John Mandrola, MD


November 10, 2016

Cardiology is on the brink of making a big mistake. Our embrace of percutaneous closure of the left atrial appendage as a means of reducing stroke in patients with nonvalvular atrial fibrillation (AF) could be one of the field's biggest reversals.

What's weird about this story is that the data are out in the open, if you do a little digging. The highest level of medical evidence comes from randomized controlled clinical trials. These trials were done for Watchman—and they showed that the device failed to reduce ischemic events. Yet we look away; or we let advocates distract us with complicated statistics.

I've avoided calling for a halt to LAA closure for as long as possible, but two recent events have changed my mind.

One is that I'm feeling pressure from colleagues to get going with a left atrial appendage closure program. Other hospitals in my city are doing and promoting the procedure.

The second reason is the irrational exuberance for this procedure. Witness the recent Transcatheter Cardiovascular Therapeutics (TCT) 2016 meeting in which a deeply flawed study was touted during a press conference and published in a leading cardiology journal. This nonrandomized, nonadjudicated series of nearly 4000 US patients implanted with the Watchman device, included as its primary end point safety data reported by device representatives present on the day of the procedure.[1]

In December 2015, I wrote that despite FDA approval of Watchman, the original proposal from the Centers for Medicare and Medicaid Services (CMS), which restricted use of Watchman to clinical studies and registries and allowed its use only in patients with contraindications to warfarin, was the correct decision. Many of my colleagues disagreed. They lobbied for broader indications. The final rule from CMS removed the requirement for inclusion into clinical trials.

What bothers me most about this story is the selective reporting and publication of data on appendage closure.

What We Know

To review, the PROTECT-AF trial[2] compared Watchman and warfarin in patients with AF (mean CHA2DS2-VASc=3.5). The results did not pass FDA muster. Although Watchman proved noninferior to warfarin for the primary end point, a composite of stroke, cardiovascular death, and systemic embolism, the FDA had concerns about excess procedural complications. These concerns, plus the confounding effects of concomitant antithrombotic drugs and subjects not receiving assigned treatments, led the FDA to ask for another trial.

At this point, the FDA was correct. If occlusion of the left atrial appendage works it should reduce ischemic stroke and systemic embolism. Yet a review of the primary end points from table 2 of the PROTECT-AF Lancet publication reveals a 50% higher rate of ischemic strokes plus systemic embolism in the Watchman arm (17/463 vs 6/244). Increasing the concern from this signal of inefficacy is the fact that 15% of patients in the Watchman arm remained on anticoagulation.

Five years later, in July 2014, came the publication of PREVAIL,[3] which compared Watchman and warfarin in another 2:1 randomization trial in 407 higher-risk patients with AF (mean CHA2DS2-VASc=3.8). Researchers studied three composite end points—safety, primary efficacy, and a late ischemic efficacy end point that excluded events in the first week after the implant. The late ischemic end point was added to show proof-of-concept that LAA occlusion prevented events.

Early safety events in PREVAIL dropped to 2.2% in the Watchman arm, which was much lower than the 7.4 per 100 patient-years in PROTECT-AF.

For the first primary efficacy end point of PREVAIL—a composite of hemorrhagic or ischemic stroke, systemic embolism, and CV/unexplained death—events rates were similar (6.4% vs 6.3%) but did not meet the criteria for noninferiority. In other words, the device was inferior to standard care. A look at the actual number of ischemic events from table 3 of the publication lends support to such clear-language wording. In the device arm, there were six ischemic strokes or systemic embolism in 269 patients vs only one in 138 patients in the control arm.

What Is Less Widely Known

The core problem with Watchman comes when we look at the late ischemic events—or, as the authors say, the "proof-of-concept" end point in PREVAIL. It's tricky because what is published and cited is different from what has been presented at meetings and to the FDA. (More on that later.)

In the July 2014 Journal of the American College of Cardiology publication of PREVAIL, in which only 28% of the patients had reached the 18-month follow-up period, the late event rate of ischemic stroke or systemic embolism was 2.5% in the Watchman arm vs 2.0% in the control arm. This met the noninferiority margin.

Later, however, at the TCT 2014 meeting and in October 2014 at the third FDA hearing , Dr Vivek Reddy presented longer-term data from PREVAIL, and this included eight new ischemic strokes in the Watchman arm. There were now 13 ischemic events in the Watchman arm vs one in the control group. With this updated data, noninferiority was not met.

Thus, in the best test of clinical medicine, the randomized controlled trial, Watchman failed to prove the concept that left atrial appendage closure prevents ischemic stroke. I reviewed the FDA transcript and found this quote from Dr Reddy (italics mine): "On the one hand we have highly significant results in PROTECT, but on the other hand we missed the two efficacy end points in PREVAIL."

When I asked Dr Reddy about this and suggested he update the original PREVAIL paper, he emailed that "all further analyses, including the analysis we presented at the third FDA panel, were post hoc analyses that were not intended by the prespecified statistical plan. Thus, the initial paper that declared noninferiority was indeed correct."

He added that the "PREVAIL-only data are not powered to make robust conclusions" and directed me to the patient-level meta-analysis published in 2015 in JACC,[4] which included all the updated data, with this caution: "Please note that this was also a post hoc analysis, but one that most people would probably agree is a reasonable one."

The top-line results of that meta-analysis showed lower rates of hemorrhagic stroke and noninferior differences in the composite end point of all-cause stroke or systemic embolism. But again, the end point that informs the proof of concept—ischemic stroke—occurred nearly twice as often in the Watchman arm. (HR 1.95, P=0.05). It's noteworthy that the majority of the third FDA panel did not believe the device to be effective.

I'm not a statistician, but updated results of a trial, results that include more patients reaching longer-term follow-up, increases our knowledge about device efficacy. These data show Watchman as inferior to warfarin, yet influential leaders in the field keep citing the incomplete PREVAIL paper. The new 2016 ESC AF treatment guidelines,[5]for instance, say the device is noninferior to vitamin K antagonist therapy, citing PROTECT AF and the incomplete 2014 PREVAIL paper. Similarly, another 2016 review paper[6] from one German and two American authors say the device is noninferior, again referencing the JACC PREVAIL paper.

A reader of the published literature would not know Watchman failed both its efficacy end points in PREVAIL. How can this be? It bothers me; it should bother you.

My Conclusion

The evidence is clear and consistent: occlusion of the left atrial appendage with Watchman fails to protect against ischemic events. This is not surprising if you think about stroke as a systemic disease. Current thinking on AF and embolic stroke must include more than just closure of the left atrial appendage. In patients with AF, vascular risk factors, nonatrial stroke mechanisms, and abnormal atrial substrate surely play a causative role for stroke. I encourage you to read this comprehensive review from the journal Stroke in which the authors call for a new model of thinking about stroke and AF.[7]

On the matter of anticoagulation-ineligible patients, some doctors believe appendage occlusion could be helpful. The answer is we don't know because these patients were not included in the clinical trials. I doubt it would be effective in this cohort. Appendage occlusion clearly does not reduce ischemic events. Proponents might point to hemorrhagic-stroke reduction with the device, but that's simply due to not using anticoagulation. And even if you posit an absolute risk reduction of 1% to 2% with the device, that's likely balanced by a 1% to 2% complication rate of implantation. And this reasoning assumes perfect closure of the appendage, which is far from reality. A contemporary series of Watchman implants found significant leaks in more 20% of cases (46/219).[8]

The most frustrating aspect of the anticoagulation-ineligible situation is that we'd be well along in having an answer to this question if the 4000 US patients who have been implanted with the device since approval were part of a mandated pragmatic trial. I'd have less moral distress implanting this device if patients were part of a trial.

Finally, the St Jude Medical Amulet LAA occluder looks to be a viable competitor to Watchman. But it would be a terrible injustice, another distraction, in fact, to do a trial of Watchman vs Amulet—given the inferiority of Watchman to warfarin. Amulet should be compared with warfarin, NOACs, or no medical therapy in ineligible patients.

Implantation of devices in the left atrial appendage should either stop or be done only as part of a clinical trial that includes a control group.



Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.