Brain Natriuretic Peptide and Discovery of Atrial Fibrillation After Stroke

Introduction

Atrial fibrillation (AF) is the most frequent cardiac arrhythmia with a prevalence up to 3% which may rise further in increasingly aging societies.¹ AF is one of the most important stroke risk factors and up to 20% of ischemic strokes are attributed to AF.² Furthermore, AF-related strokes are more severe and associated with increased mortality and recurrence rates.³ There is an ongoing discussion on whether the exact rate of thromboembolic events in paroxysmal AF (pAF) is similar^4,5 or lower compared with permanent AF.^6,7 Nevertheless, the detection of both pAF and permanent AF is of vital importance for patients with acute ischemic stroke, since recurrent events can be prevented effectively by changing the secondary preventive approach from thrombocyte inhibitors to oral anticoagulation.³

Yet, especially the diagnosis of brief and asymptomatic paroxysmal episodes of pAF can be challenging.⁸ Although it is well-documented that enhanced and prolonged Holter-ECG monitoring (EPM) significantly improves the detection rate of pAF compared with usual care,^9,10 EPM is often assumed to be uncomfortable for the patient and data analysis is very time-consuming. Furthermore, the cost-effectiveness of EPM in patients with stroke has been questioned,¹¹ and health system’s resources are limited. Thus not every patient can receive EPM in a real-life-scenario.

Hence, there is a need to identify patients at a high risk for pAF to reduce the number of patients requiring prolonged monitoring. Various scores have been developed to determine the AF risk comprising either cardiovascular risk factors, pathological echocardiographic findings, racial and geographic backgrounds, or biomarkers.^12,13 Nevertheless, their application in everyday clinical routine is limited.

Brain natriuretic peptide (BNP) is a neurohormone that is synthesized and released either by the ventricular and atrial myocardium under conditions of myocardial stretch or by the brain after ischemic stroke.^14,15 It has been found that natriuretic peptides may be elevated in patients with pAF or persistent AF in many different clinical settings as well as in the general population.^16–19 One promising approach in clinical stroke work-up could be the simple analysis of natriuretic peptides which have proven to be elevated in patients with AF.^12,20–22

Therefore, we hypothesized that BNP may provide useful information to identify ischemic stroke patients at high risk for pAF to efficaciously select patients for EPM.

Methods

The basic data for the findings of this study are available from the corresponding author upon reasonable request. Find-AF_RANDOMISED was registered. The current subanalysis was not prespecified.

Study Design and Patient Population

Find-AF_RANDOMISED was an investigator-initiated prospective, randomized, controlled, open-label multicenter study, which included acute ischemic stroke patients ≥60 years of age irrespective of the suspected stroke cause, with sinus rhythm on admission ECG. Results of the study and protocol details have been published recently.⁹

Consecutive patients were included at 4 study sites in Germany between May 2013 and August 2014. Patients were randomized 1:1 to either EPM by means of repeated 10-day Holter-ECG monitoring (at randomization, after 3 and after 6 months) or the respective in-house standard-of-care procedures of diagnostic stroke cause work-up. The primary outcome parameter was AF, which was defined as at least one episode of ≥30 seconds duration, judged by an end point committee that was blinded to all clinical data including BNP plasma levels.

For the present analysis, patients underwent BNP sampling at randomization and 3 months thereafter. The time point of the second BNP analysis was chosen to exclude higher BNP levels due to an increased release by the brain after a recent cerebral infarction. Plasma levels of BNP were measured by means of a sandwich chemiluminescence immunoassay on the ADVIA Centaur (Bayer Diagnostics, Munich, Germany). Personnel responsible for the determination of BNP levels were blinded to clinical patient data including all Holter-ECG monitoring results. Although a specific cut off for BNP was not predetermined, we chose a cut off of BNP ≥100 pg/mL for investigating the yield of prolonged monitoring according to current literature.^23,24 Furthermore, different cut offs were calculated to justify the chosen BNP cut off.

The study complies with the Declaration of Helsinki. The protocol was approved by the local ethics committee of each study site, and all patients gave written informed consent for study procedures, including Holter-ECG monitoring, BNP testing, and data collection for all data reported here.

Statistics

The statistical analyses were performed with SPSS version 24.0 (SPSS, Inc). Continuous values are given as mean ± SD, nominal variables as count and percentages. Non-normally distributed variables are expressed as median values with the corresponding interquartile range (IQR). Normally distributed data were compared by Student t test, not normally distributed data by Mann-Whitney U test. A 2-sided P≤0.05 was considered to indicate a statistically significant difference. The predictive ability of BNP for AF was checked by receiver operating characteristic curves. The area under the curve (AUC) was calculated and tested. The number needed to screen was calculated as 1/(AF rate_{(EPM arm)}−AFrate_{(usual care arm)}). To assess how meaningful the National Institutes of Health Stroke Scale (NIHSS) and BNP are for the prediction of AF, we computed C statistics for the Cox regression model predicting AF from NIHSS, BNP, and both. The 95 CI for the C statistics and their pairwise differences as well as the P values for the null hypotheses that a C statistic equals 0.5 or a difference of 2 C statistics equals zero was computed by the jackknife method.

Results

A total of 402 patients were enrolled and randomized. Four patients were randomized erroneously and in 25 patients no BNP sample was available. Thus the data of 373 patients were finally analyzed. In 275 patients (73.7%) BNP was measured at baseline (immediately after randomization) and additionally at the 3 months visit (median duration between symptom onset and baseline 3 days [IQR, 2–5 days] and for the 3 months visit 95 days [IQR, 90–102 days], respectively).

Of the 373 patients with a BNP sample at baseline, 186 patients were randomized to EPM and 187 patients to usual care. pAF was detected in 34 patients (26 patients by means of EPM and 8 patients by means of usual care, 76.5% versus 23.5%, P=0.001). Details of pAF detection by EPM have been published previously. Sixty-seven percent of pAF cases were detected within the first 10-day Holter ECG, 22% in the second and 4% in the third Holter ECG. The mean time from randomization to detection of AF was 35 days in the EPM group (SD 51 days) and 37 days in the usual care group (SD 60 days, P=0.92).

Baseline characteristics of all patients divided into groups according to the availability of BNP measurements are summarized in Table 1. No statistically significant differences were detected between patients with and without BNP determination.

The analysis of BNP values in patients with paired BNP samples (n=275), showed concentrations that ranged from <10 to 1042.0 pg/mL (median, 24.8 pg/mL, IQR, 11.2–55.5 pg/mL) at baseline and from <10 to 498.6 pg/mL (median, 31.9 pg/mL, IQR, 16.4–66.3 pg/mL) 3 months thereafter (P=0.006).

BNP values at baseline and after 3 months were significantly higher in patients with pAF compared to those without pAF. At baseline the median BNP value was 53.5 pg/mL (IQR, 29.9–175.4 pg/mL) in patients with pAF and 27.9 pg/mL (IQR, 12.0–62.0 pg/mL) in patient without AF (P=0.0002). After 3 months, it was 81.2 pg/mL (IQR, 34.3–143.5 pg/mL) in patients with pAF and 30.1 pg/mL (IQR, 16.0–59.9) in patients without AF (P=0.001).

A more detailed analysis considering pAF duration in the EPM group showed similar BNP values in patients with various durations of pAF episodes (median BNP [IQR] was 53.5 pg/mL [29.0–158.8 pg/mL] for pAF >30 seconds to <6 minutes [n=8]; 35.0 pg/mL [22.0–139.0 pg/mL] for pAF >6 minutes to <6 hours [n=5]; 51.0 pg/mL [47.5–139.5 pg/mL] for pAF >6 hours to <24 hours [n=3] and 97.0 pg/mL [52.8–198.0 pg/mL] for pAF >24 hours [n=8]).

Patients with pAF detected by EPM showed similar BNP values as compared to pAF detected by usual care (Figure 1 and Table 2).

• Download figure
• Download PowerPoint

The diagnostic accuracy of BNP in detecting pAF for the whole cohort was tested by means of a receiver operating characteristic analysis. The AUC was 0.68 (95% CI, 0.58–0.79, P=0.03) regarding BNP at baseline and 0.71 (95% CI, 0.60–0.82, P=0.01) regarding BNP measured after 3 months. We also calculated the AUC for EPM and usual care with respect of heart failure, which varied between 0.72 (EPM group) and 0.61 (usual care group) among patients with heart failure (Figure 2A through 2C and in the online-only Data Supplement).

• Download figure
• Download PowerPoint

Previous publications proposed a BNP cutoff of 100 pg/mL for preselecting patients for prolonged monitoring.²³ In patients with BNP values ≥100 pg/mL at baseline, the pAF detection rate was 32.4% in the EPM arm (11/34 patients), versus 3.6% in the usual care arm (1/28 patients, absolute difference 28.8%). This resulted in a number needed to screen of 3. In patients with BNP values below the cut off of 100 pg/mL, pAF was found in 9.9% of the patients (15/137 patients) by means of EPM and in 4.4% of the patients with pAF (7/152 patients) by means of usual care (absolute difference 5.5%), resulting in a number needed to screen of 18 (Figures 3 and 4).

• Download figure
• Download PowerPoint

• Download figure
• Download PowerPoint

NIHSS at baseline was not significantly different between those patients with detection of pAF detected opposed to those without in either group (median NIHSS 6 [4;9] versus 2 [1;4], P=0.062 in the control group and 4 [2;4] versus 3 [1;5], P=0.533 in the intervention group). Cox models for the prediction of AF by NIHSS, BNP, and both and their C statistics are displayed in Table I in the online-only Data Supplement. Both variables were associated with incident AF. Data suggest that BNP might be the ultimate predictor while NIHSS does not add predictive information (even if the formal significance in the comparison of C statistics was not reached.

Discussion

Major Findings

Within the current analysis of the Find-AF_RANDOMISED study cohort, we evaluated whether increased BNP plasma levels are helpful to select patients, who benefit from EPM with respect to the detection of pAF. We observed significantly elevated BNP plasma levels in acute ischemic stroke patients ≥60 years of age with pAF detected within 6 months after the index stroke. Although pAF detection rates were significantly lower in the usual care group as has been published previously,⁹ there was no difference between BNP plasma levels between the 2 randomization groups. This finding leads to the conclusion that pAF, which was only detected by EPM, leads to the same release of BNP as easily detectable AF, found by usual care monitoring. Patients with elevated BNP ≥100 pg/mL benefitted more from EPM than patients with BNP <100 pg/mL (number needed to screen 3 versus 18, respectively), which indicates that patients with BNP ≥100 pg/mL measured 3 days after the index stroke should definitely undergo EPM.

Optimal Time Point for BNP Determination to Identify pAF After Stroke

Several studies have found elevated levels of natriuretic peptides in stroke patients with AF as compared to patients without AF.^22,25–27 The median BNP in patients with AF in our cohort was 53.5 pg/mL (IQR, 29.9–175.4 pg/mL) at baseline and 81.2 pg/mL (IQR, 34.3–143.5 pg/mL) after 3 months which is lower than previously reported.^22,25–27 Apart from differences in comorbidities and patients’ characteristics 2 further aspects may explain this difference: first, BNP rises early after stroke in patients with or without AF, presumably explained by BNP release from the brain.²² Previous studies mostly measured BNP within 24 hours after stroke,^18,28 whereas BNP was measured 3 days after stroke in our study. BNP measured within 24 hours after stroke in patients with AF has been found to be nearly twice as high as 3 days later (173.2 versus 81.9 pg/mL).¹⁵ Given the short half-life of BNP (20 minutes),²⁹ samples taken shortly after the stroke event contain BNP not only from heart but also from the damaged brain tissue. However, after 3 days, that is, after >200 half-lives, BNP is expected to be predominantly of cardiac origin. We conclude that BNP detected after 3 days is more likely to be associated with underlying AF compared to results from very early blood sampling. This hypothesis is further supported by the fact that BNP levels after 3 months were slightly higher than those determined 3 days after stroke. If a significant amount of BNP measured 3 days after stroke derived from the brain, we would expect to have measured lower rather than higher BNP levels after 3 months. The slightly more elevated BNP levels after 3 months may partly be explained by a higher and progressive cardiovascular morbidity after hospitalization and by immobility.³⁰

A second explanation for similar levels of BNP in AF versus non-AF patients in our study could be the fact that patients with pAF identified in the EPM arm had AF at an earlier stage with shorter AF duration and were, therefore, associated with lower BNP values as compared to easier to detect AF in the control arm. However, there was no significant difference in BNP levels between AF patients in both trial arms. A similar finding was reported in previous studies^20,31 and also shown in an earlier reported study cohort of our study group.²² We, therefore, conclude that EPM identifies AF patients with a similar activation of BNP release as patients identified by standard ECG monitoring.

Discriminatory BNP Value and Cutoffs

The discriminatory value of BNP for the detection of patients with pAF was only moderate in our collective of stroke patients ≥60 years of age (mean age 72.7±7.6 years) (AUC 0.68 at baseline and 0.71 at 3 months thereafter). We obtained a similar result for the BNP AUC in our previous observational study (best AUC 0.75). In the Find-AF trial, BNP samples were collected within 24 hours after admission and elevated plasma BNP levels predicted the detection of pAF within a 7-day Holter-ECG.²² Other studies described AUCs between 0.82 and 0.89 in patients with stroke or transient ischemic attack, in whom telemetry or Holter-ECG of up to 3 days were performed during hospitalization.^25–27 However, AF detected during limited ECG monitoring probably represent more advanced stages of disease. In contrast, more prolonged ECG monitoring also reveals shorter episodes of pAF, which presumably represents earlier stages of disease that were in sinus rhythm at the time of blood sampling. EPM initiated after the measurement of elevated BNP levels of ≥100 pg/mL may reveal this hard to detect AF, which cannot be detected by usual care Holter-ECG monitoring or telemetry during hospitalization.

Analyzing the ECGs of patients divided into 2 groups with BNP values below or above the cut off of 100 pg/mL, 32.4% of all patients with AF were detected by means of EPM and only 3.6% by means of usual stroke diagnostic work-up (Figure 4). This leads to the conclusion that BNP values ≥100 pg/mL may help to identify stroke patients at special risk for pAF to select patients for EPM effectively. Currently, there is no established optimal cut off value for BNP³² and various publications recommended cut offs for BNP ranging from 65 to 144 pg/mL.^22,25–27 Our cut off of 100 pg/mL lies in the middle of this range and in line with recent consensus recommendations.²³ Yet, the choice of a cut off value for AF risk stratification regarding EPM is a compromise between screening costs for the discovery of patients with AF and risk of recurrent strokes in patients without prevention due to undiscovered AF.

Clinical Relevance of pAF Detected by Short-Term or Continuous Monitoring

Besides Find-AF_RANDOMISED⁹ 2 other randomized trials have compared prolonged monitoring with usual care ECG to detect pAF in stroke patients. The EMBRACE trial (30-Day Cardiac Event Monitor Belt for Recording Atrial Fibrillation After a Cerebral Ischemic Event) included 572 patients with cryptogenic strokes, who underwent additional ECG monitoring with either an externally applied 30-day event recorder or a conventional 24-hour Holter-ECG.³³ The CRYSTAL-AF study (Cryptogenic Stroke and Underlying Atrial Fibrillation) randomized 441 patients with cryptogenic stroke to either an implantable cardiac monitor or usual care.⁹ Significantly more cases of pAF were found in the intervention arms of all 3 studies as compared to the usual care arms. The minimum duration of clinical relevant pAF episodes probably depends on the duration and type of monitoring. We suggest that even short AF episodes of >30 seconds during a 10-day Holter-ECG monitoring are precursors of clinically relevant AF. This is supported by the finding of no additional AF cases detected within 6 to 12 months after randomization in the EPM arm in comparison to 3 cases in the usual care arm in Find-AF_RANOMISED. In CRYSTAL-AF, pAF episodes >2 minutes were detected in 30% of patients randomized to an implantable loop recorder within 3 years after randomization. However, the pAF detection rate in the control arm was only 3% after 3 years. Observation trials over 30 months in patients without prior stroke but with increased cardiovascular risk profile have established that continuous monitoring may find AF >2 minutes in up to 40% of patients. Only 10% of these episodes lasted >24 hours.³⁴ The ASSERT study (Asymptomatic Atrial Fibrillation and Stroke Evaluation in Pacemaker Patients and the Atrial Fibrillation Reduction Atrial Pacing) showed only pAF episodes exceeding 24 hours increase the risk of stroke in stroke-naïve patients with implanted pacemakers or defibrillators.³⁵ Hence, we propose a cut off of clinically relevant AF of 30 seconds in patients monitored for ≤10 days, while a cut off of 24 hours may be appropriate for stroke-naïve patients undergoing continuous monitoring.

Clinical Implications

Interestingly, BNP is currently only considered for the diagnosis of heart failure³⁶ but not mentioned in AF guidelines.^32,37 Our results indicate that prolonged AF monitoring may be useful in patients with elevated BNP levels as it may reflect undetected AF. In stroke patients, the clinical consequence of elevated BNP levels is currently unknown. BNP is a predictor of future stroke²⁸ and our analysis shows that patients with elevated BNP >100 pg/mL represent a group of patients in whom EPM is very effective. Our calculated AUC in patients with or without heart failure does not differ significantly, so that BNP with a cut off value of ≥100 pg/mL can be used to preselect stroke patients for EPM irrespective of a heart failure diagnosis.

The potential of BNPs to predict AF is also supported by a retrospective subgroup of the WARSS trial (Warfarin-Aspirin Recurrent Stroke Study), which showed that (unlike the entire cohort) a subgroup of stroke patients with elevated natriuretic peptide levels benefit from anticoagulation as compared to aspirin. Hence, in this subgroup with increased biomarker levels, the number of patients with undetected AF might have been substantially higher.³⁸

Strengths and Limitations

The strengths of the present analysis of Find-AF_RANDOMISED include the randomized, controlled, and prospective multicenter trial design, a large sample size, the real-life-scenario of patients with acute ischemic stroke included irrespective of the suspected stroke cause, the very intensive cardiac monitoring by means of three 10-day Holter-ECGs, the 2 time points of BNP sampling, and the high rate of BNP samples drawn at baseline (93.7%). The lack of generalizability to populations not included in the study—particularly patients younger than 60years of age or of ethnic groups—is a limitation. Previous observational data have shown that the number needed to screen increases with younger age.^39,40

In this study, we only analyzed BNP values, but not other natriuretic peptides, because BNP showed best diagnostic characteristics in a head-to-head comparison to NT-proBNP and NT-proANP.^20,22 However, we cannot exclude that other parameters (eg, mid-regional pro-atrial natriuretic peptide) may yield different results.

Conclusions

In summary, acute ischemic stroke patients ≥60 years of age with newly detected AF showed significantly elevated BNP levels compared to patients without AF. The subgroup with BNP levels ≥100 pg/mL measured 3 days after stroke greatly benefitted from prolonged heart rhythm monitoring in terms of pAF detection.

Footnotes