PREVAIL Yanked From ACC Program; Watchman Device Meets Safety End Point

March 09, 2013

(Updated) SAN FRANCISCO, California Preliminary data from the PREVAIL study of the Watchman (AtriTech/Boston Scientific) device for the prevention of stroke and systemic embolism in patients with nonvalvular AF appear to give some reassurance on safety issues raised previously.

But the trial has been surrounded by so much controversy, with reports of end points being changed, attempts not to present all the data, and finally an embargo break by the sponsor, leading the American College of Cardiology (ACC) to cancel the presentation completely. The slides, however, were made available to the media, but the data therein, particularly the efficacy results, appear to have many issues.

BREAKING: ACC Pulls PREVAIL After Boston Scientific Breaks Embargo

Within the hour of its presentation, the ACC announced it would be canceling the presentation of PREVAIL, following an embargo break by Boston Scientific.

An investor's notice at 6:30 am PT announced the key results in the form of an email "for immediate release." When heartwire queried the ACC and Boston Scientific about the email, Boston Scientific media spokesperson Steven Campanini responded, saying, "Due to an error with our investor-relations email system, the release went to a limited group at 6:30 am Pacific time" but insisted the embargo was still in place.

The ACC, however, disagreed and took the unprecedented step of yanking the presentation from the late-breakers and the press conference.

The trial had three co–primary end points, one for safety and two for efficacy. The safety end point was met, as was one of the efficacy end points; the other efficacy end point was narrowly missed, possibly because of a low event rate in the control group. The big question is whether this will be enough for FDA approval the second time around.

Dr David Holmes

Scheduled to present the data today at the American College of Cardiology 2013 Scientific Sessions, Dr David Holmes (Mayo Clinic, Rochester, MN) told reporters: "Despite inclusion of higher-risk patients in PREVAIL than in the previous study, there were fewer complications, and results show the device can be safely implanted by new operators. The Watchman device is therefore a real step forward, in that it offers an alternative to oral anticoagulation therapy for thromboembolic prevention in patients with nonvalvular atrial fibrillation."

The Watchman device seals off the left atrial appendage in the heart, which is the major source of stroke-causing thrombus in AF patients. It is implanted via a transseptal-catheter–based delivery system. It is already available in Europe but was turned down by the US FDA because of safety concerns raised in the previous PROTECT-AF trial. These related to a high initial rate of pericardial effusions and procedure-related strokes. The PREVAIL trial was therefore conducted to give more information on safety and to confirm the efficacy results shown in PROTECT-AF.

The PREVAIL trial enrolled 407 patients from 41 US centers who were randomized 2:1 to the device or control (warfarin). Device patients were given 45 days of warfarin therapy. The trial included new centers and operators to document that enhancements to the training program are effective. In an unusual move, the study had a Bayesian design, in concordance with a request from the FDA, Holmes said.

The device had a 95.1% implant success rate, up from the 91% rate in PROTECT.

The main safety end point--acute (seven-day) occurrence of death, ischemic stroke, systemic embolism and procedure, or device-related complications requiring major cardiovascular or endovascular intervention--occurred in six out of 269 patients (2.2%) who received the device. Holmes said this was successful, as the 95% upper confidence interval was 2.618%, below the prespecified criterion of 2.67%.

A second, broader, safety end point, including cardiac perforation, pericardial effusion with tamponade, ischemic stroke, device embolization, and other vascular complications, occurred in 4.4% of patients receiving the Watchman device in PREVAIL, compared with 8.7% in PROTECT.

Cardiac perforation requiring surgical repair occurred in 0.4% of PREVAIL patients receiving the device compared with 1.6% in PROTECT-AF. And pericardial effusion with cardiac tamponade requiring pericardiocentesis occurred in 1.5% of Watchman patients in PREVAIL vs 2.4% of those in PROTECT AF.

There were no procedure-related deaths in any of the trials, Holmes said.

The first efficacy primary end point--a comparison of the composite of stroke, systemic embolism, and cardiovascular/unexplained death in this trial vs the historical priors based on data from PROTECT--just missed showing noninferiority, with the upper 95% confidence limit slightly higher than allowed to meet the success criterion (<1.75%).

First Efficacy End Point: Rate of Stroke, Systemic Embolism, and Cardiovascular/Unexplained Death

Device, 18-mo rate Control, 18-mo rate Rate ratio (95% CI)
0.064 0.064 1.07 (0.57–1.88)

The 0.064 rate means there were 6.4 events per 100 patient-years in each group in the PREVAIL study alone. The rate ratio was calculated by combining data from both PREVAIL and PROTECT-AF, and the 1.07 figure is the median point estimate of the two trials together for the device vs warfarin (meaning the device was associated with 7% more events than warfarin).

But the upper confidence limit of 1.88 suggests that the worst-case scenario would be that the device was associated with 88% more events than warfarin.

Holmes pointed out that these results are very early, with only a limited number of patients with follow-up through 18 months thus far (58 device patients and 30 controls). He also noted that in spite of the high average CHADS2 score of 2.6, the control group had a lower rate of stroke (0.7 per 100 patient-years) compared with other published warfarin studies. "This low event rate in the control group makes it difficult to show efficacy of the device," he added.

The second efficacy end point--the comparison of ischemic stroke or systemic embolism occurring more than seven days postrandomization--did meet the criterion for noninferiority (95% CI upper bound <0.0275%).

Second Efficacy End Point: Ischemic Stroke or Systemic Embolism Occurring After Seven Days

Device, 18-mo rate Control, 18-mo rate Rate difference (95% CI)
0.0253 0.0201

0.0051 (-0.0191 to 0.0268)

Holmes said that 50% of AF patients are currently not taking anticoagulants, and there is a large unmet need for alternatives. "Many people cannot take oral anticoagulants because they are at high bleeding risk, and others don't like the inconvenience of warfarin monitoring and the nuisance bleeding with any of the anticoagulant drugs. The Watchman device offers them the alternative of undergoing one procedure instead of taking lifelong anticoagulant therapy."

All cardiologists asked by heartwire about this statement said the same thing: if it were true that fitting this device just once would give the same outcome as lifelong anticoagulant therapy, then yes, they would use it, But the data so far available are nowhere near justifying that claim.

Confusion PREVAILs

The PREVAIL trial of the Watchman device was drowning in controversy by the end of the first day of the ACC meeting.

The trial had already been the subject of many questions in the days preceding the meeting, with reports of the stipulated end points having being changed and which end points were going to be presented. Until a week ago, the entry on PREVAIL on the website stated that there was one primary end point at six months. This has now been changed to three primary end points--one safety end point at seven days and the two efficacy end points at 18 months. Boston Scientific also released a statement saying that only one end point (safety) would be presented, but after objections raised by the ACC, it agreed that all three end points would be reported.

But as it turned out, the whole presentation was dropped from the program at the last minute by the ACC because of the embargo break. However, the slides had already been made available to the press, so reports of the study are now circulating. (heartwire has posted a PDF of the full slide set at the bottom of this story.) But the complex study design, including some Bayesian analyses, which is not well understood by most cardiologists, and such a small number of patients having achieved 18 months of follow-up make the results difficult to interpret.

Statistical expert Dr Sanjay Kaul (University of California, Los Angeles) told heartwire that there was so much information missing it was hard to make sense of the trial. "A detailed study protocol is not available anywhere. That makes me worried to start with."

Large Noninferiority Margin

Kaul also pointed out that the noninferiority margins used were very large. "To have an upper margin of 1.75 mean that you are accepting that the device could be up to 75% worse than warfarin and it will be classed as noninferior. In comparison, drug trials usually have much lower noninferiority margins. Noninferiority margins need to be statistically justifiable and clinically reasonable. If a new therapy offered advantages over the standard therapy, we may be willing to accept a slightly worse efficacy and so higher noninferiority margins. But this is a device that needs an invasive procedure with a risk of serious complications to be fitted. I can't see much advantage in that, so why would be willing to accept that it could be 75% worse in efficacy?"

Kaul also pointed out that in PROTECT-AF, eight out of the 16 strokes in the device arm occurred after the 45 days of warfarin treatment had ended. Data on when the events occurred in PREVAIL have not been revealed.

Dr Steve Nissen (Cleveland Clinic, OH) was also skeptical about the data. He commented: "Warfarin has one of the largest treatment effects that has ever been seen with any drug, reducing events by more than 50% vs placebo. If you want to use something else instead, you have to be very sure it is just as good. I don't think we can say that from this data.' He added: "It is very difficult for me to see how this device could be approved by the FDA under the circumstances of this trial as we currently understand it."

Tomaselli Reassured on Safety

Commenting on the data for the media, Dr Gordon Tomaselli (Johns Hopkins University, Baltimore, MD), who has no involvement with any of the studies, said he was reassured by the safety data in PREVAIL.

"I would use this device under certain circumstances--in AF patients with a high risk of stroke who can't take anticoagulant drugs because of bleeding issues, which account for about 5% to 10% of AF patients in my practice. But I certainly wouldn't use it, on the data we have so far, in patients who could take anticoagulants. We have good data on anticoagulants in hundreds of thousands of AF patients for the prevention of stroke. This device will definitely not replace those drugs any time soon."

Statistical Gymnastics

While Tomaselli acknowledged that the studies with the Watchman device had not been conducted in patients unable to take anticoagulants (because it was compared with warfarin), he said it was not totally unreasonable to extrapolate the safety data to that group. "Yes, there is a lot of uncertainty about the efficacy results, and there does seem to be some statistical gymnastics going on, but the main concerns after the PROTECT trial had to do with safety, as there was a substantial early complication rate in that trial. After seeing the preliminary safety data from PREVAIL, I'm a little more comfortable about this."

On the efficacy data, Tomaselli said although it was not completely clear, "for me it allows the efficacy data of the PROTECT trial to prevail."

Both Holmes and the Mayo Clinic have a financial interest in technology related to this research. That technology has been licensed to Atritech.