Anticoagulation in patients with atrial fibrillation and heart failure: Insights from the NCDR PINNACLE‐AF registry

BACKGROUND
In non-valvular atrial fibrillation (NVAF) patients, congestive heart failure (CHF) confers an increased risk of stroke or systemic thromboembolism. This risk is present in both heart failure (HF) with reduced ejection fraction (HFrEF) and HF with preserved ejection fraction (HFpEF). It is unclear if clinicians account for both types of CHF in their NVAF anticoagulation practices. Accordingly, we characterized current outpatient anticoagulation trends in NVAF patients with HFpEF compared to patients with HFrEF.


METHODS
The outpatient NCDR PINNACLE-AF registry was analyzed to identify patients with NVAF and CHF. The study population was subdivided into HFpEF (ie, LVEF ≥ 40%) and HFrEF (LVEF < 40%). Anticoagulation rates by CHF group were compared and stratified by CHA2 DS2 -VASc score.


RESULTS
A total of 340 127 patients with NVAF and CHF were identified, of whom 248 136 (73.0%) were classified as HFpEF and 91 991 (27.0%) as HFrEF. Patients with HFpEF had higher mean CHA2 DS2 -VASc scores and were more likely to be female, older, and have hypertension (P < 0.001). Unadjusted anticoagulation rates were significantly lower in patients with HFpEF compared to those with HFrEF (60.6% vs 64.2%, respectively). Lower rates of anticoagulation in the HFpEF group persisted after risk adjustment (RR: 0.93 [95% CI: 0.91, 0.94]). Stratification by CHA2 DS2 -VASc score demonstrated that lower rates of anticoagulation in patients with HFpEF persisted until a score of ≥5.


CONCLUSIONS
Patients with NVAF and HFpEF have significantly lower anticoagulation rates when compared to their HFrEF counterparts. These findings suggest a potential underappreciation of HFpEF as a risk factor in patients with NVAF.


| INTRODUCTION
Atrial fibrillation (AF) is the most common cardiac arrhythmia, with a reported 2% prevalence among all medicare beneficiaries <65 years old and 9% ≥65 years old, and an estimated 2-10% annual stroke risk depending on individual risk factors. 1,2 Anticoagulation strategies for reducing the incidence of cerebral and systemic thromboembolism in the setting of non-valvular AF (NVAF) have relied on risk score systems including the CHADS 2 and the more current CHA 2 DS 2 -VASc scores. The CHA 2 DS 2 -VASc score, which has more discriminatory power in individuals characterized as low risk by CHADS 2 , was described and validated in 2009, and defines congestive heart failure (CHF) as a left ventricular ejection fraction (LVEF) ≤40%, or signs or symptoms of right or left heart failure (HF). 1,[3][4][5] The American College of Cardiology/American Heart Association/Heart Rhythm Society ABBREVIATIONS: CAD, coronary artery disease; CI, confidence interval; CHF, congestive heart failure; DM, diabetes mellitus; HFrEF, heart failure with reduced ejection fraction; HFpEF, heart failure with preserved ejection fraction; HTN, hypertension; LVEF, left ventricular ejection fraction; NVAF, non-valvular atrial fibrillation; RR, relative risk; TIA, transient ischemic attack.
(ACC/AHA/HRS) and European Society of Cardiology (ESC) all recommend using the CHA 2 DS 2 -VASc score to risk stratify individuals with AF to inform anticoagulation decisions. 2,5 Specifically, the 2014 AHA/ACC/HRSA Guidelines for the Management of Patients with AF recommend (Class I) all individuals with a history of prior stroke or TIA, or CHA 2 DS 2 -VASc score of ≥2 be considered for anticoagulation based on their bleeding risk, while the 2016 ESC Guidelines for the Management of AF recommends anticoagulation in females with a CHA 2 DS 2 -VASc score ≥ 2 and males with a score ≥ 1. 1,5 Even though congestive HF has been validated as an independent risk factor for stroke regardless of ejection fraction, the clinical significance of HF with preserved ejection fraction (HFpEF) may be underappreciated. Approximately, 50% of all patients with HF have HFpEF, with cross-sectional studies reporting rates between 40% to 71% depending on the EF cutoff (40%-50%) used to define HFpEF, and longitudinal studies showing an increasing prevalence over time. [6][7][8][9][10][11] Practice patterns regarding the prescription of anticoagulation for patients with NVAF and HFpEF or HFrEF have not been well characterized and guideline adherence remains unknown. Prior studies in the HF population have reported varying rates of anticoagulation in those with NVAF, with one meta-analysis that aggregated unadjusted data from 10 studies showing lower rates in the HFpEF compared to the HFrEF population, and another study, which showed after risk adjustment, equal rates of anticoagulation in inpatients with HFpEF or HFrEF (11,17). The aim of this study was to characterize current outpatient practice patterns regarding the use of anticoagulation in a large cohort of outpatient patients with NVAF and CHF. By characterizing rates of oral anticoagulation in patients with HFpEF compared to HFrEF, we hope to identify gaps between practice patterns and anticoagulation guidelines in patients with AF and HF.

| Definitions and study endpoints
The primary predictor variable was CHF group, with patients classified into two subgroups: HFpEF (LVEF ≥ 40%) and HFrEF (LVEF < 40%). The EF cutoff of 40% was chosen to differentiate between HFpEF and HFrEF because this was used as the objective definition of HF in the CHA 2 DS 2 -VASc score. In addition, although epidemiological studies characterizing HFpEF have had variable definitions for HFpEF (>40% vs >50%), reduced EF was consistently defined as <40%. 6-10 LVEF was obtained from electronic medical record data. CHF was defined as either symptoms or physical exam findings consistent with heart failure, a prior hospitalization with HF as a primary diagnosis, or an LVEF <40%. Patients without LVEF were excluded from the analysis as described above.
The CHA 2 DS 2 -VASc score was calculated for each patient. One point was assigned for the presence of peripheral vascular disease, CHF, diabetes mellitus (DM), hypertension (HTN) and age 6 to -74 years. Two points were assigned if prior history of stroke/TIA or age ≥ 75 years old. As per the NCDR PINNACLE-AF registry data standards, CHF was defined as symptoms of HF including dyspnea, orthopnea, rales, edema, S3, S4, ascites, or JVD, LVEF <40%, or exacerbation of CHF requiring hospitalization. In addition, vascular disease is defined as the presence of peripheral arterial disease, peripheral vascular disease, history of myocardial infarction, and prior interventional, or surgical revascularization. 12 The main outcome parameter of the present study was rates of oral anticoagulation defined as warfarin, dabigatran, rivaroxaban, apixaban, or edoxaban. To better characterize how stroke risk and guideline recommendations impact rates of anticoagulation, secondary analyses included stratification by: (a) CHA 2 DS 2 -VASc score, (b) CHA 2 DS 2 -VASc ≤1 compared to CHA 2 DS 2 -VASc ≥2, and (c) history of TIA or stroke compared to no prior TIA or stroke. 1 To test the robustness of the results, sensitivity analysis excluding those on dual antiplatelets was done. In addition, to understand trends in anticoagulation over time, temporal analysis over the study period was conducted.

| Statistical analysis
Continuous variables were reported as mean ± SD and compared using the Student's t test. Categorical variables were reported as percentages and compared using the χ 2 test or Fisher exact test. Given the large sample size, standardized differences between groups were calculated to report effective size to help with interpreting statistically significant differences. Standardized differences of <0.1 are generally considered to represent an insignificant effective size. 13 Univariate, bivariate, and multivariable logistic regression analyses were performed to assess the effect of HF type (ie, HFrEF or HFpEF),  were identified as patients with HFpEF and HFrEF, respectively. Table 1 shows all patient variables stratified by HF groups.
Patients with HFpEF were significantly older (76.0 ± 11.4 vs 73.8 ± 11.8 years), more likely to be female (45.5% vs 26.6%) and had significantly higher body mass indexes (BMI) (29.5 ± 6.6 vs 28.6 ± 6.2 kg/m 2 ) than patients with HFrEF. CHA 2 DS 2 -VASc score was significantly higher in patients with HFpEF. This was mainly driven by female gender, age, and hypertension (88.7% vs 81.4%, P < 0.001). In contrast, individuals with HFrEF were more likely to have higher rates of surrogates of ischemic heart disease including coronary artery disease, prior history of myocardial infarction, previous coronary artery bypass graft, or being on dual antiplatelet therapy. For the remaining risk factors, there was a statistically significant difference in frequency between both groups, but standardized differences were less than 0.1, suggesting a small effect.
Overall unadjusted rates of anticoagulation were higher in the HFrEF group compared to the HFpEF group (64.2% vs 60.6%), even though mean CHA 2 DS 2 -VASc score was higher in the HFpEF group (P < 0.001). After controlling for patient demographics, clinical variables, and CHA 2 DS 2 -VASc components, individuals with HFpEF still remained significantly less likely to be anticoagulated than those with HFrEF (RR 0.93, 95% CI 0.91-0.94). Age, BMI, HTN, previous cerebral or systemic thromboembolism, and prior percutaneous revascularization were all associated with increased rates of anticoagulation. Variables such as the use of dual antiplatelet therapy, prior myocardial infarction, and surgical revascularization were associated with lower rates of anticoagulation (Figure 2). These relationships persisted after exclusion of patients on dual antiplatelet therapy.

| DISCUSSION
Within our outpatient study population of individuals with NVAF and CHF, 69% of these patients had HFpEF. Although this falls within the upper range of what has been previously reported, 6-10 the outpatient study cohort used was contemporary and is consistent with prior studies suggesting that HFpEF is a growing clinical entity. Importantly, our study showed that although HFpEF patients with NVAF had higher average CHA 2 DS 2 -VASc scores, their overall rates of anticoagulation compared to their HFrEF counterparts were lower. Thus, even though data from observational studies and clinical trials show comparable rates of stroke/TIA within the reduced and preserved ejection fractions HF groups with AF, [14][15][16] there remains a discrepancy in rates of anticoagulation between these groups in the outpatient. Importantly, even though HFpEF patients are less likely to be anticoagulated than HFrEF patients, rates of anticoagulation in both groups has increased over time.
Prior studies evaluating rates of anticoagulation in HF patients with AF have shown variable anticoagulation rates depending on the FIGURE 2 Relative risks for oral anticoagulation by baseline patient characteristics. Relative risks from the variables included in the multivariable model for predicting anticoagulation. CABG, coronary artery bypass graft; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction, PCI percutaneous coronary intervention, TIA, transient ischemic attack FIGURE 3 Risk adjusted rates of anticoagulation by heart failure group and CHA2DS2-VASc score. HFpEF, heart failure with preserved ejection fraction; HFrEF heart failure with reduced ejection fraction study population evaluated and whether or not the data were risk adjusted. 6,14,15 In one observational study by Steinberg et al. which used inpatient registry data, lower unadjusted rate of anticoagulation in the HFpEF group compared to the HFrEF group resolved after risk adjustment. 6 In the Sartipy et al study which used inpatient and outpatient data unadjusted rates of anticoagulation in patients with HFrEF were higher compared to those with HFpEF. 15 Interestingly, this study found in multivariable analysis that anticoagulant use in patients with HFpEF was more likely to be associated with NVAF than in patients with HFrEF. 15 This suggests that the HFrEF population may be more likely to have additional indications for anticoagulation compared to the HFpEF group, and may be why differences in anticoagulation rates between these two groups decrease after risk adjustment.
Our study provides insight into current outpatient anticoagulation patterns in patients with HF and AF. Although it is unclear why patients with HFpEF are less likely to be anticoagulated than those with HFrEF, this difference suggests that opportunities for improving rates of antic- indication to anticoagulation or an indication to anticoagulation other than AF in this dataset. Instead, as is convention with analyses using the PINNACLE dataset, we relied on a data field indicating medical contraindication to anticoagulation to exclude individuals from analysis. Longitudinal data was also not present, and instead only a snapshot of anticoagulation rates could be provided, and no outcomes data including rates of stroke/TIA, or medication adherence was available. FIGURE 4 Observed rates of anticoagulation by heart failure group and year. This graphic shows the temporal trend in observed rates of anticoagulation in the congestive heart failure population (total and by heart failure group) with non-valvular atrial fibrillation. Comparison between the observed rates of anticoagulation between the heart failure groups by year were made. *P < 0.05, HFpEF, heart failure with preserved ejection fraction; HFrEF heart failure with reduced ejection fraction

| CONCLUSIONS
The majority of outpatients with NVAF and CHF have HFpEF, and even though these individuals have higher CHA 2 DS 2 -VASc scores, they are significantly less likely to be prescribed oral anticoagulation compared to HFrEF patients. The discrepancy in anticoagulation rate is most prevalent in the lower risk groups, and may be secondary to differences in guideline recommendations and evidence for anticoagulating individuals when bleeding is a risk. Future studies are needed to clarify why differences exist in anticoagulation by HF type; however, our data calls for educating and alerting clinicians of this important risk factor for stroke prevention.

| CLINICAL PERSPECTIVES
Congestive HF has been validated as an independent risk factor for stroke regardless of ejection fraction; however, the clinical significance of HF with preserved EF may be underappreciated. The majority of outpatients with NVAF and congestive HF have preserved ejection fraction. Even though these individuals have higher CHA2DS2-VASc scores, they are significantly less likely to be prescribed oral anticoagulation compared to outpatients with HF with reduced EF which suggests an opportunity for educating and alerting clinicians of this important risk factor for stroke prevention.

| TRANSLATIONAL OUTLOOK
Future studies are needed to clarify why differences exist in anticoagulation by HF type, and in particular why individuals with AF and HF with preserved EF are less likely to anticoagulated than their reduced EF counterparts.