Abstract
Aims
Heart failure (HF) is a risk factor for major adverse events in atrial fibrillation (AF). Whether this risk persists on non-vitamin K antagonist oral anticoagulants (NOACs) and varies according to left ventricular ejection fraction (LVEF) is debated.
Methods and results
We investigated the relation of HF in the ETNA-AF-Europe registry, a prospective, multicentre, observational study with an overall 4-year follow-up of edoxaban-treated AF patients. We report 2-year follow-up for ischaemic stroke/transient ischaemic attack (TIA)/systemic embolic events (SEE), major bleeding, and mortality. Of the 13 133 patients, 1854 (14.1%) had HF. Left ventricular ejection fraction was available for 82.4% of HF patients and was <40% in 671 (43.9%) and ≥40% in 857 (56.1%). Patients with HF were older, more often men, and had more comorbidities. Annualized event rates (AnERs) of any stroke/SEE were 0.86%/year and 0.67%/year in patients with and without HF. Compared with patients without HF, those with HF also had higher AnERs for major bleeding (1.73%/year vs. 0.86%/year) and all-cause death (8.30%/year vs. 3.17%/year). Multivariate Cox proportional models confirmed HF as a significant predictor of major bleeding [hazard ratio (HR) 1.65, 95% confidence interval (CI): 1.20–2.26] and all-cause death [HF with LVEF <40% (HR 2.42, 95% CI: 1.95–3.00) and HF with LVEF ≥40% (HR 1.80, 95% CI: 1.45–2.23)] but not of ischaemic stroke/TIA/SEE.
Conclusion
Anticoagulated patients with HF at baseline featured higher rates of major bleeding and all-cause death, requiring optimized management and novel preventive strategies. NOAC treatment was similarly effective in reducing risk of ischaemic events in patients with or without concomitant HF.
Keywords: Non-vitamin K antagonist oral anticoagulant, Edoxaban, Heart failure, Left ventricular ejection fraction, Registry, Atrial fibrillation
Graphical Abstract
Graphical Abstract.
What’s new?
The ETNA-AF-Europe subanalysis compared 2-year outcomes in edoxaban-treated patients with atrial fibrillation (AF) categorized by cardiac structural/functional impairment status and left ventricular ejection fraction (LVEF).
Ischaemic event rates were similar in patients with and without heart failure (HF). Patients with HF had higher annualized event rates of major bleeding and cardiovascular and overall mortality.
No relevant differences were observed for ischaemic or bleeding events by HF subtype (LVEF ≥40% or <40%); mortality tended to be highest in patients with HF and LVEF <40%.
The benefit of edoxaban treatment was demonstrated by the decrease in the differences in thromboembolic event rates between patients with HF (across HF subtypes) and those without HF.
The broad array of predictors of overall and cardiovascular deaths observed in this analysis highlights the importance of taking comorbidities into consideration and the requisite for comprehensive management of patients with AF and HF beyond consequent oral anticoagulant.
Introduction
More than 70% of patients with atrial fibrillation (AF) are estimated to have a cardiovascular (CV) comorbidity.1 Most comorbidities such as hypertension, coronary artery disease (CAD) and structural heart disease alongside obesity, diabetes, chronic kidney disease (CKD) and chronic obstructive pulmonary disease (COPD) are related to cardiac structural and functional abnormalities. These manifestations result in a reduction in cardiac output and/or elevated intracardiac pressure and predispose patients to the development of heart failure (HF). Many risk factors are shared between AF and HF, and there are multiple physiological mechanisms by which HF can develop in patients with AF.1 Indeed, patients with AF have up to a five-fold higher risk of HF.2 In permanent AF, HF is prevalent in more than half of the patients.3 Heart failure is accounted for in the CHA2DS2-VASc score [HF (1 point), hypertension (1 point), ≥75 years (2 points), diabetes mellitus (1 point), stroke/transient ischaemic attack (TIA)/systemic embolic events (SEE; 2 points), vascular disease (1 point), age 65–74 years (1 point), and female sex (1 point)] used to assess stroke risk.4 Further, mortality is significantly increased if HF occurs in patients with AF.5–7
Currently, non-vitamin K antagonist oral anticoagulants (NOACs) are the preferred treatment for the prevention of stroke and SEE in patients with AF.4 In randomized controlled trials (RCTs), NOACs have demonstrated efficacy in stroke reduction in subsets of patients with HF and are now commonly used in patients with concomitant HF and AF.8–10 In order to optimize the management of patients with AF and HF, patients’ risks of stroke and bleeding need to be understood and considered in the implementation of an individualized care pathway.4 However, it remains largely unknown whether HF is related to the risk of adverse events despite NOAC treatment, and, if so, whether the effect is different according to left ventricular ejection fraction (LVEF).
Data from routine clinical practice on NOAC treatment in patients with AF who have coexisting chronic structural and/or functional heart disease can enhance our understanding of the use of NOACs in this patient population and may impact treatment. The Edoxaban Treatment in routiNe clinical prActice for patients with nonvalvular AF in Europe (ETNA-AF-Europe) registry is a post-authorization observational study designed to collect safety data during routine clinical care by assessing the risks and benefits of edoxaban in unselected European patients with AF. The objective of the current analysis of the data from ETNA-AF-Europe was to compare 2-year outcomes in AF patients receiving oral anticoagulant (OAC) therapy according to cardiac structural/functional impairment status and LVEF.
Methods
Study design
Overview
The ETNA-AF-Europe registry (Clinicaltrials.gov: NCT02944019) is part of the global ETNA programme conducted in Europe, Japan, and Korea/Taiwan.11 It is a multinational, multicentre, post-authorization, observational registry spanning across 825 sites in 10 European countries (Austria, Belgium, Germany, Ireland, Italy, The Netherlands, Portugal, Spain, Switzerland, and the UK). Enrollment commenced in May 2015 in Switzerland and in August 2015 in Germany; however, this was stalled in agreement with the newly formed Pharmacovigilance Risk Assessment Committee (PRAC). Following revision of the study protocol according to the PRAC guidance, patient enrollment resumed in Germany, Ireland, The Netherlands, Switzerland, and the UK in November 2016 and in Austria, Belgium, Italy, Portugal, and Spain in the first quarter of 2017, with an overall follow-up of 4 years.12 A detailed account of the design has been reported previously.11,12 Unselected routine patients with AF treated with edoxaban were enrolled. No exclusion criteria were applied, in order to fully capture routine clinical practice.
For the current analysis, patients with and without HF were analysed over a 2-year follow-up period, including data obtained up to the data cut-off point on 26 October 2020. Patients with documented structural/functional cardiac abnormality were grouped as HF patients including patients with documented congestive HF, documented ischaemic cardiomyopathy, LVEF <40%, or frequent dyspnoea (≥1/day) without COPD and at least one of the following: documented severe valvular heart disease, documented CAD post-myocardial infarction, valve replacement, or documented hypertension treated with at least three drugs. Patients with HF were further categorized into two groups: those with HF and LVEF <40% and those with HF and LVEF ≥40% (combining patients with mildly reduced and preserved EF; Supplementary material online, Figure S1). Echocardiography was available in 7226 (64.1%) of the 11 279 patients without HF (see Supplementary material online, Figure S1). We did not perform imputation analyses to harmonize the results with the report of this study and other publications. We calculated the CHA2DS2-VASc and HAS-BLED [uncontrolled hypertension (1 point), renal disease or liver disease (1 or 2 points), stroke history (1 point), prior major bleeding or predisposition to bleeding (1 point), labile international normalised ratio (INR) (1 point), age > 65 years (1 point), medication predisposing to bleeding or alcohol usage (1 or 2 points)] scores. Alcohol consumption was categorized as ≤2 glasses/day, >2–4 glasses/day, >4 glasses/day, or unknown. Smoking status was captured using the categories current smoker, former smoker, never smoked, or unknown. Estimated creatinine clearance was calculated using the Cockcroft–Gault formula13 and categorized as ≥80, ≥50–<80, ≥30–<50, or <30 mL/min.
Outcomes
At 12 and 24 months, follow-up was performed, and clinical events were collected based on physicians’ diagnosis. Outcomes relevant to this analysis were stroke, SEE, major bleeding, clinically relevant non-major bleeding (CRNM), TIA, and all-cause death during the 2-year follow-up. We used all bleeding, including major and CRNM bleeding, in accordance with the International Society on Thrombosis and Haemostasis definition.14
The study was approved by the institutional review boards and independent ethics committees for all participating centres in compliance with the Declaration of Helsinki and Guidelines for Good Pharmacoepidemiological Practice. It is registered under clinicaltrials.gov NCT02944019. All participants provided written informed consent.
Statistical analysis
Descriptive summaries are provided as frequencies, n (%) and mean [standard deviation (SD)] for patient demographics and other baseline characteristics. Annualized event rates (AnERs), n (% per year) with 95% confidence intervals (CIs), are presented for clinical outcomes, including several composite outcomes. Cumulative event rates over 2 years were reported for any stroke or SEE, major bleeding, and all-cause mortality and are presented as Kaplan–Meier (KM) curves. P values were adjusted for age and CHA2DS2-VASc score. Multivariate Cox proportional hazards models were used for stepwise selection of predictors of ischaemic stroke/TIA/SEE, major bleeding, and all-cause death. Competing risk models for major bleeding and for ischaemic stroke/TIA/SEE confounded by all-cause death (Fine and Gray model) were calculated and are reported in Supplementary material online.15 Potential interactions of HF status and outcomes by antiplatelet therapy were also tested.
Results
Demographics
The baseline characteristics are provided in Table 1. Among the 13 133 patients with AF included in this registry, 1854 (14.1%) had HF. Heart failure with LVEF ≥40% was more frequent (56.1%). Baseline characteristics for patients with HF and missing LVEF values (n = 326) are available in Supplementary material online, Table S1. More men were reported to have HF (64.6%) vs. a more balanced sex mix in those without HF (55.4% men). Antiplatelet therapy was more commonly prescribed in patients with HF.
Table 1.
Demographic and clinical characteristics of participants in the ETNA-AF-EU 2-year follow-up analysis set, stratified by HF and EF status
| Variables | No heart failure | Heart failure | HF with EF <40% | HF with EF ≥40% |
|---|---|---|---|---|
| n = 11 279 | n = 1854 | n = 671 (43.9%) | n = 857 (56.1%) | |
| Men, n (%) | 6254 (55.4) | 1197 (64.6) | 500 (74.5) | 489 (57.1) |
| Age (years), mean (SD) | 73.5 (9.4) | 74.6 (9.8) | 72.3 (10.8) | 75.9 (8.7) |
| Weight (kg), mean (SD) | 80.9 (17.2) | 81.3 (17.5) | 81.6 (17.5) | 80.5 (16.8) |
| Systolic blood pressure (mmHg), mean (SD) | 134.2 (17.7) | 129.1 (18.4) | 125.4 (18.3) | 131.3 (17.3) |
| Heart rate (b.p.m.), mean (SD) | 75.7 (19.1) | 76.4 (18.0) | 78.9 (20.2) | 75.0 (16.8) |
| Dyslipidaemia, n (%) | 4634 (41.1) | 1010 (54.5) | 309 (46.1) | 517 (60.3) |
| Antiplatelets, n (%) | 1640 (14.5) | 360 (19.4) | 113 (16.8) | 172 (20.1) |
| Smoking, n (%) | ||||
| Current smoker | 695 (6.2) | 128 (6.9) | 56 (8.3) | 46 (5.4) |
| Former smoker | 2333 (20.7) | 525 (28.3) | 209 (31.1) | 231 (27.0) |
| Never smoker | 6278 (55.7) | 972 (52.4) | 302 (45.0) | 491 (57.3) |
| Unknown | 1973 (17.5) | 229 (12.4) | 104 (15.5) | 89 (10.4) |
| Estimated glomerular filtration rate (mL/min/1.73 m2), mean (SD) | 75.4 (30.2) | 68.2 (30.9) | 68.9 (31.1) | 67.5 (29.7) |
| CHA2DS2-VASc, mean (SD) | 3.0 (1.3) | 4.3 (1.4) | 3.8 (1.5) | 4.5 (1.3) |
| HAS-BLED, mean (SD) | 2.5 (1.1) | 2.8 (1.1) | 2.6 (1.2) | 3.0 (1.1) |
| Type of atrial fibrillation, n (%) | ||||
| Paroxysmal | 6298 (56.0) | 758 (40.9) | 268 (40.1) | 343 (40.0) |
| Persistent | 2665 (23.7) | 510 (27.5) | 202 (30.2) | 246 (28.7) |
| Long-standing persistent and permanent | 2293 (20.3) | 584 (31.5) | 199 (29.7) | 268 (31.3) |
| Sinus rhythm, n (%) | 1204 (10.7) | 136 (7.4) | 39 (5.9) | 64 (7.5) |
| Current atrial fibrillation symptoms at baseline, n (%) | 2655 (23.5) | 567 (30.6) | 200 (29.8) | 290 (33.8) |
| EHRA score, mean (SD) | 2.90 (0.84) | 3.36 (0.82) | 3.24 (0.85) | 3.47 (0.77) |
| Perceived frailtya, n (%) | 1058 (10.1) | 347 (19.9) | 122 (19.6) | 160 (19.8) |
| Diabetes mellitus, n (%) | 2299 (20.4) | 586 (31.6) | 183 (27.3) | 280 (32.7) |
| Currently treated without insulin, n (%) | 2190 (19.4) | 565 (30.5) | 174 (25.9) | 273 (31.9) |
| Currently treated with insulin, n (%) | 455 (4.0) | 150 (8.1) | 49 (7.3) | 69 (8.1) |
| Chronic obstructive pulmonary disease, n (%) | 938 (8.3) | 269 (14.5) | 108 (16.1) | 110 (12.8) |
| Coronary heart disease, n (%) | 1858 (16.5) | 894 (48.2) | 277 (41.3) | 429 (50.1) |
| Valvular disease, n (%) | 1772 (15.7) | 514 (27.7) | 170 (25.3) | 295 (34.4) |
| Prior stroke/TIA/SEE, n (%) | 1056 (9.4) | 173 (9.3) | 50 (7.5) | 87 (10.2) |
| Prior MI, n (%) | 313 (2.8) | 254 (13.7) | 72 (10.7) | 126 (14.7) |
| Prior major or non-major clinically relevant bleeding, n (%) | 213 (1.9) | 60 (3.2) | 16 (2.4) | 34 (4.0) |
| Hypertension, n (%) | 8660 (76.8) | 1469 (79.2) | 473 (70.5) | 733 (85.5) |
| Previous and/or current use of any antihypertensive treatment, n (%) | 8328 (73.8) | 1434 (77.3) | 463 (69.0) | 709 (82.7) |
| Cardiac interventions, n (%) | ||||
| Ablation, n (%) | 678 (6.0) | 97 (5.2) | 24 (3.6) | 58 (6.8) |
| Electric cardioversion, n (%) | 1916 (17.0) | 354 (19.1) | 144 (21.5) | 162 (18.9) |
| Pharmacological cardioversion, n (%) | 820 (7.3) | 147 (7.9) | 54 (8.0) | 82 (9.6) |
| Defibrillator implantation, n (%) | 61 (0.5) | 103 (5.6) | 71 (10.6) | 16 (1.9) |
| Pacemaker implantation, n (%) | 539 (4.8) | 116 (6.3) | 34 (5.1) | 66 (7.7) |
Numbers are mean (standard deviation) or number (%). Estimated glomerular filtration rate was calculated using the Cockcroft–Gault formula.
CHA2DS2-VASc score, heart failure (1 point), hypertension (1 point), ≥75 years (2 points), diabetes mellitus (1 point), stroke/TIA/SEE (2 points), vascular disease (1 point), age 65–74 years (1 point), and female sex (1 point); eGFR, estimated glomerular filtration rate; HAS-BLED, uncontrolled hypertension (1 point), renal disease or liver disease (1 or 2 points), stroke history (1 point), prior major bleeding or predisposition to bleeding (1 point), labile INR (1 point), age >65 years (1 point), and medication usage predisposing to bleeding or alcohol usage (1 or 2 points); INR, international normalised ratio; LA, left atrial diameter; SEE, systemic embolic event; TIA, transient ischaemic attack.
Surrogate measure used to inform clinical decision-making.16
A total of 2286 (17.4%) patients were lost to follow-up/discontinued from the study while living and on edoxaban. No or minor differences were observed between the baseline characteristics of these patients vs. those patients who remained in the study (see Supplementary material online, Table S2). Overall, 7.13% (937/13 133) patients died during the 2-years of follow-up.
Clinical characteristics
Patients without HF had a lower mean (SD) CHA2DS2-VASc score than those with HF (Table 1). Among patients with HF, those with LVEF <40% had relatively lower scores vs. those with LVEF ≥40%. The dominant form of AF reported in patients with and without HF was paroxysmal. Shortness of breath, perceived frailty (surrogate measure used to inform clinical decision-making16), and other comorbidities were more frequent in patients with HF vs. those without.
Outcomes
Stroke/systemic embolic events
The AnERs of any stroke or SEE were 0.86%/year and 0.67%/year in patients with and without HF, respectively (Figure 1A). The AnERs of ischaemic stroke were similar in patients with and without HF. The KM curves began to separate in the second year of the analysis (Figure 2A). Among those with HF and LVEF ≥40% and <40%, AnERs of any stroke or SEE were 0.93%/year and 0.51%/year, respectively (Figure 1B). The KM curves by EF crossed each other (Figure 2B). Besides age, stepwise selection models for stroke/TIA/SEE selected previous embolic events with an almost three-fold increased hazard ratio (HR) 2.83 (95% CI: 1.88–4.26), and a two-fold increased risk for insulin-treated diabetes mellitus, and TIA as the strongest predictors (Table 2). Heart failure was not selected by the model. No interaction by antiplatelet therapy was observed (P value of interaction: 0.65).
Figure 1.
Clinical outcomes for participants in the ETNA-AF-EU 2-year follow-up analysis set, stratified by HF and EF status. (A) Patients with or without heart failure. (B) Patients with HF and EF <40% or ≥40%. For 326 patients with HF, the EF was not documented. CI, confidence interval; CRNM, clinically relevant non-major; CV, cardiovascular; EF, ejection fraction; HF, heart failure; ICH, intracranial haemorrhage; MI, myocardial infarction; SEE, systemic embolic events.
Figure 2.
Cumulative incidence of stroke or systemic embolic event at 2-year follow-up, stratified by HF status and adjusted for age and CHA2DS2-VASc score. (A) Stratified by HF status. (B) Stratified by EF. For 326 patients with HF, LVEF was not documented. HF, heart failure; LVEF, left ventricular ejection fraction.
Table 2.
Stepwise selection multivariable Cox models for predictors of stroke/TIA/SEE, major bleeding, and mortality
| Variable | Hazard ratio | 95% confidence interval | P value |
|---|---|---|---|
| Stroke/TIA/SEE | |||
| Previous stroke/TIA/SEE | 2.83 | 1.88–4.26 | <0.0001 |
| Age, years | 1.05 | 1.03–1.07 | <0.0001 |
| Diabetes mellitus currently treated with insulin | 2.20 | 1.37–3.54 | 0.0011 |
| Previous TIA | 1.97 | 1.18–3.31 | 0.0099 |
| Major bleeding | |||
| eGFR, mL/min/1.73 m2 | <0.0001 | ||
| 50–80 vs. ≥80 | 2.00 | 1.36–2.95 | 0.0004 |
| 30–50 vs. ≥80 | 2.73 | 1.78–4.20 | <0.0001 |
| <30 vs. ≥80 | 5.77 | 3.16–10.52 | <0.0001 |
| History of major or CRNM bleeding | 2.48 | 1.44–4.28 | 0.0011 |
| Heart failure | 1.65 | 1.20–2.26 | 0.0019 |
| Chronic obstructive pulmonary disease | 1.51 | 1.03–2.20 | 0.033 |
| HAS-BLED score cont. | 1.17 | 1.03–1.33 | 0.018 |
| Smoking | 0.039 | ||
| Current smoker | 2.40 | 1.26–4.57 | 0.0078 |
| Former smoker | 1.58 | 0.93–2.67 | 0.090 |
| Never smoker | 1.31 | 0.80–2.13 | 0.28 |
| Mortality | |||
| Age, years | 1.81 | 1.64–2.00 | <0.0001 |
| Heart failure with reduced ejection fraction | 2.42 | 1.95–3.00 | <0.0001 |
| Heart failure with preserved ejection fraction | 1.80 | 1.45–2.23 | <0.0001 |
| Chronic obstructive pulmonary disease | 2.09 | 1.75–2.49 | <0.0001 |
| eGFR, mL/min/1.73 m2 vs. ≥80 | <0.0001 | ||
| 50–80 | 1.18 | 0.95–1.46 | 0.14 |
| 30–50 | 1.56 | 1.21–2.01 | 0.0005 |
| <30 | 2.66 | 1.88–3.76 | <0.0001 |
| Diabetes mellitus | <0.0001 | ||
| Currently not treated with insulin | 1.27 | 1.06–1.52 | 0.0082 |
| Currently treated with insulin | 2.31 | 1.82–2.93 | <0.0001 |
| Body mass index, reference 18.5–35 kg/m2 | <0.0001 | ||
| <18.5 kg/m2 | 2.31 | 1.53–3.51 | <0.0001 |
| ≥35 kg/m2 | 1.40 | 1.07–1.84 | 0.015 |
| Dyslipidaemiaa | 0.72 | 0.62–0.84 | 0.0001 |
| Peripheral arterial disease | 1.58 | 1.20–2.08 | 0.0010 |
| Smoking | 0.0005 | ||
| Current smoker | 1.58 | 1.14–2.17 | 0.0056 |
| Former smoker | 0.99 | 0.78–1.26 | 0.93 |
| Never smoker | 0.86 | 0.70–1.06 | 0.16 |
| Major or CRNM bleeding | 1.64 | 1.16–2.30 | 0.0046 |
| History of ischaemic stroke, TIA, and SEE | 1.38 | 1.08–1.77 | 0.0096 |
| Hepatic disease | 1.60 | 1.05–2.43 | 0.028 |
| Alcohol use ≥1 glass/day | 0.80 | 0.68–0.94 | 0.0080 |
| Female sex | 0.82 | 0.70–0.96 | 0.015 |
Provided are hazard ratios and 95% confidence intervals. If not indicated differently, hazard ratios are for the conditions present vs. not present. Smoking and alcohol is vs. unknown smoking/alcohol drinking status. P values are for multivariable Cox models.
CRNM, clinically relevant non-major; eGFR, estimated glomerular filtration rate; SEE, systemic embolic event; TIA, transient ischaemic attack.
84.3% of patients with dyslipidaemia were on lipid lowering treatment.
Major bleeding
Patients with HF reported higher AnERs of major bleeding compared with those without HF (1.73%/year vs. 0.86%/year, respectively) (Figure 1A). Annualized event rates of intracranial haemorrhage were 0.37%/year and 0.18%/year in patients with and without HF, respectively. Among those with HF, major bleeding occurred in 1.79%/year and 1.71%/year of patients with HF and EF ≥40% and <40%, respectively (Figure 1B). Separation of the KM curves for major bleeding events between the groups began at Day 30 (Figure 3A and B). In stepwise selection, HF was selected by the model as a predictor, as well as renal impairment represented by estimated glomerular filtration rate (eGFR), prior major or CRNM bleeding, COPD, smoking variables, and HAS-BLED score (Table 2). Accounting for competing risk of death did not change the results for stroke/SEE or major bleeding markedly (see Supplementary material online, Figure S2). No interaction by antiplatelet therapy was observed (P value of interaction: 0.40).
Figure 3.
Cumulative major bleeding at 2-year follow-up, stratified by HF status and adjusted for age and CHA2DS2-VASc score. (A) Stratified by HF status. (B) Stratified by LVEF. For 326 patients with HF, LVEF was not documented. HF, heart failure; LVEF, left ventricular ejection fraction.
Mortality
All-cause deaths and CV deaths (sensitivity analysis) were higher in patients with HF vs. those without HF (8.30%/year vs. 3.17%/year [P < 0.0001] and 4.87%/year vs. 1.71%/year, respectively; Figure 1A). More patients with HF and LVEF <40% died due to any as well as CV causes vs. those with HF and LVEF ≥40% (9.44%/year vs. 7.43%/year and 5.99%/year vs. 4.14%/year, respectively; Figure 1B). Distinct separation of the all-cause mortality curve between no HF and HF groups was apparent by Day 30 (Figure 4). No interaction by antiplatelet therapy was observed (P value of interaction: 0.11).
Figure 4.
Cumulative all-cause mortality at 2-year follow-up, stratified by HF status and adjusted for age and CHA2DS2-VASc score. (A) Stratified by HF status. (B) Stratified by LVEF. For 326 patients with HF, LVEF was not documented. HF, heart failure; LVEF, left ventricular ejection fraction.
Besides age and sex, stepwise selection models indicated HF with LVEF < 40% [HR (95% CI): 2.42 (1.95 to 3.00)] and HF with LVEF ≥ 40% [1.80 (1.45–2.23)] as two of the strongest predictors of death, with an additional 11 predictors identified. Annualized event rates of composite endpoints including efficacy, safety, and death parameters were higher in patients with HF vs. those without HF (see Supplementary material online, Table S3 and Figure S3A and B). Annualized event rates of composite endpoints were mostly similar in patients with HF and LVEF ≥ or <40%.
Discussion
In our contemporary cohort of patients with AF receiving OAC, overall event rates were low, with <1% attributable to ischaemic events or intracranial haemorrhage. Over 2 years of follow-up, ischaemic event rates were similar in patients with and without HF. Patients with HF had higher AnERs of major bleeding and CV and overall mortality. No relevant differences were observed for ischaemic or bleeding events by HF subtype (with LVEF ≥40% or <40%) whereas mortality tended to be highest in patients with HF and LVEF <40%. Importantly, these findings provide information across the spectrum of HF while also showing consistency with other NOAC studies.17,18
Atrial fibrillation in the presence of HF is associated with poor outcomes, warranting the use of anticoagulants.19,20 Initial NOAC trials and their secondary analyses demonstrated that event rates can be effectively and safely reduced in patients with HF compared with warfarin and support the use of NOACs as an alternative to warfarin in patients with AF and HF.21–26 In a subanalysis of the ENGAGE AF-TIMI 48 trial, edoxaban compared with warfarin was similarly effective in preventing stroke/SEE in patients with and without HF.25 Our analysis of clinical practice data outside the RCT setting can extend the knowledge from prior trials on NOACs and demonstrate that thromboembolic event rates in patients with HF can be reduced to levels observed in patients without HF independent of LVEF and baseline differences according to HF status and HF subgroups. Despite higher stroke and bleeding risk scores and, overall, more comorbidities in patients with HF and LVEF ≥40% vs. LVEF <40%, our real-world findings now show the benefit of NOAC treatment by diminishing the difference in event rates between patients with HF, across the spectrum of HF subtypes, and those without HF during the 2 years of follow-up. Although it should be mentioned, the viewpoint has been expressed as to whether the CHA2DS2-VASc criteria should be extrapolated to patients with HF and preserved EF and AF.27 The definition of HF used in the CHA2DS2-VASc criteria (recent congestive HF exacerbation without a LVEF criterion) differs from that used in this study and was based mainly on patients with HF and reduced EF, leading to questioning of the appropriateness of its use in patients with HF and preserved EF. Despite this, in the absence of RCT data in patients with HF and preserved EF and with no pathophysiological reason for why data should differ among HF patients with reduced or preserved EF, anticoagulation should be considered for both HF populations.
Compared to 6170 patients with AF treated mainly with vitamin K antagonists in Prevention of thromboembolic events - European Registry in Atrial Fibrillation (PREFER in AF),28 the stroke rates, on average, were lower in the ETNA-AF-Europe study with patients on edoxaban. In PREFER in AF, stroke incidence was also higher in patients with HF vs. those without HF (1.3% vs. 0.6%) with annual incidence linearly and inversely related to LVEF.
On OAC therapy, AF patients with HF may be prone to increased bleeding events.29 Hence, bleeding risks of different antithrombotic agents in patients with HF are a key consideration in treatment decision-making. Various secondary analyses from the NOAC RCTs evaluating the subsets of patients with HF have provided insights into the efficacy and safety of NOACs in this higher risk population and support the use of NOACs as an alternative to warfarin.24–26 A recent meta-analysis focusing on RCTs comparing the effect of NOACs with warfarin in patients with AF and HF showed that NOACs have become the preferred choice for preventing stroke/SEE and major bleeding in AF patients with HF.30 Annualized event rates in our analyses, though low, were almost twice as high for major bleeding in patients with vs. without HF and comparable across HF subgroups. We confirmed the moderate predictive ability of the variables comprised in the HAS-BLED score for bleeding.31 The score does not comprise HF as a risk indicator. Our data therefore suggest that the higher bleeding risk observed in patients with HF receiving OAC needs to be considered beyond the HAS-BLED components, and the bleeding risk assessment requires refinement.
As expected, mortality was higher in patients with AF and HF.5 This observation is explained not only by the prognostic impact of cardiac disease in itself but also by the overall higher burden of comorbidities and the fact that almost twice as many patients were perceived as frail (20% vs. 10%). In AF and HF patients, frailty is known to increase adverse outcomes,32,33 though evidence on the treatment is limited.34 It appears to be associated with slower uptake of OAC and non-recommended anticoagulant dosages but possibly also carries a higher risk of bleeding on NOACs.35–39
Whereas the number of comorbidities on average was higher among patients with LVEF ≥40%, mortality in the male-dominated HF with LVEF <40% subgroup was nominally highest. The large number of predictors of death selected in the regression analyses besides HF subtypes indicates that no single factor explains the higher mortality; however, the spectrum of comorbidities with HF is one of the strongest among them.
These data from routine clinical practice further highlight the importance of the comprehensive management of concomitant CV risk factors in patients with AF and HF beyond consequent oral anticoagulation.40–42 This approach is supported by the ESC guidelines, which recommend that risk factors and comorbidities should be well managed to reduce AF burden.4 A comprehensive care approach to AF, as defined by the Atrial Fibrillation Better Care pathway, was associated with clinical benefit across all adjudicated clinical endpoints in the ENGAGE AF-TIMI 48 trial and real-world data.42 Furthermore, the 8th AFNET/EHRA consensus conference of the Atrial Fibrillation NETwork (AFNET) and the European Heart Rhythm Association (EHRA) concluded that implementation of new evidence-based approaches to AF screening pathway and rhythm management could lead to improvement in outcomes in patients with AF.40 For example, cardiac interventions such as ablation therapy were less frequently performed in patients with reduced vs. preserved EF.43
Limitations
ETNA-AF-Europe analyses are derived from registry data with known potential of bias in treatment selection and outcomes ascertainment, as well as residual confounding. Another potential source of bias included the enrollment criterion that patients had to be continued on edoxaban.
Approximately 17% of patients were lost to follow-up or discontinued from the study while living and receiving edoxaban. The adherence to the drug was reasonable considering it was an observational study; however, it was lower than the adherence to dosing observed in a RCT setting.
The data presented are also limited by the pre-determined fields included in the Case Report Form and relied upon the accurate and complete input of data by treating physicians. Thus, patients could not be classified according to stage of HF because of limitations of the data. Since multiple comparisons were performed, no adjustment for the level of significance was implemented.
Currently, a lack of head-to-head data prevent the direct comparison between NOAC agents. Whether novel OACs such as factor XIa inhibitors can further reduce the bleeding risk in AF patients with HF needs to be demonstrated.44 However, the strength of the registry is that it comprises a large number of individuals with almost complete follow-up that help to understand the natural history, effectiveness, and safety of edoxaban by providing insights into the use of treatments during routine clinical practice. The efficacy and safety results from this registry study are consistent with those from the ENGAGE AF-TIMI 48 trial across the spectrum of HF severity and other NOACs.17,25
Conclusions
Our data suggest that in patients with AF, edoxaban treatment is effective in reducing the higher risk of ischaemic events observed with concomitant HF across the LVEF spectrum to a similar risk to patients without HF during a 2-year period. Patients with HF remain at a higher risk of major bleeding events. Therefore, HF status should be considered in addition to the HAS-BLED score to assess bleeding risk when treating patients with AF. Mortality risk also remains higher in patients with AF and HF, with highest event rates in HF with LVEF <40%. A broad spectrum of clinical mortality predictors for overall and CV deaths indicates the relevance of comorbidities and highlights the importance of comprehensive management of patients with AF and HF beyond consequent OAC.
Supplementary Material
Acknowledgements
Editorial support was provided by Sarah Wetherill and Shelley Narula of inScience Communications, Springer Healthcare Ltd, UK, and was funded by Daiichi Sankyo Europe GmbH, Munich, Germany.
Contributor Information
Renate B Schnabel, Department of Cardiology, University Clinic Hamburg-Eppendorf, University Heart and Vascular Centre Hamburg-Eppendorf, Buildung O50, Martinistrasse 52, 20246 Hamburg, Germany; DZHK (German Centre for Cardiovascular Research), partner site Hamburg/Kiel/Luebeck, Potsdamer Str, 5810785 Berlin, Germany.
Pietro Ameri, Department of Internal Medicine, University of Genova, Genova, Italy; Cardiac, Thoracic and Vascular Department, IRCCS Ospedale Policlinico San Martino, Genova, Italy.
Jolanta M Siller-Matula, Department of Cardiology, Medical University of Vienna, Vienna, Austria.
Igor Diemberger, Department of Medical and Surgical Sciences, University of Bologna, 40138 Bologna, Italy; Unit of Cardiology, IRCCS Azienda Ospedaliero-Universitaria di Bologna, 40138 Bologna, Italy.
Marianne Gwechenberger, Department of Cardiology, Medical University of Vienna, Vienna, Austria.
Ladislav Pecen, Czech Academy of Science, Institute of Computer Sciences, Prague, Czech Republic; Department of Immunochemistry Diagnostics, University Hospital Pilsen, Pilsen, Czech Republic.
Marius Constantin Manu, Daiichi Sankyo Europe GmbH, Munich, Germany.
José Souza, Daiichi Sankyo Europe GmbH, Munich, Germany.
Raffaele De Caterina, Cardiology Division, Pisa University Hospital, Pisa, Italy; Fondazione Villa Serena per la Ricerca, Pescara, Italy.
Paulus Kirchhof, Institute of Cardiovascular Sciences, University of Birmingham, Birmingham, UK; Department of Cardiology, University Heart and Vascular Centre Hamburg, University Medical Centre Hamburg Eppendorf, Hamburg, Germany; German Center for Cardiovascular Sciences (DZHK), partner site Hamburg/Kiel/Lübeck, Germany.
Supplementary material
Supplementary material is available at Europace online.
Funding
This work was supported by Daiichi Sankyo Europe GmbH, Munich, Germany. R.B.S. has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme under the grant agreement No 648131; from the European Union’s Horizon 2020 research and innovation programme under the grant agreement No 847770 (AFFECT-EU); the German Center for Cardiovascular Research (DZHK e.V.) (81Z1710103 and 81Z0710114); the Bundesministerium für Bildung und Forschung (BMBF 01ZX1408A); and ERACoSysMed3 (031L0239).
Data availability
The data underlying this article are available in the article and in its Supplementary material online.
References
- 1. Mashat AA, Subki AH, Bakhaider MA, Baabdullah WM, Walid JB, Alobudi AHet al. Atrial fibrillation: risk factors and comorbidities in a tertiary center in Jeddah, Saudi Arabia. Int J Gen Med 2019;12:71–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Odutayo A, Wong CX, Hsiao AJ, Hopewell S, Altman DG, Emdin CA. Atrial fibrillation and risks of cardiovascular disease, renal disease, and death: systematic review and meta-analysis. BMJ 2016;354:i4482. [DOI] [PubMed] [Google Scholar]
- 3. Chiang CE, Naditch-Brule L, Murin J, Goethals M, Inoue H, O'Neill Jet al. Distribution and risk profile of paroxysmal, persistent, and permanent atrial fibrillation in routine clinical practice: insight from the real-life global survey evaluating patients with atrial fibrillation international registry. Circ Arrhythm Electrophysiol 2012;5:632–9. [DOI] [PubMed] [Google Scholar]
- 4. Hindricks G, Potpara T, Dagres N, Arbelo E, Bax JJ, Blomstrom-Lundqvist Cet al. 2020 ESC guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic Surgery (EACTS): the Task Force for the diagnosis and management of atrial fibrillation of the European Society of Cardiology (ESC) developed with the special contribution of the European Heart Rhythm Association (EHRA) of the ESC. Eur Heart J 2021;42:373–498. [DOI] [PubMed] [Google Scholar]
- 5. Schrage B, Geelhoed B, Niiranen TJ, Gianfagna F, Vishram-Nielsen JKK, Costanzo Set al. Comparison of cardiovascular risk factors in European population cohorts for predicting atrial fibrillation and heart failure, their subsequent onset, and death. J Am Heart Assoc 2020;9:e015218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Wang TJ, Larson MG, Levy D, Vasan RS, Leip EP, Wolf PAet al. Temporal relations of atrial fibrillation and congestive heart failure and their joint influence on mortality: the Framingham Heart Study. Circulation 2003;107:2920–5. [DOI] [PubMed] [Google Scholar]
- 7. Mamas MA, Caldwell JC, Chacko S, Garratt CJ, Fath-Ordoubadi F, Neyses L. A meta-analysis of the prognostic significance of atrial fibrillation in chronic heart failure. Eur J Heart Fail 2009;11:676–83. [DOI] [PubMed] [Google Scholar]
- 8. Thomas I, EncisoSilva J, Schlueter M, Greenberg B. Anticoagulation therapy and NOACs in heart failure. Handb Exp Pharmacol 2017;243:515–35. [DOI] [PubMed] [Google Scholar]
- 9. Zhao L, Wang WYS, Yang X. Anticoagulation in atrial fibrillation with heart failure. Heart Fail Rev 2018;23:563–71. [DOI] [PubMed] [Google Scholar]
- 10. Xiong Q, Lau YC, Senoo K, Lane DA, Hong K, Lip GY. Non-vitamin K antagonist oral anticoagulants (NOACs) in patients with concomitant atrial fibrillation and heart failure: a systemic review and meta-analysis of randomized trials. Eur J Heart Fail 2015;17:1192–200. [DOI] [PubMed] [Google Scholar]
- 11. de Groot JR, Weiss TW, Kelly P, Monteiro P, Deharo JC, de Asmundis Cet al. Edoxaban for stroke prevention in atrial fibrillation in routine clinical care: 1-year follow-up of the prospective observational ETNA-AF-Europe study. Eur Heart J Cardiovasc Pharmacother 2021;7:f30–f9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. De Caterina R, Kelly P, Monteiro P, Deharo JC, de Asmundis C, Lopez-de-Sa Eet al. Design and rationale of the Edoxaban Treatment in routiNe clinical prActice for patients with Atrial Fibrillation in Europe (ETNA-AF-Europe) study. J Cardiovasc Med (Hagerstown) 2019;20:97–104. [DOI] [PubMed] [Google Scholar]
- 13. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976;16:31–41. [DOI] [PubMed] [Google Scholar]
- 14. Rodeghiero F, Tosetto A, Abshire T, Arnold DM, Coller B, James Pet al. ISTH/SSC bleeding assessment tool: a standardized questionnaire and a proposal for a new bleeding score for inherited bleeding disorders. J Thromb Haemost 2010;8:2063–5. [DOI] [PubMed] [Google Scholar]
- 15. Fine JP, Gray RJ. A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999;94:496–509. [Google Scholar]
- 16. Salter ML, Gupta N, Massie AB, McAdams-DeMarco MA, Law AH, Jacob RLet al. Perceived frailty and measured frailty among adults undergoing hemodialysis: a cross-sectional analysis. BMC Geriatr 2015;15:52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Dubner SJ, Teutsch C, Huisman MV, Diener HC, Halperin J, Rothman KJet al. Characteristics and 2-year outcomes of dabigatran treatment in patients with heart failure and atrial fibrillation: GLORIA-AF. ESC Heart Fail 2020;7:2679–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Savarese G, Giugliano RP, Rosano GM, McMurray J, Magnani G, Filippatos Get al. Efficacy and safety of novel oral anticoagulants in patients with atrial fibrillation and heart failure: a meta-analysis. JACC Heart Fail 2016;4:870–80. [DOI] [PubMed] [Google Scholar]
- 19. Mogensen UM, Jhund PS, Abraham WT, Desai AS, Dickstein K, Packer Met al. Type of atrial fibrillation and outcomes in patients with heart failure and reduced ejection fraction. J Am Coll Cardiol 2017;70:2490–500. [DOI] [PubMed] [Google Scholar]
- 20. McDonagh TA, Metra M, Adamo M, Gardner RS, Baumbach A, Bohm Met al. 2021 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure. Eur Heart J. 2021;42:3599–726. [DOI] [PubMed] [Google Scholar]
- 21. Patel MR, Mahaffey KW, Garg J, Pan G, Singer DE, Hacke Wet al. Rivaroxaban versus warfarin in nonvalvular atrial fibrillation. N Engl J Med 2011;365:883–91. [DOI] [PubMed] [Google Scholar]
- 22. Connolly SJ, Eikelboom J, Joyner C, Diener HC, Hart R, Golitsyn Set al. Apixaban in patients with atrial fibrillation. N Engl J Med 2011;364:806–17. [DOI] [PubMed] [Google Scholar]
- 23. Giugliano RP, Ruff CT, Braunwald E, Murphy SA, Wiviott SD, Halperin JLet al. Edoxaban versus warfarin in patients with atrial fibrillation. N Engl J Med 2013;369:2093–104. [DOI] [PubMed] [Google Scholar]
- 24. van Diepen S, Hellkamp AS, Patel MR, Becker RC, Breithardt G, Hacke Wet al. Efficacy and safety of rivaroxaban in patients with heart failure and nonvalvular atrial fibrillation: insights from ROCKET AF. Circ Heart Fail 2013;6:740–7. [DOI] [PubMed] [Google Scholar]
- 25. Magnani G, Giugliano RP, Ruff CT, Murphy SA, Nordio F, Metra Met al. Efficacy and safety of edoxaban compared with warfarin in patients with atrial fibrillation and heart failure: insights from ENGAGE AF-TIMI 48. Eur J Heart Fail 2016;18:1153–61. [DOI] [PubMed] [Google Scholar]
- 26. McMurray JJ, Ezekowitz JA, Lewis BS, Gersh BJ, van Diepen S, Amerena Jet al. Left ventricular systolic dysfunction, heart failure, and the risk of stroke and systemic embolism in patients with atrial fibrillation: insights from the ARISTOTLE trial. Circ Heart Fail 2013;6:451–60. [DOI] [PubMed] [Google Scholar]
- 27. Mulder BA, van Veldhuisen DJ, Rienstra M. What should the C (‘congestive heart failure’) represent in the CHA(2) DS(2)-VASc score? Eur J Heart Fail 2020;22:1294–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Siller-Matula JM, Pecen L, Patti G, Lucerna M, Kirchhof P, Lesiak Met al. Heart failure subtypes and thromboembolic risk in patients with atrial fibrillation: the PREFER in AF - HF substudy. Int J Cardiol 2018;265:141–7. [DOI] [PubMed] [Google Scholar]
- 29. Witt DM, Delate T, Clark NP, Martell C, Tran T, Crowther MAet al. Outcomes and predictors of very stable INR control during chronic anticoagulation therapy. Blood 2009;114:952–6. [DOI] [PubMed] [Google Scholar]
- 30. Jin H, Zhu K, Wang L, Zhou W, Zhi H. Efficacy and safety of non-vitamin K anticoagulants and warfarin in patients with atrial fibrillation and heart failure: a network meta-analysis. Thromb Res 2020;196:109–19. [DOI] [PubMed] [Google Scholar]
- 31. Gao X, Cai X, Yang Y, Zhou Y, Zhu W. Diagnostic accuracy of the HAS-BLED bleeding score in VKA- or DOAC-treated patients with atrial fibrillation: a systematic review and meta-analysis. Front Cardiovasc Med 2021;8:757087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Proietti M, Romiti GF, Raparelli V, Diemberger I, Boriani G, Dalla Vecchia LAet al. Frailty prevalence and impact on outcomes in patients with atrial fibrillation: a systematic review and meta-analysis of 1,187,000 patients. Ageing Res Rev 2022;79:101652. [DOI] [PubMed] [Google Scholar]
- 33. Dewan P, Jackson A, Jhund PS, Shen L, Ferreira JP, Petrie MCet al. The prevalence and importance of frailty in heart failure with reduced ejection fraction—an analysis of PARADIGM-HF and ATMOSPHERE. Eur J Heart Fail 2020;22:2123–33. [DOI] [PubMed] [Google Scholar]
- 34. Savelieva I, Fumagalli S, Kenny RA, Anker S, Benetos A, Boriani Get al. EHRA expert consensus document on the management of arrhythmias in frailty syndrome, endorsed by the Heart Rhythm Society (HRS), Asia Pacific Heart Rhythm Society (APHRS), Latin America Heart Rhythm Society (LAHRS), and Cardiac Arrhythmia Society of Southern Africa (CASSA). Europace 2023;25:1249–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Diemberger I, Fumagalli S, Mazzone AM, Bakhai A, Reimitz PE, Pecen Let al. Perceived vs. objective frailty in patients with atrial fibrillation and impact on anticoagulant dosing: an ETNA-AF-Europe sub-analysis. Europace 2022;24:1404–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Proietti M, Romiti GF, Vitolo M, Harrison SL, Lane DA, Fauchier Let al. Epidemiology and impact of frailty in patients with atrial fibrillation in Europe. Age Ageing 2022;51:afac192. [DOI] [PubMed] [Google Scholar]
- 37. Wilkinson C, Wu J, Clegg A, Nadarajah R, Rockwood K, Todd Oet al. Impact of oral anticoagulation on the association between frailty and clinical outcomes in people with atrial fibrillation: nationwide primary care records on treatment analysis. Europace 2022;24:1065–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Orlandi M, Dover DC, Sandhu RK, Hawkins NM, Kaul P, McAlister FA. The introduction of direct oral anticoagulants has not resolved treatment gaps for frail patients with nonvalvular atrial fibrillation. Can J Cardiol 2022;38:77–84. [DOI] [PubMed] [Google Scholar]
- 39. Joosten LPT, van Doorn S, van de Ven PM, Köhlen BTG, Nierman MC, Koek HLet al. Safety of switching from a vitamin K antagonist to a non-vitamin K antagonist oral anticoagulant in frail older patients with atrial fibrillation: results of the FRAIL-AF randomized controlled trial. Circulation 2023.. Online ahead of print. doi:10.1161/CIRCULATIONAHA.123.066485 [DOI] [PubMed] [Google Scholar]
- 40. Schnabel RB, Marinelli EA, Arbelo E, Boriani G, Boveda S, Buckley CMet al. Early diagnosis and better rhythm management to improve outcomes in patients with atrial fibrillation: the 8th AFNET/EHRA consensus conference. Europace 2023;25:6–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Patel RB, Greene SJ, Xu H, Alhanti B, Peterson P, Yancy CWet al. Intersection of atrial fibrillation and heart failure with mildly reduced and preserved ejection fraction in >400 000 participants in the Get With The Guidelines-Heart Failure Registry. Eur J Heart Fail 2023;25:63–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Patel SM, Palazzolo MG, Murphy SA, Antman EM, Braunwald E, Lanz HJet al. Evaluation of the atrial fibrillation better care pathway in the ENGAGE AF-TIMI 48 trial. Europace 2022;24:1730–8. [DOI] [PubMed] [Google Scholar]
- 43. Rillig A, Magnussen C, Ozga AK, Suling A, Brandes A, Breithardt Get al. Early rhythm control therapy in patients with atrial fibrillation and heart failure. Circulation 2021;144:845–58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Piccini JP, Caso V, Connolly SJ, Fox KAA, Oldgren J, Jones WSet al. Safety of the oral factor XIa inhibitor asundexian compared with apixaban in patients with atrial fibrillation (PACIFIC-AF): a multicentre, randomised, double-blind, double-dummy, dose-finding phase 2 study. Lancet 2022;399:1383–90. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data underlying this article are available in the article and in its Supplementary material online.