Ablation Versus Drug Therapy for Atrial Fibrillation in Heart Failure: Results from the CABANA Trial

Abstract

Background:

In patients with heart failure (HF) and atrial fibrillation (AF), several clinical trials have reported improved outcomes, including freedom from AF recurrence, quality of life (QOL), and survival, with catheter ablation. This report describes the treatment-related outcomes of the AF patients with HF enrolled in the Catheter Ablation vs Antiarrhythmic Drug Therapy for Atrial Fibrillation (CABANA) trial.

Methods:

CABANA randomized 2204 patients with AF who were ≥65 years old or <65 with ≥1 risk factor for stroke at 126 sites to ablation with pulmonary vein isolation or drug therapy including rate/rhythm control drugs. Of these, 778 (35%) had NYHA class ≥ II at baseline and form the subject of this report. The CABANA primary endpoint was a composite of death, disabling stroke, serious bleeding, or cardiac arrest.

Results:

Of the 778 HF patients enrolled in CABANA, 378 were assigned to ablation and 400 to drug therapy. Ejection fraction (EF) at baseline was available for 571 patients (73%) and 9.3% of these had an EF <40%, while 11.7% had EFs between 40–50%. In the intention-to-treat analysis, the ablation arm had a 36% relative reduction in the primary composite endpoint (hazard ratio [HR] 0.64; 95% confidence interval [CI], 0.41 to 0.99) and a 43% relative reduction in all-cause mortality (HR 0.57; 95% CI, 0.33 to 0.96) compared to drug therapy alone over a median follow-up of 48.5 months. AF recurrence was decreased with ablation (HR 0.56; 95% CI, 0.42 to 0.74). The adjusted mean difference for the AF Effect on QOL (AFEQT) summary score averaged over the entire 60-month follow-up was 5.0 points favoring the ablation arm (95% CI, 2.5 to 7.4 points), and the Mayo AF-specific Symptom Inventory (MAFSI) frequency score difference was −2.0 points favoring ablation (95% CI, −2.9 to −1.2).

Conclusions:

In patients with atrial fibrillation enrolled in CABANA who had clinically diagnosed stable heart failure at trial entry, catheter ablation produced clinically important improvements in survival, freedom from AF recurrence, and quality of life relative to drug therapy. These results, obtained in a cohort most of whom had preserved left ventricular function, require independent trial verification.

Clinical Trial Registration:

ClinicalTrials.gov Identifier: NCT0091150.

https://www.clinicaltrials.gov/ct2/show/NCT00911508

Introduction

Atrial fibrillation (AF) and heart failure (HF) often occur in the same patients and have a complex, incompletely understood interrelationship. In particular, while they have common antecedents, each also appears to promote development and progression of the other.¹ AF may lead to a decrease in ejection fraction (EF) and onset of symptomatic HF, particularly if the AF is sustained for long periods and/or produces high ventricular heart rates. Progressive heart muscle disease is also associated with a higher propensity to develop AF and to progress to more persistent forms of the disease. Optimal treatment of HF in patients with AF has been associated with improved maintenance of sinus rhythm.²

Several randomized clinical trials have reported that both AF and HF outcomes can be improved with catheter ablation.^3–11 Observational data has further suggested that ablation of AF is similarly effective in patients who have HF regardless of whether ejection fraction (EF) is preserved or reduced.^{12, 13} However, generalization from this evidence base to clinical practice is limited by important remaining uncertainties related to the modest number of randomized patients, the absence of any randomized trials in HF patients with preserved EF, and the substantial variations in methods and study cohort selection criteria employed.

The Catheter Ablation vs Antiarrhythmic Drug Therapy for AF (CABANA) trial, the largest trial to date of catheter ablation versus drug therapy in AF, found that the strategy of catheter ablation did not significantly improve the composite primary clinical outcome (death, disabling stroke, serious bleeding, or cardiac arrest) compared with drug therapy when analyzed by intention-to-treat.¹⁴ Secondary endpoints of death or cardiovascular (CV) hospitalization, and AF recurrence were significantly reduced by ablation, and quality of life (QOL) was improved out to 60 months.^14–16 In the pre-specified subgroup analyses of patients with NYHA class II or greater HF symptoms recorded at baseline, ablation reduced the primary composite endpoint by 32%.¹⁴ The purpose of this report is to provide a more complete description of this subgroup, including a more comprehensive report of outcomes by treatment group.

Methods

Trial Design and Setting

The CABANA trial design and methods have been previously reported in detail.^{14, 17} Trial registration is at ClinicalTrials.gov: NCT0091150. Each site’s Institutional Review Board and/or Ethics Committee approved the CABANA study and written informed consent was obtained from all patients. Since this was an NIH funded trial, the data utilized in our analysis and the material used to conduct the research, and the outcomes will be within the public domain within two years of the initial publication.

Study Population

Patients ≥18 years old with electrocardiographic documentation of at least 2 episodes of paroxysmal AF or 1 episode of persistent AF in the 6 months prior to enrollment, and who were suitable candidates for either catheter ablation or drug therapies were eligible for enrollment.¹⁷ To ensure sufficient event rates to detect a treatment effect, CABANA required patients to be either age ≥65, or age <65 and have at least one risk factor for stroke.¹⁷ For the purpose of the current sub-study, patients were included if they were identified by the enrolling site on the baseline case report form as having symptomatic NYHA ≥ Class II HF.

Outcomes

The primary endpoint in CABANA was a composite of all-cause mortality, disabling stroke, serious bleeding, or cardiac arrest.¹⁷ Secondary endpoints included all-cause mortality alone as well as the composite of all-cause mortality or cardiovascular hospitalization. HF-related mortality was adjudicated by the clinical events committee, and HF-related hospitalizations were designated by the site.

To capture AF recurrence, CABANA employed a proprietary monitoring system to assess follow-up rhythm,¹⁶ but not all countries were able to use that system due to regulatory issues. Of the 778 patients in this sub-study, 330 (42%) used the CABANA system, and the remainder used available local recording devices. The primary analysis of AF recurrence was pre-specified to use the subset of patients with the CABANA monitoring system data. The two main AF recurrence measures were cumulative incidence of AF, estimated from end of blanking period, and AF burden, assessed as the percentage of time spent in AF during the 6-month interval, 96-hour Holter monitor recordings.

Two QOL instruments were used as co-primary endpoints in CABANA: the AF Effect on Quality of Life (AFEQT), and the Mayo AF-Specific Symptom Inventory (MAFSI), as reported previously.¹⁵ The AFEQT is a 21-item instrument designed to assess AF-specific QOL in 3 domains: symptoms, daily activities, and treatment concerns. A summary score is calculated using 18 of the 21 items and ranges from 100 (no AF-related disability) to 0 (complete AF-related disability).¹⁸ The first AFEQT item “are you currently in AF?” is not included in the summary score and is reported separately. On an individual patient level, an AFEQT score change of 5 or more is consistent with a clinically significant change.

The MAFSI is a modification of the Bubien-Kay Symptom Checklist ¹⁹ and in the version used for CABANA includes a 10-item symptom checklist that assesses both frequency and severity of each symptom over the past month.^{19, 20} Responses for the MAFSI frequency portion were collected with a 5-item Likert scale (0=never, 5= always) and summed to generate a summary frequency score that has a theoretical range from 0 (no AF symptoms) to 40 (all 10 symptoms constant). The MAFSI severity score was collected with a 3-item Likert scale (1=mild, 3= severe) and summed to generate a summary score with a theoretical range from 0 (no AF symptoms) to 30 (all 10 symptoms at the most severe level). Patient-level benchmarks for interpretation of changes in MAFSI scales are roughly 1.6 or more points for the frequency scale and 1.3 points for the severity scale.¹⁵ QOL data were collected by structured interview at baseline, 3 and 12 months, and annually thereafter, as described previously.¹⁵

Verification of HF Classification Using Baseline QOL Data

Heart failure remains an inexact phenotypic clinical diagnosis based primarily on expert clinician integration of multiple different types of data.²¹ CABANA did not collect specific biomarker or other clinical or test data relevant to the diagnosis of HF. Left ventricular function imaging was assumed to be part of routine care and was collected as available.

We examined select baseline patient-reported functional status and symptom data relevant to the NYHA functional classification and to the clinical diagnosis of heart failure. From the DASI, we calculated the proportion of subjects who could do each of five activities of progressive workload “with no difficulty”. Using the SF-36 physical function scale, we similarly calculated the proportion of subjects who could do eight activities representing a progressive physical workload “with no limitations”. Finally, using the MAFSI frequency questions, we calculated the proportion of subjects with five different frequency levels (never to always) of “shortness of breath” and “tired/lack of energy.” For these descriptive comparisons, patients were classified as no HF/NYHA class I, NYHA class II and NYHA class III.

Statistical Analysis

Descriptive summary statistics included counts (percentages) for categorical variables, and medians (25^th and 75^th percentiles) for continuous variables. The primary statistical comparisons were performed with treatment assigned as randomized (intention to treat [ITT]).¹⁴ Kaplan-Meier (KM) cumulative event rates were estimated for each treatment group with time-to-event measured (in months) from the time of randomization.²² Treatment effect sizes for most clinical outcomes were expressed as hazard ratios (HRs) with associated 95% confidence intervals (CIs) and were estimated using a covariate adjusted Cox proportional hazards model.²³ The Cox model was constructed as a stratified model (NYHA class II or greater versus all others) using the entire CABANA cohort and was adjusted for the following list of pre-specified baseline patient characteristics: age, sex, race/ethnicity, AF type, years since onset of AF, history of heart failure, structural heart disease, CHA₂DS₂-VASc score, history of coronary artery disease, and hypertension. An interaction term treatment group x HF (defined by NYHA ≥ Class II) was included in the model. Statistical testing of treatment differences was performed with the Wald test from the Cox model.

Recurrent atrial fibrillation incidence rates were estimated using a Fine-Gray model²⁴ adjusted for baseline covariates listed above, with death treated as a competing risk.

The QOL endpoints were analyzed with a repeated-measure mixed-effects model with baseline score and month 3, 12, 24, 36, 48, and 60 responses included as outcomes and time, treatment group, and time x treatment group included as fixed effects.¹⁵ For each follow-up point, we generated point estimates for each treatment group as well as treatment group mean differences (ablation score – drug score). Precision of estimates was assessed with 95% CIs. Since the model does not require either complete data on all patients or a uniform length of follow-up, we did not impute missing values.

P values, where provided, are intended as adjunctive interpretive aids reflecting the level of unexpected observed effects under the null hypothesis.²⁵ No adjustments were made for multiple comparisons. The HF subgroup comparison was a pre-specified secondary analysis in CABANA. However, when that specification was made in the study protocol in 2009, we had no strong a priori reason to suspect that treatment benefits would be substantially larger in HF patients than other patients enrolled in CABANA.

Pre-specified and Post Hoc Sensitivity Analyses

Given the complexities involved in interpreting an ITT analysis of a procedure-based comparison where crossover is possible, we pre-specified two sensitivity analyses for the method of treatment assignment.¹⁴ “As Treated” comparisons were performed using a Cox model with catheter ablation included as a time dependent covariate. “Per Protocol” comparisons were performed in which the drug treatment arm consisted of patients randomized to drug therapy without crossover to ablation. Drug arm patients who crossed over to catheter ablation were censored at the time of the ablation. The ablation treatment arm consisted of patients randomized to ablation who received the procedure within a 6-month window following randomization. Comparisons were adjusted for baseline covariates listed above.

Because baseline ejection fraction was missing in 27% of patients, the statistical method of multiple imputation²⁶ was used to impute the missing values, under the assumption of missingness at randomization. Multiple imputation was carried-out by creating 25 imputed datasets using PROC MI in SAS v9.4 (SAS Institute, Inc., Cary, NC) with the method of fully-conditional specification. A sensitivity analysis was conducted with a covariate-adjusted model including an interaction term between treatment group and the baseline EF, with and without the imputed values.

Results

Baseline Characteristics

Of the 2204 patients randomized in CABANA, 778 had NYHA class II or greater at baseline (Supplement Figure I). Patient characteristics were well balanced between the groups, and overall had a median age of 68, 44% were female and 76% were NYHA class II (Table 1). Paroxysmal AF was present in 31.6%, persistent AF in 55.3%, and longstanding persistent in 13.1%. At enrollment, 75% of patients were taking a beta blocker and 64% were on an angiotensin converting enzyme (ACE) inhibitor or an angiotensin receptor blocker (ARB). A baseline ejection fraction was available for 571 patients (73%) (Table 1, Supplement Figure I). Of these, 79% had an EF ≥50%, 11.7% had an EF between 40% and 49%, and 9.3% had an EF <40%. Comparisons of the baseline characteristics of patients with and without HF in CABANA is provided in Supplement Table I. As shown in Supplement Table II, baseline patient-reported functional status and symptoms of dyspnea and fatigue showed a clear gradient associated with clinician-reported baseline NYHA class such that worse NYHA class was associated with greater reductions in physical functioning and more frequent dyspnea and fatigue.

Table 1.

Baseline Demographics and Clinical Characteristics in CABANA Heart Failure Patients^†

Baseline Characteristics	Ablation Group N=378 No. (%)*	Drug Group N=400 No. (%)*	All Patients N=778 No. (%)*
Baseline Characteristics
Age
Median (Q1, Q3)	68 (62, 73)	67 (62, 73)	68 (62, 73)
< 65 yrs	130/378 (34.4%)	154/400 (38.5%)	284/778 (36.5%)
65 to <75 yrs	184/378 (48.7%)	179/400 (44.8%)	363/778 (46.7%)
≥ 75 yrs	64/378 (16.9%)	67/400 (16.8%)	131/778 (16.8%)
Female Sex	171/378 (45.2%)	174/400 (43.5%)	345/778 (44.3%)
Minoritya: Hispanic or non-White	29/378 (7.7%)	32/400 (8.0%)	61/778 (7.8%)
BMI (kg/m2): Median (Q1, Q3)	31 (27, 35)	31 (27, 36)	31 (27, 35)
CCS Severity of AFb
Class 0	29/378 (7.7%)	22/399 (5.5%)	51/777 (6.6%)
Class 1	37/378 (9.8%)	43/399 (10.8%)	80/777 (10.3%)
Class 2	135/378 (35.7%)	143/399 (35.8%)	278/777 (35.8%)
Class 3	144/378 (38.1%)	159/399 (39.8%)	303/777 (39.0%)
Class 4	33/378 (8.7%)	32/399 (8.0%)	65/777 (8.4%)
NYHA Classificationc
Class II	277/378 (73.3%)	315/400 (78.8%)	592/778 (76.1%)
Class III	99/378 (26.2%)	85/400 (21.3%)	184/778 (23.7%)
Class IV	2/378 (0.5%)	0/400 (0.0%)	2/778 (0.3%)
Medical History
Hypertension (> 140/90 mmHg)	316/378 (83.6%)	349/400 (87.3%)	665/778 (85.5%)
Diabetes (Glucose ≥126 mg/dl)	97/378 (25.7%)	98/400 (24.5%)	195/778 (25.1%)
Prior CVA or TIA	39/378 (10.3%)	40/400 (10.0%)	79/778 (10.2%)
Coronary artery disease	80/378 (21.2%)	90/400 (22.5%)	170/778 (21.9%)
History of CHF	111/378 (29.4%)	118/399 (29.6%)	229/777 (29.5%)
Sleep apnea	72/378 (19.0%)	82/400 (20.5%)	154/778 (19.8%)
Chronic Kidney Disease (eGFR < 60 mL/min/1.73m2)	91/369 (24.7%)	90/386 (23.3%)	181/755 (24.0%)
LVEF, Median (Q1, Q3)	55 (50, 60)	56 (50, 62)	55 (50, 61)
LVEF ≤ 35%	22/285 (7.7%)	23/286 (8.0%)	45/571 (7.9%)
CHA2DS2-VASc Scored
Median (Q1, Q3)	3 (2, 4)	3 (2, 4)	3 (2, 4)
0–1	50/378 (13.2%)	55/400 (13.8%)	105/778 (13.5%)
2	83/378 (22.0%)	90/400 (22.5%)	173/778 (22.2%)
3	110/378 (29.1%)	112/400 (28.0%)	222/778 (28.5%)
4	73/378 (19.3%)	71/400 (17.8%)	144/778 (18.5%)
≥5	62/378 (16.4%)	72/400 (18.0%)	134/778 (17.2%)
Arrhythmia History
Years since onset of AF: Median (Q1, Q3)	1.1 (0.2, 3.7)	1.2 (0.3, 4.2)	1.1 (0.3, 4.1)
Type of AF at enrollment
Paroxysmal	110/378 (29.1%)	136/400 (34.0%)	246/778 (31.6%)
Persistent	221/378 (58.5%)	209/400 (52.3%)	430/778 (55.3%)
Longstanding Persistent	47/378 (12.4%)	55/400 (13.8%)	102/778 (13.1%)
Prior hospitalization for AF	170/378 (45.0%)	186/400 (46.5%)	356/778 (45.8%)
Prior direct current cardioversion for AF	135/378 (35.7%)	160/400 (40.0%)	295/778 (37.9%)
History of AFL	38/371 (10.2%)	49/393 (12.5%)	87/764 (11.4%)
Prior ablation for AFL	11/377 (2.9%)	22/398 (5.5%)	33/775 (4.3%)
Rhythm Control Therapye
0 Rhythm control drugs	200/354 (56.5%)	191/379 (50.4%)	391/733 (53.3%)
1 Rhythm control drug	127/354 (35.9%)	150/379 (39.6%)	277/733 (37.8%)
≥ 2 Rhythm control drugs	27/354 (7.6%)	38/379 (10.0%)	65/733 (8.9%)

^*unless otherwise noted

^†No baseline demographics or clinical characteristics demonstrated a statistically significant difference between the treatment groups presented in this table.

Q1 and Q3=quartiles (25th and 75th percentiles), BMI=body mass index, CCS=Canadian Cardiovascular Society, AF=atrial fibrillation, NYHA=New York Heart Association, CVA=cerebral vascular accident, TIA=transient ischemic attack, CHF=congestive heart failure, eGFR=estimated glomerular filtration rate calculated using CKD-EPI creatinine equation, LVEF=left ventricular ejection fraction, AFL=atrial flutter

^aRace/Minority was determined by the site investigator in conjunction with the patient based on predefined categories as required by the National Institute of Health (NIH) using NIH-specified categories.

^bOn a scale of 0 to 4 in which 0 is the least severe and 4 is the most severe symptom of AF

^cOn a scale of I to IV in which I is the least severe and IV is the most severe symptom of heart failure

^dOn a scale of 0 to 9 in which 0 is the lowest risk of stroke and 9 is the highest risk of stroke

^eCurrent or past use of Rhythm Control Therapy reported at the time of enrollment

Treatment Data

In the ablation group, 344 (91.0%) HF patients underwent ablation at a median of 24 days following randomization, while 34 (9.0%) patients did not receive ablation. Among the catheter ablation patients with HF and post-blanking follow-up, 155/330 (47%) were on a rhythm control drug at the end of the blanking period (Supplement Table III) and 76/325 (23%) were on a rhythm control drug at the latest of one or more follow-up contacts.

In the drug therapy alone group, 89 (22.3%) received an ablation procedure at a median of 351 days following randomization (25^th percentile 162, 75^th percentile 725). At the end of the blanking period, 307/383 (80%) were on a rhythm control drug (Supplement Table III) and 188/376 (50%) were receiving one of these drugs at the latest of one or more follow-up contacts.

The most common treatment-related adverse events in the ablation arm included hematoma (3.2%), pseudo aneurysm (1.2%), esophageal ulcer (1.2%), and severe pericardial chest pain (0.6%). The most common treatment-related adverse events in the drug therapy arm included hyper- or hypothyroidism (2.5%), gastrointestinal abnormality excluding moderate/severe diarrhea (1.3%), major proarrhythmic event (0.8%), and liver injury/failure (0.5%).

Clinical Outcome Comparisons by Intention-to-Treat

The CABANA primary outcome event (death, disabling stroke, serious bleeding, or cardiac arrest) occurred in 34/378 (9.0%) HF patients in the catheter ablation group and in 49/400 (12.3%) HF patients in drug therapy (HR for ablation vs drug therapy 0.64; 95% CI, 0.41 to 0.99) (Figure 1). Death from any cause occurred in 23/378 (6.1%) HF patients in the ablation arm and 37/400 (9.3%) HF patients in the drug therapy arm (HR 0.57; 95% CI, 0.33 to 0.96) (Figure 2). Death from cardiovascular causes occurred in 12/378 (3.2%) patients in the ablation arm and 14/400 (3.5%) patients in the drug therapy arm (HR 0.70; 95% CI, 0.31 to 1.57). Deaths due to HF occurred in 6 patients in the ablation arm and 4 patients in the drug therapy arm. Death or CV hospitalization occurred in 212/378 (56.1%) patients in the ablation arm and 245/400 (61.3%) patients in the drug therapy arm (HR 0.84; 95% CI, 0.70 to 1.02). HF hospitalization occurred in 34/378 (9.0%) patients in the ablation arm and 37/400 (9.3%) patients in the drug therapy arm (HR 0.89; 95% CI, 0.56 to 1.44).

Figure 1:

Primary Composite End Point (Death, Disabling Stroke, Serious Bleeding, or Cardiac Arrest) Kaplan-Meier Curves by Intention-to-Treat Among CABANA Heart Failure Patients

Figure 2:

All-Cause Mortality Kaplan-Meier Curves by Intention-to-Treat Among CABANA Heart Failure Patients

Treatment Assignment Sensitivity Analyses

In a pre-specified “as treated” analysis, the ablation arm showed a 42% reduction in the primary composite endpoint (HR 0.58; 95% CI, 0.37 to 0.90) in HF patients. Reductions were also seen in all-cause mortality (HR 0.50; 95% CI, 0.30 to 0.85); the composite of death or CV hospitalization (HR 0.84; 95% CI, 0.70 to 1.01); and the composite of death or HF hospitalization (HR 0.59; 95% CI, 0.41 to 0.87). No deaths occurred within the first 30 days after initiation of either therapy. One disabling stroke occurred within the first 30 days of treatment after initiation of drug therapy.

In a per-protocol analysis, patients in the ablation arm who received catheter ablation within 6 months showed reduction in the primary composite endpoint (HR 0.60; 95% CI, 0.38 to 0.94) and all-cause mortality (HR 0.52; 95% CI, 0.31 to 0.90).

Subgroup Analyses

Analyses of pre-specified subgroups in the HF cohort by ITT using the primary composite endpoint were consistent with the overall CABANA trial (Figure 3).

Figure 3:

Forest Plot of Pre-specified Subgroup Comparisons in CABANA Heart Failure Patients

After employing multiple imputation to impute missing baseline EF values, 9.8% had an EF <40%, 15.6% had an EF 40–49%, and 74.6% had an EF ≥50%. In a post hoc analysis, ablation reduced mortality by 60% relative to drug therapy in the patients with EF ≥50% (HR 0.40 (95% CI 0.18 to 0.88) with 4-year KM mortality rates of 3.3% vs. 8.6% (Supplement Table IVa and IVb, Supplement Figure II). Analysis with complete EF data, showed (HR 0.51, 95% CI 0.23 to 1.12) and 4-year KM mortality rates (4.2% vs. 8.3%). In EF 40% to 49% which included patients with imputed EFs, the HR for the ablation effect on mortality was 0.43 (95% CI 0.09 to 2.13). Not enough patients were in this subgroup excluding the imputed EF patients or in the subgroup with EF<40% (with or without imputation) to reliably estimate a treatment effect on mortality (Supplement Table IV).

AF Recurrence

Of the 778 HF patients, 330 (42.4%) used the CABANA recording system to detect AF recurrence following the blanking period. By 12 months, 37% of the HF ablation arm patients and 58% of the HF drug therapy arm patients recorded a recurrence of any AF (Figure 4). At 5 years, the corresponding values were 56% (ablation arm) and 72% (drug arm). Overall, the ablation arm had a 44% relative reduction in first AF recurrence when compared to the drug arm (HR 0.56; 95% CI, 0.42 to 0.74). AF burden at baseline showed an average of 57.8% of the CABANA Holter recording time was spent in AF. At 12 months, AF burden averaged 7% in the ablation arm and 18% in the drug therapy arm. At 5 years, the corresponding percentages were 17% and 26%, respectively. At all follow-up time points, the AF burden was lower in the ablation arm relative to the drug therapy arm (Figure 5). Ablation also had lower burden regardless of AF type recorded at baseline (Supplement Figure IIIa and IIIb).

Figure 4:

Cumulative Incidence Curves of First Recurrence of AF in the Post Blanking Period Among CABANA Heart Failure Patients Who Used the CABANA ECG Recording System

Figure 5:

Atrial Fibrillation Burden by Time and Randomization Assignment Among CABANA Heart Failure Patients Who Used the CABANA ECG Recording System

AF-Related Quality of Life Outcomes

Mean AFEQT summary scores were equivalent at baseline in the two treatment arms (median 57 ablation arm, 56 drug therapy arm) and higher (more favorable) at each follow-up assessment out to 60 months in the ablation arm (Table 2). The adjusted mean difference at 12 months was 5.7 points favoring the ablation arm (95% CI, 2.8 to 8.7 points). The adjusted mean difference averaged over the entire 60-month follow-up was 5.0 points favoring the ablation arm (95% CI, 2.5 to 7.4 points) (Table 2, Supplement Figure IVa and IVb).

Table 2:

Quality of Life Outcomes in CABANA Heart Failure Patients

	Catheter Ablation (N=378)			Drug Therapy (N=400)
Timepoint	Median [q1 q3]	Mean (SD)	N	Median [q1 q3]	Mean (SD)	N	Adjusted Difference (95% CI)
AFEQT Summary Score
Scale: 0 = complete disability, 100 = no disability
Baseline	57 (44, 73)	57.6 (19.9)	371	56 (44, 73)	57.7 (19.9)	394	-0.1 (−2.9 to 2.7)
3 month	79 (63, 91)	75.7 (18.9)	329	72 (55, 90)	70.6 (21.5)	356	4.9 (1.9 to 7.9)*
12 month	86 (68, 97)	80.6 (19.8)	310	80 (59, 93)	75.0 (19.6)	317	5.7 (2.8 to 8.7)*
24 month	86 (70, 97)	80.9 (18.8)	285	78 (60, 92)	74.4 (21.1)	280	5.9 (2.8 to 9.0)*
36 month	87 (70, 96)	81.8 (17.7)	204	82 (67, 94)	77.8 (20.1)	203	4.0 (0.6 to 7.3)
48 month	85 (68, 97)	80.7 (18.5)	144	82 (61, 93)	76.1 (20.1)	146	4.6 (0.9 to 8.4)
60 month	86 (69, 96)	80.5 (18.9)	106	81 (58, 94)	75.0 (22.3)	101	4.5 (−0.3 to 9.3)
All follow-up	85 (68, 96)	79.7 (19.0)	1378	78 (59, 93)	74.3 (20.8)	1403	5.0 (2.5 to 7.4)
MAFSI Frequency Score
Scale: 0 = never symptoms, 40 = always symptoms
Baseline	13 (9, 17)	13.3 (5.9)	371	13 (8, 18)	13.3 (6.4)	389	-0.0 (−0.9 to 0.9)
3 month	8 (4,12)	8.5 (6.2)	300	10 (5, 15)	10.7 (6.9)	328	-2.1 (−3.1 to −1.1)*
12 month	7 (2,12)	8.1 (6.7)	291	10 (5, 14)	9.9 (6.5)	295	-1.9 (−3.0 to −0.9)*
24 month	7 (2, 12)	7.9 (6.5)	271	10 (5, 15)	10.5 (7.1)	245	-2.2 (−3.3 to −1.1)*
36 month	8 (3, 13)	8.2 (6.3)	195	9 (4, 15)	10.0 (6.9)	189	-1.9 (−3.1 to −0.7)
48 month	7 (2, 12)	7.8 (6.6)	138	9 (5, 13)	9.8 (7.0)	133	-1.7 (−3.1 to −0.3)
60 month	6 (2, 12)	7.8 (6.7)	102	9 (6, 15)	10.2 (6.8)	93	-2.3 (−4.0 to −0.7)
All follow-up	7 (3, 12)	8.1 (6.4)	1297	10 (5, 15)	10.2 (6.8)	1283	-2.0 (−2.9 to −1.2)*
MAFSI Severity Score
Scale: 0 = mild symptoms, 30 = extreme symptoms
Baseline	10 (7, 13)	10.5 (4.7)	372	10 (7, 14)	10.5 (5.1)	385	-0.0 (−0.7 to 0.7)
3 month	6 (3, 11)	6.8 (4.9)	297	8 (4, 13)	8.6 (5.6)	328	-1.7 (−2.5 to −0.9)*
12 month	5 (2, 10)	6.5 (5.4)	291	8 (4, 12)	8.0 (5.2)	295	-1.5 (−2.3 to −0.7)*
24 month	5 (2, 9)	6.3 (5.2)	270	7 (4, 12)	8.1 (5.5)	243	-1.5 (−2.3 to −0.6)*
36 month	6 (2, 10)	6.6 (5.1)	193	8 (3, 12)	7.9 (5.7)	189	-1.5 (−2.5 to −0.5)
48 month	6 (2, 10)	6.2 (5.1)	137	7 (3, 10)	7.3 (5.3)	133	-0.9 (−2.0 to 0.2)
60 month	5 (2, 9)	6,5 (5.7)	102	7 (4, 11)	7.9 (5.3)	93	-1.3 (−2.7 to 0.1)
All follow-up	6 (2, 10)	6.5 (5.2)	1290	7 (4, 12)	8.1 (5.5)	1281	-1.4 (−2.1 to −0.7)*

^*p values <0.001

AFEQT Summary Score interpretive guidance: <70=severely symptomatic, 70–89=mildly to moderately symptomatic, ≥90=minimally symptomatic or asymptomatic; clinically important change 5.0

MAFSI Frequency Score interpretive guidance: >9=severely symptomatic, 4–9=mildly to moderately symptomatic, <4=minimally symptomatic or asymptomatic; clinically important change 1.6

MAFSI Severity Score interpretive guidance; clinically important change 1.3

Mean MAFSI Frequency scores were equivalent at baseline (median 13 in each arm) (Table 2). At 12 months MAFSI scores were more favorable (lower) in the ablation arm (Figure 6a) (adjusted mean treatment group difference −1.9; 95% CI, −3.0 to −0.9). Over the 60 months of follow-up, the average adjusted MAFSI score difference was −2.0 points favoring ablation (95% CI, −2.9 to −1.2) (Figure 6b) (Table 2). A similar pattern was seen for the MAFSI Severity score (Table 2).

Figure 6:

Quality of Life Outcomes in CABANA Heart Failure Patients by MAFSI Scoring

Discussion

Catheter ablation in CABANA trial patients with AF and ≥ Class II HF produced clinically consequential reductions in all-cause mortality (43% relative reduction, 3.1 per 100 absolute reduction at 5 years) as well as a lower AF recurrence rates (44% reduction in time to first recurrence) and AF burden. Ablation patients also demonstrated substantial and sustained improvements in quality of life out to 5 years. While these treatment benefits appear plausible, they clearly need to be confirmed with an adequately sized clinical trial of ablation in HF subjects.^27–31

CABANA is the first large randomized trial to describe an important mortality benefit from AF treatment in HF subjects who predominately have preserved systolic function. One of the main objectives of the CABANA trial was to test whether effectively treating the AF state could reduce the excess mortality risk associated with AF.¹⁷ CABANA was originally powered to detect a 30% relative mortality reduction. While the effect of ablation on all-cause mortality relative to drug therapy was indeterminate in the overall 2204 patient CABANA cohort (HR 0.85, 95% CI 0.60 to 1.21), initial subgroup analyses (of the trial primary composite outcome measure) suggested that the treatment effect in the subset of 778 patients with HF (NYHA class II or III) was substantially larger.¹⁴ The present paper, part of the pre-specified CABANA research program, provides substantial new information on the HF subgroup treatment effects.

Atrial Fibrillation and Risk of Mortality

The morbidity of AF has been recognized for many decades but the recognition that the development of and presence of AF is also associated with a higher mortality risk is more recent.^32–35 Some of this excess mortality risk has been attributed to associated heart diseases leading to the AF, such as valvular or coronary disease, but AF appears to have important adverse effects on survival even in the absence of such associations and for reasons other than inadequate anticoagulation leading to major stroke.^{36, 37}

The Women’s Health Study, involving almost 35,000 subjects initially free of CV disease who were followed for a median of 15 years, found that new AF was associated with a 2-fold increase in adjusted risk for all-cause death and a 4-fold increase in adjusted risk for cardiovascular death.³⁸ Accounting for non-fatal events (myocardial infarction, stroke, HF) modestly reduced the size of these risks but did not eliminate the associations. The question of whether AF causes some or all of this excess mortality or is simply a marker for factors that are the true causes remains unsettled. Clinical trial evidence supporting the ability of ablation to reduce mortality relative to drug therapy is needed to resolve the issue. CABANA adds important new evidence in this area, pointing particularly to the need for a larger confirmatory trial specifically in the HF population.

Prior Clinical Trial Evidence of Effects of Ablation in Heart Failure

A number of relatively small randomized trials have been published comparing catheter ablation with drug therapy in AF patients with HF.^3–11 Most were limited to subjects with systolic dysfunction, with predominately NYHA class II or III symptoms and persistent AF. Only three, CASTLE-HF, AATAC and AMICA,^{4, 7, 9} randomized more than 100 patients. Summary effect sizes included about a 50% reduction in all-cause mortality, a 40% reduction in HF hospitalizations and a 9-point improvement in Minnesota Living with HF scores. Impact on BNP, 6-minute walk, and VO2 have been variable.^{4–6, 8–10} AF recurrence was decreased by 34% with ablation. These trials also suggested that ablation improved EF by an absolute 7%, although no difference with drug therapy was seen in the AMICA trial.⁷ No evidence was found for an increase in serious adverse effects with ablation.

Effects of Ablation in Heart Failure with Preserved EF

No randomized controlled trials have tested ablation versus drug therapy in HF patients with preserved EF. The available evidence consists of a few, mostly small, observational reports of ablation outcomes in this population.^{13, 39, 40} A retrospective single center cohort study of 230 HF patients (97 with reduced EF, 133 with preserved EF, 62%−63% in both subgroups with non-paroxysmal AF) found effectiveness of ablation in producing freedom from recurrent AF and improvement in QOL did not vary as a function of baseline ventricular function.¹² Analysis of a large US claims database of 289,366 patients with AF and HF (identified by ICD codes) reported that of the 7465 who received ablation, 57% had preserved ejection fraction.⁴¹ Comparison of ablation versus drug therapy in HF patients with versus without reduced EF showed no difference in treatment effect size. In CABANA, 79% of HF patients with EF measured at baseline had values ≥ 50%.

AF Recurrence in Heart Failure

Freedom from recurrent AF in the HF subgroup was very similar to that previously reported in the overall CABANA trial both when assessed as time to first recurrence (HR 0.56 for HF subgroup, 0.52 for overall trial) and as AF burden (at 12 months 7% in ablation patients in HF subgroup versus 6% in overall trial, 18% in drug arm of HF subgroup versus 14% in overall trial). Thus, in the context of the HF patients selected for CABANA with predominantly preserved LV function, and the subset of those who had access to the CABANA ECG recording system, HF was not associated with inferior technical results with ablation. Ablation had superior results at 12 months in patients with both paroxysmal and with persistent/longstanding persistent AF. More attenuation of treatment differences appears present in the latter subgroup (Supplement Figures IIIa and IIIb) but samples are small and precision relatively low.

Quality of Life in Heart Failure

AFEQT Summary scores, reflecting the impact of AF on symptoms, activities and treatment concerns showed greater impairments in HF patients relative to the overall trial cohort (baseline median scores 56 to 57 points in HF patients versus 63 in overall CABANA cohort).¹⁵ As reported previously, both treatment groups in CABANA demonstrated a substantial improvement over the first 12 months of study participation after which QOL values at the cohort level showed little change. In addition to these changes, mean treatment group differences were larger in favor of ablation at all assessment points out to 60 months. Similar patterns were seen for the MAFSI frequency and severity scores. These findings complement the clinical outcome data by providing a patient perspective on the treatment effects studied in CABANA. The results reported in the present paper provide new evidence of a substantial, clinically important incremental benefit from ablation on quality of life in a cohort of HF patients with predominantly preserved left ventricular function.

Adverse Effects of Treatment

Catheter ablation has the potential to harm subjects either by major procedural complications that occur in temporal proximity to the procedure or later due to adverse cardiac remodeling. Early complications can include perforation by catheters or excessive ablative energy-related injury to adjacent tissues. Such complications were infrequent in CABANA, with no procedure-related deaths recorded.¹⁴ Further, in the present sub-study we found no evidence that HF patients were at increased risk for these procedure-related events.

Catheter ablation, as well as surgical MAZE procedures, work by creating scars that block pulmonary vein triggers of AF and in some cases interrupt the reentrant circuits that allow AF to perpetuate. While the elimination of AF usually improves cardiac performance and clinical outcomes, in theory aggressive or repeated catheter ablation procedures could provoke deterioration in a vulnerable subset by increasing the total amount of atrial fibrosis thereby worsening left atrial (LA) compliance.³⁹ The “stiff left atrial syndrome” complication, marked by symptomatic pulmonary hypertension, has been reported as a rare complication after catheter ablation treatment of AF,^{42, 43} and extensive scarring of the LA has been one proposed mechanism to explain the phenomenon.⁴⁴ The need for multiple repeat ablation procedures is believed to be a risk factor, presumably due to cumulative fibrosis burden, although the diagnosis is still largely one of exclusion. No data yet exist to indicate that HF patients are at higher risk for this complication.

Limitations

Our results should be considered in light of several important limitations. First, for the purpose of this study, HF was defined phenotypically by the enrolling clinicians (NYHA class II or III symptoms), and we did not require confirmatory diagnostic testing. Without measurement of left ventricular filling pressures, however, the diagnosis of heart failure still depends largely on clinical judgment. The core clinical features are effort intolerance and symptoms of dyspnea and fatigue with activity. As shown in Supplement Table II, in our study cohort the NYHA classification was quite concordant with levels of patient-reported physical functional impairments and symptoms of dyspnea and fatigue. Second, we did not require baseline echocardiography to define ejection fraction in all subjects. These data were available in ~75% of patients. Using the subset of data with complete EF as well as the full HF cohort with imputed EF when missing did not suggest any clear variation in all-cause mortality benefit from ablation according to status of left ventricular function. These results are concordant with prior observational studies that suggest similar relative clinical benefits in phenotypical HF regardless of baseline EF. Third, in patients who had a baseline EF measured, only 9% had value less than 40%. Supporting the classification of most of these HF patients as having heart failure with preserved left ventricular function, hypertension and/or left ventricular hypertrophy was present at baseline in 92%. Thus, our estimates of the treatment effects of ablation in AF patients with heart failure address a different part of the heart failure spectrum from earlier randomized ablation trials, such as CASTLE-AF. Finally, ablation had a quantitatively greater effect on overall mortality than on cardiovascular mortality or HF hospital admissions. Too few heart failure deaths occurred in CABANA to assess a treatment effect. Replication in a larger sample is required to determine the clinical significance of these patterns.

Conclusions

In CABANA AF patients with clinically defined HF, ablation provided clinically important reductions in mortality, recurrent AF, and improved QOL relative to drug therapy. These results, in a cohort with predominantly preserved LV function, complement and extend recent trial results showing survival and QOL benefits of ablation in HF subjects with reduced LV function. If these findings can be confirmed in adequately sized replication trials, clinicians would have a powerful new strategy for reducing the patient suffering and premature mortality that result when AF and HF occur together.

Clinical perspectives:

What is new?

• This study provides new randomized trial information regarding the benefits of catheter ablation in atrial fibrillation patients who have the clinical phenotype of heart failure.

• Specifically, we found that in the CABANA Trial there were substantial clinical outcome benefits with ablation over drug therapy in patients with New York Heart Class II or III at trial entry, most of whom did not have a reduced ejection fraction.

• Benefits were evident for both all-cause mortality and AF reduction.

• However, the effects on heart failure hospitalizations were small and not significant.

What are the clinical implications?

• This study suggests that catheter ablation may provide important prognostic and symptomatic benefits relative to drug therapy in patients with symptomatic AF, where symptoms and functional impairments are due to the combined effects of AF and heart failure.

• These results should not be viewed as practice changing until they are reproduced in a confirmatory trial of ablation in the same population.

Abbreviations:

AAD

antiarrhythmic drug

AF

atrial fibrillation

AFL

atrial flutter

AT

atrial tachycardia

CI

confidence interval

HF

heart failure

HR

hazard ratio

IR

interquartile

NYHA

New York Heart Association