Cryptogenic Stroke and High-Risk Patent Foramen Ovale: The DEFENSE-PFO Trial

Authors, Journal, Affiliations, Type, DOI

• Authors: Pil Hyung Lee and Jae-Kwan Song (senior author, Asan Medical Center, Seoul), with Jong S. Kim, Ran Heo, Sahmin Lee, Dae-Hee Kim, Jong-Min Song, Duk-Hyun Kang, Sun U. Kwon, Dong-Wha Kang, Dongwhane Lee and Hyuk Sung Kwon (neurology), Sung-Cheol Yun (biostatistics), Byung Joo Sun, Jae-Hyeong Park, Jae-Hwan Lee, Hye Seon Jeong, Hee-Jung Song and Jei Kim (Chungnam National University), Seung-Jung Park
• Affiliations: Asan Medical Center, University of Ulsan College of Medicine, Seoul, and Chungnam National University Hospital, Daejeon, South Korea
• Journal: Journal of the American College of Cardiology. 2018;71(20):2335–2342; received 11 Feb 2018, accepted 27 Feb 2018
• Type: Investigator-initiated, multicentre, randomised, open-label, superiority trial at 2 sites; September 2011 – October 2017; NCT01550588; terminated early
• DOI: https://doi.org/10.1016/j.jacc.2018.02.046

Overview

The trial that asked whether PFO anatomy should decide who gets closed. Rather than enrolling everyone with a cryptogenic stroke and a PFO — the design that had produced three negative trials — DEFENSE-PFO screened 1,715 patients by transoesophageal echocardiography and randomised only those meeting a morphological definition of high-risk: atrial septal aneurysm, septal hypermobility, or a PFO ≥2 mm.

Sixty patients per arm, mean age 51.8 years. Every primary endpoint event occurred in the medication-only group — 0 of 60 versus 6 of 60 (2-year event rate 12.9%, log-rank P=0.013), with ischaemic stroke 0 versus 5 (10.5%, P=0.023) and a number needed to treat of 10 at two years. No patient randomised to closure had a stroke.

The trial's importance is not its size — it is small, stopped early, and open-label — but its demonstration that morphology is a treatment modifier, which is the empirical basis for the anatomy half of the PASCAL classification and the reason ESO 2024 can specify that patients with a large shunt and an aneurysm derive most benefit.

Keywords

Cryptogenic stroke · echocardiography · patent foramen ovale · atrial septal aneurysm · device closure

Key Takeaways

The high-risk PFO definition

• Patients were eligible with an ischaemic stroke within the previous 6 months and no identifiable cause other than a high-risk PFO with right-to-left shunt.
• Cryptogenic status was established stringently: CTA/MRA/ultrasound excluding ≥50% stenosis or occlusion of any major vessel; exclusion of small-vessel disease (<1.5 cm deep infarct or lacunar syndrome); Holter or prolonged monitoring to exclude paroxysmal AF; exclusion of hypercoagulable states requiring anticoagulation and of arterial dissection.

The screening cascade

• 1,715 patients underwent TEE for evaluation of a cardiac embolic source.
• 1,265 had no PFO; 450 had a PFO; 175 (38.9%) met the high-risk definition.
• 50 declined and 5 met an exclusion criterion → 120 randomised.
• Nearly four in ten cryptogenic-stroke patients with a PFO had high-risk morphology — the practical yield of doing TEE rather than TTE in this population.

Results

• NNT 10 at 2 years to prevent one stroke.
• The events in the medication arm were not all strokes and not all PFO-related: 5 ischaemic strokes, 1 cerebral haemorrhage, 2 TIMI major bleeds and 1 TIA. The major bleeds were one spontaneous intracerebral haemorrhage on warfarin and one haemorrhagic transformation on dual antiplatelet therapy — two of the six primary events were caused by the treatment, not by the PFO.
• Of the 5 recurrent strokes, 4 occurred in patients with an aneurysm or hypermobility, and 3 were in a vascular territory different from the index lesion — consistent with, but not proof of, a continuing embolic source.
• Silent infarction on 6-month MRI did not differ significantly (8.8% vs 18.4%, P=0.24) in the 72 patients who had the scan.

Procedural outcomes and crossovers

• Closure was successful in all 60 assigned patients, using the Amplatzer PFO Occluder; residual shunt was <10 microbubbles in all but 4 the next day, and the one severe residual shunt had resolved by 2 months.
• Nonfatal complications: AF in 2, pericardial effusion 1, puncture-site pseudoaneurysm 1.
• 7 of 60 randomised to closure declined the procedure; 4 of 60 in the medication arm underwent closure during follow-up — 109 patients in the per-protocol analysis, in which the primary endpoint still occurred only in the medication group (log-rank P=0.016).

Why the trial stopped early

• Target enrolment was 210 patients (105 per arm) for 80% power to detect 4% versus 15% at 24 months. Recruitment was slower than expected.
• The publication of CLOSE — which used similar morphological selection criteria and showed an HR of 0.03 — prompted the trial's neurologists to raise safety concerns, and the steering committee stopped enrolment in November 2017 without knowledge of the event rates in their own trial. Follow-up of enrolled patients continued.

Limitations of the document

• Stopped early, underpowered for the hazard ratio, and 120 of a planned 210 patients. The authors state that this precludes a reliable effect estimate — the p-value is driven by an event count of six against zero, not by a stable rate.
• Open-label with no sham procedure, and the medication arm was not standardised: warfarin was used in 25% of the medication group and continued in over 20% at 12 months, so the comparator was not "antiplatelet therapy" but "investigator-chosen antithrombotic therapy." This makes DEFENSE-PFO closer in design to the earlier negative trials than to CLOSE.
• Two centres in one country, with potential selection bias acknowledged by the authors.
• The primary composite mixes efficacy and safety. Two of the six events in the medication arm were bleeding complications of treatment — including one on warfarin, which the closure arm would not have received. Including them in the primary endpoint inflates the apparent benefit.
• The morphology criterion is inconsistently stated within the paper. The Methods and Abstract define PFO size as ≥2 mm during Valsalva; the Figure 1 legend describes a high-risk PFO as ">3 mm." Baseline mean PFO size was 3.2 ± 1.5 mm, so a substantial minority would not have met a 3 mm threshold.
• The composite endpoint was not powered for silent infarction, which did not differ significantly and could not exclude a modest effect.
• No long-term follow-up beyond a median of 2.8 years, in a population whose expected remaining lifespan is 30–40 years with a device in situ.
• Does not address the modern question. Warfarin was permitted in the control arm; contemporary practice would compare closure with a DOAC, which no closure trial has done.

Key Concepts Mentioned

concepts/Patent-Foramen-Ovale · concepts/PFO-Closure-Selection · concepts/PFO-Detection-and-Anatomy

Key Entities Mentioned

entities/Ischemic-Stroke · concepts/Atrial-Septal-Defect

Named but unlinked: Amplatzer PFO Occluder · RoPE · PASCAL · CLOSE · TIMI bleeding criteria · atrial septal aneurysm · septal hypermobility

Wiki Pages Updated

• sources/pfo-jacc-2018 — created
• concepts/Patent-Foramen-Ovale — created
• concepts/PFO-Detection-and-Anatomy — created
• concepts/PFO-Closure-Selection — created
• entities/Ischemic-Stroke — high-risk morphology and stroke recurrence
• wiki/wikiindex · wiki/sourceindex