Ejection fraction, B‐type natriuretic peptide and risk of stroke and acute myocardial infarction among patients with heart failure

Background Real‐world data on the clinical outcomes of heart failure (HF) across the spectrum of ejection fraction (EF) and the prognostic value of B‐type natriuretic peptide (BNP) have not been well examined. Hypothesis The real‐world association between the clinical outcomes of HF and EF or BNP levels may differ across different EF or BNP values. Methods The Optum Integrated Claims‐Clinical data (07/2009‐09/2016) was used to identify adult patients with ≥1 HF diagnosis during hospitalization or emergency room visit. Three EF cohorts were formed: reduced (rEF; EF < 40%), mid‐range (mrEF; EF 40%‐49%), and preserved EF (pEF; EF ≥ 50%). Stratifications by BNP levels were performed using median BNP as cutoff between high vs low BNP (H‐BNP vs L‐BNP). Results In total, 7005 HF patients with EF measurements (2456 patients with both HF and BNP measurements) were identified. rEF patients had higher risk of stroke (hazard ratio [HR] = 1.57, P = 0.010) and acute myocardial infarction (AMI) (HR = 2.42, P < 0.001) compared to pEF patients. H‐BNP was associated with a significantly higher risk of mortality (P < 0.001). rEF patients with H‐BNP had a significantly higher risk of stroke than those with L‐BNP. Conclusions Patients with rEF had a significantly higher rate of stroke and AMI vs pEF patients, as did patients with H‐BNP vs L‐BNP. The present study is the first to show the real‐world association of EF and BNP (alone and in combination) with clinical outcomes, further supporting the recommendation to use these markers in clinical practice. These results may help to guide future recommendations and improve the clinical management of HF.

(2.2%) individuals had HF in the United States. 3,4 The burden of this disease is substantial, and approximately half of patients with HF die from complications ensuing from HF within 5 years following initial diagnosis. 4 HF patients are also at increased risk of cardiovascular events, including ischemic stroke and acute myocardial infarction (AMI). 3,5,6 Comorbidities associated with HF, such as diabetes and coronary artery disease (CAD), are risk factors that may be present in a substantial proportion of HF cases. 7 Current clinical management includes beta-blockers, angiotensin-converting enzyme inhibitors, mineralocorticoid receptor antagonists, and/or diuretics to limit fluid accumulation. 1,8,9 The left-ventricular ejection fraction (EF) is a measurement of the systolic function of HF patients 10 that has been shown to predict cardiovascular risks and mortality. 11 The European Society of Cardiology (ESC) and the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines/Heart Failure Society of America (ACC/AHA/HFSA) recommend using different treatment approaches for HF with reduced (rEF; EF <40%,), mid-range (mrEF; EF 40%-49%), and preserved (pEF; EF ≥50%) EF. 1,9 However, HF diagnosis can be challenging, 12,13 and the prognostic potential of EF appears to be reduced for values above 40% to 45%, 14,15 thereby further complicating the diagnosis of HF for patients with mrEF or pEF. 1,12,13,16 Moreover, the effectiveness of HF-approved therapies has mainly been demonstrated in rEF patients. 1 Thus, additional predictors are needed to more accurately stratify HF patients and improve clinical decision making.
B-type natriuretic peptide (BNP) is a peptide hormone that is now the gold standard diagnostic and prognostic biomarker for HF. 17 Indeed, a systematic review of 19 studies reported that, for every 100 pg/mL rise in BNP concentration, there is a corresponding 35% increase in the risk of death, and recent updates of the ACC/ AHA/HFSA and ESC guidelines recommend using BNP levels in the risk stratification of HF. 1,9 The use of BNP levels in stratifying HF patients is supported by three prospective studies, which collectively demonstrate that BNP levels correlate with EF, 18 and that the prognostic value of BNP is equal or even higher than that of EF. 19,20 However, to the best of our knowledge, the clinical outcomes of HF patients stratified using EF or BNP levels have not been studied in a real-world setting. Furthermore, there are limited data pertaining to the use of BNP to predict events, such as AMI and ischemic stroke in patients with HF. In order to fill this knowledge gap, this US retrospective claims study was conducted to evaluate the association of cardiovascular events (ie, ischemic stroke, AMI) with EF levels and to assess the prognostic value of BNP in a real-world setting.

| Outcome definition
Study outcomes included a primary diagnosis of ischemic stroke or AMI resulting in hospitalization. All-cause mortality was also evaluated as a secondary outcome to account for the competing risk of death in patients with HF. The observation period was defined as the shortest time frame between the 1-year period following the index date and the period spanning from the index date up to the earliest date among end of data availability (September 30, 2016), end of insurance coverage, or death. Of note, for the all-cause mortality analyses, the end of the eligibility period was used as a proxy of the date of death in patients indicated as deceased but without an associated date of death. More specifically, in these patients, the date of death was defined as the last day of the month during which end of eligibility occurred.

| Sensitivity analyses
Sensitivity analyses were performed for HF patients diagnosed with

| Statistical analysis
Baseline characteristics were evaluated using means, medians, and standard deviations (SDs) for continuous variables, and using frequencies and percentages for categorical variables. Kaplan-Meier (KM) rates and log-rank tests evaluated during the observation period were used to compare the cumulative incidence of study outcomes (ie, stroke and AMI) among the different EF cohorts and BNP subgroups. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were calculated at 12 months using Cox proportional hazards models (ie, time-to-event analysis) adjusting for the following covariates evaluated during 18-month baseline period: age, gender, region, race, insurance type, year of index date, baseline hospitalizations, AF, Quan-Charlon comorbidity index (Quan-CCI) score, and CHA 2 DS 2 -VASc score (Appendx S3).

| Outcomes of patients stratified by EF (N = 7005)
Patients with rEF had a significant 1.6-fold higher risk of ischemic stroke compared to patients with pEF during the observation period (ie, HR = 1.57, P = 0.010; Figure 1A). The risk of stroke was not significantly different between rEF vs mrEF and between mrEF vs pEF ( Figure 1A).
Patients with rEF had a significant 2.4-fold higher risk of AMI compared to patients with pEF (ie, HR = 2.42, P < 0.001; Figure 1B).
Similarly, there was not any significant difference in AMI risk between rEF vs mrEF, but there was one between mrEF vs pEF cohorts (ie, HR = 1.83, P < 0.001; Figure 1B Relative to patients with pEF, patients with rEF had a slightly higher risk of all-cause mortality (ie, HR = 1.19, P = 0.015; Figure 1C).
Statistical significance was not reached for the rEF vs mrEF or mrEF vs pEF comparisons for this outcome ( Figure 1C).

| Outcomes of patients stratified by EF and BNP levels (N = 2456)
Among rEF patients, the risk of ischemic stroke was significantly higher for H-BNP patients compared to L-BNP patients (ie, HR = 5.03, P = 0.013; Table 2). Although the risk of ischemic stroke was numerically higher among pEF and mrEF patients with H-BNP, the differences did not reach statistical significance (Ps > 0.05;

| CAD and diabetes subgroups
In   The ESC guidelines also highlight challenges inherent to the diagnosis and treatment of HF patients with pEF. 1 It has been reported that the prevalence of HF with pEF is substantial (~50%) and seemingly increases over time. 26   HF patients without AF also have unmet needs, which may help in guiding the design of future studies and trials.

| Limitations
The present study is subject to a number of limitations. First, due to the inherent nature of insurance claims databases, coding inaccuracies or omissions in procedures and diagnoses could have occurred. Second, despite adjusting for many baseline covariates, the impact of unmeasured confounders cannot be ruled out. Third, this analysis does not differentiate between the etiologies and heterogeneity of HF, which could play a role in the clinical outcomes and patient care.
Fourth, although the closest value to the index date was used to define EF, this measure was allowed to be recorded up to 90 days before or after the index date, a time frame during which EF may vary.
Similarly, the most recent BNP value prior to discharge date of the index hospitalization or ER was used, however, BNP measures could also vary throughout the same hospitalization or ER visit. Finally, the requirement for diagnostic tests like EF and/or BNP may lead to a selection bias for patients with characteristics that may be different from the overall population of patients with HF.

| CONCLUSION
The prognostic value of EF and BNP levels among HF patients remains understudied in the real world. In this retrospective cohort study that used data from a large US insurance claims database, HF patients with rEF were found to have significantly worse clinical outcomes compared to patients with pEF, thereby confirming the reliability of EF in this subpopulation. BNP levels were also a reliable predictor of mortality. Moreover, the prognosis of rEF patients was significantly worse among those who had high BNP levels with respect to stroke risk, and mortality. The present work also contributes to documenting the prognosis of mrEF patients, who were found to have a higher risk of AMI vs pEF patients. This study is the first to show the real-world association of EF and BNP with clinical outcomes, further supporting the recommendation to use these markers in clinical practice. These results may help guide future recommendations and improve the clinical management of HF.

ACKNOWLEDGMENTS
Medical writing assistance was provided by Samuel Rochette, an employee of Analysis Group, Inc. This research was funded by Janssen Scientific Affairs, LLC, Titusville, NJ, United States.