Clinical Question

PFO closure for cryptogenic stroke?

Q: PFO Closure for Cryptogenic Stroke

Three RCTs published in NEJM 2017 established benefit of PFO closure for cryptogenic stroke in patients <60 years. CLOSE required atrial septal aneurysm or large shunt (NNT ~20 over 5 y). RESPECT long-term follow-up showed HR 0.55 for stroke recurrence (NNT ~42). REDUCE used a clean antiplatelet comparator and showed HR 0.23 (NNT ~28). All three trials demonstrated excess atrial fibrillation (typically transient and periprocedural) as a consistent trade-off. The earlier CLOSURE-I (2012) and original RESPECT (2013) were neutral; refined patient selection (RoPE score, shunt size, septal aneurysm) was needed for benefit.

Synthesises 7 trialsFrom the early trials that missed their primary endpoint (CLOSURE-I, PC, RESPECT 2013) to the positive 2017 cluster and DEFENSE-PFO; AF excess is the trade-off

Clinical Synthesis

Whether PFO closure helps depends far less on the presence of a PFO than on how likely that PFO caused the stroke: pooled trial data show a roughly 60% relative reduction in recurrent stroke on average (adjusted HR 0.41), but the benefit concentrates in patients whose PFO is the probable culprit, and the atrial fibrillation excess falls hardest on the patients in whom it is not.

Bottom line

Select on causal likelihood, not on the mere presence of a PFO. In patients 18 to 60 with nonlacunar cryptogenic stroke, a PFO, and no competing etiology after vessel imaging, prolonged cardiac monitoring, and echocardiography with bubble study, closure plus long-term antiplatelet is reasonable as a decision made jointly with cardiology and the patient (Class IIa, Level B-R), and the case is strongest when the RoPE score is high or the PFO carries high-risk anatomy. When causal likelihood is low, expect no stroke reduction and a larger atrial fibrillation excess: that is the group to leave alone.

Sources

Trials in this question· 7

CLOSURE I2012

Primary not metHR 0.78 (0.45–1.35), P=0.37

First major PFO closure RCT; STARFlex device did not reduce stroke/TIA recurrence vs medical therapy. Negative.

Primary not metHR 0.78 (0.45–1.35), P=0.37

PC Trial2013

Primary not metHR 0.63 (0.24–1.62), P=0.34

Second major PFO closure RCT; Amplatzer device did not reduce composite events vs medical therapy. Underpowered; negative primary.

Primary not metHR 0.63 (0.24–1.62), P=0.34

RESPECT (Original 2013)2013

ITT P=0.08HR 0.49 (0.22–1.11)

Original 2013 RESPECT primary publication; ITT primary NOT met at median 2.1y (HR 0.49, P=0.08 borderline). The 2017 extended follow-up later converted the result.

ITT P=0.08HR 0.49 (0.22–1.11)

CLOSE2017

NNT 20HR 0.03 (0.00–0.26)

In cryptogenic stroke <60 with PFO + ASA or large shunt, PFO closure abolished recurrent stroke vs antiplatelet alone.

NNT 20HR 0.03 (0.00–0.26)

REDUCE2017

NNT 28HR 0.23 (0.09–0.62)

In cryptogenic stroke 18–59 with PFO and clean antiplatelet comparator, Gore HELEX/Cardioform closure beat antiplatelet alone on clinical stroke.

NNT 28HR 0.23 (0.09–0.62)

RESPECT (long-term)2017

NNT 42 (exploratory)HR 0.55 (0.31–0.999)

In cryptogenic stroke 18–60 with PFO, Amplatzer Occluder beat medical therapy on recurrent stroke over a median 5.9-year follow-up.

NNT 42 (exploratory)HR 0.55 (0.31–0.999)

DEFENSE-PFO2018

NNT 8 over 2y (6 events)0% vs 12.9% at 2y (P=0.013)

In cryptogenic stroke with HIGH-RISK PFO anatomy (ASA, hypermobility ≥10 mm, or PFO size ≥2 mm), closure markedly reduced 2-year stroke / vascular death / major bleeding vs medical therapy.

NNT 8 over 2y (6 events)0% vs 12.9% at 2y (P=0.013)

Guideline timeline

How the PFO recommendations moved, 2018 to 2025

A patent foramen ovale is one finding, but the advice depends entirely on why you are asking. Three separate questions run through this sequence: whether to close it after a stroke, what to give if it is not closed, and whether to close it for migraine. Read across them rather than down one.

2018BMJ Rapid Recs
Antithrombotic choice after cryptogenic stroke
Recommends itWeak, low-quality evidence
BMJ Rapid Recommendation on PFO closure, antiplatelet therapy or anticoagulation for cryptogenic stroke. Verbatim: "We make a weak recommendation for anticoagulant therapy rather than antiplatelet therapy." The recommendation rests on low-quality evidence; the panel estimated a 7.1% absolute reduction in ischaemic stroke over 5 years alongside a 1.2% absolute increase in major bleeding. This is the reason the statement "no guideline recommends anticoagulation over antiplatelet therapy" is false as usually written and must not be published in that form. The defensible framing, which every surface uses: no major neurology or cardiology society recommends anticoagulation over antiplatelet therapy; the AAN treats the two as equally acceptable (Level C); and a BMJ Rapid Recommendation panel, which is not a neurology or cardiology society, made a weak recommendation for anticoagulation in 2018 on an evidence base that predates RE-SPECT ESUS, which no society has adopted.
The most favourable position anticoagulation has been given. A rapid-recommendation panel came down weakly in its favour on low-quality evidence, before RE-SPECT ESUS reported. No society adopted it, and the later evidence did not confirm it.
A BMJ Rapid Recommendation panel, not a neurology or cardiology society.
Patent foramen ovale closure, antiplatelet therapy or anticoagulation therapy alone for management of cryptogenic stroke? A clinical practice guideline (BMJ Rapid Recommendations, Kuijpers et al., BMJ 2018;362:k2515)
2020AAN
Antithrombotic choice after cryptogenic stroke
Either option acceptableLevel C for the choice; Level B where anticoagulation is separately indicated
AAN 2020 Practice Advisory Update on patent foramen ovale and secondary stroke prevention, endorsed by SCAI, the AHA/ASA and the European Academy of Neurology. Conclusion: "For patients with cryptogenic stroke and PFO, anticoagulation medication and antiplatelet medication are possibly equally effective at reducing recurrent stroke (HR 0.73, 95% CI 0.45 to 1.17). Of note, the high end of the CI does rule out a clinically important benefit for aspirin." Statement 3a (Level C): "In patients who opt to receive medical therapy alone without PFO closure, clinicians may recommend either an antiplatelet medication such as aspirin or anticoagulation (using a vitamin K antagonist, a direct thrombin inhibitor, or a factor Xa inhibitor)." Statement 3b (Level B): "In patients who would otherwise be considered good candidates for PFO closure but require long-term anticoagulation because of suspected or proven hypercoagulability (defined thrombophilia, unprovoked deep venous thrombosis, or unprovoked pulmonary embolism), clinicians should counsel the patient that the efficacy of PFO closure in addition to anticoagulation cannot be confirmed or refuted." NNT WARNING, binding: this advisory computed an NNT of 21 with a confidence interval of 19 to -60, which crosses infinity and includes net harm. That interval is the reason NNT must not be displayed anywhere on this question; if 21 is ever shown it must carry the full interval and an explicit statement that the interval includes harm. Trial figures carried by this advisory and attributable to it rather than to the primary publications: PICSS "recurrent stroke at 2 years" 2 of 42 (4.8%) versus 8 of 56 (14.3%) with HR 0.52 (95% CI 0.16 to 1.67), which conflicts with the composite 4 of 42 versus 10 of 56 that Turc and colleagues attach to the identical hazard ratio and interval; CLOSE anticoagulation versus antiplatelet HR 0.44 (95% CI 0.11 to 1.48) with "There was no significant difference in stroke recurrence rate" and 93% of the anticoagulation arm receiving a vitamin K antagonist; the RE-SPECT ESUS PFO-stratum HR 0.88 (95% CI 0.45 to 1.71); and a 47-patient randomized comparison of aspirin against warfarin attributed to Shariat and colleagues, "HR combined stroke and TIA favoring aspirin 3.03, 95% CI 0.59 to 16.7".
Two years later the neurology position was equipoise rather than preference, and a 2022 to 2026 search located no later society guideline on this question. Note the sentence most often misread: the interval also rules out a clinically important benefit for ASPIRIN. The advisory did not conclude that aspirin is better, only that neither has been shown to be.
AAN Practice Advisory Update Summary: Patent Foramen Ovale and Secondary Stroke Prevention (Messé et al., Neurology 2020;94(20):876-885)
2021AHA/ASA
Closure after cryptogenic stroke
Decide jointlyClass 2a, Level B-R
In patients 18 to 60 years of age with a nonlacunar ischemic stroke of undetermined cause despite a thorough evaluation and a PFO, the decision of percutaneous PFO closure should be made jointly by the patient, the cardiologist, and the neurologist, taking into account the probability of a causal role for the PFO. (Class 2a, Level B-R.)
The stroke indication diverges here, and the recommendation is about HOW the decision is made rather than what to decide: jointly by the patient, the cardiologist and the neurologist. It is scoped tightly, to patients 18 to 60 years of age with a nonlacunar stroke of undetermined cause after a thorough evaluation.
2021 AHA/ASA Secondary Prevention Guideline · PFO closure (Section 5.2.2)
2022SCAI
Closure for migraine
Recommends against itConditional against routine use, moderate certainty
In persons experiencing migraines without a prior PFO-associated stroke, the SCAI guideline panel suggests against the routine use of PFO closure for the treatment of migraine (conditional recommendation, moderate certainty of evidence). Patients, particularly those with debilitating migraines who have failed to benefit from conventional medical therapy, who place a high value on the uncertain benefits … may reasonably choose PFO closure. (Conditional recommendation against, moderate certainty of evidence.)
The migraine indication goes the other way, and it is the interventional cardiology society that says so. Read the scope carefully: the recommendation is against ROUTINE use, and it preserves an explicit carve-out for a patient with debilitating migraine refractory to conventional therapy who places a high value on the uncertain benefit and chooses closure after shared decision-making.
SCAI Guidelines for the Management of Patent Foramen Ovale · PFO closure for the treatment of migraine
2023VA/DoD
Closure for migraine
Recommends against itWeak against
We suggest against patent foramen ovale closure for the treatment or prevention of migraine. (Recommendation 40; Strength: Weak against.)
A headache guideline reaches the same answer independently. This one is worth naming precisely, because it is often cited as finding the evidence insufficient. It did not: that guideline maintains a separate insufficient-evidence category and deliberately did not place PFO closure in it, so this is an active recommendation against.
VA/DoD Clinical Practice Guideline for the Management of Headache, Version 2.0 · Recommendation 40 (PFO closure for migraine)
2025Asia-Pacific panel
Closure for migraine
Recommends against itExpert consensus, ungraded
PFO closure should not be considered as a first-line/routine treatment for migraine. Patients undergoing PFO closure for other indications may experience a reduction in migraine symptoms. Rationale: complete resolution of migraine was only observed in observational studies, not RCTs … These findings provide insufficient support for PFO closure as a standalone treatment for migraine. (Expert consensus; Abbott-funded summit, no industry involvement in content.)
The most recent word, and the one that is hardest to dismiss. The summit was funded by the manufacturer of the device used in PRIMA and PREMIUM, and the panel still declined to endorse closure as a first-line or routine migraine treatment.
An expert consensus statement, not a graded practice guideline. It carries no class, level or certainty rating.
Asian-Pacific Expert Opinion on Patent Foramen Ovale Closure: Review of New Evidence and Focused Update · PFO closure for migraine

What to carry away

For stroke, in a patient 18 to 60 with a nonlacunar stroke and no other cause found, closure is a decision to make jointly with cardiology and the patient. If it is not closed, the neurology position is that antiplatelet and anticoagulation are equally acceptable. For migraine the direction reverses: SCAI 2022 and VA/DoD 2023 both recommend against, as does a 2025 Asian-Pacific expert consensus that is not a graded guideline, and the SCAI recommendation is against ROUTINE use rather than an absolute bar. No AHA/ASA or AAN position on closure for migraine was located, which is not the same as knowing there is none.

Score this at the bedside

RoPE Score

PFO-attributable fraction in cryptogenic stroke.