CLOSURE I: First PFO/Stroke Study Is Best Answer So Far

Shelley Wood

March 15, 2012

March 14, 2012 (Boston, Massachusetts) — A device that many hoped might change the landscape of stroke prevention will likely never get the chance to prove itself properly for a host of reasons now being lamented by experts, as the first randomized clinical trial in this arena finally makes its way to print [1]. Published online today in the New England Journal of Medicine, the CLOSURE I trial of patent foramen ovale (PFO) closure for stroke/transient ischemic attack (TIA) showed no differences in outcomes between the device and medical therapy. Despite its much-discussed flaws, the trial is in the minds of many the best available evidence to date and should lead to a slowdown in the use of PFO-closure devices.

As previously reported by heartwire , principal investigator for the trial, Dr Anthony Furlan (Case Western Reserve University, Cleveland, OH), first presented the results in November 2010 at the American Heart Association Scientific Sessions.

Speaking with heartwire , Furlan recounted the hurdles the trialists faced in getting to the finish line, in particular the excruciatingly slow enrollment, hampered by rampant off-label use of the devices in patients whose physicians felt could not run the risk of being randomized to medical management.

"No randomized trial is perfect," Furlan said. "We move our understanding of the problem forward incrementally. There is a reason nobody ever did a randomized trial previously. . . . It took us nine years to get this into print, and that doesn't even include the year it took to design the trial. So 10 years of stick-to-it-ness to get to this discussion."

Power vs altruism

The trial has also been accused of being radically underpowered to show a difference between medical management and device closure or wildly optimistic about the benefit of closing a PFO to prevent stroke. "Accuse yourselves, everybody. We're underpowered because it took us years and years to get to this number of patients," Furlan says.

In an accompanying editorial [2], Dr S Claiborne Johnston (University of California, San Francisco) also points to the decision to include TIA in the end point (something the recently completed but not-yet-reported RESPECT trial excluded) and to enroll patients who likely had initial strokes of non-PFO–related causes. This last "may have reduced the overall event rate and the ability to detect a treatment effect," Johnston writes.

Furlan doesn't dispute this, but calls it "this whole feasibility vs power issue, or what I call stroke realpolitik vs scientific altruism."

"I have felt so bad for these companies," Johnston told heartwire . "Some of these device companies have been around forever, and it's just not fair that it should have been so hard." But that said, he acknowledges these same companies must clearly have been marketing their products outside of what was "proven" about their efficacy. "So they've made some money, and that part makes me mad, but they really wanted to get those trial answers."

In his editorial, Johnston, using the US government's Health Care Cost and Utilization Project (HCUP) database, estimates that in the nine years it took CLOSURE I to produce its disappointing results, approximately 80 000 patients had PFOs closed, at an average cost of $10 000 per procedure.

"Even if only half these patients were treated by this method for the purpose of preventing stroke, it would suggest that during that period of time $400 million was spent on a procedure that had no apparent benefit, to say nothing of the potential clinical risks involved," Johnston writes.

As such, Johnston hopes CLOSURE I and the other trials in this field will send a strong message to insurance companies, whose decision to reimburse an unproven therapy is a key reason that physicians continued to implant these devices. Had reimbursement for PFO-closure devices been permitted only for patients participating in the PFO trials, he believes the many thousands of patients would have resulted in a clear-cut answer.

As it is, Johnston and Furlan are eager to see the results of RESPECT, anticipated to be released later this year. And if RESPECT is also negative, Johnston, for one, predicts Gore — the third company with a randomized clinical trial in this space — to pull the plug on its PFO/stroke study, REDUCE.

 . . . which doesn't mean it doesn't work

Of note, both Johnston and Furlan seem to agree, CLOSURE I, while failing to show a benefit of PFO closure, does not rule out a possibility that it might, in fact, help some patients.

"This is not the last answer, this is part of the answer; we need to refine our diagnostic criteria for cryptogenic stroke, and we need to define the patient subgroups [in which] the device might still have a role," Furlan said.

In fact, Furlan told heartwire that he and others actually met with the FDA, proposing a registry with the aim of drilling down to a "kind of purified population" that might benefit. "So for example, patients under the age of 45 with absolutely no vascular risk factors and only a cortical infarct — sort of the ideal patient [one who couldn't or didn't want to be on lifelong anticoagulation], but we couldn't convince the FDA that our subgroup analyses showed a strong enough trend to support a registry."

And both Furlan and Johnston agree it is highly unlikely another company will be conducting a trial in this field any time soon, given the seemingly restricted group in whom the device might — might — be the best option for stroke prevention.

Key points from CLOSURE I, Furlan stresses, are that physicians were "closing too many holes" and also that they need do a better job screening patients for so-called "cryptogenic" stroke.

"I'll tell you how it's changed my own practice," Furlan said. "Every patient I see now with cryptogenic stroke — whether or not they have a PFO — they get a 30-day event monitor. . . . I think we've probably underdiagnosed AF in these patients: that's one of the take-home messages of this trial."

It's too soon to say whether the presentation of the CLOSURE I results in 2010 curbed enthusiasm for the use of the devices: Furlan points to the fact that NMT Medical, which funded the trial, has since gone out of business. Johnston, however, thinks physicians simply "ignored" the results of CLOSURE I, saying "this is the first trial, it has problems, let's wait for more." And, he adds, many people likely were waiting to see the published results.

Closure details

As previously reported, CLOSURE I enrolled 909 patients randomized equally to PFO closure using the STARFlex closure device (NMT Medical) with 24 months of aspirin and six months of clopidogrel or to best medical therapy — aspirin or warfarin or a combination.

During two years of follow-up, the cumulative incidence of the primary end point (a composite of stroke or TIA, death during the first 30 days, or death from neurologic causes between day 31 and two years, expressed as Kaplan-Meier estimates) was not statistically different between the two groups. No between-group differences were seen according to baseline features. Of note, an alternative source of stroke, unrelated to paradoxical embolism, was seen in most of the recurrent TIA or strokes in the study.

Both major vascular complications and atrial fibrillation, mostly periprocedural, were significantly more common in the intervention group, but other safety end points were no different between study arms. Procedural and technical success rates (no or trace residual leaking) were high, with 86% percent of PFOs closed at one year.

Efficacy and safety: Intention-to-treat population

End point Device (%) Medical therapy (%) Hazard ratio (95% CI) p
Composite end point 5.5 6.8 0.78 (0.45–1.35) 0.37
Stroke 2.9 3.1 0.90 (0.41–1.98) 0.79
TIA 3.1 4.1 0.75 (0.36–1.55) 0.44
Major vascular complications 3.2 0.0 <0.001
Atrial fibrillation 5.7 0.7 <0.001



The published conclusions, says Furlan, "haven't changed one whit" since the conclusions presented in 2010, although investigators delved a bit deeper into who should and shouldn't be included in intention-to-treat analyses — information included in an online appendix.

"The final results that we're reporting in the NEJM do not deviate from what we presented in public," Furlan stated. "There have been trials in the past where the presentation and the publication aren't quite in sync," he added, possibly alluding to MIST. The NMT Medical migraine trial of the same device studied in CLOSURE I, MIST was roundly criticized after somewhat rosier conclusions were presented publicly, only to be replaced by robustly negative conclusions when the study was ultimately published in Circulation.

Asked why CLOSURE I took almost a year and a half to make it into print, Furlan explained, "When NMT declared chapter 11, we lost all supporting resources. In addition, NEJM requested several manuscript revisions. It is a credit to the executive committee and [the Harvard Clinical Research Institute] HCRI that we nonetheless persevered."


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.