BNP but Not s-cTnln Is Associated with Cardioembolic Aetiology and Predicts Short and Long Term Prognosis after Cerebrovascular Events

Background We analyzed the prognostic value of b-type natriuretic peptide (BNP) and sensitive cardiac Troponin (s-cTnI) in patients with ischemic stroke or transient ischemic attack (TIA) and their significance in predicting stroke aetiology. Methods In a prospectively enrolled cohort we measured BNP and s-cTnI levels upon admission. Primary endpoints were mortality, unfavorable functional outcome and stroke recurrence after 90 days and after 12 months. Secondary endpoint was cardioembolic aetiology. Results In 441 patients BNP but not s-cTnI remained an independent predictor for death with an adjusted HR of 1.2 (95% CI 1.1–1.4) after 90 days and 1.2 (95% CI 1.0–1.3) after one year. The comparison of the Area under Receiver Operating Characteristic (AUROC) of model A (age, NIHSS) and model B (age, NIHSS, BNP) showed an improvement in the prediction of mortality (0.85 (95% CI 0.79–0.90) vs. 0.86 (95% CI 0.81–0.92), Log Rank p = 0.004). Furthermore the category free net reclassification improvement (cfNRI) when adding BNP to the multivariate model was 57.5%, p<0.0001. For the prediction of functional outcome or stroke recurrence both markers provided no incremental value. Adding BNP to a model including age, atrial fibrillation and heart failure lead to a higher discriminatory accuracy for identification of cardioembolic stroke than the model without BNP (AUC 0.75 (95% CI 0.70–0.80) vs. AUC 0.79, (95% CI 0.75–0.84), p = 0.008). Conclusion BNP is an independent prognostic maker for overall mortality in patients with ischemic stroke or TIA and may improve the diagnostic accuracy to identify cardioembolic aetiology. Trial Registration ClinicalTrials.gov NCT00390962


Introduction
Several prognostic [1] and some aetiological biomarkers have been evaluated in ischemic stroke but so far most failed to show incremental information. Aetiological classification of ischemic stroke and TIA is critical to discern the best treatment for specific secondary prevention. In a European population-based study the highest recurrence rate after stroke or TIA was found among patient with cardioembolic aetiology when compared to all other types of aetiology [2]. Therefore, early identification of cardioembolic aetiology of stroke is crucial.
B-type natriuretic peptide (BNP) is a neurohormone secreted predominantly from the myocardium in response to wall stretch or increased intracardiac pressure [3][4][5][6]. High BNP levels have been shown to predict atrial fibrillation within the general population as well as in stroke patients [7]. BNP levels were associated with cardioembolic stroke aetiology [8,9]. Furthermore, high plasma levels of BNP were an independent predictor of short and longterm mortality [10,11] and of functional outcome at 6 months after ischemic stroke [12,13]. New published data show that BNP levels are associated with TIA recurrence after a TIA [14]. However, these data base on a small group of patients and need to be verified in a larger cohort. Recently, a meta-analysis found that BNP levels were associated with all cause mortality but not with functional outcome. The incremental prognostic value concerning overall mortality, however, was limited. The authors of this metaanalysis concluded with the remark that further well designed cohort studies are needed to assess the predictive value of BNP in stroke patients [15].
Troponin is a highly sensitive and specific marker of myocardial necrosis and has an important role in the diagnosis of myocardial infarction [16]. Similar to BNP, also elevated s-cTnI has been suggested as marker for increased risk of mortality after an acute stroke [17][18][19]. However data on Troponin in the setting of stroke are more spars. To our knowledge, there are no data on the association of s-cTnI levels and stroke recurrence.
The purpose of this study was, first, to further validate BNP and s-cTnI as prognostic markers after stroke and TIA and, second, to evaluate their ability to identify cardioembolic aetiology.

Study design and setting
The study design of this prospective cohort study has been described in detail elsewhere (ClinicalTrials. gov number, NCT00390962) [20]. Briefly, between November 2006 and November 2007, all consecutive patients (n = 605) presenting at the emergency department of the University Hospital Basel, Switzerland, a 660-bed tertiary care hospital, with a suspected cerebrovascular event were evaluated. All patients presenting with an ischemic stroke or a TIA were included after obtained written informed consent from the patient or the patients' next of kin. We defined initial ischemic stroke according to the World Health Organization criteria as an acute focal neurological deficit lasting longer than 24 hours [21]. TIA was defined accordingly as a transient episode of neurological dysfunction caused by focal brain ischemia with symptoms lasting 24 hours or less. Exclusion criteria were missing informed consent or any diagnosis different from ischemic stroke or TIA (i.e. stroke/TIA mimics) The Ethics Committee of Basel, Switzerland, approved the trial protocol; informed consent was obtained from all patients.
The protocol for this trial and supporting STROBE Checklist are available as supporting information (see Checklist S1 and Protocol S1).

Clinical variables
On admission, the following data were collected with a standardized bed-side interview and complete chart review: vital signs, co-morbidities as assessed by the Charlson Comorbidity Index (CCI) adjusted for stroke, medication prior to ischemic stroke or TIA and cardiovascular risk factors (i.e. age, gender, smoking habits, history of hypercholesterolemia, hypertension, diabetes mellitus, previous TIA or stroke, positive family history for myocardial infarction, stroke or TIA and history of coronary heart disease).
Severity of stroke was prospectively assessed on admission by a neurologist certified in the use of the National Institute of Health Stroke Scale (NIHSS) [22]. Risk stratification for patients with a TIA was performed according to the ABCD2-score [23].
Stroke aetiology was determined according to the criteria of the TOAST classification [24], which distinguishes large-artery atherosclerosis, cardioembolism, small-artery occlusion, other aetiology and undetermined aetiology. All patients underwent the necessary diagnostic work-up to classify aetiology according to the TOAST criteria [24] such as standard 12-lead electrocardiography and at least 24-hour electrocardiography, echocardiography and neurosonographic study of the extra-and intracranial arteries as well as routine laboratory testing.
The 90 days and the 12 months follow-up was performed using structured telephone interviews from a blinded stroke physician to assess mortality from any cause, functional outcome using the modified Rankin Scale (mRS) and recurrent cerebrovascular events.

Blood sampling
In all patients blood samples were obtained on admission within 72 hours from symptom onset. BNP and sensitive cardiac Troponin T were measured in a batch analysis blinded to clinicians. After centrifugation, samples were frozen at 270uC until they were assayed in a blinded fashion in a dedicated core laboratory. S-cTnI was measured in EDTA-plasma using the recently refined s-cTnI assay (Abbott-Architect, Abbott Laboratories, Abbott Park, IL, USA) [25][26][27]. For this test, a limit of detection (LoD) of 0.01 mg/L, a 99 th -percentile cut-off of 0.028 mg/L and a coefficient of variation (CV) of ,10% of 0.032 mg/L have been reported by the manufacturer. Three s-cTnI strata were predefined: below the limit of detection (, 0.01 mg/L, undetectable), detectable but still in the normal range (0.01-0.027 mg/L), and increased ($0.028 mg/L, above the 99th percentile of healthy individuals). The concentration of BNP was measured in EDTA-plasma using the Architect BNP assay (Abbott Laboratories, Abbott Park, IL, USA). The analytical range as reported by the manufacturer extends from 15 to 20'000 pg/ml for the AxSYM assay.

Endpoints
The primary endpoints of this analysis were a) mortality within 90 days and 1 year in stroke as well as TIA patients b), functional outcome after 90 days and at 1 year in stroke patients, c) stroke or TIA recurrence within 90 days.
Patients and or their caregivers were questioned by structured telephone interviews from blinded stroke physicians at 3 months and 1 year after the index event to assess mortality from any cause, functional outcome using the modified Rankin Scale (mRS) and recurrent cerebrovascular events. Poor outcome was defined as mRS.2 points. A recurrent cerebrovascular event was defined as an as an acute ischemic lesion in the brain not attributed to infection, tumor, demyelination or a degenerative neurologic disease but due to an occlusive vascular disorder. Further criteria were rapid onset of a focal neurological deficit occurring for at least 24 hours in conjunction with brain imaging consistent with acute ischemic stroke. The CT or MRI may either show a new infarct or no change from the study performed at entry, i.e. the diagnosis is clinical and does not require CT/MRI confirmation.
The secondary endpoint was cardioembolic aetiology of stroke diagnosed during hospitalization according to TOAST criteria [24] by stroke physicians, which were blinded to BNP and Troponin measurements.

Neuroimaging
CCT was performed in all patients on admission mainly to exclude intracranial hemorrhage. Additionally MRI was per-formed in 197 (45%) patients on a clinical 1.5 T MR Avant system (SIEMENS, Erlangen, Germany) using a stroke protocol, including T1-, T2-, and diffusion-weighted (DWI) sequences, as well as apparent diffusion coefficient (ADC) maps and a MR angiography. Lesions were ranked into three sizes to represent typical stroke patterns: (1) small lesion with a volume of less than 10 ml, (2) medium lesion of 10 to 100 ml and (3) large lesion with a volume of more than 100 ml [28].

Statistical analysis
Discrete variables are expressed as frequency (percentage), continuous Gaussian variables as means with standard deviation SD and non-Gaussian variables as medians with interquartile ranges [IQR]. Comparisons between groups were made using chi-square test, Mann-Whitney U test and Kruskal-Wallis test as appropriate.
Univariate Cox Proportional Hazard Analysis was utilized to compute hazard ratios and 95% confidence intervals (CI) for potential predictors of death during the follow-up period. All significant variables with a p-value ,0.001 (i.e. BNP levels, age, NIHSS, and lesion size) were tested in a multivariate model using the Forward Stepwise Method to avoid over fitting of the model.
To estimate association of BNP and Troponin with functional outcome and re-events, we used logistic regression models. In this analysis we tested all significant variables in the multivariate analysis with a p-value ,0.05 (i.e. BNP levels, age, NIHSS, Charlson Comorbidity Index, CRP, heart failure, atrial fibrillation and lesion size). Due to a greater number of endpoints the risk of over fitting was minimal. Receiver operating characteristic (ROC) curves were constructed for different models to assess the discriminative value in terms of sensitivity and specificity for mortality, outcome and aetiology. Comparison was made using the DeLong test [29] and likelihood ratios for nested models as appropriate. Furthermore we calculated the continuous (category free) net reclassification improvement (NRI) [30] for the primary endpoint, i.e. mortality to assess the incremental value of BNP over the multivariate prognostic model, which included the following variables: age and NIHSS. For the secondary endpoint, i.e. cardioembolic aetiology, the NRI was calculated to assess the incremental value of BNP over the multivariate model including age, atrial fibrillation and heart failure. Analyses were performed using SPSS (SPSS 21, Chicago, IL, USA) and MedCalc (MedCalc 10.4, Mariakerke, Belgium), p-values less than 0.05 were considered to indicate statistical significance.

Patients
From 605 screened patients, 469 patients were eligible for our study and 466 completed follow up after 90 days or one year, respectively. In 362 of the 469 patients ischemic stroke was diagnosed, while 107 out of the 469 had a TIA. 28 patients were excluded from the final analysis because BNP and s-cTnI levels were missing (Figure 1). Baseline demographic and cardiovascular risk factors including NIHSS and lesion size were not different when comparing patients with and without available measurements.

Baseline characteristics
The median age was 74.6 (IQR 62.6-81.9) years and 42% were women. On admission, the median NIHSS in stroke patients was 5 points (IQR 2-10) and median ABCD2 Score in TIA patients was 3 (IQR 3-5). After 90 days, overall 42 patients had died and the mortality rate was thus 9.5%. After the one-year follow-up, 67 patients had died and the mortality rate was 15.2%. After 90 days 138 (31.3%) of patients with ischemic stroke had an unfavorable outcome after one year 141 patients (32.0%) of 336 patients with a completed follow up had an unfavorable outcome (mRS.2). Re-events were identified in 27 (6.1%) patients with ischemic stroke or TIA.
The principal baseline characteristics of all patients are provided in Table 1. There were no significant differences between stroke and TIA patients concerning demographic and risk factor variables besides the etiologies. In patients with TIA we found more undetermined etiologies (50.1%) than in patients with an ischemic stroke (24.0%) (P,0.0001) and in patients with an ischemic stroke we found more patients with a cardioembolic source (37.4%) than in patients with a TIA (8%) (p,0.0001). In multivariate cox regression analysis BNP but not s-cTnI remained an independent predictor for death within 90 days with a hazard ratio (per 100pg/ml unit's increase) of 1.2 (95% CI 1.1-1.4). The other independent predictors for death after 90 days were age and lesion size (see Table 2). Also for long term mortality (within 1 year) only BNP levels (HR 1.2 (95% CI 1.0-1.3), age (per 10 years increase) (HR 3.43 (95% CI 1.6-7.3) and lesion size (HR 1.2 (95% CI 1.0-1.3) remained independent predictors (see Table 3). In a multivariate analysis including coronary heart disease, atrial fibrillation and heart failure instead of the CCI, the hazard ratio of BNP remained similar for 90 days and for one year (HR 1.2 (95% CI 1.0-1.3) and (HR 1.2 (95% CI 1.0-1.3), respectively) while the other parameters did not remain significant in the multivariate model. BNP was predictive also across all subgroups stratified by presence of atrial fibrillation (HR of BNP among pat with AF 1.16, p = 0.01 after 90 days and HR 1.14, p = 0.005 after one year respectively, without AF 1.19, p,0.001 after 90 days and HR 1.18, p,0.001 after one year respectively) or heart insufficiency (HR of BNP in patients with heart insufficiency 1.12, p,0.001 after 90 days and HR 1.12, p,0.001 after one year respectively, HR of BNP in patients without heart insufficiency 1.24, p,0.001 after 90 days and HR 1.16, p,0.001 after one year respectively), thus we found no relevant effect modification.

Cardiac biomarkers and functional outcome
For functional outcome only patients with an ischemic stroke (n = 344) were considered. BNP levels in patients with an unfavorable outcome after 90 days and after one year were significantly higher than those in patients with a favorable outcome (150.0 [IQR 57. 3 Table 4) and 0.008 [IQR 0.001-0.0285] vs. 0.003 [IQR 0.000-0.0105] mg/L, (p = 0.0008) for 1 year) (see Table 5). However, after adjustment for all other significant outcome predictors, the only independent outcome predictors were the NIHSS, age and lesion size (see Table 4 and Table 5).

Cardiac biomarkers and cerebrovascular recurrence
BNP levels in patients with a cerebrovascular re-event were similar as compared to patients without a re-event (123. 3

Discussion
In this prospective study we evaluated both, BNP and s-cTnI levels, in a large cohort of ischemic stroke and TIA patients for prediction of mortality, functional outcome and re-events after 90 days and 1 year as well as for stroke aetiology.
Our data show the following main findings: First, BNP and s-cTnI levels on admission were elevated in patients with ischemic stroke or TIA who had died or had an unfavorable outcome 90 days and one year after the event. After adjustment for other risk factors BNP but not s-cTnI was an independent predictor of death but not of functional outcome or recurrence. Second, BNP improved the identification of patients with cardioembolic strokes or TIAs when compared to clinical information on admission alone (i.e. age, CHF and the history of atrial fibrillation).
Our data showing an independent association of BNP with mortality is mostly in line with findings in the literature. Studies reported high levels of natriuretic peptides and an association with mortality in acute stroke patients at admission [11], during hospitalization [10] and within 6 months [13] after stroke and a recent meta-analysis concluded that natriuretic peptides are independent predictors for all cause mortality after stroke [15]. However a recent study found no incremental value of BNP for the prediction of mortality compared to clinical variables such as age and NIHSS [11]. This discrepant finding might be due to methodological differences. The cohort included hemorrhages and the primary endpoint was mortality or neurological worsening in the stroke unit, thus within the first days after stroke. There was no assessment of several potentially influencing medical comorbidities or of stroke aetiology. In contrast, in our study we assessed only ischemic stroke and TIA patients and the primary endpoint was mortality within 90 days and 1 year. Our data contained more moderate strokes and thus our cohort had a lower level of NIHSS at admission. We took into account stroke aetiology and several medical comorbidities which are known to affect mortality and functional outcome in stroke patients [32]. Importantly, we adjusted for renal insufficiency (within the CCI), which is known to affect BNP levels [33].
Data regarding the predictive value of BNP for functional outcome in patients after a cerebrovascular event are more controversial. Some studies found an association with functional outcome [13,34], other studies showed that especially after adjustment in multivariate analyses BNP did not remain an independent prognostic marker for functional outcome [11,35]. This is in line with our results. It is biologically plausible since functional outcome is probably more prominently influenced by other factors such as brain plasticity. On the other hand, hemodynamic changes and underlying cardiac disease are major contributors to mortality after stroke [36]. Data supporting the association of natriuretic peptides with cardioembolic stroke cumulate. Elevated BNP levels in acute cerebral infarction were associated with cardioembolic aetiology in several studies [8,37]. However, the incremental value of BNP when added to clinical variables, which are known to be associated with cardioembolic events such as age, has not been evaluated in these studies. Our data show that the combination of BNP with age and atrial fibrillation (i.e. history of or a newly diagnosed atrial fibrillation) had a higher discriminatory accuracy in comparison to the same model without BNP. BNP may therefore have a role in risk stratification predicting mortality on one hand and may support early specific implementation of secondary prevention such as oral anticoagulants on the other hand.
Midregional proANP (MR-proANP), another natriuretic peptide, was assessed for its predictive role concerning mortality, functional outcome and cardioembolic aetiology in patients with an ischemic stroke but not TIA [38]. As BNP, MR-proANP was an independent predictor of mortality after 90 days and improved the prognostic value of the NIHSS. Moreover it was also associated with cardioembolic aetiology improving the diagnostic accuracy of already known clinical information, such as age, known heart failure, and known AF on admission. However, BNP compared to MRproANP is more widely available and clinically established in most countries and physicians are already familiar with it. In this study we evaluated BNP with further statistical methods such as the cfNRI and found a significant improvement in the prediction even if added to models including not only the NIHSS but all significant predictors. However a direct comparison is needed in the future to establish which natriuretic peptide is better for risk classification and identification of cardioembolic strokes. In our cohort, s-cTnI was not an independent marker for all three endpoints. Recently higher Troponin T levels were found in older patients with a more severe ischemic stroke and with comorbidities such as heart failure and/or renal insufficiency [39]. Another study showed that elevated levels of high sensitive Troponin T are frequently present in patients with acute ischemic stroke but, in accordance to our results, did not provide additional prognostic information in these subjects [40]. Interestingly s-cTnI did also not provide additional information for the identification of cardioembolic aetiology in our study. It may be that s-cTnI is too specific for myocardial infarction; it is therefore possible that by measuring s-cTnI levels patients with underlying cardiac disease not directly due to myocardial damage are missed.
To our knowledge, this is the first study that assesses BNP and s-cTnI for prediction of stroke recurrence. Our results show that neither BNP nor Troponin levels were associated with a re-event after ischemic stroke or TIA. Possibly, these markers have no role in stroke recurrence. Alternatively, our study was underpowered to find a significant association with recurrence. Nevertheless our findings suggest that even if there is an association it is of rather moderate clinical relevance.