Positive Results for PFO Closure Come With Caveats

John Mandrola, MD


September 27, 2017

Two opposing factors figure in the decision to close a patent foramen ovale (PFO) in patients with stroke of undetermined cause (ie, cryptogenic stroke [CS]).

One is the high threshold of evidence we must have before performing an invasive procedure that leaves a permanent foreign body in the heart. The other is the seriousness of stroke when it occurs in a young person. Stroke is not supposed to occur in young people without vascular risk factors.

Also relevant is the fact that PFO is estimated to occur in one in four human beings.[1] This leads to the obvious question: in patients with cryptogenic stroke, is the PFO incidental or causative?

The New England Journal of Medicine recently published not one but three randomized clinical trials comparing PFO closure and medical therapy in patients with cryptogenic stroke.[2,3,4] All three favored PFO closure.

These decidedly positive results represent a medical reversal of sorts: three previous clinical trials of PFO closure had found no statistically significant benefits for closure over medical therapy.[5,6,7]

The Three Recent Trials

Two of the latest publications were new trials. In the multinational REDUCE trial,[2] patients with PFO and CS were assigned to closure (with the Helex Septal Occluder or Cardioform Septal Occluder, Gore Medical) plus antiplatelet therapy (n=441) or to antiplatelet therapy alone (n=223). Only patients who had a documented stroke on MRI or CT scan, a careful workup to exclude non-PFO causes of stroke, and right-to-left shunting through a PFO on a transesophageal echo were enrolled in REDUCE.

After a median follow-up of 3.2 years, ischemic stroke occurred in 1.4% of the PFO-closure group and in 5.4% of the antiplatelet arm (hazard ratio 0.23; 95% CI 0.09–0.62; P=0.002). Serious device-related adverse events occurred in 1.4%, and 29 patients (6.6%) developed atrial fibrillation.

In the CLOSE trial,[3] investigators in France and Germany randomized patients with CS who had an associated atrial septal aneurysm or large interatrial shunt to either PFO closure (mostly Amplatzer, St Jude Medical [Abbott]) plus antiplatelet therapy, antiplatelet therapy alone, or oral anticoagulation. The authors performed two analyses: one was PFO closure (n=238) vs antiplatelet therapy (n=235) and the other was anticoagulation (n=187) vs antiplatelet therapy (n=174).

After a follow-up of 5.3 years, zero strokes occurred in the PFO-closure group vs 14 (5.9%) in the antiplatelet arm. Procedural complications occurred in 14 patients. Stroke rates did not significantly differ between the anticoagulation- vs antiplatelet-treated patients—although this comparison was underpowered.

The third trial is one we've read about before. The RESPECT investigators enrolled 980 patients with CS to undergo PFO closure (Amplatzer) or medical therapy. The medical-therapy arm allowed four different regimens: aspirin, warfarin, clopidogrel, or aspirin plus dipyridamole.

The first report from RESPECT,[7] which included 2.1 years of follow-up, showed a 51% relative reduction of stroke with PFO closure, but this did not reach statistical significance (P=0.08) in the intention-to-treat comparison. In the on-treatment analysis, which counts only those who got their assigned treatment, PFO closure resulted in a statistically significant 73% relative reduction in events (P=0.007).

The latest publication from RESPECT with longer-term follow-up changed the results.[4] Over time, significantly more events occurred in the medical-therapy arm than in the PFO-closure arm. Now, the intention-to-treat analysis found recurrent stroke in 18 patients in the PFO-closure group and in 28 patients in the medical-therapy group. This 45% relative decrease barely reached statistical significance (P=0.046). Notably, the absolute difference was only 0.49 fewer events per 100 patient-years. Procedure- or device-related adverse events occurred in 4.2% of the patients in the PFO-closure group.


Taken together, the results of these three trials are clear: when used in young patients (mean age mid-forties) with PFOs who have undergone careful evaluation to exclude other causes of stroke, PFO closure reduces the rate of recurrent stroke compared with medical therapy—which mostly consisted of antiplatelet drugs.

Why were these trials positive and the previous three were not?

I think the main reason for the medical reversal was the careful selection of patients and end points.

Previous negative trials like CLOSURE I[5] and PC[6] used transient ischemic attacks (TIA) in both the inclusion criteria and the primary end point. This is a fatal flaw. TIAs are far too subjective to include in an outcomes trial. What is more, CLOSURE I enrolled patients with lacunar strokes; PFO closure can help only people whose stroke was due to PFO.

The newer trials, REDUCE[2]and CLOSE[3] enrolled only patients with cryptogenic stroke and PFOs that were likely to be causal, not incidental. This makes recruitment harder but it ends up being a better test of PFO closure.

What about the RESPECT[4,7] trial—why did longer follow-up turn the results statistically significant for PFO closure?

Two-year follow-up data from RESPECT numerically favored PFO closure but did not reach statistical significance. If the trend is real then more time and events should favor PFO. That is exactly what happened in the long-term report. In addition, a pooled analysis of three previous trials also favored PFO closure over medical therapy.[8]

Caveats and Areas of Uncertainty

A major limitation for these studies concerns the comparator. Most of the patients assigned to the medical arm in these trials received antiplatelet therapy. That is a problem. If the most likely mechanism of PFO-related stroke is paradoxical embolism of a venous clot, antiplatelet drugs are not the best drugs to prevent that. No one treats venous thrombosis with antiplatelet therapy; we use anticoagulation.

This is not a knock on the trialists; the PFO trials began well before the advent of direct-acting oral anticoagulants (DOACs). But given recent data[9]  showing similarly low bleeding rates for full-dose apixaban vs aspirin in older patients with atrial fibrillation, the question now is how would PFO closure compare with DOAC therapy?

I am worried this important study will never be funded. Device makers already have three positive trials, and DOAC makers likely view this population as too small to justify the cost of a study.

Another area of uncertainty is determining whether the PFO was causal or incidental. This is crucial, because PFO closure is an invasive procedure that comes with a risk of harm. Complication rates will likely rise as the procedure expands to lower-volume centers.

Patient selection for PFO closure will be key. By email, Dr David Spence (Western University, Ontario) suggested two ways to sort out the causality of a PFO in patients with stroke of undetermined origin.

He first emphasized important clinical clues to paradoxical embolism. These include prolonged sitting, dyspnea, or a Valsalva maneuver at the onset of stroke; previous history of DVT, PE, or varicose veins; waking up with stroke; and sleep apnea.[10] Spence described a few cases of PFO-related paradoxical stroke he had seen in tax preparers on the last due date for returns—likely related to prolonged sitting.

He also noted the value of transcranial Doppler (TCD) for detecting right-to-left shunting. In an observational study of 334 patients from his center, TCD looked to be more sensitive than transesophageal echocardiography (TEE) for finding right-to-left shunting[11]. The stroke neurologist at my hospital agreed on the utility of TCD but warned that it was not easily obtained at our hospital because we do not own the device or have on-site staff trained to use it.


In 2017, the young person with stroke of unknown origin who has a PFO will require a skilled clinician(s) to help with decision-making.

The first step is a careful evaluation to exclude other potential causes of stroke. The three trials were positive likely because of their strict selection criteria.

The next step will be determining whether the PFO is causal—don't rely on echocardiography alone, but rather a robust collaboration with neurology. The transcranial Doppler looks to be something cardiologists should become more familiar with.

Communicating the absolute benefits, harms, and uncertainties from the trials may be the toughest challenge. The trials were positive, but the event rates were low, the absolute differences small, the procedure includes harm, and the lack of comparison with modern anticoagulants adds doubt to the decision.

As it is in so much of modern cardiology, patients need much more than a proceduralist; they need a doctor, or in this case, a team of doctors.


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.
Post as: