Abstract
Aims
Atrial fibrillation (AF) is a common comorbid condition in heart failure with preserved ejection fraction (HFpEF). The effect of AF on heart failure (HF) exacerbation in HFpEF has not been well described. This study investigated how AF modifies the clinical trajectory of HFpEF patients after hospitalization for decompensated HF.
Methods and results
We stratified HFpEF subjects by AF diagnosis and performed longitudinal analysis to compare risk for HF hospitalization after index hospitalization for decompensated HF. All‐cause mortality, 30 day all‐cause readmissions, and response to inpatient diuresis were also evaluated. Of 90 subjects enrolled, 35.6% (n = 32) had AF. Subjects with AF were older (72.5 vs. 60.5 years; P < 0.01), more often male (46.9% vs. 24.1%; P = 0.03), and had greater left atrial diameter (4.9 vs. 3.8 cm; P < 0.01) compared with those without AF. Subjects with AF had a higher risk for HF hospitalization than their counterparts without AF (P = 0.02); this relationship remained significant following multivariable competing risk regression with propensity score weighting (hazard ratio 2.53, P = 0.04 and hazard ratio 2.91, P = 0.04, with overlap and inverse probability weighting, respectively). Although having AF appeared to increase the risk of all‐cause hospital readmission within 30 days of discharge (37.5% vs. 17.5%; P = 0.036), this relationship failed to remain significant following propensity score adjustment for clinical covariates.
Conclusions
Atrial fibrillation is an independent risk factor for HF rehospitalization in HFpEF. Further understanding of the interplay between AF and HFpEF will be critical to guide the selection of appropriate rhythm management strategies in this population.
Keywords: Atrial fibrillation, Heart failure with preserved ejection fraction, Heart failure hospitalization, Survival analysis, All‐cause hospital readmissions, All‐cause mortality
Background
An estimated 6 million Americans will have heart failure with preserved ejection fraction (HFpEF) by 2030, representing at least half of all individuals in the USA carrying the diagnosis of congestive heart failure (HF). 1 Although HFpEF is one of the fastest growing epidemics within cardiology, it is also one of the most poorly understood. 2 Indeed, despite concerted efforts by our medical scientific community to develop our understanding of this pathology, there remain exceedingly limited therapies with proven benefit. As these efforts continue, the median survival rate for HFpEF patients following first HF hospitalization, remains poor. 3
Intrinsic to HFpEF are numerous medical comorbidities, many of which affect the clinical trajectory of these patients. At least one‐third of all HFpEF patients have atrial fibrillation (AF), which has been associated with significantly reduced exercise tolerance and poorer survival in retrospective review. 4 , 5 , 6 , 7 , 8 , 9 The presence of AF has been associated with increased right‐sided and left‐sided pressures on right heart catheterization. 10 However, understanding the effect of AF on HF exacerbation in HFpEF patients has been limited by the many potential confounders associated with carrying a concomitant diagnosis of AF. We posit that the higher right‐sided and left‐sided pressures seen in HFpEF patients with concomitant AF may lead to increased risk for HF exacerbation requiring hospitalization.
To address this hypothesis, we evaluated the impact of AF on response to therapy during index HF hospitalization and subsequent risk for rehospitalization due to HF exacerbation. Our primary outcome was time from discharge from index HF hospitalization to readmission for HF exacerbation. Secondary outcomes included response to intravenous diuretic therapy, and length of stay during the index hospitalization, all‐cause hospitalizations within 30 days of discharge, and death. Improving our understanding of the effect of AF as a risk factor for adverse outcomes in the HFpEF population is crucial to guiding selection of appropriate AF treatment strategies and carries the potential to improve the clinical trajectory of a significant subset of the HFpEF population.
Materials and methods
Study population
We performed secondary analysis of the ROPA‐DOP study, which was a prospective, randomized, single‐blinded trial conducted at Johns Hopkins Hospital. HFpEF patients (n = 90) hospitalized for treatment of acute HF were enrolled within 24 h of admission. The diagnosis of acute HF was based on the presence of at least one symptom (dyspnoea, orthopnoea, or oedema) and one sign of HF (rales, jugular venous distension, positive hepatojugular reflex, peripheral oedema, ascites, or pulmonary vascular congestion on chest radiography). Patients were included if they had an established left ventricular ejection fraction (EF) ≥ 50% within 12 months of admission without interval history, electrocardiography changes, or cardiac biomarkers to suggest a cardiac event (myocardial infarction, myocarditis, pericarditis) that may have resulted in a decline in cardiac function. Further detailed inclusion and exclusion criteria have been described previously. 11 All participants provided informed consent, and the study was approved by the Johns Hopkins Institutional Review Board. Laboratory studies including N‐terminal pro‐B type natriuretic peptide (NT‐proBNP), renal function, and exercise capacity were monitored and collected at baseline, 72 h into enrolment, and upon discharge. Clinical information for follow‐up was ascertained during active and passive surveillance following discharge from index hospitalization.
ROPA‐DOP participants were stratified by whether they carried a diagnosis of AF at the time of study enrolment: an AF diagnosis was defined as previously documented AF leading up to their index admission electrocardiogram. The primary endpoint under investigation was the time to the first hospitalization for acute HF exacerbation following discharge from index hospitalization. Secondary endpoints included all‐cause hospitalizations within 30 days of discharge from index hospitalization, all‐cause mortality, and length of stay of index hospitalization.
Statistical analysis
Simple summary statistics were used to tabulate demographic and clinical parameters stratified by AF history. Continuous data from each subgroup were analysed using the Student's t‐test or Wilcoxon rank sum test and reported as means with standard deviation. All results demonstrating non‐linearity in distribution were described using median and interquartile range. Categorical data were analysed using the χ 2 test and reported as a number (%). Univariable logistic regression was performed to assess the effect of AF on diuresis volume and on 30 day readmission risk. Potential predictor variables were categorized into the following groups: demographic parameters (i.e. age), clinically comorbid conditions [i.e. glomerular filtration rate (GFR)], and biomarkers indicating effective decongestion using intravenous diuretics (total urine output). A two‐sided P‐value < 0.05 was considered significant.
Time‐to‐event analysis for heart failure exacerbation
Time to hospitalization for HF exacerbation was estimated by taking the difference between the day of hospital admission for acute HF and the day of discharge from index hospitalization. Administrative censoring was set on 30 May 2020. Kaplan–Meier methods were used to assess unadjusted differences between study subgroups, which were compared using the log‐rank test. Factors associated with the primary endpoint were identified using Cox proportional hazard analysis. Variables with P < 0.05 in the univariable model were included in the multivariable model. In addition, history of prior hospitalizations for HF exacerbation as well as left atrial diameter were both included in the multivariable model as indirect measures of duration of AF and HF diagnoses, in order to account for lead‐time bias. 12 , 13 , 14 Interaction terms were used to assess whether there was difference between diuretic dosing strategies (continuous infusion vs. intermittent bolus) and, separately, to assess whether any differences arose from use of dopamine during inpatient decongestion.
To account for confounders associated with AF, our covariate of interest, two separate propensity score weighting techniques, overlap weighting and inverse probability treatment weighting (IPTW), were used to conduct weighted multivariable competing risk regression with robust standard errors. 15 , 16 , 17 , 18 Both of these propensity score weighting methods balance covariates based on the probability of the individual having the covariates of interest (AF), using multivariable logistic regression to adjust for covariates of age, sex, race, prior HF admissions, beta‐blocker, aldosterone, dose of loop diuretic, hypertension, hyperlipidaemia, diabetes mellitus, obstructive sleep apnoea, history of coronary artery disease, active tobacco use, 6 min walk distance upon admission for acute decompensated HF, estimated glomerular filtration rate on admission (eGFR), initial NT‐proBNP, New York Heart Association (NYHA) class, and body mass index (BMI).
To reduce the variability of IPTW weights and give individuals with extreme weights less influence, a stabilization technique was implemented by multiplying the treatment and comparison weights by a constant, equal to the expected value of being in the treatment or comparison groups. 19 , 20 Model discrimination index (c‐statistics 0.91) and Hosmer–Lemeshow goodness‐of‐fit test (P = 0.47) were used to confirm appropriate predictive power of the propensity score model prior to implementation in these doubly robust multivariable longitudinal regression models.
Odds for all‐cause hospital readmission within 30 days of index hospitalization discharge
To compare risk for all‐cause readmission to the hospital within 30 days, multivariable logistic regression was performed with sequential addition of demographic parameters to start given small sample size to avoid over‐adjustment. Interaction terms were used to assess whether there was any interaction between continuous dopamine infusion and AF status, and whether there was any interaction between diuresis‐dosing strategy (continuous vs. intermittent) and AF status. Doubly robust logistic regression was then performed alternately with overlap weighting and IPTW models as described above to adjust for clinical covariates associated with AF.
All statistical analyses were performed with STATA (STATACORP, Austin, TX, USA, Version 15.2) and R software 3.1.3 [R Foundation for Statistical Computing (http://www.r‐project.org)].
Results
Demographics and comorbidities
Out of a total of 90 HFpEF subjects enrolled, 35.6% (n = 32) had documented history of AF: 56% of these subjects had paroxysmal AF (18/32), and 44% had persistent AF (14/32); 81% (26/32) were in AF at the time of study enrolment. Subjects with AF were significantly older (72.5 vs. 60.5 years; P < 0.01), more often male (46.9% vs. 24.1%; P = 0.03), had greater left atrial diameter by echocardiography (4.9 cm [4.5, 5.3] vs. 3.8 cm [3.4, 4.3]; P < 0.01), and trended towards higher NT‐proBNP levels on admission (1918 pg/mL [991, 3900] vs. 961.5 pg/mL [169, 2417]; P = 0.10) compared with those without AF (Table 1 ). Unadjusted univariable analysis of study subgroups suggested that AF was not associated with a significantly higher prevalence of other common clinical comorbidities, such as hypertension and diabetes mellitus. Biomarkers of comorbid disease severity, including haemoglobin A1c and eGFR, were also not significantly different between HFpEF subjects with and without AF. The mean follow‐up for study subjects was 870 days until the subjects were censored, from reaching the primary endpoint of HF rehospitalization, death, or administrative censoring.
Table 1.
Baseline demographics and clinical findings stratified by history of atrial fibrillation
| HFpEF with AF (n = 32) | HFpEF without AF (n = 58) | P‐value a | |
|---|---|---|---|
| Demographics and clinical comorbidities | |||
| Age (years) | 72.5 [66, 82] | 60.5 [52, 70] | <0.01 |
| Gender (% female) | 53.1 (17/32) | 75.9 (44/58) | 0.03 |
| Race (% African American) | 62.5 (20/32) | 62.1 (36/58) | 0.97 |
| NYHA class (I–IV) | 3 | 3 | 0.33 |
| Hypertension (%) | 96.9 [89.8, 103.4] | 93 [86.1, 99.8] | 0.46 |
| Diabetes mellitus (%) | 53.3 (17/32) | 60.4% (35/58) | 0.71 |
| Former smoking (%) | 12.5 (4/32) | 15.7 (9/58) | 0.56 |
| Obstructive sleep apnoea (%) | 40.6 [21.4, 58.6] | 32.7 [19.1, 44.0] | 0.46 |
| Body mass index (kg/m2) | 35.5 [27.7, 45.9] | 41.6 [34.1, 49.1] | 0.19 |
| Systolic blood pressure (mmHg) | 123.5 [113.5, 141.5] | 137.5 [115, 160] | 0.22 |
| Heart rate (b.p.m.) | 71 [62.5, 78] | 80 [70, 90] | 0.03 |
| 6MWT at enrolment (feet) | 83.8 [16.2, 142.6] | 85 [15.2, 155.07] | 0.86 |
| Global well‐being assessment score (1–100) | 25 [10, 76] | 40 [26, 59] | 0.42 |
| Left atrial diameter (cm) | 4.9 [4.5, 5.3] | 3.8 [3.4, 4.3] | <0.01 |
| LVEF (%) | 60 [55, 65] | 65 [60, 70] | 0.17 |
| LVEDD (cm) | 4.75 [4.2, 5.3] | 4.7 [4.3, 5.1] | 0.89 |
| LVMI (kg/m2) | 97.9 [74.2, 125.7] | 91 [73.4, 114.28] | 0.37 |
| NT‐proBNP (pg/mL) | 1918 [991, 3900] | 961.5 [169, 2417] | 0.10 |
| GFR (mL/min per 1.73 m2) | 45.95 [33.9, 80.6] | 62.5 [42.5, 88.4] | 0.38 |
| Beta‐blocker (%) | 72 (23/32) | 58 (34/58) | 0.18 |
| Aldosterone (%) | 25 (8/32) | 14 (8/58) | 0.17 |
| ACE inhibitors (%) | 28 (9/32) | 34 (20/58) | 0.57 |
| Insulin (%) | 25 (8/32) | 36 (21/58) | 0.28 |
| Clinical treatment outcomes | |||
| Change in weight at discharge (kg) | 4.35 [1.5, 10] | 2.70 [1.10, 5.40] | 0.06 |
| Change in NT‐proBNP at discharge (pg/mL) | 620 [204, 1095] | 563 [68, 1314] | 0.93 |
| Change in 6MWT at discharge (feet) | 0 [0, 0] | 0 [0, 0] | 0.31 |
| Per cent change in creatinine at 72 h | 8.71 [1.66, 18.18] | 7.29 [0, 25] | 1.00 |
| Length of inpatient hospitalization (days) | 8 [6, 11] | 6 [4, 8] | 0.67 |
6MWT, 6 min walk test; ACE, angiotensin‐converting enzyme; AF, atrial fibrillation; GFR, glomerular filtration rate; HFpEF, heart failure with preserved ejection fraction; LVEDD, left ventricular end‐diastolic diameter; LVEF, left ventricular ejection fraction; LVMI, left ventricular mass index; NT‐proBNP, N‐terminal pro‐B type natriuretic peptide; NYHA, New York Heart Association.
Left atrial diameter is measured using the parasternal long‐axis view. Entries are listed as median values with 95% confidence interval. Continuity corrected χ 2 testing was performed for covariates with non‐normal distributions identified using Shapiro–Wilk's testing. Categorical variables were reported as percentage (number of subjects/total subjects observed in AF subcategory).
P‐values based on unequal variance t‐test for continuous variables and Fisher's exact test.
Response to decongestion with intravenous diuretic therapy during index hospitalization
In general, NT‐proBNP on admission was higher in subjects with AF; however, there was no significant difference in length of stay, volume of decongestion, renal function, nor change in NT‐proBNP from admission to discharge following intravenous diuresis (Figure 1 ). HFpEF subjects with AF had comparable exercise capacity to their counterparts without AF on admission to the hospital, and neither subgroup showed significant improvement in exercise capacity between admission and discharge, as measured by 6 min walk test distance. Subjects with AF endorsed similar symptoms as their counterparts without AF upon admission and also had comparable improvement in their overall symptomatology, as measured by interval changes in global well‐being and dyspnoea scores taken on admission and discharge from index hospitalization.
Figure 1.
Response to inpatient decongestion with intravenous diuretic therapy, stratified by AF status. No significant difference was seen in either AF subgroup in response to intravenous diuretic therapy during index hospitalization for decompensated heart failure, with regard to (A) total length of stay, (B) change in weight from admission to discharge, (C) change in NT‐proBNP from admission to discharge, (D) per cent change in NT‐proBNP from admission to discharge, (E) creatinine over the course of hospitalization, and (F) prescribed loop diuretic dose over 24 h. AF, atrial fibrillation; HFpEF, heart failure with preserved ejection fraction; NT‐proBNP, N‐terminal pro‐B type natriuretic peptide.
Time to heart failure hospitalization
Following discharge from index hospitalization, the subgroup of HFpEF patients with AF had a significantly higher hazard for HF hospitalization than their counterparts without AF, with a total of 18 out of 32 individuals with AF and 43 out of 58 individuals without AF requiring inpatient admission for HF exacerbation (Figures 2 B and 3 B, P = 0.02). The univariable associations of clinical comorbidities with AF were assessed. Using a stepwise forward Cox regression model, AF remained significantly associated with higher hazard for HF decompensation after adjusting for AF, age, sex, prior history of HF hospitalization, GFR, and left atrial diameter (hazard ratio = 2.18 [1.09, 4.33]; P = 0.03; Table 2 A ). All‐cause mortality was also noted during longitudinal follow‐up, affecting a total of 3 out of 32 individuals with AF and 8 out of 58 individuals without AF (Figure 3 A and 3 C, P = 0.98). Multivariable competing risk regression was subsequently performed to account for mortality events that preceded HF hospitalizations, which demonstrated that AF remained a significant risk factor for HF hospitalization (hazard ratio = 1.77 [1.04, 3.02]; P = 0.04; Table 2 B ). Further covariate balance was achieved with overlap weighting and alternately with IPTW. Multivariable competing risk regression using either of the two propensity score weighting strategies revealed AF to be a strongly significant risk factor for HF hospitalization. In sum, carrying a diagnosis of AF was associated with at least a 2.5‐fold greater hazard for HF hospitalization, even after doubly robust regression strategies to balance contribution from clinical confounders.
Figure 2.
Unadjusted risk for (A) all‐cause and (B) heart failure hospitalizations following index hospitalization in HFpEF patients, stratified by diagnosis of AF. Patients carrying both diagnoses of HFpEF and AF carried a significantly higher risk for readmission within 1 month of hospital admission. AF, atrial fibrillation; HFpEF, heart failure with preserved ejection fraction.
Figure 3.
Kaplan–Meier curves showing time‐to‐event analysis following index hospitalization for (A) composite outcome of heart failure hospitalization or all‐cause mortality, (B) heart failure exacerbation requiring hospitalization, and (C) all‐cause mortality. Groups are stratified by concomitant diagnosis of AF, with number of subjects in each subgroup at risk listed in the tables. AF, atrial fibrillation.
Table 2.
Unadjusted and adjusted longitudinal data analysis showing both (A) Cox proportional regression and (B) competing risk regression for heart failure hospitalization following index hospitalization
| (A) | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Cox regression | Unweighted | Inverse probability weighting | Overlap weighting | ||||||
| HR | CI | P‐value | HR | CI | P‐value | HR | CI | P‐value | |
| Treatment group | |||||||||
| AF vs. no AF | 2.18 | [1.09, 4.33] | 0.027 | 3.19 | [1.10, 9.24] | 0.033 | 2.69 | [1.07, 6.78] | 0.035 |
| Covariates | |||||||||
| Age | 1.01 | [0.99, 1.03] | 0.44 | 1.01 | [0.98, 1.04] | 0.46 | 1 | [0.98, 1.04] | 0.55 |
| GFR | 0.99 | [0.98, 1.00] | 0.1 | 0.99 | [0.98, 1.00] | 0.06 | 0.99 | [0.97, 1.00] | 0.11 |
| Prior HF admission | 1.75 | [0.99, 3.10] | 0.05 | 2.28 | [1.07, 4.87] | 0.033 | 2.16 | [1.02, 4.60] | 0.045 |
| Left atrial diameter | 0.76 | [0.52, 1.10] | 0.15 | 0.63 | [0.37, 1.07] | 0.09 | 0.68 | [0.44, 1.04] | 0.08 |
| (B) | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Competing risk regression | Unweighted | Inverse probability weighting | Overlap weighting | ||||||
| HR | CI | P‐value | HR | CI | P‐value | HR | CI | P‐value | |
| Treatment group | |||||||||
| AF vs. no AF | 1.77 | [1.04, 3.02] | 0.035 | 2.91 | [1.04, 8.19] | 0.04 | 2.53 | [1.03, 6.22] | 0.04 |
| Covariates | |||||||||
| Age | 1 | [0.98, 1.02] | 0.99 | 1 | [0.97, 1.03] | 0.96 | 1 | [0.97, 1.03] | 0.92 |
| GFR | 0.99 | [0.98, 1.00] | 0.1 | 0.99 | [0.98, 1.00] | 0.08 | 0.99 | [0.98, 1.00] | 0.12 |
| Prior HF admission | 1.77 | [1.00, 3.12] | 0.05 | 2.02 | [0.98, 4.19] | 0.06 | 2.03 | [0.99, 4.13] | 0.05 |
| Left atrial diameter | 0.74 | [0.54, 1.03] | 0.08 | 0.69 | [0.42, 1.14] | 0.15 | 0.74 | [0.49, 1.10] | 0.14 |
AF, atrial fibrillation; CI, confidence interval; GFR, glomerular filtration rate; HF, heart failure; HR, hazard ratio.
Competing risk regression was analysed against risk for death, with and without propensity score weighting, adjusted for AF status, age, GFR, prior history of HF, and echocardiographic proxy measure for duration of AF/HF with preserved ejection fraction. Adjusted HRs were performed serially with only propensity score weighting and AF as the other multivariable covariate, given limited sample size. Adjusted AF hazard is listed after adjusting for all listed covariates. From this table, it is evident that both assessed with history of AF and evidence of AF on admission EKG remained statistically significant after doubly robust multivariable regression.
Thirty‐day all‐cause hospital readmissions
Univariable analysis showed that carrying a concomitant AF diagnosis in HFpEF was associated with higher rates of 30 day hospital readmission (Figure 2 A , 37.5% vs. 17.5%, n = 89; P = 0.04), with 7 out of 32 participants with AF and 7 out of 58 participants without AF requiring rehospitalization. The presence of AF remained an independent predictor of 30 day hospital readmission (odds ratio 4.13 [1.23, 13.55]; P = 0.02) after adjusting for age, gender, and race in multivariable logistic regression analysis. However, unlike the sustained risk AF posed for HF readmissions, this relationship failed to remain significant after further covariate balance was sought using propensity score weighting, suggesting that this increased risk was likely attributable to clinical covariates associated with carrying an AF diagnosis (Table 3 ).
Table 3.
Crude and adjusted logistic regression of 30 day readmission on atrial fibrillation status, demographics, and biomarkers of effective inpatient diuresis
| Crude | Multivariable logistic regression | Overlap weighting + Multivariable logistic regression | |||||||
|---|---|---|---|---|---|---|---|---|---|
| OR | 95% CI | P‐value | OR | 95% CI | P‐value | OR | 95% CI | P‐value | |
| AF vs. no AF | 3.61 | 1.29, 10.08 | 0.01 | 4.17 b | 1.26, 13.55 | 0.019 | 2.54b | 0.51, 12.68 | 0.254 |
| Demographics | |||||||||
| Age (years) | 1.02 | 0.98, 1.06 | 0.82 | 0.99 a | 0.95, 1.04 | 0.807 | 1.00 a | 0.95, 1.06 | 0.874 |
| Male vs. female | 1.16 | 0.39, 3.44 | 0.785 | 1.50 a | 0.46, 4.88 | 0.499 | 3.31 a | 0.61, 17.89 | 0.16 |
| Race/ethnicity | |||||||||
| Black vs. other | 0.86 | 0.34, 2.12 | 0.736 | 0.91 a | 0.34, 2.42 | 0.848 | 0.96 a | 0.29, 3.10 | 0.94 |
| Biomarkers of effective diuresis | |||||||||
| Urine output from admission to discharge | 1.00 | 0.99, 1.01 | 0.253 | 1.00 a | 0.99, 1.00 | 0.201 | 1.00 a | 0.99, 1.00 | 0.27 |
| 6 min walk test (feet) | 0.99 | 0.98, 1.00 | 0.199 | 0.99 a | 0.99, 1.00 | 0.335 | 1.00 a | 0.99, 1.01 | 0.93 |
| Glomerular filtration rate (mL/min per 1.73 m2) | 0.99 | 0.98, 1.02 | 0.801 | 1.00 a | 0.98, 1.02 | 0.869 | 1.00 a | 0.98, 1.03 | 0.70 |
| Change in pro‐BNP between admission and discharge | 1.00 | 0.99, 1.00 | 0.652 | 1.00 a | 0.99, 1.00 | 0.532 | 0.99 a | 0.99, 1.00 | 0.98 |
AF, atrial fibrillation; CI, confidence interval; OR, odds ratio.
Adjusted: The adjusted odds for biomarkers of effective diuresis was performed serially with only AF status as the other multivariable covariate, given limited sample size. Adjusted AF odds remained significantly higher after correcting for each biomarker of effective diuresis.
AF is adjusted for all demographic parameters along with change in pro‐BNP.
Discussion
Atrial fibrillation has previously been identified as a risk factor for decreased exercise capacity, 7 stroke, 21 and mortality in HFpEF. 8 , 22 , 23 , 24 However, little is known regarding the impact of AF on response to inpatient therapies for decompensated HF and the risk for subsequent HF hospitalization. We assessed the impact of AF as a risk factor for HF and all‐cause readmission in hospitalized HFpEF patients, using doubly robust regression techniques to isolate the contribution of AF from associated clinical comorbidities. We found that HFpEF patients with AF are at significantly higher risk for HF readmission and showed that this risk was in spite of adjustment for other clinical risk factors. Our findings add to the growing investigation surrounding the interplay between AF and HFpEF. Previous studies on risk for HF hospitalization in AF were largely focused on outcomes in HF with reduced EF, with very few studies reporting effects on risk for decompensation in HFpEF. 24 , 25 , 26
Atrial fibrillation did not significantly impact diuretic response in HFpEF patients admitted for HF exacerbation. HFpEF subjects with and without AF had comparable hospitalization days, urine output, change in body weight, and similar decreases in NT‐proBNP over the course of their hospitalization. We postulate that the similarities in response to inpatient therapy arise from two possible sources. First, inpatients are subject to frequent dose adjustments and titrations of decongestive therapies in response to serially checked parameters for kidney function and urine output. These frequent adjustments may mask underlying differences in response to fluid fluctuations and decongestion that may exist between HFpEF patients with and without AF on the inpatient front and lends further credence to our explanation for their divergent clinical courses following hospital discharge. Second, while we know that over 85% of the patients who had a history of AF presented with this arrhythmia on the day of admission for their index hospitalization, we do not have information on their AF burden—the proportion of time they were in AF—over the course of their index hospitalization. As the interquartile ranges shown for various decongestion parameters are consistently wider for HFpEF patients with AF, it may be that responsiveness to decongestion is related to overall AF burden. Understanding the relationship between AF burden as a continuous variable and its impact on clinical risk factors in the HFpEF population will be critical in guiding selection of AF therapies and requires additional investigation.
Following discharge from index hospitalization, HFpEF patients with AF were at significantly higher risk for redeveloping decompensated HF requiring hospitalization. This may be because HFpEF patients with AF have a narrower euvolemic window, with higher baseline right‐sided filling pressures that place them at higher risk for HF decompensation in the setting of natural outpatient fluctuations in fluid status. Previous publications have highlighted baseline haemodynamic differences in left‐sided and right‐sided filling pressures for HFpEF patients with AF. 7 , 10 From a physiologic perspective, elimination of atrial filling in end‐diastole decreases cardiac output and simultaneously leads to increases in pulmonary pressures. These effects may lead to greater intolerance to fluid fluctuations, decreased exercise capacity, and increased susceptibility to developing volume overload and decompensated HF.
While clinical comorbidities associated with AF may certainly contribute to risk for HF exacerbation, we demonstrate a sustained significant association between AF and risk for decompensated HF. These findings suggest that HFpEF patients with AF may benefit from closer outpatient follow‐up. As such, further investigation on the frequency of follow‐up for HFpEF patients with AF should be performed to try to prevent readmissions.
We found that AF was associated with significantly higher rates of all‐cause readmission within 30 days and remained an independent predictor of 30 day readmission after adjusting for age, gender, and race in multivariable logistic regression analysis. However, AF did not remain a significant risk factor after propensity score weighting adjustments. This suggests that all‐cause readmissions in HFpEF, unlike HF readmissions, are driven largely by the multiple comorbidities common to many HFpEF patients, rather than AF alone. However, unlike many other clinical risk factors associated with poor outcomes in HFpEF, there are many different therapies readily available for managing this arrhythmia. Demonstrating that AF serves as an independent risk factor for HF exacerbation, even after balancing for other clinical comorbidities, raises the important question of whether we are able to modify AF to abrogate risk for HF decompensation in the HFpEF population.
Given recent randomized control studies assessing rhythm control strategies in HF with reduced EF (i.e. CASTLE‐AF and CAMERA‐MRI), further studies on rhythm control in AF for HFpEF patients need to be conducted and carry the potential for great clinical impact. 27 , 28 While none of the study subjects underwent catheter ablation for AF management, other observational retrospective studies and post hoc analyses have suggested promising initial results, 29 , 30 which will need to be verified with prospective randomized control studies dedicated to investigating the impact of AF ablation in the HFpEF population. In preparation for these randomized control studies, additional research should be done to enhance our understanding of the relationship between continuous AF burden and risk for HF hospitalization, which will be critical in identifying which subgroup of HFpEF patients with AF may benefit most from rhythm control therapies.
This study should be taken in the context of several limitations. First, this was a relatively small, single‐centre data study, and therefore limited the power necessitating use of propensity score methods to balance the potential confounders, and external generalizability. Although the study design attempted to mitigate potential confounding diagnoses leading to the inclusion of individuals with diastolic dysfunction arising from valvular and infiltrative aetiologies, there is intrinsic heterogeneity that underlies the HFpEF population. As such, one major limitation of this study arises from the residual confounding present within the many contributing entities leading to diastolic dysfunction that is inherent to any research done on this HF subpopulation. Second, lead‐time bias was not able to be completely addressed: the index hospitalization was not confirmed to be the first HF hospitalization for each patient, nor was the duration since diagnosis of HFpEF or AF available for analysis. As time since diagnosis of HFpEF, of AF, and the time since first acute HF hospitalization were not available for analysis, we attempted to adjust for these factors using indirect covariates that have been previously correlated with duration with HF and AF, such as left atrial size. Finally, our study is limited by potential residual confounding from risk factors that may not have been fully addressed by multivariable regression and propensity score weighting techniques. Despite the limitations inherent to the data source available, we felt that the robust longitudinal clinical data with relatively little patient drop‐out allowed for more granular adjustment of confounders known to be predictive of clinical outcomes. Finally, 6% of individuals who did not carry an AF diagnosis upon index hospitalization were noted to have developed AF during longitudinal follow‐up; the development of new onset AF may be one of the explanations to why the initially significant differences noted in short‐term follow‐up became less pronounced over time.
Conclusions
In a hospitalized HFpEF cohort with acute decompensated HF, we demonstrated that a concomitant diagnosis of AF was associated with increased risk for recurrent HF exacerbation. While the increased risk for readmission within 30 days may largely be attributable to other comorbid clinical risk factors that are associated with AF, AF serves as a strong, independent risk factor for HF exacerbation. Our data suggest that HFpEF patients with AF may benefit from closer outpatient follow‐up given significantly higher risk for redeveloping decompensated HF. These findings also underscore the need to determine whether AF is a modifiable risk factor in HFpEF: in light of recent randomized control studies showing benefit in rhythm control in HF with reduced EF, 27 , 28 further studies investigating the potential benefit of reducing AF burden in HFpEF patients will be imperative to guiding treatment of individuals with both of these cardiovascular conditions.
Conflict of interest
J.V. declared no relevant disclosures or conflicts of interest. E.S. declared no relevant disclosures or conflicts of interest. J.L. declared no relevant disclosures or conflicts of interest. S.S. declared no relevant disclosures or conflicts of interest. H.C. declared no relevant disclosures or conflicts of interest. R.B. declared no relevant disclosures or conflicts of interest. S.R. declared no relevant disclosures or conflicts of interest. K.S. is an advisory board member and consultant for AstaZeneca, Bayer, Boehringer Ingelheim, Bristol Myers Squibb, Janssen, Novartis and NovoNordisk and receives honoraria. She receives research funding support from the American Heart Association and Amgen.
Funding
E.Y. was funded by the NIH LRP‐CR, NOT‐OD‐20‐133, and The Lou and Nancy Grasmick Endowed Research Fellowship.
Yang, E. , Vaishnav, J. , Song, E. , Lee, J. , Schulman, S. , Calkins, H. , Berger, R. , Russell, S. D. , and Sharma, K. (2022) Atrial fibrillation is an independent risk factor for heart failure hospitalization in heart failure with preserved ejection fraction. ESC Heart Failure, 9: 2918–2927. 10.1002/ehf2.13836.
References
- 1. Virani SS, Alonso A, Benjamin EJ, Bittencourt MS, Callaway CW, Carson AP, Chamberlain AM, Chang AR, Cheng S, Delling FN, Djousse L, Elkind MSV, Ferguson JF, Fornage M, Khan SS, Kissela BM, Knutson KL, Kwan TW, Lackland DT, Lewis TT, Lichtman JH, Longenecker CT, Loop MS, Lutsey PL, Martin SS, Matsushita K, Moran AE, Mussolino ME, Perak AM, Rosamond WD, Roth GA, Sampson UKA, Satou GM, Schroeder EB, Shah SH, Shay CM, Spartano NL, Stokes A, Tirschwell DL, VanWagner L, Tsao CW, American Heart Association Council on Epidemiology and Prevention Statistics Committee and Stroke Statistics Subcommittee . Heart Disease and Stroke Statistics—2020 update: a report from the American Heart Association. Circulation 2020; 141: e139–e596. [DOI] [PubMed] [Google Scholar]
- 2. Sharma K, Kass DA. Heart failure with preserved ejection fraction: mechanisms, clinical features, and therapies. Circ Res 2014; 115: 79–96. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Steinberg BA, Zhao X, Heidenreich PA, Peterson ED, Bhatt DL, Cannon CP, Hernandez AF, Fonarow GC. Trends in patients hospitalized with heart failure and preserved left ventricular ejection fraction: prevalence, therapies, and outcomes. Circulation 2012; 126: 65–75. [DOI] [PubMed] [Google Scholar]
- 4. McManus DD, Hsu G, Sung SH, Saczynski JS, Smith DH, Magid DJ, Gurwitz JH, Goldberg RJ, Go AS, Cardiovascular Research Network PRESERVE Study . Atrial fibrillation and outcomes in heart failure with preserved versus reduced left ventricular ejection fraction. J Am Heart Assoc 2013; 2: e005694. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Dries DL, Exner DV, Gersh BJ, Domanski MJ, Waclawiw MA, Stevenson LW. Atrial fibrillation is associated with an increased risk for mortality and heart failure progression in patients with asymptomatic and symptomatic left ventricular systolic dysfunction: a retrospective analysis of the SOLVD trials. Studies of Left Ventricular Dysfunction. J Am Coll Cardiol 1998; 32: 695–703. [DOI] [PubMed] [Google Scholar]
- 6. Wang TJ, Larson MG, Levy D, Vasan RS, Leip EP, Wolf PA, D'Agostino RB, Murabito JM, Kannel WB, Benjamin EJ. Temporal relations of atrial fibrillation and congestive heart failure and their joint influence on mortality: the Framingham Heart Study. Circulation 2003; 107: 2920–2925. [DOI] [PubMed] [Google Scholar]
- 7. Kaye DM, Silvestry FE, Gustafsson F, Cleland JG, van Veldhuisen DJ, Ponikowski P, Komtebedde J, Nanayakkara S, Burkhoff D, Shah SJ. Impact of atrial fibrillation on rest and exercise haemodynamics in heart failure with mid‐range and preserved ejection fraction. Eur J Heart Fail 2017; 19: 1690–1697. [DOI] [PubMed] [Google Scholar]
- 8. Elshazly MB, Senn T, Wu Y, Lindsay B, Saliba W, Wazni O, Cho L. Impact of atrial fibrillation on exercise capacity and mortality in heart failure with preserved ejection fraction: insights from cardiopulmonary stress testing. J Am Heart Assoc 2017; 6: e006662. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Olsson LG, Swedberg K, Ducharme A, Granger CB, Michelson EL, McMurray J, Puu M, Yusuf S, Pfeffer MA, CHARM Investigators . Atrial fibrillation and risk of clinical events in chronic heart failure with and without left ventricular systolic dysfunction: results from the Candesartan in Heart failure‐Assessment of Reduction in Mortality and morbidity (CHARM) program. J Am Coll Cardiol 2006; 47: 1997–2004. [DOI] [PubMed] [Google Scholar]
- 10. Reddy YNV, Obokata M, Verbrugge FH, Lin G, Borlaug BA. Atrial dysfunction in patients with heart failure with preserved ejection fraction and atrial fibrillation. J Am Coll Cardiol 2020; 76: 1051–1064. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Sharma K, Vaishnav J, Kalathiya R, Hu JR, Miller J, Shah N, Hill T, Sharp M, Tsao A, Alexander KM, Gupta R, Montemayor K, Kovell L, Chasler JE, Lee YJ, Fine DM, Kass DA, Weiss RG, Thiemann DR, Ndumele CE, Schulman SP, Russell SD, Osler Medical Housestaff . Randomized evaluation of heart failure with preserved ejection fraction patients with acute heart failure and dopamine: the ROPA‐DOP trial. JACC Heart Fail 2018; 6: 859–870. [DOI] [PubMed] [Google Scholar]
- 12. Suarez GS, Lampert S, Ravid S, Lown B. Changes in left atrial size in patients with lone atrial fibrillation. Clin Cardiol 1991; 14: 652–656. [DOI] [PubMed] [Google Scholar]
- 13. Goldberg A, Kusunose K, Qamruddin S, Rodriguez LL, Mazgalev TN, Griffin BP, van Wagoner DR, Zhang Y, Popović ZB. Left atrial size and function in a canine model of chronic atrial fibrillation and heart failure. PLoS ONE 2016; 11: e0147015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Blume GG, Mcleod CJ, Barnes ME, Seward JB, Pellikka PA, Bastiansen PM, Tsang TSM. Left atrial function: physiology, assessment, and clinical implications. Eur J Echocardiogr 2011; 12: 421–430. [DOI] [PubMed] [Google Scholar]
- 15. Robins JM, Hernan MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology 2000; 11: 550–560. [DOI] [PubMed] [Google Scholar]
- 16. Thomas LE, Li F, Pencina MJ. Overlap weighting: a propensity score method that mimics attributes of a randomized clinical trial. JAMA 2020; 323: 2417. [DOI] [PubMed] [Google Scholar]
- 17. Austin PC, Schuster T. The performance of different propensity score methods for estimating absolute effects of treatments on survival outcomes: a simulation study. Stat Methods Med Res 2016; 25: 2214–2237. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Austin PC. The performance of different propensity score methods for estimating marginal hazard ratios. Stat Med 2013; 32: 2837–2849. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Harder VS, Stuart EA, Anthony JC. Propensity score techniques and the assessment of measured covariate balance to test causal associations in psychological research. Psychol Methods 2010; 15: 234–249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Xu S, Ross C, Raebel MA, Shetterly S, Blanchette C, Smith D. Use of stabilized inverse propensity scores as weights to directly estimate relative risk and its confidence intervals. Value Health 2010; 13: 273–277. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Oluleye OW, Rector TS, Win S, McMurray JJV, Zile MR, Komajda M, McKelvie RS, Massie B, Carson PE, Anand IS. History of atrial fibrillation as a risk factor in patients with heart failure and preserved ejection fraction. Circ Heart Fail 2014; 7: 960–966. [DOI] [PubMed] [Google Scholar]
- 22. Goyal P, Almarzooq ZI, Cheung J, Kamel H, Krishnan U, Feldman DN, Horn EM, Kim LK. Atrial fibrillation and heart failure with preserved ejection fraction: insights on a unique clinical phenotype from a nationally‐representative United States cohort. Int J Cardiol 2018; 266: 112–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Sartipy U, Dahlström U, Fu M, Lund LH. Atrial fibrillation in heart failure with preserved, mid‐range, and reduced ejection fraction. JACC Heart Fail 2017; 5: 565–574. [DOI] [PubMed] [Google Scholar]
- 24. Cikes M, Claggett B, Shah AM, Desai AS, Lewis EF, Shah SJ, Anand IS, O'Meara E, Rouleau JL, Sweitzer NK, Fang JC, Saksena S, Pitt B, Pfeffer MA, Solomon SD. Atrial fibrillation in heart failure with preserved ejection fraction: the TOPCAT trial. JACC Heart Fail 2018; 6: 689–697. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Zareba W, Steinberg JS, McNitt S, Daubert JP, Piotrowicz K, Moss AJ. Implantable cardioverter‐defibrillator therapy and risk of congestive heart failure or death in MADIT II patients with atrial fibrillation. Heart Rhythm 2006; 3: 631–637. [DOI] [PubMed] [Google Scholar]
- 26. Hagens VE, Crijns HJ, van Veldhuisen D, van den Berg M, Rienstra M, Ranchor AV, Bosker HA, Kamp O, Tijssen JG, Veeger NJ, van Gelder I, RAte Control versus Electrical cardioversion for persistent atrial fibrillation study group . Rate control versus rhythm control for patients with persistent atrial fibrillation with mild to moderate heart failure: results from the RAte Control versus Electrical cardioversion (RACE) study. Am Heart J 2005; 149: 1106–1111. [DOI] [PubMed] [Google Scholar]
- 27. Marrouche NF, Brachmann J, Andresen D, Siebels J, Boersma L, Jordaens L, Merkely B, Pokushalov E, Sanders P, Proff J, Schunkert H, Christ H, Vogt J, Bänsch D. Catheter ablation for atrial fibrillation with heart failure. N Engl J Med 2018; 378: 417–427. [DOI] [PubMed] [Google Scholar]
- 28. Prabhu S, Taylor AJ, Costello BT, Kaye DM, McLellan AJA, Voskoboinik A, Sugumar H, Lockwood SM, Stokes MB, Pathik B, Nalliah CJ, Wong GR, Azzopardi SM, Gutman SJ, Lee G, Layland J, Mariani JA, Ling LH, Kalman JM, Kistler PM. Catheter ablation versus medical rate control in atrial fibrillation and systolic dysfunction: the CAMERA‐MRI study. J Am Coll Cardiol 2017; 70: 1949–1961. [DOI] [PubMed] [Google Scholar]
- 29. Kelly JP, DeVore A, Wu J, Hammill BG, Sharma A, Cooper LB, Felker GM, Piccini JP, Allen LA, Heidenreich PA, Peterson ED, Yancy CW, Fonarow GC, Hernandez AF. Rhythm control versus rate control in patients with atrial fibrillation and heart failure with preserved ejection fraction: insights from get with the guidelines‐heart failure. J Am Heart Assoc 2019; 8: e011560. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Packer DL, Piccini JP, Monahan KH, al‐Khalidi HR, Silverstein AP, Noseworthy PA, Poole JE, Bahnson TD, Lee KL, Mark DB, For the CABANA Investigators . Ablation versus drug therapy for atrial fibrillation in heart failure: results from the CABANA trial. Circulation 2021; 143: 1377–1390. [DOI] [PMC free article] [PubMed] [Google Scholar]