PDF: NAVIGATE ESUS
(2018)Objective
Evaluate whether rivaroxaban is superior to aspirin for secondary prevention in patients with embolic stroke of undetermined source (ESUS).
Study Summary
• Rivaroxaban 15 mg daily did not reduce recurrent stroke or systemic embolism compared to aspirin and led to significantly higher rates of major bleeding. The trial was terminated early for futility and safety concerns.
Intervention
Rivaroxaban 15 mg daily vs. aspirin 100 mg daily in 7,213 patients with recent ESUS. Randomized, double-blind trial at 459 sites across 31 countries. Median follow-up: 11 months.
Study Design
Arms: Array
Outcome
• Primary efficacy: Stroke/systemic embolism — 5.1% vs. 4.8%/year; HR 1.07 (95% CI 0.87–1.33); p=0.52
• Ischemic stroke — 4.7% vs. 4.7%/year; HR 1.01 (95% CI 0.81–1.26)
• Hemorrhagic stroke — 0.4% vs. 0.1%/year; HR 6.50 (95% CI 1.47–28.8)
• Major bleeding (ISTH) — 1.8% vs. 0.7%/year; HR 2.72 (95% CI 1.68–4.39); p<0.001
• Symptomatic intracranial hemorrhage — 0.6% vs. 0.1%/year; HR 4.02 (95% CI 1.51–10.7); p=0.003
• No significant difference in disabling stroke, CV death, MI, or all-cause death
• Ischemic stroke — 4.7% vs. 4.7%/year; HR 1.01 (95% CI 0.81–1.26)
• Hemorrhagic stroke — 0.4% vs. 0.1%/year; HR 6.50 (95% CI 1.47–28.8)
• Major bleeding (ISTH) — 1.8% vs. 0.7%/year; HR 2.72 (95% CI 1.68–4.39); p<0.001
• Symptomatic intracranial hemorrhage — 0.6% vs. 0.1%/year; HR 4.02 (95% CI 1.51–10.7); p=0.003
• No significant difference in disabling stroke, CV death, MI, or all-cause death
Bottom Line
Rivaroxaban was not superior to aspirin in preventing recurrent stroke in ESUS patients and led to significantly more major bleeding, leading to early termination of the trial.
Major Points
- NAVIGATE ESUS was the first large randomized trial testing anticoagulation for Embolic Stroke of Undetermined Source (ESUS), a concept proposed by Hart et al. in 2014 as a therapeutic target within cryptogenic stroke — hypothesizing that most ESUS was cardioembolic and would respond to anticoagulation.
- 7,213 patients across 459 centers in 31 countries — the largest ESUS-specific trial. Used rivaroxaban 15mg once daily (lower dose than AF trials' 20mg) to balance efficacy vs bleeding in a population without confirmed AF.
- Trial stopped early by the data safety monitoring board after a median of only 11 months follow-up (range 1–33; IQR 5–17) due to futility (no efficacy signal) and safety concerns (excess bleeding in rivaroxaban arm).
- Primary endpoint (recurrent stroke or systemic embolism): 5.1%/yr rivaroxaban vs 4.8%/yr aspirin (HR 1.07, 95% CI 0.87–1.33, p=0.52) — completely neutral, with point estimate favoring aspirin. No trend toward benefit at any time point.
- Major bleeding was significantly higher with rivaroxaban: 1.8% vs 0.7%/yr (HR 2.72, 95% CI 1.68–4.39, P<0.001). Life-threatening or fatal bleeding was also more frequent (35 vs 15 events; HR 2.34, 95% CI 1.28–4.29, P=0.004) — a striking safety signal that contributed to early termination.
- Symptomatic intracranial hemorrhage was markedly increased with rivaroxaban (20 vs 5 events; HR 4.02, 95% CI 1.51–10.7; P=0.003), driven predominantly by intracerebral hemorrhage (12 vs 3; HR 4.01, 95% CI 1.13–14.2; P=0.02).
- ESUS definition required: (1) non-lacunar brain infarct on CT/MRI, (2) no extracranial or intracranial atherosclerosis >50% stenosis in the artery supplying the ischemic area, (3) no major cardioembolic source (≥20 hours of cardiac rhythm monitoring required to rule out atrial fibrillation lasting ≥6 minutes; no mechanical prosthetic valve, LV thrombus, or severe mitral stenosis), (4) no other identified specific cause — yet this broad definition likely captured heterogeneous mechanisms.
- Only 34% of patients had cardiac monitoring for ≥48 hours before enrollment (median 24h, IQR 24–48) — meaning occult paroxysmal AF was likely present in a substantial proportion. AF was ultimately identified during follow-up in 3% of participants at a median of 5 months after entry.
- Together with RE-SPECT ESUS (dabigatran, also negative in 2019), definitively closed the door on empiric anticoagulation for unselected ESUS — shifting the field toward ESUS subtyping: PFO closure for shunt-related ESUS, prolonged monitoring for AF detection, and targeted therapy based on mechanism.
- Post-hoc analyses (published separately) suggested patients with left atrial enlargement or elevated NT-proBNP (markers of atrial cardiopathy) may have had a trend toward benefit — spawning the ARCADIA trial testing apixaban specifically in ESUS with atrial cardiopathy biomarkers. (Not part of the primary NEJM publication.)
Study Design
- Study Type
- Phase 3, international, randomized, double-blind, event-driven trial
- Randomization
- Yes
- Blinding
- Double-blind (participants and investigators)
- Sample Size
- 7213
- Follow-up
- Median follow-up of 11 months (range 1–33; IQR 5–17)
- Centers
- 459
- Countries
- 31 countries globally
Primary Outcome
Definition: Time to first recurrent ischemic, hemorrhagic, or undefined stroke, or systemic embolism (time-to-event analysis)
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 4.8% annual rate (160/3604) | 5.1% annual rate (172/3609) | 1.07 (0.87–1.33) | 0.52 |
Limitations & Criticisms
- Stopped early at median 11 months — may have missed late-emerging benefit, though the point estimate (HR 1.07) favored aspirin, making late reversal implausible.
- ESUS is a heterogeneous 'wastebasket' diagnosis — lumping PFO-related, atrial cardiopathy, cancer-associated, and aortic arch disease into one category may have diluted a real treatment effect in a subgroup that would benefit from anticoagulation.
- Only 34% of participants had cardiac rhythm monitoring for ≥48 hours before enrollment (median 24h, IQR 24–48) — inadequate in many patients to exclude paroxysmal AF. CRYSTAL AF showed 30% AF detection at 3 years with implantable monitors, meaning occult AF was likely present in some enrollees.
- Used 15 mg rivaroxaban (not the AF dose of 20 mg) — potentially subtherapeutic for truly cardioembolic ESUS, though higher bleeding without efficacy suggests dose was not the issue.
- No renal dose adjustment was permitted (unlike ROCKET AF's 15 mg for CrCl 30–49) — patients with CrCl <30 were excluded entirely.
- No imaging-based stratification: cortical infarcts (likely embolic) vs deep/subcortical infarcts (possibly small vessel) were treated identically, though different mechanisms may warrant different therapies.
- Broad enrollment window (7 days to 6 months post-stroke) — very early randomization may have included patients whose stroke mechanism would have been identified with more time.
- Industry-sponsored (Bayer/Janssen) — same sponsors as ROCKET AF, raising questions about commercial interest in expanding rivaroxaban's indications.
- Life-threatening or fatal bleeding signal was concerning — 35 vs 15 events (HR 2.34) underscores the harm of empiric anticoagulation in a non-AF population.
Citation
Hart RG, Sharma M, Mundl H, et al. Rivaroxaban for Stroke Prevention after Embolic Stroke of Undetermined Source. N Engl J Med. 2018;378(23):2191–2201.