible H-FABP release of nonischemic origin from injured and functionally impaired myocytes in acutely symptomatic but nonmyocardial infarction cardiac conditions.22, 23 The prevalence of concomitant tachyarrhythmia, hypertensive crisis, and congestive heart failure was 16% in the subgroup of patients without AMI and in 18% of the patients diagnosed with nonischemic chest pain (NS compared to AMI group but significantly greater [p <0.05] compared to control group without AMI). Third, H-FABP release of noncardiac origin might play a role in patients with acute chest pain and explain our finding of elevated H-FABP serum levels in the nonischemic chest pain subgroup compared to controls without AMI. H-FABP is also present in the skeletal muscle and other tissues at low concentrations,6, 24 although relative cardiospecificity is generally reported to be much greater than for myoglobin or creatinine kinase-MB. No study data on recent physical activity, injury, noncardiac surgery, or intramuscular injections were collected to address this issue. The H-FABP enzyme-linked immunosorbent assay used in the present study precluded cross-reactivity with other types of fatty acid binding protein types (human intestinal-type FABP, human liver-type FABP).
Results similar to ours have led other researchers10 to suggest that cTnT and H-FABP should be seen as complementary biomarkers for the early prediction or to rule out AMI. When using H-FABP and cTnT serum admission levels as a combined test in the present data, a significantly greater early sensitivity could be reached compared with cTnT alone. The specificity, however, was significantly lower than for cTnT alone. This resulted in a lower combined overall test accuracy and greater misclassification than for cTnT alone. Analysis of the misclassification rate revealed that by trying to rule out AMI at admission by combining the results of elevated H-FABP and cTnT serum levels, a desirable overall reduction in the false-negative rate from 26% to 3% was accompanied by a marked increase in the overall false-positive rate to 35% (Table 4). Therefore, we have concluded that greater sensitivity for an early diagnosis of AMI using a combined test can only be achieved with a marked loss in overall diagnostic test efficiency and a greater misclassification rate. Despite a gain in power, the complementary use of H-FABP and cTnT serum levels led to a high degree of imprecision, making diagnostic and therapeutic decisions prone to type I errors. Confirmatory H-FABP testing 1 hour after admission has been suggested25 and could possibly minimize the false-positive diagnosis of AMI.
In the present study, we assessed the global test characteristics of biomarkers using receiver operating characteristic curves accounting for sensitivities and specificities at different cutoff values (Figure 3 and Table 5). In acute STEMI, we found no significant difference between the diagnostic test characteristics of H-FABP and cTnT as measured by the AUC. If patients with AMI presented with a symptom duration of <4 hours, H-FABP performed significantly better than cTnT as a marker to differentiate between acute STEMI and non-AMI conditions. When analyzing patients with non-STEMI, however, we found that H-FABP performed particularly poorly, irrespective of symptom duration. Because non-STEMI is a diagnostic model reflecting a thrombembolic, often repetitive, microvascular occlusion from an upstream thrombus-laden ruptured plaque, misclassification by H-FABP testing could be attributed to the small amounts of H-FABP being rapidly cleared from serum in contrast to the large molecular cTnT, which cumulates to finally maintain elevated serum levels. A gain in sensitivity of a rapid-release biomarker in the early phase of STEMI could be explained if the affected myocardial mass is large enough.
How do the biomarker findings of the present study translate into clinical reasoning in a cohort of unselected patients with chest pain in the electrocardiographic-equipped emergency department? Diagnostic decision making for AMI is more about minimizing misclassification than about optimizing sensitivity. This was the rationale for using decision algorithms with binary recursive partitioning (Figure 4). This method relies on minimizing misclassification and node impurity and is tailored to finding the best fit on each possible split on each predictor variable. Confining electrocardiographic interpretation to discriminating normal from abnormal initial electrocardiographic tracings, H-FABP proved a usable predictor of AMI in cTnT-negative patients with abnormal electrocardiographic findings and early presentation within the first 4 hours after symptom onset. However, the false-positive rate was greater and the decision algorithm complexity and depth was greater for the H-FABP–containing algorithm. The American Heart Association/American College of Cardiology guidelines for non-STEMI/unstable angina pectoris based on cTnT recommend a wait-and-resample attitude, with the possibility of delaying therapeutic intervention in 11% of our patient population (false-negative rate). The additional use of H-FABP reduced this possible delay by a factor of 3.5 to 3% but is prone to an anticipated risk of overtreatment in 5% of cases (false-positive rate). In patients with STEMI, the additive H-FABP test results provided no additional decisive information to the specific electrocardiographic interpretation.
The present study had some limitations. First, the sample size of the total study population was too small to allow for a generalization of the results. However, the study results are in accordance with the reported test characteristics of H-FABP,6, 10 and the investigation was aimed at testing the potential role of H-FABP in early risk stratification of acute ischemic-type chest pain in the clinical setting of a real world emergency department in unselected patients. Another limitation was that this investigation only studied the potential benefit from a single measurement of H-FABP at admission. Sequential measurements were not performed. Furthermore, a multimarker approach (including biomarkers of concomitant cardiac conditions, such as proBNP, myoglobin, creatinine kinase) was not taken, which could have further elucidated the false-positive rate in H-FABP testing. However, the present study was designed with cTnT as the only comparator, and priority was given to determine whether H-FABP added significant information to aid in the clinical assessment of patients with acute chest ischemic-type pain.
Acknowledgment
The present study was conducted with the laboratory assistance of Ilka Renneberg, PhD, 8sens.biognostic GmbH, Berlin, Germany. Parts of this study were included in Sigune Peiniger's doctoral thesis. We thank Martin Dörner, MD, for statistical review and Joanne Davies, BSc, for manuscript review (Institute for Heart and Circulation Research, University of Witten/Herdecke, Dortmund, Germany).
References
- . Biomarkers of myocardial necrosis—past, present, and future. In: Morrow DA editors. Cardiovascular Biomarkers: Pathophysiology and Disease Management. Totowa, NJ: Humana Press; 2006;p. 3–25
- . Universal definition of myocardial infarction. Circulation. 2007;116:2634–2653
- . The sensitivity of cardiac markers: an evidence-based approach. Clin Chem Lab Med. 1999;37:1097–1106
- . Competency in interpretation of 12-lead electrocardiograms: a summary and appraisal of published evidence. Ann Intern Med. 2003;138:751–760
- . AHA/ACCF/HRS recommendations for the standardization and interpretation of the electrocardiogram. Part VI (Acute ischemia/infarction: a scientific statement from the American Heart Association Electrocardiography and Arrhythmias Committee, Council on Clinical Cardiology; the American College of Cardiology Foundation; and the Heart Rhythm Society. Endorsed by the International Society for Computerized Electrocardiology). J Am Coll Cardiol. 2009;53:1003–1011
- . Do we need additional markers of myocyte necrosis: the potential value of heart fatty-acid-binding protein. QJM. 2004;97:187–198
- . Pre-hospital detection of acute myocardial infarction with ultra-rapid human fatty acid-binding protein (H-FABP) immunoassay. Int J Cardiol. 2007;119:349–354
- . Human heart-type fatty acid-binding protein for on-site diagnosis of early acute myocardial infarction. Int J Cardiol. 2008;133:420–423
- . Human heart-type fatty acid-binding protein as a point-of-care test in the early diagnosis of acute myocardial infarction. QJM. 2007;100:203–210
- . Novel biomarkers in early diagnosis of acute myocardial infarction compared with cardiac troponin T. Eur Heart J. 2008;29:2843–2850
- . The use of human heart-type fatty acid-binding protein as an early diagnostic biochemical marker of myocardial necrosis in patients with acute coronary syndrome, and its comparison with troponin-T and creatine kinase-myocardial band. Heart Vessels. 2006;21:309–314
- . Clinical assessment of heart-type fatty acid binding protein in early diagnosis of acute coronary syndrome. Eur J Emerg Med. 2008;15:140–144
- . Prediction of the need for intensive care in patients who come to the emergency departments with acute chest pain. N Engl J Med. 1996;334:1498–1504
- . Combination of Goldman risk and initial cardiac troponin I for emergency department chest pain patient risk stratification. Acad Emerg Med. 2001;8:696–702
- . Increased fatty acid-binding protein concentration in plasma of patients with chronic renal failure. Clin Chem. 1997;43:193–195
- . Influence of renal function on serum and urinary heart fatty acid-binding protein levels. J Cardiovasc Surg. 2001;42:735–740
- . Myocardial infarction redefined—a consensus document of the joint European Society of Cardiology/American College of Cardiology Committee for the redefinition of myocardial infarction. J Am Coll Cardiol. 2000;36:959–969
- . Early diagnostic efficiency of cardiac troponin I and troponin T for acute myocardial infarction. Acad Emerg Med. 1997;4:13–21
- . Accuracy of biomarkers to diagnose acute cardiac ischemia in the emergency department: a meta-analysis. Ann Emerg Med. 2001;37:478–494
- . Human heart-type fatty acid-binding protein as an early diagnostic and prognostic marker in acute coronary syndrome. Cardiology. 2003;99:96–104
- . Use of a whole blood rapid panel test for heart-type fatty acid-binding protein in patients with acute chest pain: comparison with rapid troponin T and myoglobin tests. Am J Med. 2003;115:185–190
- . Prognostic value of elevated circulating heart-type fatty acid binding protein in patients with congestive heart failure. J Card Fail. 2005;11:56–60
- . Predictive value of heart-type fatty acid-binding protein for left ventricular remodelling and clinical outcome of hypertensive patients with mild-to-moderate aortic valve diseases. J Hum Hypertens. 2007;21:551–557
- . Fatty acid-binding proteins as plasma markers of tissue injury. Clin Chim Acta. 2005;352:15–35
- . A superior early myocardial infarction marker: human heart-type fatty acid-binding protein. Z Kardiol. 2004;93:388–397
PII: S0002-9149(09)02210-3
doi:10.1016/j.amjcard.2009.08.645
© 2010 Elsevier Inc. All rights reserved.
