SEARCH

SEARCH BY CITATION

Keywords:

  • acute chest pain;
  • anti-apolipoprotein A-1 autoantibodies;
  • anti-phosphorylcholine antibodies;
  • myocardial infarction;
  • NSTEMI diagnosis

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References

Abstract.  Keller P-F, Pagano S, Roux-Lombard P, Sigaud P, Rutschmann OT, Mach F, Hochstrasser D Vuilleumier N (Geneva University Hospitals, Geneva). Autoantibodies against apolipoprotein A-1 and phosphorylcholine for diagnosis of non-ST-segment elevation myocardial infarction. J Intern Med 2012; 271: 451–462.

Objectives.  To explore the diagnostic accuracies of anti-apolipoproteinA-1 (anti-ApoA-1) IgG and anti-phosphorylcholine (anti-PC) IgM alone, expressed as a ratio (anti-ApoA-1 IgG/anti-PC IgM), and combined with the Thrombolysis In Myocardial Infarction (TIMI) score for non-ST-segment elevation myocardial infarction (NSTEMI) (NSTEMI-TIMI score) to create a new diagnostic algorithm – the Clinical Autoantibody Ratio (CABR) score – for the diagnosis of NSTEMI and subsequent cardiac troponin I (cTnI) elevation in patients with acute chest pain (ACP).

Methods.  In this single-centre prospective study, 138 patients presented at the emergency department with ACP without ST-segment elevation myocardial infarction. Anti-ApoA-1 IgG and anti-PC IgM were assessed by enzyme-linked immunosorbent assay on admission. Post hoc determination of the CABR score cut-off was performed by receiver operating characteristics analyses.

Results.  The adjudicated final diagnosis was NSTEMI in 17% (24/138) of patients. Both autoantibodies alone were found to be significant predictors of NSTEMI diagnosis, but the CABR score had the best diagnostic accuracy [area under the curve (AUC): 0.88; 95% confidence interval (CI): 0.82–0.95]. At the optimal cut-off of 3.3, the CABR score negative predictive value (NPV) was 97% (95% CI: 90–99). Logistic regression analysis showed that a CABR score >3.3 increased the risk of subsequent NSTEMI diagnosis 19-fold (odds ratio: 18.7; 95% CI: 5.2–67.3). For subsequent cTnI positivity, only anti-ApoA-1 IgG and CABR score displayed adequate predictive accuracies with AUCs of 0.80 (95% CI: 0.68–0.91) and 0.82 (95% CI: 0.70–0.94), respectively; the NPVs were 95% (95% CI: 90–98) and 99% (95% CI: 94–100), respectively.

Conclusion.  The CABR score, derived from adding the anti-ApoA-1 IgG/anti-PC IgM ratio to the NSTEMI-TIMI score, could be a useful measure to rule out NSTEMI in patients presenting with ACP at the emergency department without electrocardiographic changes.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References

Acute chest pain is one of the most common reasons to attend a hospital emergency department (ED). In this setting, timely identification of patients with acute coronary syndrome (ACS) is of paramount importance not only for patient management but also to optimize patient flow through the ED [1–6]. Based on the clinical presentation, electrocardiographic (ECG) features and elevation in cardiac troponin (cTn) levels (a measure of myocardial necrosis), ACS is divided into ST-segment elevation myocardial infarction (STEMI) and non-ST-segment elevation ACS (NST-ACS) [1–3]. Whereas diagnosis of STEMI is straightforward, NST-ACS diagnosis is more difficult and relies mostly on ECG changes and cTn elevation to discriminate between non-ST-segment elevation myocardial infarction (NSTEMI) and unstable angina or chest pain of noncardiac origin. However, mainly because of the kinetics of cTn release into the circulation during the development of myocardial necrosis, it is well known that the sensitivity of cTn within the first 6 h following the onset of symptoms is low [3, 4]. Because of this, prolonged monitoring and repeated blood sampling over a period of 6 h are often required before NSTEMI can be diagnosed. It has been suggested that postponing diagnosis in this way not only increases the risk of complications associated with this condition [2–5], but also contributes to ED overcrowding, the costs of which have been estimated at several billion US dollars each year [6]. Simultaneous assessment of multiple emergent cardiac biomarkers reflecting different underlying pathophysiological processes has produced promising improvements in NSTEMI diagnosis [7–10]. There is also a growing body of evidence to suggest that some autoantibodies could represent possible candidates for cardiovascular (CV) risk stratification, some acting to increase risk and others being protective [11]. Among these candidates, both high levels [above an optical density (OD) of 0.6] of anti-apolipoprotein A-1 (anti-ApoA-1) IgG and low levels (below 27 U mL−1) of anti-phosphorylcholine (anti-PC) IgM have been shown to be independently associated with increased risk of CV disease [12–16]. From a pathophysiological point of view, anti-apoA-1 IgG antibodies have been shown in vitro to act as positive chronotropic agents in cardiomyocytes [12], to directly promote the release from macrophages of mediators of atherogenesis and plaque vulnerability, such as pro-inflammatory cytokines and matrix metalloproteinase-9 [13, 14], and to negatively affect atherogenesis and atherosclerotic plaque vulnerability in apoE -/- mice [14]. By contrast, anti-PC IgM antibodies are considered to be protective, reducing atherogenesis mainly by preventing uptake of oxidized low-density lipoprotein (LDL), which is a key step in the formation of foam cells [15]. Nevertheless, to date, their respective potential contribution to the diagnosis of NSTEMI in the context of acute chest pain has not been evaluated. Therefore, in the present explorative study, we assessed the accuracy of anti-ApoA-1 IgG, anti-PC IgM and the ratio of anti-ApoA-1 IgG to anti-PC IgM for i) NSTEMI diagnostic prediction in patients presenting at the ED with acute chest pain and ii) subsequent cTn positivity prediction following an initial negative cTn sample.

We also studied the relationship between these autoantibodies and the validated prognostic clinical Thrombolysis In Myocardial Infarction (TIMI) score for NSTEMI (NSTEMI-TIMI score) to predict patient outcome at 14 days [17]. Finally, we investigated whether the combination of the autoantibody ratio with the NSTEMI-TIMI score could improve the predictive accuracy of the NSTEMI-TIMI score for NSTEMI prediction.

Material and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References

The research ethics committee of Geneva University Hospitals approved the study protocol, and all patients gave written informed consent before enrolment.

Patient population and study design

As no data were available regarding the prevalence of high titres of anti-ApoA-1 IgG and low titres of anti-PC IgM in patients with acute chest pain, we computed the sample size extrapolating our previously published results [18] on anti-ApoA-1 IgG and myocardial infarction (MI) to this explorative study. Accordingly, assuming that the prevalence of high titres of anti-ApoA-1 IgG should be 10% in the NSTEMI group and 1% in patients with diagnoses other than NSTEMI at discharge [18], a sample of 133 patients was needed to achieve a power of 90% with an alpha error of 5% to detect a difference in the prevalence of anti-ApoA-1 IgG positivity between patients with NSTEMI and those with other diagnoses.

Inclusion criteria consisted of chest pain lasting more than 5 min, regardless of age and gender, without ST-segment elevation on ECG defined by the absence of ST/T abnormalities or dynamic changes, such as nonpersistent ST-segment elevation, ST depression, T-wave abnormalities or no ECG changes. Exclusion criteria consisted of STEMI, chest pain for a duration of less than 5 min, prior hospitalization within 48 h, known autoimmune diseases such as rheumatoid arthritis (RA), systemic lupus erythematosous (SLE) or anti-phospholipid syndrome (APS), known HIV or clinically patent signs of heart failure.

Between January and April 2009, 159 patients were screened at the ED of Geneva University Hospital (a primary care hospital) for acute chest pain. Twenty-one patients were excluded: eight patients with STEMI, four with RA, one with APS, five with patent signs of heart failure, one with HIV and two with chest pain for <5 min; thus, 138 patients were eligible for analyses. Medical history was obtained at admission.

Study endpoints

Two predetermined endpoints were considered for this explorative study.

The primary endpoint was a discharge diagnosis of NSTEMI versus chest pain related to unstable angina or to ‘other diagnoses’. Chest pain aetiology was adjudicated by two senior cardiologists blinded to the participants’ biochemical data.

Diagnosis of NSTEMI was established using the universal criteria of type 1 acute myocardial infarction (AMI) based on dynamic changes in cTnI levels in the appropriate clinical context [19], excluding persistent STEMI. Patients were considered to have diagnoses other than NSTEMI when cTnI values were negative, following further investigation by coronary angiography for those with a higher pretest probability of ischaemic origin or noninvasive testing, including treadmill test, cardiac magnetic resonance imaging, stress echocardiography or myocardial perfusion scintigraphy in ambulatory settings, for patients at lowest risk. Patients with possible UA were considered as low-risk NST-ACS and were accordingly attributed to the ‘other diagnoses’ group [2, 3]. If patients did not fulfil the universal criteria of AMI [19] in the presence of cTnI elevation, a nonischaemic aetiology was concluded only after exclusion of ischaemia using myocardial scintigraphy or cardiac magnetic resonance imaging, or after exclusion of a significant culprit coronary lesion by coronary angiography.

The secondary endpoint was subsequent cTnI elevation (<0.09 ng mL−1) when the first cTnI result was negative (≤0.09 ng mL−1).

Biochemical analyses

Venous blood samples were collected on patient admission to the ED, prior to treatment initiation. Samples were immediately processed for assessment of cTnI and other standard biochemical parameters, and then frozen and stored at −80 °C until required for anti-ApoA-1 IgG and anti-PC IgM analyses.

Quantification of cTnI, C-reactive protein, creatinine, total cholesterol, triglycerides and HDL

Plasma concentrations of cTnI, C-reactive protein, creatinine, total cholesterol, triglycerides and HDL were quantitatively assessed using autoanalysers (DXI™; Beckman Coulter, Brea, CA, USA) at the accredited Service of Laboratory Medicine of Geneva University Hospitals. LDL levels were calculated according to the Friedwald formula. Glomerular filtration rate was computed using the Cockroft-Gault formula. For cTnI, we used the cut-off value of 0.09 ng mL−1, corresponding to the 99th percentile of normal distribution with a coefficient of variation of 10% as recommended by the recent universal AMI definition proposed by the European Society of Cardiology and the American Heart Association [19]. This cut-off is routinely used at our institution to detect myocardial injury in the setting of ACS.

Anti-ApoA-1 IgG assessment by enzyme-linked immunosorbent assay

Anti-ApoA-1 IgG antibodies were measured as previously described [12–14]. Briefly, Maxisorp plates (Nunc™, Roskilde, Denmark) were coated with purified, human-derived delipidated ApoA-1 (20 μg mL−1; 50 μL well−1) for 1 h at 37 °C. After washing, the wells were blocked for 1 h with phosphate-buffered saline (PBS) with 2% bovine serum albumin (BSA) at 37 °C. Then, serum samples diluted 1/50 were incubated for 1 h. Patient serum samples were also added to a noncoated well to assess the individual nonspecific binding. After washing six times, 50 μL well−1 alkaline phosphatase-conjugated anti-human IgG (Sigma-Aldrich, St Louis, MO, USA) diluted 1/1000 in PBS/BSA solution was incubated for 1 h at 37 °C. After washing again six times, the phosphatase substrate p-nitrophanylphosphate disodium (Sigma-Aldrich) dissolved in diethanolamine buffer (pH 9.8) was added. Each sample was tested in duplicate, and absorbance in OD was determined at 405 nm, after incubation for 20 min at 37 °C, using a plate reader (Molecular Devices Versa Max™; Molecular Device, Sunnyvale, CA, USA). The corresponding nonspecific binding was subtracted from the mean absorbance for each sample. The cut-off value for positivity was prospectively defined and set at 0.6 OD and 37% of the positive control value, as previously described [12–14]. At the cut-off level, the intra- and inter-assay coefficients of variation were 16% (= 10) and 12% (= 8), respectively.

Anti-PC IgM assessment by enzyme-linked immunosorbent assay

Anti-PC IgM levels were assessed using a commercially available enzyme-linked immunosorbent assay kit (CVDefine™; Athera Biotechnologies, Uppsala, Sweden), with purified PC as antigen, and performed according to the manufacturer’s instructions. Results are expressed in U mL−1, based on a standard curve build on six points. Samples were run in duplicate. Inter-assay coefficients of variation were 2.4% at 12.5 U mL−1 and 2.0% at 25 U mL−1 (n = 4). The intra-assay coefficient of variation at 65 U mL−1 was 5% (n = 8). As the anti-PC IgM cut-off for CV disease risk prediction varies from 17 to 29 U mL−1 according to published reports [15, 16], and because no data are currently available for NSTEMI prediction with this autoantibody, we defined the cut-off based on post hoc receiver operating curve (ROC) analysis.

Anti-ApoA-1 IgG/anti-PC IgM ratio determination

The anti-ApoA-1 IgG/anti-PC IgM ratio was determined by dividing the anti-ApoA-1 IgG value (index expressed as percentage of positive control) by anti-PC IgM units (U mL−1). Results of the ratio are therefore reported as arbitrary units (AU). As for anti-PC IgM, the cut-off was defined post hoc on ROC analysis.

Determination of the clinical autoantibody ratio score

The clinical autoantibody ratio (CABR) score was computed by simple addition of the NSTEMI-TIMI score to the anti-ApoA-1 IgG/anti-PC IgM ratio. The optimal CABR score cut-off for NSTEMI was defined post hoc by ROC analysis.

Statistical analyses

Analyses were performed using statistica™ software (StatSoft, Tulsa, OK, USA). Fisher’s bilateral exact test and Mann–Whitney U-test were used where appropriate. Associations between anti-ApoA-1 IgG, anti-PC IgM, anti-ApoA-1 IgG/anti-PC IgM ratio and study endpoints are presented as the odds ratio (OR) and corresponding 95% confidence interval (95% CI). Multivariable analyses with logistic regression were used to assess associations between variables. In this model, endpoints were set as dependent variables, and NSTEMI-TIMI score [17] (allowing for adjustment for major CV determinants of patient outcome at 14 days within a single continuous variable) was set as the unique confounder because of the limited sample size. The log-normal variation of this model was also used to assess the interdependence of the anti-ApoA-1 IgG/anti-PC IgM ratio and traditional CV risk factors. ROC analyses were performed using analyse-it™ software for Excel (Microsoft, Redmond, WA, USA) to (i) confirm the cut-off values prospectively chosen for anti-ApoA-1 IgG, (ii) determine the best cut-off for anti-PC IgM and for the anti-ApoA-1 IgG/anti-PC IgM ratio and (iii) determine which marker alone, in combination (ratio) or together with the NSTEMI-TIMI score yields the best area under the curve (AUC). AUC comparisons were performed according to the nonparametric approach proposed by DeLong et al. [20]. To further support the ROC curve analyses, reclassification statistics using the integrated discrimination index (IDI) compared the predictive performances of the CABR score with the anti-ApoA-1 IgG/anti-PC IgM ratio and the NSTEMI-TIMI score, as recommended by Pencina et al. [21]. The predicted risk of NSTEMI diagnosis according to the values of variables was assessed by a logistic regression model (the goodness-of-fit was checked by the Hosmer–Lemeshow test), and the IDI was derived from the mean predicted risks. The IDI is interpreted as the difference in the mean risk in patients with the event predicted by two variables minus the difference in the mean risk in patients without the event. It reflects the average gain in sensitivity (SE) minus the average loss in specificity (SP). We report IDI in relative percentage, expressing the CABR-related improvement in discrimination when compared to the NSTEMI-TIMI score and the autoantibody ratio [21].

SE, SP, positive predictive value (PPV), negative predictive value (NPV), positive and negative likelihood ratios (LR+ and LR−, respectively) with the respective 95% CIs are given. Ranked Spearman correlations were performed to establish correlations between variables. A value of < 0.05 was considered statistically significant.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References

Patient demographic characteristics are shown in Table 1. At discharge, 17% (24 of 138) of the patients were diagnosed with NSTEMI. Other diagnoses were considered to account for the symptoms of the remaining patients, as shown in Table 1. One patient with pulmonary embolism classified in the ‘other diagnoses’ group on discharge had elevated cTnI levels on admission.

Table 1.   Patient demographic characteristics
 Acute chest pain (= 138)NSTEMI (= 24)Other diagnoses (= 114)P
  1. All continuous variables are expressed as median [interquartile range (IQR); and range].

  2. For continuous variables, Mann–Whitney U-test was used for group comparisons and bilateral exact Fisher’s test for proportion comparisons between groups.

  3. CHD, coronary heart disease; ACE, angiotensin-converting enzyme; AT-1, angiotensin-1 receptor; cTnI, cardiac troponin I; CRP, C-reactive protein; GFR, glomerular filtration rate; CABR score, Clinical AutoantiBody Ratio score, computed as the anti-ApoA-1 IgG/anti-PC IgM ratio added to the NSTEMI-TIMI score.

Age (years)58 (48–71; 23–93)69 (62–74; 46–89)56 (46–69; 23–93)0.001
Gender
 Male, % (n)62 (86)83 (20)58 (66)0.02
 Female, % (n)38 (52)17 (4)42 (48)
Cardiovascular risk factors
 Diabetes, % (n)18 (25)38 (9)14 (16)0.01
 Smoking, % (n)23 (31)29 (7)21 (24)0.42
 Dyslipidaemia, % (n)36 (50)58 (14)40 (46)0.11
 Obesity, % (n)14 (19)13 (3)14 (16)1
 Hypertension, % (n)44 (61)71 (17)39 (44)0.005
 CHD, % (n)29 (40)54 (13)24 (27)0.005
 Stroke, % (n)5 (7)4 (1)5 (6)1
 Family history, % (n)12 (16)29 (7)8 (9)0.008
Blood pressure (mmHg)
 Systolic130 (120–148; 95–200)130 (117–149; 95–164)131 (120–148; 97–200)0.37
 Diastolic75 (90–70; 50–110)70 (70–80; 60–90)79 (70–90; 50–110)0.31
 Heart rate (bpm)75 (66–84; 40–170)72 (66–80; 47–170)75 (66–85; 40–130)0.59
 Body mass index (kg m−2)26.1 (23.9–29.4; 16.4–38.3)28.1 (25.8–30.1; 20.8–32.9)25.7 (23.8–29.4; 16.4–38.3)0.27
 NSTEMI-TIMI score at admission2 (1–3; 1–6)4 (3–5; 2–6)2 (1–3; 1–6)<0.0001
Medical treatment on admission
 Aspirin, % (n)40 (55)63 (15)35 (40)0.02
 Clopidogrel, % (n)8 (11)8 (2)8 (9)1
 β-blockers, % (n)28 (39)42 (10)25 (29)0.14
 ACE inhibitors, % (n)23 (32)42 (10)19 (22)0.03
 AT-1 blockers, % (n)10 (14)8 (2)11 (12)1
 Insulin,% (n)5 (7)13 (3)4 (4)0.1
 Oral antidiabetic agents, % (n)20 (27)33 (8)17 (19)0.09
 Diuretics, % (n)10 (14)25 (6)7 (8)0.01
 Calcium channel blockers, % (n)10 (14)17 (4)9 (10)0.26
 Statins, % (n)31 (43)46 (11)28 (32)0.1
Biological parameters on admission
 Total cholesterol (mmol L−1)4.4 (4.0–5.0; 2.7–6.7)4.4 (3.9–5.2; 2.8–6.1)4.4 (4.0–5.0; 2.7–6.7)0.96
 HDL (mmol L−1)1.04 (0.79–1.28; 0.65–2.29)0.98 (0.79–1.28; 0.65–2.29)1.1 (0.8–1.3; 0.7–1.5)0.81
 LDL (mmol L−1)2.6 (2.1–3.2; 0.7–4.6)2.9 (2.1–3.2; 0.7–3.6)2.5 (2.2–2.9; 1.5–4.6)0.59
 Triglycerides (mmol L−1) 1.4 (0.9–20.7; −0.3–5.1)1.5 (0.9–2.1; 0.3–3.0)1.3 (0.9–2.0; 0.4–5.1)0.96
 GFR (mL min−1)66 (60–108; 14–202)60 (57–88; 29–167)68 (60–110; 14–202)0.41
 CRP (mg L−1)3 (1–10; <1–274)6.5 (3.5–16.5; 3–237)2 (1–8; <1–274)0.02
 Initial cTnI value (ng mL−1)0.02 (0.01–0.04; 0–23)0.16 (0.07–2.18; 0.01–23)0.02 (0.01–0.03; 0–0.34)<0.0001
 Initial Elevated cTnI, % (n)15 (20)54 (13)6 (7)<0.0001
 Initial cTnI−, second cTnI+7 (10)38 (9)1 (1)<0.0001
 Anti-ApoA-1 IgG, index16 (10–24; 0–92)23 (18–44; 11–67)14 (9–23; 0–92)0.0001
 Anti-ApoA-1 IgG, OD0.25 (0.14.0.39; 0–1.57)0.37 (0.26–0.74; 0.15–1.14)0.22 (0.13–0.33; 0–1.57)0.0001
 Anti-ApoA-1 IgG positivity, % (n)12 (16)38 (9)6 (7)0.0002
 Anti-PC IgM, U mL−143.1 (26–63.9; 8.5–1714)32.5 (20.5–50.6; 10–1714)46.9 (29–69.6; 8.5–216.5)0.02
 Anti-ApoA-1 IgG/anti-PC IgM ratio, Arbitrary units0.36 (0.16–0.62; 0–6.44)0.69 (0.48–1.13; 0.02–6.4)0.28 (0.13–0.52; 0–3.32)<0.0001
 CABR score2.4 (1.5–4.1; 1.0–12.4)4.9 (3.9–5.9; 2.3–12.4)2.2 (1.3–3.4; 1–6.4)<0.0001
Diagnosis at discharge, % (n)
 NSTEMI17 (24)
 Other diagnoses83 (114)
 Parietal aetiology7 (9)
 Gastroenterological aetiology5 (7)
 Pulmonary aetiology3 (4)
 Supraventricular arrythmia4 (5)
 Pulmonary embolism3 (4)
 Pericarditis2 (2)
 Psychogenic3 (4)
 Malignant hypertension3 (4)
 Undetermined (pulmonary embolism and aortic dissection ruled out)54 (75)

Traditional risk factors and their association with NSTEMI

Patients with NSTEMI at discharge were older, more likely to be men, and had a higher prevalence of diabetes, hypertension and known coronary heart disease and related treatment when compared to patients with non-NSTEMI-related diagnoses (Table 1).

Autoantibody association with NSTEMI-TIMI score

Ranked Spearman correlation demonstrated a significant association between anti-ApoA-1 IgG/anti-PC IgM ratio and NSTEMI-TIMI score (r = 0.29; = 0.005). Individually, anti-ApoA-1 IgG and anti-PC IgM were less strongly but still significantly correlated to NSTEMI-TIMI score (r = 0.24, = 0.005 and r = −0.15, = 0.04, respectively).

Anti-ApoA-1 IgG, anti-PC IgM levels and anti-ApoA-1 IgG/anti-PC IgM ratio and their association with diagnosis of NSTEMI at discharge

The frequency of anti-ApoA-1 IgG positivity was 12% (16/138), and NSTEMI patients had higher median levels of anti-ApoA-1 IgG (index: 23 vs. 14; < 0.0001), lower levels of anti-PC IgM (32.5 vs. 46.9 U mL−1; = 0.02), a higher anti-ApoA-1/anti-PC IgM ratio (0.69 vs. 0.28 AU; < 0.0001) and a higher CABR score (4.9 vs. 2.2; < 0.0001) on admission, compared with patients with other diagnoses at discharge.

ROC curve analyses for prediction of NSTEMI diagnosis based on autoantibodies

ROC curve analyses confirmed that both anti-ApoA-1 IgG and anti-PC IgM levels on admission of patients presenting to the ED with acute chest pain were significant predictors of an NSTEMI diagnosis at discharge, with a better AUC for anti-ApoA-1 IgG when compared to anti-PC IgM (Table 2). These analyses also confirmed that the prospectively defined anti-ApoA-1 IgG cut-off (OD > 0.6 and index > 37%) validated in patients with AMI and RA [13, 14] can also be useful to predict NSTEMI in patients with acute chest pain, with an SP of 93%, an SE of 38%, a PPV of 53% and an NPV of 88%. The LR+ and LR− values were 5.43 and 0.67, respectively (Table 2).

Table 2.   Predictive accuracy of anti-ApoA-1 IgG, anti-PC IgM, anti-ApoA-1 IgG/anti-PC IgM ratio and CABR score for (a) NSTEMI diagnosis and (b) subsequent cTnI positivity in patients with acute chest pain without ST-segment elevation
Auto-antibodies and CABR score for NSTEMI prediction
 Anti-ApoA-1 IgGAnti-PC IgMAnti-ApoA-1 IgG/anti-PC IgM ratioCABR score
  1. aAUC values were computed using continuous values for anti-ApoA-1 IgG and anti-PC IgM.

  2. SP, specificity; E, sensitivity; PPV, positive predictive value; NPV negative predictive value; LR+, positive likelihood ratio; LR−, negative likelihood ratio; NA, not applicable (anti-PC IgM was not found to be a significant predictor according to AUC).

*AUC (95% CI) P value0.75 (0.64–0.85) <0.00010.65 (0.53–0.77) <0.0070.79 (0.70–0.89) <0.00010.88 (0.82–0.95) <0.0001
 Anti-ApoA-1 IgG positivityAnti-PC IgM cut-off <29 U mL−1Anti-ApoA-1 IgG/anti-PC IgM ratio >0.64CABR score >3.3
SP, % (95% CI)93 (87–97)76 (67–84)83 (75–90)74 (66–97)
SE, % (95% CI)38 (19–59)48 (27–69)57 (35–77)87 (65–82)
PPV, % (95% CI)53 (28–77)29 (15–46)41 (24–59)40 (26–55)
NPV, % (95% CI)88 (80–92)88 (80–94)91 (83–95)97 (90–99)
LR + (95% CI)5.34 (1.98–11.4)2.02 (1.24–3.59)3.39 (1.97–5.85) 3.30 (2.34–4.67)
LR − (95% CI)0.67 (0.54–2.00)0.68 (0.43–1.01)0.52 (0.33–0.84) 0.18 (0.06–0.51)
Auto-antibodies and CABR score for subsequent cTnI positivity prediction
*AUC (95% CI) P value0.80 (0.68–0.91) <0.00010.52 (0.33–0.70) 0.430.74 (0.61–0.88) 0.00030.82 (0.70–0.94) <0.0001
 Anti-ApoA-1 IgG positivityAnti-PC IgM cut-off <29 U mL−1Anti-ApoA-1 IgG/anti-PC IgM ratio >0.64CABR score >3.3
SP, % (95% CI)90 (83–95)NA79 (71–86)67 (58–75)
SE, % (95% CI)40 (12–74)NA56 (21–86)89 (52–100)
PPV, % (95% CI)24 (7–50)NA16 (5–33)16 (7–29)
NPV, % (95% CI)95 (90–98)NA96 (91–99)99 (94–100)
LR + (95% CI)3.94 (1.57–9.86)NA2.63 (1.34–5.17)2.71 (1.93–3.80)
LR − (95% CI)0.67 (0.40–1.10)NA0.56 (0.27–1.17)0.17 (0.03–1.06)

For anti-PC IgM, ROC curve analyses indicated that the best cut-off for NSTEMI prediction was 29 U mL−1, which is the same cut-off previously suggested to be appropriate for MI prediction in a population-based study [17]. At this cut-off value, SP was 76%, SE was 48%, PPV was 29% and NPV was 88%. The LR+ and LR− values were 2.02 and 0.68, respectively (Table 2).

Finally, ROC curve analyses indicated that the anti-ApoA-1 IgG/anti-PC IgM ratio outperformed anti-PC IgM alone for NSTEMI prediction with an AUC of 0.79. Indeed, AUC comparisons using the nonparametric approach indicated that the anti-ApoA-1 IgG/anti-PC IgM ratio had a higher diagnostic accuracy when compared to anti-PC IgM alone (= 0.005) but was not superior to anti-ApoA-1 IgG alone (= 0.46). ROC analyses demonstrated that the best cut-off for the anti-ApoA-1 IgG/anti-PC IgM ratio was 0.64 AU. At this cut-off, SP was 83%, SE was 57%, PPV was 41% and NPV was 91%. The LR+ and LR− values were 3.39 and 0.52, respectively (Table 2).

ROC curve analyses and reclassification statistics for NSTEMI prediction based on NSTEMI-TIMI score, the anti-ApoA-1 IgG/anti-PC IgM ratio and the CABR score

ROC curve analyses indicated that the clinical NSTEMI-TIMI score was a significant predictor of NSTEMI diagnosis at discharge with an AUC of 0.86; that is, patients with a diagnosis of NSTEMI at discharge had a higher NSTEMI-TIMI score on admission (Table 1, Fig. 1). By the addition of anti-ApoA-1 IgG/anti-PC IgM ratio to the NSTEMI-TIMI score, we derived the CABR score, with values ranging from 1.0 to 12.4 (Table 1). Using this CABR score for ROC analysis, we observed an increase in the accuracy of NSTEMI prediction with an AUC of 0.88 (Table 2). According to the nonparametric approach, this increase from an AUC of 0.86 (NSTEMI-TIMI score alone) to 0.88 for the CABR score was significant (= 0.01), and the CABR score tended to outperform the anti-ApoA-1 IgG/anti-PC IgM ratio (= 0.05). Further analysis indicated that the best cut-off for the CABR score was 3.3. At this cut-off, SP was 74%, SE was 87%, PPV was 40%, NPV was 97%, LR+ was 3.30 and LR− was 0.18 (Table 2). Fifty (36%) patients had a CABR score above 3.3, and 88 (64%) had a CABR score below or equal to 3.3. These results were further corroborated by reclassification statistics indicating that the mean predicted risk of NSTEMI diagnosis in patients with diagnoses other than NSTEMI on discharge was 11.5% for the CABR score, 14.5% for the autoantibody ratio and 12.1% for the NSTEMI-TIMI score. In patients with NSTEMI diagnosis on discharge, the predicted risks of NSTEMI diagnosis were 42.9% for the CABR score, 28.0% for the autoantibody ratio and 40.1% for the NSTEMI-TIMI score. The IDI values comparing the CABR score to the autoantibody ratio alone and to the NSTEMI-TIMI score alone were 18% (= 0.002) and 3% (= 0.008), respectively, indicating that the CABR score provides significant incremental diagnostic ability when compared to either the NSTEMI-TIMI score or the autoantibody ratio.

image

Figure 1.  Respective diagnostic accuracies for NSTEMI prediction.

Download figure to PowerPoint

Anti-ApoA-1 IgG, anti-ApoA-1 IgG/anti-PC IgM ratio and the CABR score as significant predictors of subsequent cTnI elevation

As an important part of the delay in diagnosing NSTEMI can be attributed to the results of cTn measurement in patients who present at the ED within the first 6 h of symptom onset [3, 4], we investigated whether autoantibodies measured at the time the first cTnI value was negative could predict subsequent cTnI positivity. For this purpose, we performed ROC curve analyses in patients with a first negative cTnI result followed by cTnI elevation measured in the second blood sample. Overall, 10 patients had a normal initial cTnI level followed by an elevated level on the second assessment (>0.09 ng mL−1). The adjudicated final diagnosis was NSTEMI in these 10 patients (Table 1).

ROC curve analyses indicated that anti-ApoA-1 IgG and anti-ApoA-1 IgG/anti-PC IgM ratio (but not anti-PC IgM) on admission were significant predictors of subsequent cTnI positivity with respective AUC values of 0.80 and 0.74 (Table 2). At the cut-off used in the present study both anti-ApoA-1 IgG and the anti-ApoA-1 IgG/anti-PC IgM ratio provided a modest but significant LR+ (3.94; and 2.63, respectively), whereas the LR− values were not significant as the 95% CIs included 1 (Table 2). For subsequent cTnI elevation prediction, ROC curve analyses indicated that the CABR score was also the best discriminator with an AUC of 0.82 (Table 2), which, according to the nonparametric approach, did not significantly outperform the AUCs of anti-ApoA-1 IgG and the ratio of autoantibodies (= 0.77 and = 0.06, respectively), but was superior to the NSTEMI-TIMI score alone (= 0.01). ROC curves for the anti-ApoA-1 IgG/anti-PC IgM ratio, NSTEMI-TIMI score and CABR score are shown in Fig. 2.

image

Figure 2.  Respective predictive accuracies for subsequent cardiac troponin I elevation.

Download figure to PowerPoint

At the chosen cut-off value of 3.3, the CABR score had a sensitivity of 89%, an NPV of 99% and an LR− of 0.17 (Table 2). By contrast, SP and PPV were much too low to be clinically meaningful (Table 2).

High anti-ApoA-1 IgG levels, anti-ApoA-1 IgG/anti-PC IgM ratio and CABR score are associated with an increased risk of NSTEMI diagnosis at discharge

Because ranked Spearman correlation indicated that anti-ApoA-1 IgG/anti-PC IgM ratio was only modestly associated with NSTEMI-TIMI score (r = 0.29) and because adding the ratio of autoantibodies significantly improved the NSTEMI-TIMI AUC, we investigated whether the autoantibody ratio could predict the risk of subsequent NSTEMI diagnosis independently of NSTEMI-TIMI score. First, the proportion of patients that tested positive for anti-ApoA-1 IgG was higher in the NSTEMI group compared with patients without ischaemia (38% vs. 6%; = 0.002; Table 1). In univariate analyses, anti-ApoA-1 IgG positivity and anti-ApoA-1 IgG/anti-PC IgM ratio > 0.64 on admission were respectively associated with a 10-fold (OR: 9.8, 95% CI: 3.2–30.5) and a seven-fold (OR: 6.5; 95% CI: 2.5–17.0) increased risk of NSTEMI diagnosis at discharge, which remained significant after adjustment for the NSTEMI-TIMI score (OR: 6.4, 95% CI: 1.72–24.2 and OR: 5.4, 95% CI: 1.7–17.0, respectively; Table 3). Univariate analysis indicated that anti-PC IgM levels below 29 U mL−1 on admission were associated with a three-fold (OR: 2.9, 95% CI: 1.2–7.5) increased risk of NSTEMI diagnosis at patient discharge; however, this association did not remain significant after adjustment for the NSTEMI-TIMI score (OR: 2.2, 95% CI: 0.70–6.7; Table 3). As shown in Table 3, the CABR score was most strongly associated with the subsequent NSTEMI risk. Indeed, CABR score above 3.3 increased the risk of NSTEMI 19-fold (OR: 18.7; 95% CI: 5.2–67.3), and further confirmed the usefulness of combining the autoantibody ratio with the NSTEMI-TIMI score to markedly increase predictive performance, as suggested by ROC curve analyses.

Table 3.   Autoantibody-based risk analysis for NSTEMI prediction
NSTEMI predictionUnivariate odds ratio (95% CI)Odds ratio adjusted for NSTEMI-TIMI score (95% CI)
Anti-ApoA-1 IgG positivity 9.8 (3.2–30.5)6.4 (1.72–24.2)
Anti-PC IgM <29 U mL−1 2.9 (1.2–7.5)2.2 (0.7–6.7)
Anti-ApoA-1/anti-PC IgM ratio > 0.64 6.5 (2.5–17.0)5.4 (1.7–17.0)
CABR score > 3.318.7 (5.2–67.3)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References

The novel and important finding of this study is that an autoantibody-based panel relying on the anti-ApoA-1 IgG/anti-PC IgM ratio could be of clinical value, especially when combined with the NSTEMI-TIMI score, for NSTEMI diagnosis in patients presenting at the ED with acute chest pain but without ST-segment elevation. Indeed, from our ROC curve analyses and reclassification statistics using the IDI, we have demonstrated that the CABR score was not only the strongest predictor of NSTEMI diagnosis (AUC: 0.88), but also provided modest but significantly increased diagnostic ability when compared to the NSTEMI-TIMI score (IDI: 3%, = 0.008). It is interesting that the CABR score was also found to be a significant predictor of subsequent cTnI positivity when the first cTnI assessment was negative (AUC: 0.82). At the chosen cut-off of 3.3, the CABR score appeared to be a good test to rule out the probability of an NSTEMI diagnosis with an NPV of 97%. Nevertheless, SP, PPV and the LR+ were too low to be clinically useful. Furthermore, at the same cut-off, the CABR score reached the optimal NPV of 99% with a negative LR− of 0.17, clearly indicating that this test could be useful to rule out myocardial ischaemia in patients with chest pain and a first negative cTnI sample regardless of changes in ECG, such as nonpersistent ST-segment elevation, ST depression or T-wave abnormalities or no ECG changes. This in turn could favourably impact patient flow through the ED, by potentially obviating the need for serial blood sampling in patients with a CABR score ≤3.3, or the need to wait for the results of the second cTnI assessment in cases of a normal initial cTnI level. Given the difficulty and costs of rapidly ruling out the diagnosis of NSTEMI, particularly in the absence of ST/T abnormalities or dynamic changes [2–6], these preliminary results need to be confirmed in larger prospective cohorts before any clinical recommendations can be made.

On the other hand, anti-ApoA-1 IgG at the previously validated cut-off [12–14] appeared to be the candidate with the best LR+ of 5.34 and 3.94 for NSTEMI prediction and subsequent cTnI elevation, respectively. Even though these LR+ values did not exceed the recommended standard of 10 to be clinically meaningful as a rule-in test [22], they were above the values for cTn used in the context of pulmonary embolism risk stratification [23]. Whether measurement of anti-ApoA-1 IgG could serve as a rule-in test should also be further investigated. Moreover, given the well-known difficulty of improving the AUC with single biomarkers of CV risk, despite a strong association with CV disease [23–26], observing such an increase in AUC and risk with the CABR score compared with the autoantibodies or NSTEMI-TIMI score separately is a good indication that these autoantibodies could reflect other pro-atherogenic properties in addition to those of traditional CV risk factors, as previously suggested [12–14, 27]. Indeed, the rationale for examining the anti-ApoA-1 IgG/anti-PC IgM ratio (with or without the NSTEMI-TIMI score) in the context of acute chest pain is that both autoantibodies have been independently associated with increased CV risk [12–16]. High levels of anti-ApoA-1 IgG have been shown to independently predict major adverse CV events in patients with MI or RA and to be associated with more vulnerable atherosclerotic plaque in patients with severe carotid stenosis [12–14]. By contrast, low levels of anti-PC IgM have been consistently associated with a higher risk of CV disease [15, 16]. From a pathophysiological point of view, anti-ApoA-1 IgG antibodies have been shown in vitro and in vivo to promote inflammation, development of atherosclerosis and plaque vulnerability, whereas anti-PC IgM reduces atherosclerosis [28]. The exact mechanisms of action of these autoantibodies are still under investigation, but appear to involve antagonistic effects on the innate immune receptors of macrophages, with anti-apoA-1 IgG being pro-inflammatory through the engagement of the CD14/Toll-like receptor 2 complex (S. Pagano, N. Satta , D. Werling, V. Offord, P. de Moerloose, E. Charbonney, D. Hochstrasser, F. Mach, P. Roux-Lombard, N.Vuilleumier, unpublished data), whereas anti-PC IgM prevent scavenger receptor-mediated foam cell formation [15]. Whether acute dynamic changes in anti-apoA-1 IgG and anti-PC autoantibody levels precede AMI remains unknown. Nevertheless, based on published data indicating that anti-apoA-1 IgG and anti-PC IgM modulate atherogenesis in the long term but to a moderate degree [12–14, 17, 28], and because autoantibody levels are relatively stable over time, we consider that these autoantibodies are most likely to contribute to an elevated baseline CV disease risk rather than precipitating the acute event itself. However, this hypothesis needs to be proven.

Another interesting finding of this study is that the NSTEMI-TIMI score, in addition to its well-established short-term CV prognostic value [17], could also be important in terms of diagnosis of myocardial necrosis with an AUC to predict myocardial necrosis of 0.86 (95% CI: 0.79–0.93). These results are in line with a recently published meta-analysis including 17 625 patients showing that the TIMI score could be a useful diagnostic option for myocardial ischaemia, with SP and SE values above 95% depending on the cut-off used, in patients presenting at the ED with acute chest pain [29]. However, the results of several studies have indicated that TIMI score alone is not sufficient for diagnostic purposes as some patients with low TIMI score are still at increased risk of CV disease [30, 31]. In this respect, our results indicate that an autoantibody-based ratio reflecting both pro- and anti-atherogenic processes as well as plaque vulnerability could significantly increase the diagnostic accuracy of clinical parameters for myocardial ischaemia and could therefore represent a different approach for NSTEMI diagnosis in the future.

This study has several limitations. First, despite being appropriately powered, this single-centre study has a limited sample size, as reflected by broad confidence intervals. However, as patient demographics were comparable with those in recent studies including consecutive patients with acute chest pain [29–31], we consider our cohort to be representative of unselected patient samples presenting at the ED with suspicion of NSTEMI. Furthermore, the design of this study does not allow us to quantify the potential benefit of the CABR score in terms of patient flow through the ED and costs associated with a more rapid exclusion of NSTEMI provided by its use. A second limitation of this work relates to analytical issues. Current anti-ApoA-1 IgG and anti-PC IgM assessments take several hours and have too a long turn-around time to meet the conventional 60-min requirement for emergency tests; however, this problem could easily be overcome in the near future by the development of fully automated tests [3]. Third, as most of our cut-off values (except for anti-ApoA-1 IgG) have been defined in a post hoc fashion based on ROC curve analyses, these preliminary results and cut-off values must be further confirmed and validated in larger multicentre clinical trials before any clinical recommendations can be made. Nevertheless, the optimal anti-PC IgM cut-off defined in the present study was identical to the value described for MI prediction in a general population-based study [16], suggesting that our post hoc cut-off definition was adequate. Fourth, whether the CABR score can outperform or provide incremental information in addition to other promising biomarkers, such as brain natriuretic peptide, high-sensitivity cTn or copeptin for early rule-out of myocardial ischaemia [8–10] remains to be demonstrated and also warrants further study. Finally, given the exclusion criteria of this study, the potential of the CABR score with regard to NSTEMI diagnosis may not apply to patients with SLE, APS, RA or HIV.

In conclusion, the results of this hypothesis-generating study demonstrate that the CABR score based on an autoantibody ratio coupled to the NSTEMI-TIMI score seems to be a promising complementary partner to cTnI for rapid rule-out of NSTEMI, provided that anti-apoA-1 IgG and anti-PC IgM assays have a turn-around time suitable for emergency tests. This combination of the anti-ApoA-1 IgG/anti-PC IgM ratio and the NSTEMI-TIMI score improved the predictive accuracy for NSTEMI and for subsequent cTnI elevation, when compared to either candidate alone. Because of its good NPV for NSTEMI diagnosis (97%) and subsequent cTnI elevation (99%), a CABR score ≤3.3 could accelerate patient discharge from the ED by removing the need for prolonged monitoring and blood sampling for the majority (60–70%) of patients. These preliminary data should be confirmed in larger multicentre prospective trials, and further cost-effectiveness studies are required to determine the impact of CABR on both patient management in the ED and its related costs. These preliminary results also further support the growing body of evidence indicating that these two autoantibodies could be of clinical value for NSTEMI diagnosis.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References

Methodological support for the assessment of the IDI was provided by C. Combescure from the Clinical Research Centre, (University of Geneva and Geneva University Hospitals). This work was supported by the Lucie and Ernst Schmidheiny, Telemaque, Gustave and Simone Prevot and DeReuters foundations (to NV).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Material and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. Conflicts of interest
  9. References
  • 1
    Thygesen K, Alpert JS, White HD. Universal definition of myocardial infarction. J Am Coll Cardiol 2007; 50: 217395.
  • 2
    Alpert JS, Thygesen K, Antman EM, Bassand JP. 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: 95969.
  • 3
    Bassand JP, Hamm CW, Ardissino D et al. Guidelines for the diagnosis and treatment of non-ST-segment elevation acute coronary syndromes. Eur Heart J 2007; 28: 1598660.
  • 4
    Macrae AR, Kavsak PA, Lustig V et al. Assessing the requirement for the 6-hour interval between specimens in the American Heart Association classification of myocardial infarction in epidemiology and clinical research studies. Clin Chem 2006; 52: 8128.
  • 5
    Pope JH, Aufderheide TP, Ruthazer R et al. Missed diagnoses of acute cardiac ischemia in the emergency department. N Engl J Med 2000; 342: 116370.
  • 6
    Forberg JL, Henriksen LS, Edenbrandt L, Ekelund U. Direct hospital costs of chest pain patients attending the emergency department: a retrospective study. BMC Emerg Med 2006; 6: 28.
  • 7
    Scirica BM, Sabatine MS, Jarolim P et al. Assessment of multiple cardiac biomarkers in non-ST-segment elevation acute coronary syndromes: observations from the MERLIN-TIMI 36 Trial. Eur Heart J 2010; 31: 19932005.
  • 8
    Apple FS, Pearce LA, Chung A, Ler R, Murakami MM. Multiple biomarker use for detection of adverse events in patients presenting with symptoms suggestive of acute coronary syndrome. Clin Chem 2007; 53: 87481.
  • 9
    Reichlin T, Hochholzer W, Stelzig C et al. Incremental value of copeptin for rapid rule out of acute myocardial infarction. J Am Coll Cardiol 2009; 54: 608.
  • 10
    Reichlin T, Hochholzer W, Bassetti S et al. Early diagnosis of myocardial infarction with sensitive cardiac troponin assays. N Engl J Med 2009; 361: 85867.
  • 11
    Roux-Lombard P, Pagano S, Montecucco F, Satta N, Vuilleumier N. Auto-antibodies as emergent prognostic markers and possible mediators of ischemic cardiovascular diseases. Clin Rev Allergy Immunol 2010. [Epub ahead of print].
  • 12
    Vuilleumier N, Rossier MF, Pagano S et al. Anti-apolipoprotein A-1 IgG as an independent cardiovascular prognostic marker affecting basal heart rate in myocardial infarction. Eur Heart J 2010; 31: 81523.
  • 13
    Vuilleumier N, Bas S, Pagano S et al. Anti-apolipoprotein A-1 IgG predicts major cardiovascular events in patients with rheumatoid arthritis. Arthritis Rheum 2010; 62: 264050.
  • 14
    Montecucco F, Vuilleumier N, Pagano S et al. Anti-apolipoprotein A-1 auto-antibodies are active mediators of atherosclerotic plaque vulnerability. Eur Heart J 2011; 32: 41221.
  • 15
    de Faire U, Su J, Hua X et al. Low levels of IgM antibodies to phosphorylcholine predict cardiovascular disease in 60-year old men: effects on uptake of oxidized LDL in macrophages as a potential mechanism. J Autoimmun 2010; 34: 739.
  • 16
    Gronlund H, Hallmans G, Jansson JH et al. Low levels of IgM antibodies against phosphorylcholine predict development of acute myocardial infarction in a population-based cohort from northern Sweden. Eur J Cardiovasc Prev Rehabil 2009; 16: 3826.
  • 17
    Antman EM, Cohen M, Bernink PJ et al. The TIMI risk score for unstable angina/non-ST elevation MI: a method for prognostication and therapeutic decision making. JAMA 2000; 16: 284.
  • 18
    Vuilleumier N, Charbonney E, Fontao L et al. Anti-apolipoprotein A-1 IgG are associated with high oxidised low-density lipoprotein levels in acute coronary syndrome. Clin Sci (London) 2008; 115: 2533.
  • 19
    Thygesen K, Alpert JS, White HD. Joint ESC/ACCF/AHA/WHF task force for the redefinition of myocardial infarction. Universal definition of myocardial infarction. Eur Heart J 2007; 28: 252538.
  • 20
    DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 1988; 44: 83745.
  • 21
    Pencina MJ, D’Agostino Sr RB, D’Agostino Jr RB, Vasan RS. Evaluating the added predictiveability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med 2008; 27: 15772.
  • 22
    Nendaz MR, Perrier A. Théorème de Bayes et rapports de vraisemblance. Rev Mal Respir 2004; 21: 3947.
  • 23
    Jiménez D, Uresandi F, Otero R et al. Troponin-based risk stratification of patients with acute non-massive pulmonary embolism: systematic review and meta-analysis. Chest 2009; 136: 97482.
  • 24
    Hanley JA, Mc Neil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982; 143: 2936.
  • 25
    Pepe MS, Jane H, Longton G, Leisenring W, Newcomb P. Limitations of the odds ratio in gauging the performance of a diagnostic, prognostic or screening marker. Am J Epidemiol 2004; 159: 88290.
  • 26
    Cook NR. Methods for evaluating novel biomarkers – a new paradigm. Int J Clin Pract 2010; 64: 17237.
  • 27
    Vuilleumier N, Bratt J, Alizadeh R, Jogestrand T, Hafström I, Frostegård J. Anti-ApoA-1 IgG and oxidized LDL are raised in rheumatoid arthritis (RA): potential associations with cardiovascular disease and RA disease activity. Scand J Rheumatol 2010; 39: 44753.
  • 28
    Caligiuri G, Nicoletti A, Poirier B, Hansson GK. Protective immunity against atherosclerosis carried by B cells of hypercholesterolemic mice. J Clin Invest 2002; 109: 74553.
  • 29
    Hess EP, Agarwal D, Chandra S et al. Diagnostic accuracy of the TIMI risk score in patients with chest pain in the emergency department: a meta-analysis. CMAJ 2010; 182: 103944.
  • 30
    Campbell CF, Chang AM, Sease KL et al. Combining thrombolysis in myocardial infarction risk score and clear-cut alternative diagnosis for chest pain risk stratification. Am J Emerg Med 2009; 27: 3742.
  • 31
    Vorlat A, Claeys MJ, De Raedt H et al. TIMI risk score underestimates prognosis in unstable angina/non-ST segment elevation myocardial infarction. Acute Card Care 2008; 10: 269.