Cardiology is on the brink of making a big mistake. We have embraced a new procedure called left atrial appendage occlusion.
You may be seeing the ads for a device called Watchman. Like this one>
The appendage-closure idea was a good one: during atrial fibrillation (AF), blood can pool in the left atrial appendage, and this promotes clot formation. (The LA appendage has many nooks and crannies.) So… if we could put a device in there, see image, this would block clots from getting out and causing stroke. Also, once the device has been in for months, the body walls it off and the patient can stop the anticoagulant drug (warfarin, or one of the new drugs, called NOACs).
Here is the problem: The Watchman device does not prevent strokes. When compared in the best test of medicine, the randomized controlled trial (RCT), the device was inferior to warfarin.
Another problem: advocates for the device have used selective reporting and publishing of trial results to alter the way patients and doctors perceive reality.
This week I published a detailed critical appraisal of the evidence for Watchman.
Here is the piece: Left Atrial Appendage Closure Should Stop Now.
The three take-home messages from that post are:
Take-home #1: In the first trial of Watchman v warfarin, called the PROTECT-AF study, complications were high. This finding, plus other irregularities led the FDA to reject the device and call for another trial.
The second trial of Watchman v warfarin, called the PREVAIL study, showed lower complication rates (good), but higher rates of ischemic stroke (the kind caused by clots) in the Watchman arm.
Take-home #2: The tricky part of the selective reporting of Watchman began with the PREVAIL trial.
Here, the authors of PREVAIL published the paper (2014) with incomplete results–only 28% of patients had reached the 1.5 years of follow-up. In this very early look at the data, the results looked nearly equivalent, though there were still more strokes in the Watchman group.
Two months after that paper was published, the PREVAIL investigators reported longer-term follow-up from the study at a medical meeting and to the FDA. (Neither of which shows in PubMed searches.)
This new data showed 8 additional strokes in the Watchman arm and none in warfarin group. Now the PREVAIL study totals were 13 strokes in Watchman versus 1 in the warfarin group. With this new data, the endpoints of PREVAIL were NOT noninferior. Or, in clear language, inferior.
Yet, crucially, this updated data, which would turn the pivotal trial for this device negative, have not been published. Instead, the authors decided to publish a meta-analysis of multiple studies of Watchman, including softer registry studies.
By combining this mishmash of studies (meta-analysis) and by using a composite endpoint, the authors noted noninferiority of the device. But even in that meta-analysis, if you look at strokes due to clots, the results show more with Watchman.
It gets worse.
Not only is the more complete PREVAIL data not published, but leaders in the field of cardiology, including influential guideline writers, continue to cite the 2014 study that is published as noninferior, but is really inferior–if it had been updated. A reader of the published literature, therefore, would think the device is equivalent when in fact it is not. Even the majority of the FDA reviewers felt the device did not meet its efficacy endpoints.
A literature search in PubMed for two of the more prominent authors of the non-updated PREVAIL study and “Watchman,” shows 10 papers since the 2014 incomplete paper. One wonders why this important data has not been published. Could it be that it would turn the pivotal trial for Watchman into a negative trial?
In my post, the primary investigator of the study, Dr. Vivek Reddy, explains why that data is not published. It’s a complicated statistical explanation.
Take-home #3: What about patients who can’t take anticoagulant drugs?
Proponents of the technology say we should use the device for patients who cannot take anticoagulant drugs because of bleeding. It’s true; for patients who have had AF, stroke, and bleeding, there are no good options.
But these patients were not included in trials with Watchman. That means we don’t know the device would work. The signals thus far suggest the device would not help these patients. For one, it doesn’t protect against stroke relative to warfarin. For another, even if it did reduce stroke by 1-2%, this benefit would be balanced by the 2-3% complication rate of implant.
What I believe should have happened for these difficult patients is the device been approved but only as part of a study. Nearly 4000 patients in the US have had this device, and if they were in a clinical trial, we’d be well along in knowing whether this device helps people who cannot tolerate anticoagulant drugs.
At the root of our embrace of this technology is the desire to help patients. But that benevolent drive should not blind us to the evidence.
Here’s the slide I usually start my critical appraisal talks with: JMM