← Back
NeuroTrials.ai
Neurology Clinical Trial Database

CATALYST

Collaboration on the optimal timing of anticoagulation after ischaemic stroke and atrial fibrillation: a systematic review and prospective individual participant data meta-analysis of randomised controlled trials (CATALYST)

Share Share Copied! Compare

Year of Publication: 2025

Authors: Hakim-Moulay Dehbi, Urs Fischer, Signild Åsberg, ..., David J Werring

Journal: The Lancet

Citation: Lancet 2025; 406: 43–51. Published online June 23, 2025. https://doi.org/10.1016/S0140-6736(25)00439-8

Link: https://doi.org/10.1016/S0140-6736(25)00439-8

PDF: https://www.thelancet.com/action/showPdf...%2825%2900439-8

Clinical Question

In patients with acute ischaemic stroke and atrial fibrillation, does early DOAC initiation (within 4 days of stroke onset) reduce the composite of recurrent ischaemic stroke, symptomatic intracerebral haemorrhage, or unclassified stroke at 30 days compared to later initiation (≥5 days)?

Bottom Line

Early DOAC initiation (within 4 days) after acute ischaemic stroke with atrial fibrillation reduced the 30-day composite of recurrent ischaemic stroke, symptomatic ICH, and unclassified stroke (OR 0.70; p=0.039) and specifically reduced recurrent ischaemic stroke (OR 0.66; p=0.029), without increasing symptomatic ICH or major extracranial bleeding. These findings support routine early DOAC use within 4 days in this population.

Major Points

  • CATALYST is a prospective individual patient data meta-analysis (IPDMA) pooling data from four RCTs — TIMING, ELAN, OPTIMAS, and START — comprising 5,441 participants with acute ischaemic stroke and atrial fibrillation. The collaboration was initiated prospectively in August 2022 before trial results were known, and PROSPERO-registered (CRD42024522634).
  • Early DOAC arm: n=2,691 (median time to DOAC start 3.0 days, IQR 2.1–3.7). Later DOAC arm: n=2,750 (median time to DOAC start 7.2 days, IQR 5.8–8.4). Primary outcome data available for 5,429 participants.
  • Primary composite outcome (recurrent ischaemic stroke, symptomatic ICH, or unclassified stroke at 30 days): 57/2683 (2.1%) early vs 83/2746 (3.0%) later; OR 0.70 (95% CI 0.50–0.98, p=0.039). Absolute risk difference: −0.93% (95% CI −1.82% to −0.04%). NNT=108 (95% CI 55–2500).
  • Early DOAC specifically reduced recurrent ischaemic stroke: 45/2683 (1.7%) vs 70/2746 (2.6%); OR 0.66 (95% CI 0.45–0.96, p=0.029).
  • No increase in symptomatic intracerebral haemorrhage with early initiation: 10/2683 (0.37%) vs 10/2746 (0.36%); OR 1.02 (95% CI 0.43–2.46, p=0.96). No signal of harm for major extracranial haemorrhage either: 12/2678 (0.45%) vs 15/2738 (0.55%); OR 0.82 (95% CI 0.38–1.75, p=0.60).
  • No significant heterogeneity across prespecified subgroups: stroke severity (NIHSS 0–4, 5–10, 11–15, ≥16), reperfusion treatment, prior anticoagulant use, sex, and age (≤75 vs >75 years). Interaction p-values all non-significant (NIHSS 0.54, reperfusion 0.36, prior anticoagulation 0.89, sex 0.75, age 0.46).
  • At 90 days, the composite primary outcome showed a numerically lower rate with early DOAC (83/2678 [3.1%] vs 96/2738 [3.5%]; OR 0.88, 95% CI 0.65–1.19, p=0.40) — not statistically significant, likely because both arms are on effective DOAC by 90 days and non-cardioembolic events dilute the treatment effect.
  • 30-day all-cause mortality was numerically lower with early DOAC (3.68% vs 4.36%; OR 0.83, 95% CI 0.63–1.09, p=0.19) but not statistically significant. No difference in 90-day mortality (7.41% vs 7.57%; OR 0.97, p=0.78).
  • OPTIMAS contributed 67% of participants (3,621), TIMING 16% (886), ELAN 14% (754), and START 3% (180). ELAN participants with major stroke and START participants with severe stroke were excluded because their timing arms did not map cleanly to the ≤4-day vs ≥5-day definition used in CATALYST.
  • The study demonstrates that the common practice of waiting 1–2 weeks to start anticoagulation after ischaemic stroke in AF patients, driven by concerns about haemorrhagic transformation, is not supported by pooled RCT evidence across a wide range of stroke severities (24% had NIHSS >10).

Design

Study Type: Prospective individual participant data meta-analysis (IPDMA) of randomised controlled trials; systematic review conducted according to PRISMA guidelines for IPDMA

Randomization: 1

Blinding: Open-label DOAC initiation in individual component trials; all trials used blinded, independent outcome event adjudication panels

Enrollment Period: Varied by component trial (TIMING enrolled 2017–2020; ELAN 2017–2022; OPTIMAS 2018–2022; START 2018–2022). CATALYST data collection finalised after all trial results were known; collaboration policy signed March 26, 2024.

Follow-up Duration: Primary: 30 days; Secondary: 90 days

Centers: Multiple centres across all 4 component trials (114 in TIMING; 103 in ELAN; 100+ in OPTIMAS; multiple US sites in START)

Countries: Sweden (TIMING), Switzerland, Japan, Germany, and 14 other countries (ELAN), United Kingdom (OPTIMAS), United States (START)

Sample Size: 5441

Analysis: One-stage individual patient data meta-analysis using generalised linear mixed-effects model with logit link function and uncorrelated random slopes and intercepts accounting for between-trial heterogeneity. Intention-to-treat. Sensitivity analyses: two-stage frequentist and Bayesian models. Complete case analysis (missing data minimal).

PROSPERO Registration: CRD42024522634

Statistical Software: R version 4.2.2 (packages: lme4, marginaleffects)

Inclusion Criteria

  • Pre-registered, randomised controlled trial comparing DOAC timing after ischaemic stroke
  • Participants with acute ischaemic stroke and documented atrial fibrillation
  • Assignment to early (≤4 days from onset) or later (≥5 days from onset) DOAC initiation in approved doses
  • Clinical outcomes investigated (not surrogate endpoints)
  • For ELAN: only participants with moderate stroke severity (not minor or major) where timing arms mapped to CATALYST definition
  • For START: only participants with mild-to-moderate strokes randomised to Day 3 (early arm) vs Day 6, 10, or 14 (later arm)

Exclusion Criteria

  • Participants who opted out of data sharing
  • ELAN participants with minor stroke (timing arms were ≤48h vs 3–4 days — both potentially ≤4 days) and major stroke (arms 6–7 vs 12–14 days — neither within 4 days)
  • START participants with severe stroke (randomised to days 6, 10, 14, 21 — all later than 4 days)
  • OPTIMAS participants found to be ineligible post-randomisation (n=27 removed to form modified ITT set)
  • START participants lost to follow-up (n=19) or who revoked consent (n=1)
  • TIMING participants who withdrew consent (n=2)
  • ELAN participants from Norway who could not share data (n=42)
  • From original trials — participants with large space-occupying parenchymal haematoma type 2 (PH2) were excluded by all component trials
  • START excluded patients with >50% infarction in MCA territory

Baseline Characteristics

Early DOAC (n=2691):

  • Age Mean (SD): 77.7 (10.1) years
  • Female: 1227 (45.6%)
  • Male: 1464 (54.4%)
  • NIHSS Median (IQR): 6 (3–11)
  • NIHSS 0–4 (Mild): 1131 (42.0%)
  • NIHSS 5–10 (Moderate): 862 (32.0%)
  • NIHSS 11–15: 342 (12.7%)
  • NIHSS ≥16 (Severe): 340 (12.6%)
  • NIHSS Missing: 16 (0.6%)
  • Hypertension: 1836 (68.2%)
  • No Hypertension: 840 (31.2%)
  • Diabetes: 553 (20.5%)
  • No Diabetes: 2131 (79.2%)
  • Mechanical Thrombectomy: 312 (11.6%)
  • Intravenous Thrombolysis: 737 (27.4%)
  • Any Reperfusion Treatment: 882 (32.8%)
  • Previous Anticoagulation: 873 (32.4%)
  • No Previous Anticoagulation: 1769 (65.7%)
  • Median Time to DOAC: 3.0 days (IQR 2.1–3.7)

Later DOAC (n=2750):

  • Age Mean (SD): 77.7 (10.0) years
  • Female: 1245 (45.3%)
  • Male: 1505 (54.7%)
  • NIHSS Median (IQR): 5 (3–10)
  • NIHSS 0–4 (Mild): 1196 (43.5%)
  • NIHSS 5–10 (Moderate): 886 (32.2%)
  • NIHSS 11–15: 306 (11.1%)
  • NIHSS ≥16 (Severe): 338 (12.3%)
  • NIHSS Missing: 24 (0.9%)
  • Hypertension: 1923 (69.9%)
  • No Hypertension: 818 (29.7%)
  • Diabetes: 548 (19.9%)
  • No Diabetes: 2192 (79.7%)
  • Mechanical Thrombectomy: 334 (12.1%)
  • Intravenous Thrombolysis: 706 (25.7%)
  • Any Reperfusion Treatment: 886 (32.2%)
  • Previous Anticoagulation: 904 (32.9%)
  • No Previous Anticoagulation: 1715 (62.4%)
  • Median Time to DOAC: 7.2 days (IQR 5.8–8.4)

Overall (n=5441):

  • Age Mean (SD): 77.7 (10.0) years
  • Female: 2472 (45.4%)
  • NIHSS Median (IQR): 5 (3–10)
  • NIHSS 0–4: 2327 (42.8%)
  • NIHSS 5–10: 1748 (32.1%)
  • NIHSS 11–15: 648 (11.9%)
  • NIHSS ≥16: 678 (12.5%)
  • Mild-to-moderate stroke (NIHSS 0–10): 4075 (75%)
  • Moderate-to-severe stroke (NIHSS ≥11): 1326 (24%)
  • Hypertension: 3759 (69.1%)
  • Diabetes: 1101 (20.2%)
  • Mechanical Thrombectomy: 646 (11.9%)
  • Intravenous Thrombolysis: 1443 (26.5%)
  • Any Reperfusion Treatment: 1768 (32.5%)
  • Previous Anticoagulation: 1777 (32.7%)

Arms

FieldEarly DOACControl
InterventionDirect oral anticoagulant (DOAC) initiated within 4 days of ischaemic stroke onset. Specific DOAC type and dose per individual trial protocol (approved doses for stroke prevention in AF). In TIMING: rivaroxaban or dabigatran or apixaban or edoxaban within 4 days. In ELAN: DOACs within 48h (minor/moderate). In OPTIMAS: any approved DOAC within 4 days. In START: DOAC on Day 3.Direct oral anticoagulant (DOAC) initiated 5 days or more after ischaemic stroke onset. In TIMING: 5–10 days. In ELAN moderate stroke: 6–7 days. In OPTIMAS: 7–14 days. In START mild-to-moderate: Days 6, 10, or 14.
n26912750
Median time to start3.0 days (IQR 2.1–3.7)7.2 days (IQR 5.8–8.4)

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Composite of recurrent ischaemic stroke, symptomatic intracerebral haemorrhage, or unclassified stroke within 30 days of randomisationPrimary0.70.039
Recurrent ischaemic stroke at 30 daysSecondary0.660.029
Symptomatic intracerebral haemorrhage (sICH) at 30 daysSecondary1.020.96
Major extracranial haemorrhage at 30 daysSecondary0.820.60
Composite of sICH or major extracranial haemorrhage at 30 daysSecondary0.90.72
All-cause mortality at 30 daysSecondary0.830.19
Primary composite outcome or all-cause mortality at 30 daysSecondary0.810.066
Primary composite outcome (recurrent ischaemic stroke, sICH, unclassified stroke) at 90 daysSecondary0.880.40
Primary composite outcome or all-cause mortality at 90 daysSecondary0.970.73
Recurrent ischaemic stroke at 90 daysSecondary0.850.32
Symptomatic intracerebral haemorrhage at 90 daysSecondary1.220.62
All-cause mortality at 90 daysSecondary0.970.78

Criticisms

  • The upper bound of the 95% CI for the primary outcome OR is 0.98 — just statistically significant at p=0.039, which is a modest margin. Adding the competing risk of mortality shifts the point estimate to 0.81 (95% CI 0.64–1.01), crossing the null.
  • NNT of 108 (95% CI 55–2500) reflects the wide confidence interval — the upper end of NNT is extremely large, reflecting imprecision in absolute benefit estimation.
  • Not all participants from the 4 trials were included: ELAN excluded minor- and major-stroke arms; START excluded severe-stroke arms; Norway ELAN patients (n=42) could not share data. This reduces generalisability and introduces potential selection effects within trials.
  • Only START required radiographical confirmation of all recurrent ischaemic strokes; TIMING, ELAN, and OPTIMAS used conventional clinical definitions with available imaging — introducing potential misclassification of stroke recurrence.
  • The rate of symptomatic intracerebral haemorrhage was low (0.37–0.36%), limiting statistical power to detect a difference in this key safety outcome even in this pooled dataset.
  • Very severe strokes (NIHSS >20) and large space-occupying haemorrhagic transformation (PH type 2) were largely excluded — results cannot be extrapolated to these higher-risk patients.
  • Ethnicity data were only available from one of the four trials (OPTIMAS), precluding subgroup analysis by race/ethnicity.
  • The 4-day cutoff was determined by OPTIMAS and TIMING trial designs rather than being derived analytically — it may not represent the true optimal initiation point. Future analyses will examine DOAC timing as a continuous variable.
  • Cannot estimate formal heterogeneity statistics (I²) because only 4 trials were included, which is insufficient for reliable between-study variance estimation.
  • This is an IPDMA, not a prospectively planned mega-trial; each component trial had slightly different definitions for stroke severity, DOAC selection, and outcome adjudication.

Funding

CATALYST collaboration facilitated by British Heart Foundation grant for OPTIMAS (CS/17/6/33361); National Institute for Health and Care Research University College London Hospitals Biomedical Research Centre; Swiss National Science Foundation for ELAN (32003B_197009; 32003B_169975). TIMING funded by Swedish Research Council (grant 2015-00881). START funded by State of Texas legislature via Lone Star Stroke Research Consortium.

Based on: CATALYST (The Lancet, 2025)

Authors: Hakim-Moulay Dehbi, Urs Fischer, Signild Åsberg, ..., David J Werring

Citation: Lancet 2025; 406: 43–51. Published online June 23, 2025. https://doi.org/10.1016/S0140-6736(25)00439-8

Content summarized and formatted by NeuroTrials.ai.