PDF: CLOSURE I
(2012)Objective
To determine whether closure of a patent foramen ovale (PFO) with the STARFlex device is superior to medical therapy in preventing recurrent stroke or TIA in patients with cryptogenic stroke or TIA.
Study Summary
• PFO closure with the STARFlex device did not significantly reduce the risk of recurrent stroke or TIA compared to medical therapy alone over a 2-year follow-up, and was associated with procedure-related complications..
Intervention
STARFlex device closure with clopidogrel (75 mg/day for 6 months) + aspirin (81–325 mg/day for 2 years) vs. medical therapy (warfarin, aspirin, or both at investigator’s discretion).
Inclusion Criteria
Patients aged 18–60 with cryptogenic ischemic stroke or TIA within 6 months and a PFO confirmed by TEE with right-to-left shunting on Valsalva. Excluded if other causes of stroke (e.g., AF, significant carotid disease, LV dysfunction) were present.
Study Design
Arms: PFO Closure with STARFlex vs. Medical Therapy Alone
Patients per Arm: Closure: 447, Medical Therapy: 462
Outcome
• Primary endpoint (stroke/TIA/death at 2 years): 5.5% (closure) vs 6.8% (medical), HR 0.78 (95% CI 0.45–1.35), P=0.37. <br>• Stroke: 2.9% vs 3.1% (P=0.79), TIA: 3.1% vs 4.1% (P=0.44). <br>• No neurologic deaths. <br>• Procedural success in 89%, with effective closure in 86% at 6 months. <br>• Device-related complications included atrial fibrillation and device-associated thrombus..
Bottom Line
Device closure of a PFO did not significantly reduce the risk of recurrent stroke or TIA compared to medical therapy in patients with cryptogenic stroke or TIA.
Major Points
- CLOSURE I was the FIRST randomized trial of PFO closure for cryptogenic stroke — and it was NEGATIVE (HR 0.78, p=0.37). This result delayed widespread PFO closure by 5 years until RESPECT long-term, CLOSE, and DEFENSE-PFO finally showed benefit in 2017.
- 909 patients aged 18–60 with cryptogenic stroke/TIA and PFO on bubble TEE, randomized at 87 US/Canadian centers. Primary composite (stroke, TIA, death within 30d, neurologic death 31d–2yr): 5.5% closure vs 6.8% medical (HR 0.78, 95% CI 0.45–1.35, p=0.37).
- Stroke alone was nearly identical: 2.9% closure vs 3.1% medical (HR 0.90, p=0.79) — the result that seemed to close the door on PFO closure.
- Device-related atrial fibrillation was the major safety concern: 5.7% closure vs 0.7% medical (p<0.001) — an 8-fold increase that offset any potential stroke prevention benefit.
- The STARFlex device (NMT Medical) used in CLOSURE I was a DIFFERENT closure device than the Amplatzer PFO Occluder used in RESPECT and CLOSE — the STARFlex had higher residual shunt rates and is no longer available, raising questions about whether device choice drove the negative result.
- Medical therapy was heterogeneous: aspirin alone, warfarin alone, or both — reflecting clinical equipoise about optimal medical management for PFO-associated stroke. Later trials required antiplatelet therapy specifically.
- 27.4% of patients had TIA (not stroke) as the qualifying event — inclusion of TIA patients (who have very low recurrence rates) diluted the population and reduced event rates below what was needed for statistical power.
- Short 2-year follow-up — later trials (RESPECT at 5.9 years) showed that PFO closure benefit emerges gradually over time as recurrent paradoxical embolism accumulates. The 2-year window may have been too short.
- KEY LESSON: CLOSURE I taught the field that patient selection matters — later positive trials selected younger patients with larger shunts, atrial septal aneurysms, and stricter cryptogenic stroke definitions (excluding patients with alternative stroke mechanisms).
- Despite the negative result, CLOSURE I was essential for the field: it informed the design of RESPECT, CLOSE, PC Trial, and DEFENSE-PFO, which collectively established PFO closure as standard of care in properly selected patients by 2017.
Study Design
- Study Type
- Randomized, open-label, multicenter trial
- Randomization
- Yes
- Blinding
- Open-label
- Sample Size
- 909
- Follow-up
- 2 years
- Centers
- 87
- Countries
- United States, Canada
Primary Outcome
Definition: Composite of stroke or TIA, death from any cause within 30 days, or neurologic death between 31 days and 2 years
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 6.8% | 5.5% | 0.78 (0.45–1.35) | 0.37 |
Limitations & Criticisms
- STARFlex device — this specific closure device had higher residual shunt rates and more AF than the Amplatzer PFO Occluder used in later positive trials (RESPECT, CLOSE). Device inadequacy may have driven the negative result.
- Underpowered: 909 patients with only 2-year follow-up and low event rates (~3%/yr) — needed ~2,000+ patients or longer follow-up to detect the modest benefit later confirmed by RESPECT at 5.9 years.
- 27.4% TIA patients included — TIA has much lower recurrence risk than stroke, diluting the at-risk population and reducing statistical power. Later trials excluded or minimized TIA patients.
- Device-related AF in 5.7% was a major concern — the NNH for AF was lower than any potential NNT for stroke prevention, making the risk-benefit unfavorable. Amplatzer device has lower AF rates (~3%).
- Short 2-year follow-up — PFO closure benefit is cumulative over time (paradoxical embolism risk is ongoing). RESPECT showed benefit emerging at 3+ years, with HR 0.55 at 5.9 years.
- Heterogeneous medical therapy (aspirin, warfarin, or both) — lack of standardized medical arm introduces variability. If some medical patients received optimal anticoagulation, closure benefit would be harder to detect.
- Liberal stroke mechanism classification — some 'cryptogenic' events may have had unrecognized alternative causes (AF, small vessel disease), diluting the true PFO-attributable stroke population.
- Open-label design — knowledge of device implantation could influence reporting of neurological symptoms and threshold for imaging, potentially biasing TIA ascertainment.
- No RoPE Score available — the Risk of Paradoxical Embolism score (developed later) identifies patients most likely to benefit from closure. CLOSURE I enrolled an unselected PFO population.
Citation
Furlan AJ, et al. N Engl J Med. 2012;366(11):991–999.