PDF: ELAN
(2023)Objective
To estimate the safety and efficacy of early versus later initiation of direct oral anticoagulants (DOACs) in patients with acute ischemic stroke and atrial fibrillation, using imaging-based stroke severity classification.
Study Summary
• Recurrent ischemic stroke was numerically lower with early treatment (1.4% vs 2.5% at 30 days; 1.9% vs 3.1% at 90 days) with no increase in symptomatic ICH (0.2% in both groups).
• Findings support that early DOAC initiation is a reasonable approach with no excess bleeding risk when guided by imaging-based stroke severity classification.
Intervention
Early DOAC initiation (within 48h for minor/moderate stroke; day 6-7 for major stroke) vs. later DOAC initiation (day 3-4 for minor, day 6-7 for moderate, day 12-14 for major stroke)
Inclusion Criteria
Acute ischemic stroke confirmed on CT/MRI with nonvalvular atrial fibrillation (permanent, persistent, paroxysmal, or newly diagnosed), stroke severity classified by imaging (minor ≤1.5 cm, moderate = cortical branch territory, major = larger territorial or brainstem/cerebellar >1.5 cm), no therapeutic anticoagulation at onset, no parenchymal hematoma at randomization
Study Design
Arms: Early DOAC initiation vs. later DOAC initiation (1-3-6-12 day rule)
Patients per Arm: 2013 total (Early: 1006, Later: 1007)
Outcome
• Recurrent ischemic stroke at 30d: 1.4% vs 2.5% (OR 0.57, 95% CI 0.29-1.07); at 90d: 1.9% vs 3.1% (OR 0.60, 95% CI 0.33-1.06). sICH: 0.2% in both groups at 30 and 90 days.
• Exploratory 90-day composite: 3.7% vs 5.6% (OR 0.65, 95% CI 0.42-0.99). Functional independence (mRS 0-2) similar: 62.6% vs 62.6% at 30d; 66.6% vs 65.8% at 90d.
Bottom Line
Early DOAC initiation showed a trend toward fewer composite events (2.9% vs 4.1%) at 30 days with no increase in symptomatic intracranial hemorrhage (0.2% in both groups). Recurrent ischemic stroke was numerically lower with early treatment. No formal hypothesis was tested; the trial was designed to estimate treatment effects and their precision. Results support that early DOAC initiation is reasonable when guided by imaging-based severity classification.
Major Points
- Trial designed to estimate treatment effects and precision -- no superiority or noninferiority hypothesis was tested.
- Primary composite outcome at 30 days: 2.9% early vs 4.1% later (OR 0.70, 95% CI 0.44-1.14; risk difference -1.18 percentage points, 95% CI -2.84 to 0.47).
- Recurrent ischemic stroke at 30 days: 1.4% vs 2.5% (OR 0.57, 95% CI 0.29-1.07); at 90 days: 1.9% vs 3.1% (OR 0.60, 95% CI 0.33-1.06).
- Symptomatic ICH was 0.2% in both groups at both 30 and 90 days -- no excess bleeding risk with early initiation.
- Major extracranial bleeding at 30 days: 0.3% early vs 0.5% later (OR 0.63, 95% CI 0.15-2.38).
- Systemic embolism at 30 days: 0.4% early vs 0.9% later (OR 0.48, 95% CI 0.14-1.42).
- Exploratory 90-day composite: 3.7% early vs 5.6% later (OR 0.65, 95% CI 0.42-0.99).
- Post hoc analysis: 98% probability that early treatment increases primary-outcome risk by no more than 0.5 percentage points.
- Imaging-based stroke severity classification used: minor (≤1.5 cm), moderate (cortical superficial branch territory), major (larger territorial or brainstem/cerebellar >1.5 cm).
- 37% had minor stroke, 40% moderate, 23% major. Median NIHSS at randomization was 3.
- Approximately 50% were receiving aspirin at screening. Thrombolysis given in ~39%, thrombectomy in ~22%.
- Functional independence (mRS 0-2) similar between groups at 30 and 90 days (~63% and ~66%).
- Any SAE by 90 days: 13.9% early vs 15.8% later.
- No apparent heterogeneity across prespecified subgroups (age, infarct size, NIHSS), though trial not powered for subgroup analysis.
Study Design
- Study Type
- International, multicenter, randomized, open-label, assessor-blinded trial (estimation framework -- no hypothesis testing)
- Randomization
- Yes
- Blinding
- Open-label with blinded outcome assessment. Telephone assessors unaware of group assignment. Independent clinical events committee blinded to assignment adjudicated all events. Stroke severity classification not centrally adjudicated.
- Sample Size
- 2013
- Follow-up
- 90 days
- Centers
- 103
- Countries
- Switzerland, Germany, Austria, United Kingdom, Ireland, Belgium, France, Norway, Sweden, Slovakia, Portugal, Italy, Greece, Israel, India, Japan
Primary Outcome
Definition: Composite of recurrent ischemic stroke, systemic embolism, major extracranial bleeding, symptomatic intracranial hemorrhage, or vascular death within 30 days after randomization
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 41/1007 (4.1%) | 29/1006 (2.9%) | - (0.44-1.14 (OR); risk difference 95% CI -2.84 to 0.47) |
Limitations & Criticisms
- No formal hypothesis testing (no superiority or noninferiority design) -- the trial provides estimates of treatment effect and precision but cannot confirm or refute a definitive benefit.
- Low median NIHSS at randomization (3), suggesting a relatively mild stroke population, which limits applicability to more severe strokes.
- Only 23% had major strokes; early treatment for major stroke was defined as day 6-7 (same as the later group for moderate stroke), so the 'early' window for major strokes was not truly ultra-early.
- Patients with parenchymal hematoma (Heidelberg PH-1 or PH-2) at randomization were excluded, so safety of early anticoagulation in this higher-risk group remains unknown.
- Stroke severity classified by site investigators without central adjudication, introducing potential misclassification.
- Predominantly European population (>85%); limited generalizability to non-White populations. No data on race/ethnicity collected.
- Patients already on therapeutic anticoagulation at stroke onset were excluded.
- Treatment compliance was imperfect: 949/1006 (94.3%) in early group and 935/1007 (92.8%) in later group received treatment per protocol.
- One site (13 patients) excluded for GCP nonadherence.
- Approximately 50% were on aspirin at screening, which may have reduced early recurrence risk in the later-treatment group.
- Open-label design with potential performance bias, though outcome assessment was blinded.
- Confidence intervals for all outcomes overlap null -- individual component analyses are underpowered.
Citation
N Engl J Med 2023;388:2411-21. DOI: 10.1056/NEJMoa2303048