GRAND DEBATE | Is catheter ablation superior to drugs in the treatment of patients with AF and HF?
In the second part of the Grand Debate on Drugs or interventions - The beat goes on, the value of catheter ablation in comparison to drug for treatment of patients with atrial fibrillation (AF) and heart failure (HF) was discussed. The session was chaired by Eugene Braunwald (Boston, MA, US) and Frank Ruschitzka (Zurich, CH), the latter of whom gave a brief introduction to the topic.
In the case of AF, managing patients is pretty straightforward. Management of patients with both HF and AF is a bit more complicated, as some, but not all HF drugs, work in AF patients. The results of several ongoing outcome trials in this patient group with both conditions are eagerly awaited. The 2016 ESC HF Guidelines state that ‘AF ablation may be considered’, a recommendation which, according to Ruschitzka, is close to ‘we don’t know’. The 2016 ESC AF Guidelines included patient preference in the treatment algorithm. The results of the CASTLE-AF and CABANA trials likely provide new insights on the best management of patients with AF and HF.
Whenever treating patients, it should be noted that HF is not one disease; multiple different processes can play a role in the syndrome, such as dilative, hypertrophic or restrictive cardiomyopathy, non-compaction, myocarditis, right ventricular dysfunction, mitral regurgitation, left atrial disease, or amyloidosis. In the combination of HF and AF, much goes wrong in the left heart: both conditions contribute to atrial remodeling, which leads to hypertrophy, fibrosis, apoptosis and necrosis. This can ultimately result in left atrium (LA) dilation, LA pressures and LA hypocontractility.
In the discussion on the best therapy for patients with both AF and HF, Nassir Marrouche (Salt Lake City, UT, US) defended the PRO position on the statement that catheter ablation is superior to drugs. The first data he showed in line with this position was a graph that showed no effect of the antiarrhythmic drug dofetilide on mortality in patients with AF and HF. A study in these type of patients that does not get a lot of attention is the rhythm vs. rate control study (Roy et al., NEJM 2008), in which no difference was seen on survival between the two approaches. Most patients in this study were in sinus rhythm, but this did not prolong their life.
Studies have, however, shown that ablation is effective and superior. The PABA-CHF trial compared pulmonary vein isolation (PVI) with AV node ablation (AVNA) and biventricular ICD. PVI was superior to AVNA in improving ejection fraction, 6-minute walking distance and quality of life. The CAMTAF trial (2014) showed a better effect of catheter ablation as compared with medical treatment of AF in HF patients on LV function and Minnesota Living with HF Score. The AATAC compared ablation with amiodarone (De Biase, 2016) in patients with persistent AF and congestive HF and an implanted device. A larger proportion of patients in the catheter ablation group was free of arrhythmia than in the amiodarone-treated patients. Unplanned hospitalization and death were more frequent in the amiodarone groups. The CAMERA-MRI (Prabhu, 2017) showed a greater improvement of ejection fraction (MRI) with catheter ablation than with medical rate control, and left atrial volume was reduced, and LV stroke volume increased. Patients who underwent ablation also showed a significant decline in NYHA class as compared with medical rate control, as well as a decrease of BNP. Recently, subanalyses of the CABANA trial (Packer, 2018) in AF patients with a history of congestive HF or with NYHA class II and higher at baseline, suggested that catheter ablation is effective in reducing the risk of the primary endpoint of all-cause mortality, disabling stroke, serious bleeding and cardiac arrest. Hence, Marrouche concluded; study after study shows that ablation improves not only ejection fraction, but also AF burden and quality of life, as well as the hospitalization rate and mortality. Or in short: it works.
Milton Packer (Dallas, TX, US) was invited to defend the CONTRA position. He also presented the CABANA trial results, presented recently at the Heart Rhythm Association Congress in Boston, but now to prove the opposite point. He emphasized that the prespecified criterion of success was a benefit on the primary endpoint by intention-to-treat (ITT) analysis, based on the original assumption of 25-30% crossover. This primary endpoint was not met, as the P-value was 0.3. The original endpoint was all-cause mortality, under the same ITT assumptions, which showed a P-value of 0.377: ‘not even close to a success’, according to Packer. Ablation may change symptoms, but it does not change the course of disease.
He was critical about the electrophysiology field, blaming them for presenting on-treatment analyses, rather than ITT. In posthoc ‘on treatment’ analyses, the trial was converted to an observational study, which meant that if a patient was randomized to drug therapy and responded favorably for many years but received ablation towards the end of the trial, the entire period of benefit on drug therapy was ignored, and the patient was counted in the ablation group.
He pointed out that all trials on catheter ablation in HF and AF before CASTLE AF, as presented by Marrouche, were small, open-label trials with missing data and subjective endpoints. According to Packer, the open-label CASTLE-AF was plagued by similar problems: the reported benefits of catheter ablation on ejection fraction, left atrial dimensions and exercise tolerance were assessed in a small (and non-representative) fraction of randomized patients, they were subject to bias of unblinding, and these benefits were typically not sustained.
He is of the opinion that the CASTLE-AF investigators did not adhere to established principles of unbiased analyses, as not all randomized patients were included in the primary analysis and the number of primary endpoint events was too small to yield reliable results. We know little about the patients who were enrolled an how they were treated. Patients had to be symptomatic, but it is unclear what this means in these patients, as it is hard to tell if AF contributes to dyspnea. HF was treated according to a 2005 guideline, and care was not provided by specialists. Agents and doses of HF and antiarrhythmic drugs were not recorded.
Enrollment in the CASTLE-AF study was slow, due to which it was difficult to preserve the integrity of randomization. As a consequence, significantly more control patients had ischemic heart disease at baseline and were treated with digitalis: even before treatment assignment, the control group was at higher risk. Packard was also critical about the primary endpoint for all-cause mortality or hospitalization for HF: the latter is not a hard endpoint in an open-label trial, and it may lead to reporting bias in the description of the admission by the investigators who know about the treatment assignment.
Furthermore, many patients, especially in the ablation group, were excluded from analysis or lost to follow-up. He noted that 397 patients were randomized, but only 363 were included in the ITT analysis. 21 ablated patients and 13 control patients were excluded from the primary ITT analysis. The difference between these numbers is disconcerting. If one would assume that all these patients had an event, the results change substantially. The number of counted events was about 30% less than the number specified in the protocol. The trial was stopped for futility.
In general, we know that studies with a small number of events generally do not yield reliable results. Several of such studies have shown effects in the opposite direction when repeated. Thus, Packer concluded his critic of CASTLE-AF by saying that it is at most hypothesis-generating. Its hypothesis should be evaluated in large, definitive, well-executed and well-analyzed trials.
In his rebuttal, Marrouche repeated that it is about the patient; and the CASTLE-AF data showed that many patients stayed out of the hospital. In the spirit of the debate, he said that Packer did not show all of the slides of this trial, while he clicked through some positive results of CASTLE-AF, including the primary composite endpoint and all-cause mortality, and an ITT analysis that showed a significant reduction of events, and as treated analysis, and per protocol analysis, all showing a significant benefit of ablation. He pointed out that Packer once wrote in a Medpage Blog post that ‘the intention-to-treat analysis works because it has rules. Its major rule is that all patients who are randomized in the trial must be analyzed according to their assigned group regardless of what happens during the course of follow-up. If you violate the rule, you are going to get invalid or unreliable results.'
In turn, in his rebuttal, Packer also blamed Marrouche for not showing everything, as he did not show the 33 patients lost to follow up, nor were they included in the sensitivity analyses. Not all randomized patients were included. He showed anecdotal information of a patient, and then claims the improvement is due to ablation. Did the patient not get HF drugs? We do not know why he improved, Packer warned.
The audience voted and showed that Packer convinced most attendants. This debate was another good example that it is important to carefully look at methods, patient populations, analyses, when interpreting the results of a trial.