← Back
NeuroTrials.ai
Neurology Clinical Trial Database

START

Optimal Delay Time to Initiate Anticoagulation After Ischemic Stroke in Atrial Fibrillation: A Pragmatic, Response-Adaptive Randomized Clinical Trial

Share Share Copied! Compare

Year of Publication: 2025

Authors: Warach SJ, Davis LA, Lawrence P, ..., Milling TJ

Journal: JAMA Neurology

Citation: Warach SJ, et al. JAMA Neurol. 2025;82(5):470-476. doi:10.1001/jamaneurol.2025.0285

Link: https://doi.org/10.1001/jamaneurol.2025.0285

Clinical Question

Is there an optimal day within the first 14 days to initiate a direct oral anticoagulant after atrial fibrillation-associated ischemic stroke that minimizes the composite risk of ischemic or hemorrhagic events?

Bottom Line

No statistically superior start day was identified, but the day 3-4 group had 0 ischemic events and the highest posterior probability of being optimal (0.41), suggesting earlier DOAC initiation is preferable to later within the first 2 weeks after AF-related ischemic stroke.

Major Points

  • First US-based randomized clinical trial addressing timing of DOAC initiation after AF-related ischemic stroke, and the first to apply response-adaptive randomization to treatment delay timing. Conducted at 10 Texas stroke centers (Lone Star Stroke Research Consortium), June 2017 to June 2023.
  • 200 patients randomized to 4 DOAC start-time groups: day 3-4 (n=54), day 6 (n=53), day 10 (n=46), day 14 (n=47). Median age 75 years (IQR 65-81), 50% female, 17.5% Asian/Black/multiracial, 16.5% Hispanic, median NIHSS 6.5 (IQR 4-14), median lesion diameter 3.1 cm (IQR 2.0-4.4 cm). Apixaban was the DOAC in 89% of patients.
  • Primary composite outcome at 30 days (ischemic stroke/systemic embolism OR symptomatic ICH/major systemic hemorrhage) occurred in 10/200 patients (5%) total: 7 ischemic events (all stroke) and 3 hemorrhagic events (2 intracranial, 1 systemic). No group exceeded the prespecified superiority threshold of posterior probability >0.75.
  • Group 1 (day 3-4): 0 ischemic events, 1 hemorrhagic event (1.9%). Posterior probability of being optimal = 0.41. Posterior probability for optimal ischemic prevention >0.99. Group 2 (day 6): 3 ischemic (5.7%), 1 hemorrhagic (1.9%), posterior probability = 0.26. Group 3 (day 10): 2 ischemic (4.3%), 1 hemorrhagic (2.2%), posterior probability = 0.17. Group 4 (day 14): 2 ischemic (4.3%), 0 hemorrhagic, posterior probability = 0.15.
  • Expected ischemic event rates derived from Bayesian model (posterior mean): Group 1 = 2.8% (95% CI 1.0-5.6%), Group 2 = 3.2% (1.3-6.0%), Group 3 = 3.9% (1.6-7.3%), Group 4 = 5.3% (2.0-10.9%). Expected hemorrhagic rates: Group 1 = 2.4% (0.7-5.9%), Group 2 = 2.0% (0.4-5.0%), Group 3 = 1.8% (0.3-4.7%), Group 4 = 1.5% (<0.1-4.3%). Utility scores: Group 1 = -0.053, Group 2 = -0.052, Group 3 = -0.057, Group 4 = -0.067.
  • Response-adaptive randomization performed as designed: after 100 patients, posterior randomization probabilities updated to Group 1 = 0.31, Group 2 = 0.28, Group 3 = 0.21, Group 4 = 0.20. After 150 patients: Group 1 = 0.27, Group 2 = 0.29, Group 3 = 0.22, Group 4 = 0.22 — demonstrating progressive allocation shift toward better-performing groups.
  • 30-day all-cause mortality 3.3% overall (6/180). By group: Group 1 = 2.0% (1/49), Group 2 = 4.3% (2/47), Group 3 = 2.3% (1/43), Group 4 = 4.9% (2/41). 90-day all-cause mortality 6.0% overall (11/184): Group 1 = 4.0% (2/50), Group 2 = 10.0% (5/48), Group 3 = 2.3% (1/44), Group 4 = 7.1% (3/42). Median mRS = 2 (IQR 1-3) at both 30 and 90 days.
  • Of 613 hospital admissions screened at the lead site with ischemic stroke and AF, 21.5% were eligible and 15.8% were enrolled. Most common exclusion reasons: insufficient lesion size, no intention to prescribe a DOAC, and life expectancy <90 days.
  • Compared to contemporaneous European trials: TIMING (n=888, median NIHSS 4) found early initiation (≤4 days) noninferior to delayed (5-10 days) at 90 days. ELAN (n=2013, median NIHSS 3) found early initiation noninferior at 30 days. OPTIMAS (n=3621, median NIHSS 4) found early initiation (≤4 days) noninferior to delayed (7-14 days) at 90 days. START had a higher median NIHSS (6.5) and was the only US trial; it evaluated 4 discrete start days rather than a binary early-vs-late design.
  • Key limitations: sample size reduced from 1000 to 200 due to slow accrual and funding; event rate ~5% vs assumed 10%, substantially limiting power; open-label; variable adherence (Group 4 only 54% started within window); 10-14% lost to follow-up; Texas-only population; physician discretion for DOAC choice and antiplatelet co-therapy introduced within-group heterogeneity.

Design

Study Type: Phase 2, pragmatic, prospective, multicenter, parallel-group, response-adaptive randomized clinical trial

Randomization: 1

Blinding: Open-label (participants, caregivers, and investigators unblinded); outcome adjudicator blinded to randomization assignment and DOAC status at time of event

Enrollment Period: June 22, 2017 – June 2, 2023

Follow-up Duration: 90 days (primary outcome at 30 days; secondary outcomes at 90 days)

Centers: 10

Countries: USA

Sample Size: 200

Original Planned Sample Size: 1000

Protocol Amendment: Sample size reduced from 1000 to 200 after 83 patients enrolled, converting trial to phase 2; made due to slower-than-anticipated enrollment and funding limitations

Analysis: Primary: Bayesian posterior probability using composite utility function (negative sum of ischemic and hemorrhagic event rates, modeled separately); intention-to-treat; missing 30-day outcomes imputed as no event

Adaptive Design: Response-adaptive randomization: first 100 patients enrolled 1:1:1:1 (probability 0.25 each); allocation ratios updated at n=100 and n=150 based on posterior probabilities favoring better-performing groups

Statistical Software: FACTS version 7.0 (Berry Consultants) and R version 4.3.2 (R Foundation for Statistical Computing)

Randomization Platform: REDCap (HIPAA-compliant secure web platform)

Reporting Guideline: CONSORT

Trial Registration: ClinicalTrials.gov NCT03021928

Consortium: Lone Star Stroke Research Consortium

Inclusion Criteria

  • New disabling neurological deficit attributable to new ischemic stroke.
  • Minimum lesion diameter of 1.5 cm on CT or MRI performed less than 48 hours from stroke onset (time last known well).
  • If lesion not visible on imaging: NIHSS score greater than 4.
  • If treated with thrombolytic or endovascular therapy for current stroke: qualifying scan performed after treatment to rule out clinically significant hemorrhagic transformation.
  • Nonvalvular atrial fibrillation (paroxysmal, persistent, or permanent).
  • Not currently being treated with anticoagulants and will not be treated with anticoagulants prior to starting the DOAC at the randomized initiation time (exception: DVT prophylaxis anticoagulation permitted).
  • Treating physician planned to treat the patient with an FDA-approved DOAC.
  • Available to be randomized and begin treatment within 96 hours from symptom onset (time last known well) — the end of the time window for Group 1.
  • Written informed consent obtained from the patient or a legally authorized representative.

Exclusion Criteria

  • Clinical or imaging evidence of spontaneous intracranial hemorrhage within the previous 6 months. (Hemorrhagic transformation of current or previous ischemic stroke, or chronic microbleeding, were allowed at investigator discretion.)
  • Estimated infarct volume greater than 50% of the middle cerebral artery territory on qualifying scan.
  • If full extent of lesion was not visible on imaging: NIHSS score greater than 23.
  • Anticipated need for major surgery within the next 30 days requiring delay, discontinuation, or extended suspension of anticoagulant treatment for longer than 5 days.
  • Symptomatic cerebral edema expected based on the size and location of the ischemic stroke.
  • Current decreased level of consciousness or anticipated decreased level of consciousness.
  • Life expectancy less than 90 days.
  • Follow-up for 90 days (in person or by telephone) not feasible.

Arms

FieldGroup 1: DOAC initiation Day 3 or 4Group 2: DOAC initiation Day 6Group 3: DOAC initiation Day 10Group 4: DOAC initiation Day 14
InterventionDirect oral anticoagulant initiated on day 3 or day 4 after stroke onset (treatment window 48-96 hours from time last known well). Choice of DOAC, dose, and any concomitant antiplatelet therapy left entirely to treating physician discretion. Apixaban was prescribed in 85% (46/54). Of 45 who confirmed starting: 37 started within the 48-96h window; 2 started within 24h of the window.Direct oral anticoagulant initiated on day 6 after stroke onset (treatment window 120-144 hours). Choice of DOAC, dose, and any concomitant antiplatelet therapy left entirely to treating physician discretion. Apixaban was prescribed in 91% (48/53). Of 41 who confirmed starting: 30 started within the 120-144h window; 7 started within 24h of the window.Direct oral anticoagulant initiated on day 10 after stroke onset (treatment window 216-240 hours). Choice of DOAC, dose, and any concomitant antiplatelet therapy left entirely to treating physician discretion. Apixaban was prescribed in 93% (43/46). Of 41 who confirmed starting: 28 started within the 216-240h window; 6 started within 24h of the window.Direct oral anticoagulant initiated on day 14 after stroke onset (treatment window 312-336 hours). Choice of DOAC, dose, and any concomitant antiplatelet therapy left entirely to treating physician discretion. Apixaban was prescribed in 87% (41/47). Of 35 who confirmed starting: 19 started within the 312-336h window; 10 started within 24h of the window.
DurationOngoing indefinitely for secondary stroke preventionOngoing indefinitely for secondary stroke preventionOngoing indefinitely for secondary stroke preventionOngoing indefinitely for secondary stroke prevention
n54534647

Outcomes

OutcomeTypeControlInterventionHR / OR / RRP-value
Composite of recurrent ischemic stroke or systemic embolism within 30 daysPrimaryLater initiation (4-14 days): 8/229 (3.5%)Early initiation (≤3 days): 6/213 (2.8%)Posterior probability of superiority: 0.77Bayesian design; no traditional P-value. Stopped for expected futility at interim analysis.
All-cause mortality at 30 days | Overall: 3.3% (6/180 non-missing outcomes); Group 1 (Day 3-4): 2.0% (1/49); Group 2 (Day 6): 4.3% (2/47); Group 3 (Day 10): 2.3% (1/43); Group 4 (Day 14): 4.9% (2/41)Secondary
All-cause mortality at 90 days | Overall: 6.0% (11/184 non-missing outcomes); Group 1 (Day 3-4): 4.0% (2/50); Group 2 (Day 6): 10.0% (5/48); Group 3 (Day 10): 2.3% (1/44); Group 4 (Day 14): 7.1% (3/42)Secondary
Modified Rankin Scale score at 30 days | Overall: Median 2 (IQR 1-3)Secondary
Modified Rankin Scale score at 90 days | Overall: Median 2 (IQR 1-3)Secondary
Composite events at 90 days | Note: Adjudicated using same criteria as primary 30-day outcome; reported in supplementary eTableSecondary

Subgroup Analysis

No formal prespecified subgroup analyses reported in primary paper. Exploratory analyses: (1) Ischemic-only analysis: Group 1 posterior probability >0.99 for being optimal for ischemic event prevention; Groups 2-4 each <0.01. (2) Hemorrhagic-only analysis: Group 4 posterior probability 0.54 for being optimal; Group 3 = 0.19, Group 2 = 0.15, Group 1 = 0.12. Authors noted intent to separately publish secondary analyses of each event type. Subgroup analyses by sex, age, NIHSS, stroke severity, race/ethnicity, and reperfusion treatment status not reported in primary manuscript.

Criticisms

  • Sample size was reduced from the planned 1000 to 200 (protocol amendment after 83 patients enrolled) due to slower-than-expected enrollment and funding limitations. The study was therefore underpowered and became a phase 2 feasibility trial rather than a definitive phase 3 efficacy trial.
  • Event rate was approximately half of that assumed during trial planning (5% observed vs 10% assumed), further reducing statistical power and ability to identify a superior timing group.
  • Open-label design: participants, caregivers, and investigators were all unblinded to group assignment, introducing potential performance and detection bias. Only the independent outcome adjudicator was blinded.
  • Pragmatic design with no monitoring of actual DOAC adherence: a substantial number of patients did not start the DOAC within the assigned window. Group 2: 6 patients did not start, 6 unknown status. Group 4: 5 did not start, 7 unknown status; only 19 of 35 confirmed starters (54%) actually initiated within the 312-336h window.
  • Slower-than-expected enrollment despite ischemic stroke being common, few exclusions, and minimal protocol requirements beyond standard care — suggests many eligible patients were treated outside the trial due to physician preference or resource limitations, introducing potential selection bias.
  • High rates of loss to follow-up: 30-day outcomes missing for 20/200 (10%), 90-day outcomes missing for 28/200 (14%), primarily because patients could not be contacted.
  • Exclusively Texas-based hospitals with a US-specific patient population — limits geographic and demographic generalizability.
  • Choice of DOAC agent, dose, and concomitant antiplatelet therapy was entirely at physician discretion, introducing within-group heterogeneity in actual treatment received.
  • Trial enrolled from 2017 to 2023 — during this period, larger European trials (TIMING 2022, ELAN 2023, OPTIMAS 2024) were published, which may have influenced physician decisions about enrolling patients and managing anticoagulation timing.
  • Phase 2 designation means the primary purpose was feasibility assessment and hypothesis generation — results should not be used as definitive guidance for clinical practice without confirmation in a larger phase 3 trial.

Funding

Texas legislature (awarded to the Lone Star Stroke Research Consortium). The Texas legislature had no role in study design, data collection, management, analysis, interpretation, manuscript preparation, review, approval, or the decision to submit for publication.

Based on: START (JAMA Neurology, 2025)

Authors: Warach SJ, Davis LA, Lawrence P, ..., Milling TJ

Citation: Warach SJ, et al. JAMA Neurol. 2025;82(5):470-476. doi:10.1001/jamaneurol.2025.0285

Content summarized and formatted by NeuroTrials.ai.