Dr Harriette Van Spall (HVS):
I'm Harriette Van Spall, clinical trialist from Canada, and I am delighted to have with me Dr Saibal Kar, cardiologist at the Los Robles Medical Center, who is here at ACC 2026 to present the late-breaking results of the CHAMPION-AF clinical trial. Welcome, Dr Kar.

Dr Saibal Kar:
Thank you, Harriette.

HVS:
We are excited about your trial, especially on the backbone of the recent CLOSURE-AF trial that was similar in its goals. I wonder if you could give us a brief overview of those trial results, and how your trial, on the background of CLOSURE-AF, is predicated.

SK:
Harriette, that is a great question to start with. I think there were two important trials that were recently published. One was the OPTION trial and the second is CLOSURE-AF, both in the New England Journal of Medicine.

Starting with the OPTION trial, the population was kind of similar to our study, which is patients who had atrial fibrillation ablation. One group had anticoagulants continued and the other group had left atrial appendage closure using the WATCHMAN device. Some procedures were done concomitantly and some were done at a later time. What they found is that they met non-inferiority for their composite of stroke, death and systemic embolization, and they were superior for non‑procedural major bleeding and clinically relevant non‑major bleeding, and that is where they met their endpoints, which makes sense.

What is nice about that study is that the ischaemic stroke rates were exactly the same in both arms, which was good.

The second study that you just referred to is the recently published CLOSURE-AF trial. I think it is a very important and pragmatic study. The importance of that study is that they took high‑risk patients; the average CHA\(_2\)DS\(_2\)-VASc score was 5 and the HAS‑BLED score was approximately 3, and if you look in more detail at their demographics, these were quite sick patients, including patients with chronic renal failure. They had a composite endpoint of stroke, systemic embolism, death and major bleeding together, and the whole point was to show non‑inferiority.

They did not meet non‑inferiority, based on the fact that they had more bleeding and some increase in deaths. But what was most impressive in the data, which was not emphasized, is that the stroke reduction was similar in both groups.

So putting these two studies together, the first thing I am learning is that left atrial appendage closure does reduce stroke, and now we have at least three or four trials – the original WATCHMAN trials, PRAGUE‑17, CLOSURE-AF and OPTION – all showing results in the same direction.

The difference, however, in CLOSURE-AF was bleeding. Bleeding was expected to be less, but the bleeding was more in the device arm, and I personally think it was related to procedure‑related bleeding, because they actually had a very sick group of patients and most of the bleeding events took place early. If you look at the procedural complications, they were actually quite high: there was about a 1% pericardial effusion rate, there was a bleeding rate that was quite high, and there was actually even a death within 7 days.

So I think what it tells me is that patients who are very high risk – not just from a stroke perspective, but very high‑risk overall – may not always benefit from an invasive procedure because of the side effects.

Now, you cannot extrapolate those data to all devices, because in the CLOSURE-AF trial they used four different types of devices and some of them are not even used nowadays. They also had a complicated post‑procedure regimen that is not often used, such as dual antiplatelet therapy for 3 to 6 months, which is also different. So these may have led to higher bleeding complications.

HVS:
Sure. Very insightful results, because bleeding and periprocedural complications are clinically relevant and sometimes have life‑threatening sequelae. On the background of this, what was the question that you tried to answer?

SK:
So I think it is very interesting that our study has been presented at the time when the CLOSURE-AF results came out, and that is why we did separate these two endpoints. It was important to separate the effectiveness endpoint from the bleeding endpoints.

In our study, the CHAMPION-AF trial, first of all, this is the largest randomized, prospective study comparing left atrial appendage closure with the WATCHMAN FLX versus NOAC therapy. The data that are being presented are the 3‑year endpoints.

There are two co‑primary endpoints. The first is the Kaplan–Meier estimated rate of a composite of cardiovascular death, stroke and systemic embolization, and the second is a safety endpoint for superiority for non‑procedural major or clinically relevant non‑major bleeding over 3 years. You had to meet both these endpoints to call it a successful study.

There is a third endpoint which is just examining ischaemic events – ischaemic stroke and systemic embolization – which is powered at 5 years. In addition, there are two secondary endpoints: one for major bleeding (procedural and non‑procedural) over 3 years, and the second is a net‑benefit endpoint, quite similar to CLOSURE-AF, which is a combination of stroke, death, systemic embolization and major or clinically relevant non‑major post‑procedural bleeding.

HVS:
You had three co‑primary endpoints, two of which you will present at ACC, and perhaps we can focus on those two. The first one was a non‑inferiority hypothesis, and the second one was superiority. Tell us about the way in which you set the non‑inferiority margin. What did you use? Was it a relative risk or an absolute risk that drove the setting of your margins?

SK:
First of all, Harriette, that is a very important question. Regarding the first endpoint, which is the non‑inferiority design, it was based on the event rates from prior trials. The prior trials that we used were the four WATCHMAN clinical trials and the four large NOAC trials. We found that the average event rate around 3 years was about 12%.

Based on that, we set a non‑inferiority margin of 4.8%, which would mean a relative risk of 1.4. So we set it at 4.8%. We also estimated that there would be a 12.5% attrition rate. Putting all of that together, we thought that 3,000 patients would provide 97% power to assess the primary analysis.

HVS:
Okay, fantastic. Tell us about the study methods.

SK:
Basically, 3,000 eligible patients, who had to have a CHA\(_2\)DS\(_2\)-VASc score of 2 or more for females or 3 or more for males and had indications for anticoagulant therapy, were randomized 1:1 to either receiving the WATCHMAN FLX device or continuation of NOAC therapy.

If they were randomized to the device arm, they received the device and then an antithrombotic regimen for 3 months – either a NOAC, a NOAC plus aspirin, or dual antiplatelet therapy for 3 months. After 3 months, they had imaging done at about 4 months, but they were continued after 3 months on single antiplatelet therapy (aspirin) for the rest of the study.

If you were randomized to the NOAC arm, you were just continued on a NOAC at the dose based on the FDA label throughout the conduct of the study. Patients were allowed to cross over from the medical arm to the device arm if they reached an endpoint – for example, if they had a stroke or a major bleeding event, which is the current indication, they could cross over.

The study analysis was intention‑to‑treat. That means if you were assigned to the medical arm and later crossed over, you still stayed in the medical arm for analysis. Similarly, in the device arm, if an attempted procedure was done and you could not implant the device, they were still followed in that arm under intention‑to‑treat analysis.

HVS:
And are you presenting both the intention‑to‑treat and the on‑treatment analyses, which are important for a non‑inferiority design?

SK:
We do have them in the manuscript, but in the 10‑minute presentation we are just going to show the intention‑to‑treat analysis.

HVS:
Okay. Tell us about your study population and the baseline characteristics.

SK:
The baseline characteristics were well matched in both groups. The average age was about 71 years, which is about 5 years older than in the CLOSURE-AF trial. The average CHA\(_2\)DS\(_2\)-VASc score was over 3, with over 70% of patients having a score of 3 or more.

The HAS‑BLED score average was 1.3. This is very important because these are patients with a lower bleeding risk, and the reason for choosing these patients is that they had to be candidates for long‑term NOACs. That put our study at risk, because we might not have “won” on bleeding, as we basically excluded patients with a very high bleeding risk.

Another important point in the demographics is that about 40–50% in both arms had an AF ablation in the last one year.

HVS:
Okay, and were there any procedural complications in the LAAC group?

SK:
For procedural complications through 7 days, the death rate was zero. There was only one person who had a stroke. There were no cases of device embolization. Pericardial effusion requiring intervention was 0.6% – that is 8 patients – and other major bleeds were only 0.4%, which is 5 patients.

Putting it all together, this event rate is quite low and actually much lower than what was observed in CLOSURE-AF, and this dataset is consistent with current clinical practice in the US.

HVS:
What determines the antiplatelet or anticoagulation regimen in the device patients?

SK:
Harriette, this is a great question. Both these arms – NOACs and dual antiplatelet therapy – are currently indicated after implantation. It was the choice of the physician to choose either a NOAC or DAPT. We modified it slightly because we truncated therapy to only 3 months, not 6 months.

So NOACs or DAPT were given for 3 months. More importantly, later in the study we encouraged physicians not to use a combination of a NOAC plus aspirin because it just increases bleeding risk without any benefit. As expected, about 85% of patients received NOACs for 3 months and about 12% received DAPT.

HVS:
And the imaging protocol?

SK:
Harriette, this is also an extremely important question, because the imaging protocol is slightly different from current practice. We decided to use 4 months, not the standard 45 days, because we wanted the patients to be off their intensive antithrombotic regimen before we imaged them.

For example, if we imaged them early while they were still on DAPT, you would get a false sense of security that there is no thrombus. That is why we treated them, stopped the more powerful regimens, put them on single antiplatelet therapy, and imaged them at 4 months.

At 4 months, if there was a significant leak – greater than 5 mm – or a large device‑related thrombus that required treatment, they would then be reinstituted on a DOAC or any other therapy that the physician thought appropriate. As you can see in the slide set, there was no case of a leak greater than 5 mm.

There was a group of patients who had device‑related thrombus – overall it was 4.8%. We tried to be conservative: any layer 3 mm or more was defined by the core lab as device‑related thrombus, but a pedunculated, large thrombus was only 1.8%. Only a small percentage of patients actually needed resumption of NOAC therapy: 1.8% of patients with DRT had resumption of NOACs, which is 24 out of 63 patients.

HVS:
Okay. What was your primary efficacy treatment effect?

SK:
In the NOAC arm, the Kaplan–Meier rate of the composite endpoint was 4.8%. In the left atrial appendage closure arm, it was 5.7%. The absolute difference was 0.9%, with a 97.5% upper confidence bound of 2.6%, which was lower than the non‑inferiority margin of 4.8%. That met the hypothesis of non‑inferiority, with a p‑value of less than 0.001.

HVS:
Right, and was your proportional hazards assumption met for this analysis?

SK:
We did check for the proportional hazards assumptions and they were met, and that is why we could use that test for non‑inferiority.

HVS:
Okay. So, in summary, for the first co‑primary endpoint, the LAAC strategy was non‑inferior to a strategy of NOAC therapy in these patients. Were you able to glean what the events were driven by?

SK:
Cardiovascular death was equal in both groups at 2.7%. Systemic embolization was 0.1% in the NOAC group and zero in the left atrial appendage closure group.

Regarding stroke, there was no difference in haemorrhagic stroke, which was 0.4% in both groups. For ischaemic stroke, it was 2% in the NOAC group and 3.2% in the left atrial appendage closure group. Therefore, putting the ischaemic events together, the annualized average difference in ischaemic stroke and systemic embolization rates was 0.33% per year or greater.

An important point is that 87% of patients in the NOAC group were adherent to their medications for the majority of follow‑up, and 76% in the LAAC group. This is important because we know that compliance with medications over 3 years is rarely 87%; it is usually much lower. In spite of that, stroke rates were quite low, and that probably explains the low event rates.

HVS:
Okay. And how about the non‑procedural bleeding?

SK:
Non‑procedural major or clinically relevant non‑major bleeding was 19% in the NOAC arm and 10.9% in the left atrial appendage closure arm. The hazard ratio was 0.55, with a confidence interval of 0.45 to 0.67, and it met the superiority criteria with a p‑value of less than 0.001.

HVS:
And perhaps the procedural bleeding is even more relevant, because one might deduce that non‑procedural bleeding would be worse with NOAC therapy. Tell us about the combined procedural and non‑procedural bleeding, which I think you had as a secondary endpoint.

SK:
For the secondary safety endpoint of major bleeding (procedural and non‑procedural) over 3 years, there was a slight increase in the first few months in the LAAC group and then the curves crossed over in favour of left atrial appendage closure. The rate was 5.9% in the left atrial appendage closure arm and 6.4% in the NOAC arm, and this met the criteria for non‑inferiority with a p‑value of less than 0.001.

We also did a post‑hoc analysis of all major and modified clinically relevant non‑major bleeding. In that analysis, in the first 6 months there was a slight increase in bleeding in the LAAC group because of procedural events, but beyond 6 months the curves crossed over, and at 3 years it was 12.8% in the LAAC group and 19% in the NOAC group. This was not pre‑specified, but we did the analysis and it is consistent.

HVS:
So early complications with the LAAC group, as one might expect, and then bleeding events accrue over time in the NOAC group. Any flags in your subgroup analyses that you would like to highlight?

SK:
Regarding the pre‑specified subgroups, for both primary efficacy and safety, the effectiveness seems to be similar across all the pre‑specified subgroups, both for efficacy and safety.

HVS:
Okay. Any limitations – methodological or related to adherence? You did not get into the on‑treatment analysis, but if you have any highlights from that analysis that are relevant, we can discuss those as well.

SK:
As expected, all trials have limitations. The trial results may not apply to all left atrial appendage closure devices and all post‑implant medication regimens, because you actually pointed out the differences in post‑implant medications.

Second, this may not be generalizable to all atrial fibrillation patients. For example, patients with advanced heart failure or ejection fractions less than 30% were excluded.

Third, the non‑inferiority margin for the primary efficacy endpoint was defined in absolute and relative terms, and the lower‑than‑expected observed primary efficacy event rate in the NOAC arm facilitated achieving statistical non‑inferiority. These are the major limitations.

In addition, we did not include patients of very advanced age or with very high CHA\(_2\)DS\(_2\)-VASc scores, similar to CLOSURE-AF.

HVS:
Right, and where we set margins and how we go about setting them is really a major point of contention in non‑inferiority trials. With absolute margins, you are relying on event rates, and when event rates are lower or higher than expected, it affects our analyses and the results and therefore the inferences we make. So these are important points to raise.

Did your intention‑to‑treat analysis remain consistent with an on‑treatment analysis approach?

SK:
We did a per‑protocol sensitivity analysis because there were some crossovers, and the results were similar.

HVS:
Such an interesting trial, with many methodological nuances, some of which we did not get into, but we are going to keep this discussion within our time allocation.

I congratulate you on these findings from your important trial. They really give us a lot to think about in terms of how we select patients for this intervention, how we use thrombotic risk but also bleeding risk to help guide care and ensure that patients experience the best outcomes possible.

Thank you for highlighting, through your work, that in patients with non‑valvular AF who are relatively low risk for bleeding, a strategy of left atrial appendage closure is non‑inferior to long‑term NOAC use and provides benefit in terms of long‑term bleeding risk. Thank you so much, congratulations, and I cannot wait to see your presentation.

SK:
Thank you very much. This was very helpful and I enjoyed every moment of it.

The transcript has been edited with the assistance of AI.