Left atrial ablation for management of atrial fibrillation: CABANA vs. real-world data. Apples and oranges?

This editorial refers to ‘Atrial fibrillation ablation in practice: assessing CABANA generalizability’†, by P.A. Noseworthy et al., on page 1257. Almost every clinical cardiac electrophysiologist, the doctors who specialize in the management of arrhythmias such as atrial fibrillation, is convinced that it is better for the patient to be in sinus rhythm than to exist with atrial fibrillation managed by ventricular rate control and anticoagulation. No doubt every patient who suffers with atrial fibrillation would prefer to be in sinus rhythm, as would their doctors. However, randomized trials that have been designed to explore whether patients do better in sinus rhythm than in atrial fibrillation have not supported the conviction of the doctors and their patients.1 , 2 Searching for an explanation for these counter-intuitive results, investigators identified the potential adverse effects of antiarrhythmic agents, particularly proarrhythmia and negative inotropic properties, as the most likely reasons. However, it has now been documented that left atrial ablation is superior to antiarrhythmic drugs at maintaining sinus rhythm,3 although it took some time to be certain that this also applied to patients that were naïve to, and therefore not, ‘by definition’, refractory to antiarrhythmic drug treatment.4 , 5 It took much longer to show that ablation therapy might lead to a reduction of irreversible outcomes such as death, stroke, and myocardial infarction, something that had never been convincingly demonstrated with regard to antiarrhythmic drug treatment, although the ATHENA trial with dronedarone had resulted in a significant reduction of its primary composite endpoint of cardiovascular hospitalizations or all-cause mortality,6 and subgroup and post-hoc exploratory analyses also documented a possible reduction of myocardial infarction and stroke.7 , 8 Lately these findings were supported by an analysis of real-world evidence from the IQVIA database in Germany.9 Some previous studies where ablation proved superior to antiarrhythmic therapy in suppressing atrial fibrillation had also demonstrated that secondary outcome events, including mortality, might be reduced by left atrial ablation.10 However, it is only recently that more convincing, but not incontrovertible, evidence has emerged from randomized controlled trials suggesting that, in comparison with ‘placebo’ or any other conventional approach, left atrial ablation reduced major outcome events. In CASTLE-AF (Catheter Ablation versus Standard conventional Therapy in patients with LEft ventricular dysfunction and Atrial Fibrillation), left atrial ablation led to a reduction of a composite endpoint of cardiovascular hospitalizations or all-cause mortality in symptomatic patients with heart failure and a reduced ejection fraction.11 A secondary endpoint of all-cause mortality alone was also significantly reduced by the ablation procedure. However, the trial was criticized because of its prolonged recruitment period, its highly selected population, and the relatively large number of patients who were lost to the trial between randomization and the end of the study. The Catheter ABlation vs ANtiarrhythmic drug therapy in atrial Fibrillation (CABANA) trial is published coincident with this publication, but some data are already available from presentations made by the investigators at scientific meetings.12 A total of 2204 patients with atrial fibrillation associated with cardiovascular risk were recruited and randomized 1:1 to left atrial ablation (1108) or to a clinical alternative (1096) such as rhythm control with antiarrhythmic drugs (the majority) or rate control. Patients were anticoagulated over the long term according to clinical practice standards and were followed for at least 48 months. The results were not straightforward. Although the intention-to-treat analysis showed a numerical reduction of the primary outcome (all-cause mortality, stroke, major bleeding, or cardiac arrest), the absolute rate reduction was small (1.2%) and not statistically significant. There were 301 (27.5%) crossovers from those assigned to drug therapy to the ablation treatment arm and, for a variety of reasons, 102 (9.2%) patients randomized to atrial ablation were not ablated. Remembering the comment from C. Everett Koop, a past US Surgeon General, that ‘drugs don’t work in patients who don’t take them’, they reasoned that there should be a per protocol analysis, or better still an ‘as-treated’ approach. The former involved censoring crossed-over patients so that only patients who stuck to their randomized treatment would be counted in the analysis, and the latter compared all those who received ablation (1307 patients) with all of those who did not (897 patients). Of course, both analyses, especially the ‘as-treated’ approach, disrupted the value of the randomization, but both showed statistically significant reductions of the primary endpoint associated with ablation. Technically the trial ‘failed’, but the investigators and many commentators prefer to accept the per protocl or as treated analyses. The study reported in this issue of the European Heart Journal was conceived, planned in detail, and undertaken before the results of CABANA were known.13–15 It was not intended to bolster the results of the trial, but to test the generalizability of its results, whatever they were. It is a study based on ‘real-world’ data retrieved from a very large US database. It explores the outcomes of patients with atrial fibrillation who would have qualified as subjects in the CABANA trial and underwent an ablation procedure to manage their arrhythmia in clinical practice and not in a randomized trial. Their outcomes are compared with those of other patients with atrial fibrillation who were not treated with ablation. Of course, since the participants were not randomized to these treatment groups, their baseline characteristics were very different (Table 1). Using an impeccable statistical approach, it was possible to compensate for many differences, but as usual hidden confounders could not be addressed. Nevertheless, to be relevant, these confounders must have had a major effect given the statistical strength and effect size of the differences between groups. The results suggested that the better outcome of the ablation group compared with the non-ablation group were very similar to the improved outcome of the ‘as-treated’ ablation group compared with the drug therapy group in the CABANA trial. Thus, the hazard ratio (HR) for the primary outcome was 0.67 [95% confidence interval (CI) 0.50–0.89, P = 0.006] in favour of ablation in CABANA and 0.70 (95% CI 0.63–0.77, P <0.001) in the real-world study, and the HRs for all-cause mortality were 0.60 (95% CI 0.42–0.86, P = 0.005) and 0.67 (95% CI 0.61–0.74, P <0.001), respectively (Take home figure). Several important sensitivity analyses were made, and the results were not unduly affected. Therefore, the authors conclude that the results of CABANA are generalizable, as least as far as the US population is concerned. Further, they propose that the inclusion/exclusion criteria for entry into the CABANA trial might be used in future guidelines to select patients for left atrial ablation to improve outcomes. Drug and ablation event rates and hazard ratios for all-cause mortality and the composite endpoint of all-cause mortality, stroke, major bleeding, and cardiac arrest in the observational study of CABANA-eligible patients and in both the intention-to-treat (ITT) and as-treated (AT) analysis of CABANA outcomes. Some baseline characteristicss of patients in the observational study of Noseworthy et al.13 and the CABANA trial (taken from slide presentations) Please note that the definitions of heart failure and stroke may differ. The raw data show substantial differences between the drug and ablation groups in the observational study. There are also marked differences between the characteristics of the weighted ablation group from the observational study and the patients randomized to the ablation group in CABANA. Prior cerebrovascular accident or tranisent ischaemic attack in CABANA, and prior ischaemic cerebrovascular accident for the observational study. Some baseline characteristicss of patients in the observational study of Noseworthy et al.13 and the CABANA trial (taken from slide presentations) Please note that the definitions of heart failure and stroke may differ. The raw data show substantial differences between the drug and ablation groups in the observational study. There are also marked differences between the characteristics of the weighted ablation group from the observational study and the patients randomized to the ablation group in CABANA. Prior cerebrovascular accident or tranisent ischaemic attack in CABANA, and prior ischaemic cerebrovascular accident for the observational study. There is an unwritten implication that the results of the present study reinforce the superiority of left atrial ablation over other treatments as demonstrated in the as-treated analysis in CABANA, and to some extent legitimizes the as-treated analysis of the outcomes. This conclusion should be rejected, although in any guideline process the results of well-conducted observational studies can certainly add strength to any recommendations. The main comparison, which was very carefully made in this real-world study, was between CABANA-eligible patients treated with ablation and similar patients not treated with ablation whose data were drawn from an American healthcare database. We may be tempted to accept the similarity of the outcome hazard ratios in CABANA and the real-world study as evidence that the results concur, and therefore that both must be ‘true’. However, we must also examine the similarity between the characteristics of the CABANA and observational study populations, and we must look at the outcome event rates (Take home figure). Judged from what is known, the CABANA study ablation patients were younger on average but had a higher portion of patients aged ≥75 years, fewer were female, but more had heart failure (the definition of heart failure may of course differ between these studies) than the weighted population of eligible CABANA patients in the observational study. The adjusted mortality and the composite of primary endpoint event rates were lower in the observational study than in either the intention-to-treat or the as-treated population in CABANA (Take home figure). This does not detract from the main conclusion of the authors that the CABANA results may be generalizable, but reduces our temptation to assume that the study necessarily validates the CABANA as-treated outcome analysis. Apples and oranges are both edible fruit, but they certainly do not taste the same. Conflict of interest: A.J.C. has received institutional research grants from Daiichi Sankyo and Boehringer Ingelheim; and personal fees from Acesion, Huya, Incarda, Merck, Menarini, Milestone, Sanofi, Servier, Bayer, Boehringer Ingelheim, Bristol-Myers Squibb, Daiichi Sankyo, Pfizer, Portola, Boston Scientific, Abbott, Biotronik, and Medtronic.