← Back
NeuroTrials.ai
Neurology Clinical Trial Database

COCROACH

Effects of oral anticoagulation in people with atrial fibrillation after spontaneous intracranial haemorrhage (COCROACH): prospective, individual participant data meta-analysis of randomised trials

Year of Publication: 2023

Authors: Rustam Al-Shahi Salman, Jacqueline Stephen, Jayne F Tierney, ..., H Bart van der Worp

Journal: The Lancet Neurology

Citation: Lancet Neurol 2023; 22: 1140–49

Link: https://doi.org/10.1016/S1474-4422(23)00315-0


Clinical Question

What are the effects of starting versus avoiding oral anticoagulation on stroke and cardiovascular death in people with atrial fibrillation and spontaneous intracranial haemorrhage, and do effects vary by subgroup (age, sex, ICH location, timing, CHA2DS2-VASc score)?

Bottom Line

Oral anticoagulation (99% DOACs) significantly reduced ischaemic major adverse cardiovascular events (HR 0.27) in ICH survivors with atrial fibrillation, but had uncertain effects on the composite of any stroke or cardiovascular death (HR 0.68, not significant), haemorrhagic events, and functional outcome. No significant subgroup interactions were found, including by ICH location. The absolute ischaemic benefit appeared to exceed the potential haemorrhagic risk increase, but net benefit remains unconfirmed pending completion of 5 ongoing trials.

Major Points

  • Prospective IPD meta-analysis of 412 participants from 4 completed RCTs (SoSTART, APACHE-AF, NASPAF-ICH, ELDERCARE-AF)
  • 99% of participants assigned to start OAC received a DOAC (apixaban 130, edoxaban 61, dabigatran 12, rivaroxaban 6); only 3 received warfarin
  • 33% of the avoid group received antiplatelet monotherapy; 67% received no antithrombotic therapy
  • Primary outcome (any stroke or cardiovascular death): 14% vs 22%, pooled HR 0.68 (95% CI 0.42–1.10, p=0.12, I²=0%)
  • Ischaemic MACE significantly reduced: 4% vs 19%, pooled HR 0.27 (95% CI 0.13–0.56, p=0.0004, I²=0%)
  • Ischaemic stroke specifically reduced: 4% vs 16%, pooled HR 0.37 (95% CI 0.17–0.83, I²=0%)
  • Haemorrhagic MACE not significantly increased: 7% vs 5%, pooled HR 1.80 (95% CI 0.77–4.21, I²=0%)
  • All-cause death numerically higher with OAC: 18% vs 15%, pooled HR 1.29 (95% CI 0.78–2.11, I²=50%)
  • No significant subgroup interaction by ICH location (lobar vs non-lobar), age, sex, timing, or CHA2DS2-VASc score
  • ENRICH-AF Data Monitoring Committee stopped enrolment of lobar ICH and convexity SAH participants due to excess haemorrhagic stroke risk with edoxaban, contrasting with COCROACH findings
  • Five ongoing trials (STATICH, A3-ICH, ASPIRE, ENRICH-AF, PRESTIGE-AF) may add ~2000 participants by 2028
  • Death or dependence at 1 year was similar: 53% vs 51%, pooled OR 1.12 (95% CI 0.70–1.79, I²=0%)

Design

Study Type: Prospective individual participant data meta-analysis of randomised controlled trials

Randomization: 1

Blinding: Three of four included trials were open-label (SoSTART, APACHE-AF, NASPAF-ICH); ELDERCARE-AF was placebo-controlled. All trials had independent outcome adjudication blinded to treatment assignment.

Enrollment Period: Included trials recruited between 2016 and 2020

Follow-up Duration: Maximum 2–6 years across included trials

Centers: 67

Countries: UK, Netherlands, Canada, Japan

Sample Size: 412

Analysis: Two-stage IPD meta-analysis. Cox regression models for time-to-event outcomes within each trial, then pooled using fixed-effect inverse-variance model. Firth's method used when zero outcomes in one treatment group. Sensitivity analyses with random-effects models. Subgroup interactions assessed using the deft approach. SAS version 9.4 and STATA version 16.0.


Inclusion Criteria

  • Participants from registered randomised trials
  • Spontaneous intracranial haemorrhage (ICH, IVH, SAH, or SDH)
  • Atrial fibrillation
  • Randomly assigned to start long-term oral anticoagulation vs avoid oral anticoagulation
  • Trials eligible regardless of setting, epoch, sample size, follow-up, publication status, or language
  • Subgroups of eligible participants from broader trials could be included
  • Participants who had not opted out of data sharing

Exclusion Criteria

  • No exclusion criteria for trials
  • Participants who opted out of data sharing were excluded (n=2 from SoSTART)

Arms

FieldStart oral anticoagulationControl
InterventionLong-term oral anticoagulation, predominantly DOACs (apixaban, edoxaban, dabigatran, rivaroxaban) at full dose except ELDERCARE-AF which used low-dose edoxaban 15 mg daily. Three participants received warfarin.No antithrombotic therapy (67%) or antiplatelet monotherapy (33%). ELDERCARE-AF used placebo; other three trials were open label.
DurationLong-term (varied by trial, maximum follow-up 2–6 years)Long-term (varied by trial)

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Any stroke (non-fatal symptomatic ischaemic stroke, haemorrhagic stroke, or stroke of unknown type) or cardiovascular death (death within 30 days of stroke, other ICH, extracranial haemorrhage, MI, or systemic embolism; sudden cardiac death; death from another vascular cause; or death of unknown cause)Primary43/200 (22%)29/212 (14%)0.680.12
Ischaemic major adverse cardiovascular events (ischaemic stroke, systemic arterial embolism, PE, or MI)Secondary38/200 (19%)9/212 (4%)0.270.0004
Ischaemic strokeSecondary31/200 (16%)9/212 (4%)0.37
Pulmonary embolismSecondary4/200 (2%)0/212 (0%)
Myocardial infarctionSecondary4/200 (2%)0/212 (0%)
Haemorrhagic major adverse cardiovascular events (spontaneous ICH or major extracranial haemorrhage)Secondary9/200 (5%)15/212 (7%)1.80.17
Intracranial haemorrhageSecondary5/200 (3%)12/212 (6%)
Major extracranial haemorrhageSecondary4/200 (2%)3/212 (1%)
Death from any causeSecondary29/200 (15%)38/212 (18%)1.290.32
Cardiovascular deathSecondary12/200 (6%)17/212 (8%)
Death from any other causeSecondary17/200 (9%)21/212 (10%)
Death or dependence (mRS 3–6) at 1 yearSecondary74/145 (51%)78/147 (53%)1.12

Subgroup Analysis

Subgroup analyses were performed on the primary outcome using data from SoSTART and APACHE-AF (302 participants). No significant treatment effect modification was found by: age (<75 vs ≥75, overall p-interaction=0.42), sex (male vs female, p-interaction=0.34), ICH location (lobar vs non-lobar, p-interaction=0.98), time from ICH to randomisation (<8 weeks vs ≥8 weeks, p-interaction=0.97), or CHA2DS2-VASc score (≤4 vs >4, p-interaction=0.10). A post-hoc suggestion of possible greater benefit in women vs men was noted. Post-hoc sensitivity analyses excluding ELDERCARE-AF (HR 0.72, 95% CI 0.43–1.20), restricting to confirmed ICH (HR 0.78, 95% CI 0.45–1.33), and restricting to fatal outcomes (HR 1.32, 95% CI 0.60–2.90, showing qualitatively different direction) were performed.


Criticisms

  • Small total sample size (412 participants) insufficient to provide definitive evidence of superiority, non-inferiority, or heterogeneity in subgroups
  • Three of four trials were open-label, conferring risk of performance bias
  • Baseline covariates for ELDERCARE-AF were not fully available due to deidentification policy (sex, ethnicity removed)
  • ELDERCARE-AF used low-dose edoxaban (15 mg daily) rather than full therapeutic dose, differing from the other 3 trials
  • 33% of the control group received antiplatelet monotherapy, which may have attenuated effect sizes
  • Incomplete adherence to OAC in open-label trials may have attenuated treatment effects
  • Under-representation of Black and non-White ethnic groups (~76% White overall)
  • No as-treated or per-protocol analyses performed
  • Contrasting findings with ENRICH-AF safety signal in lobar ICH/convexity SAH raise uncertainty about ICH location effects
  • Direction of effect reversed for fatal outcomes (HR 1.32) vs overall primary composite (HR 0.68), raising concern about severity of events in the OAC group
  • Moderate heterogeneity (I²=50%) observed for all-cause death outcome
  • Two participants from SoSTART opted out of data sharing
  • Only 4 trials completed; 5 ongoing trials with ~2000 additional participants needed for definitive conclusions

Funding

British Heart Foundation (CS/18/2/33719)

Based on: COCROACH (The Lancet Neurology, 2023)

Authors: Rustam Al-Shahi Salman, Jacqueline Stephen, Jayne F Tierney, ..., H Bart van der Worp

Citation: Lancet Neurol 2023; 22: 1140–49

Content summarized and formatted by NeuroTrials.ai.