SoSTART
(2021)Objective
To establish whether starting oral anticoagulation is non-inferior to avoiding oral anticoagulation for survivors of spontaneous intracranial haemorrhage who have atrial fibrillation
Study Summary
• Weak evidence suggested starting OAC might be superior for preventing composite major vascular events (12% vs 24%, HR 0.51, p=0.060)
• Trial was inconclusive due to lower-than-expected event rates; further adequately powered trials are needed
Intervention
Starting full-dose oral anticoagulation (DOAC or VKA) vs avoiding oral anticoagulation (antiplatelet or no antithrombotic) for atrial fibrillation after spontaneous intracranial haemorrhage
Inclusion Criteria
Adults ≥18 years who survived ≥24h after spontaneous intracranial haemorrhage, with atrial fibrillation or flutter and CHA2DS2-VASc score ≥2, where clinician and patient were uncertain about starting OAC
Study Design
Arms: Start oral anticoagulation vs Avoid oral anticoagulation
Patients per Arm: 101 start vs 102 avoid
Outcome
• Any major vascular event: 12% (start) vs 24% (avoid), HR 0.51 (95% CI 0.26–1.03, p=0.060)
• Any stroke: 11% (start) vs 22% (avoid), HR 0.53 (95% CI 0.25–1.09, p=0.084)
Bottom Line
The trial was inconclusive for non-inferiority of starting vs avoiding oral anticoagulation for recurrent intracranial haemorrhage. Event rates were lower than expected, limiting precision. However, weak evidence from composite secondary outcomes suggested starting OAC might be superior for preventing major vascular events, warranting further definitive trials.
Major Points
- SoSTART is the largest published RCT of OAC for AF after intracranial haemorrhage (n=203)
- Non-inferiority was not demonstrated: 8% (start) vs 4% (avoid) had recurrent ICH (adjusted HR 2.42, 95% CI 0.72–8.09, p=0.152)
- 7 of 8 ICH recurrences in the start group were fatal vs 0 of 4 in the avoid group
- Composite secondary outcomes showed a non-significant trend favoring starting OAC: any major vascular event 12% vs 24% (HR 0.51, p=0.060), any stroke 11% vs 22% (HR 0.53, p=0.084)
- More total deaths in the start group (22 vs 11), largely driven by non-cardiovascular causes (14 vs 9)
- 96% of participants were intended to receive a DOAC if allocated to start OAC; 59% prespecified apixaban
- 84% of participants had used OAC before their qualifying intracranial haemorrhage
- Median time from ICH to randomisation was 115 days (IQR 49–265)
- Adherence to allocated treatment was high (98% at discharge, 96% at 1 and 2 years)
- Trial established feasibility for a definitive 6-year, 60-site, 800-patient main phase trial
Study Design
- Study Type
- Prospective, randomised, open-label, assessor-masked, parallel-group, pilot-phase, non-inferiority trial
- Randomization
- Yes
- Blinding
- Open-label treatment allocation known to participants, clinicians, and local investigators. Outcome event adjudicator was masked to participant identity, treatment allocation, and drug use. Brain imaging assessors were masked to treatment allocation.
- Sample Size
- 203
- Follow-up
- Median 1.2 years (IQR 0.97–1.95); minimum 1 year after randomisation
- Centers
- 67
- Countries
- UK
Primary Outcome
Definition: Recurrent symptomatic spontaneous intracranial haemorrhage (adjudicated by assessor masked to treatment allocation)
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| 4/102 (4%) | 8/101 (8%) | 2.42 (0.72–8.09) | 0.152 |
Limitations & Criticisms
- Primary outcome event rates were lower than assumed in the sample size calculation, reducing precision of the effect estimate
- Pilot-phase trial underpowered for definitive conclusions (n=203 vs planned definitive trial of 800)
- Non-inferiority margin of 12% (HR 3.2) was wide
- Open-label design, though outcomes were adjudicated by masked assessor
- More non-cardiovascular deaths in the start group (14 vs 9), representing a competing risk that may have reduced observed ICH recurrence in the start group
- 42% of approached patients declined participation, raising potential selection bias
- Women were under-represented in the trial
- 93% of participants were White, limiting generalisability to other ethnic groups
- All sites were in UK state-funded healthcare, limiting international generalisability
- Only 29% of eligible patients were recruited in the feasibility phase
- Time to OAC initiation after ICH was longer than in observational cohort studies
- Model non-convergence due to low event numbers prevented adjustment for all 6 minimisation variables (only 2 of 6 included)
- Use of antiplatelet agents in the avoid group could confound interpretation
Citation
Lancet Neurol 2021; 20: 842–853