OPTIMAS — Early vs. Delayed DOAC After Ischemic Stroke

Clinical Question (PICO)

In adults with acute ischemic stroke and atrial fibrillation (AF), does early initiation of a direct oral anticoagulant (DOAC) (within 4 days of stroke) versus delayed initiation (day 7–14), affect the risk of recurrent ischemic stroke, symptomatic intracranial hemorrhage (ICH), or systemic arterial embolism — and is that effect modified by AF time of diagnosis (known pre-stroke vs. detected post-stroke) or AF subtype (persistent vs. paroxysmal)?

Bottom Line

Within each pre-specified subgroup — AF known before vs. detected after the index stroke, and persistent vs. paroxysmal AF — early DOAC initiation was non-inferior to delayed initiation for the composite primary outcome; the interaction P-values were not significant. Separately, persistent AF (vs. paroxysmal) was independently associated with roughly twice the odds of the primary outcome (adjusted OR 2.10; 95% CI 1.19–3.68).

Design

Trial type: Randomized, parallel-group, open-label with blinded outcome assessment (PROBE design).
N: 3,619 participants.
- Early DOAC: 1,813.
- Delayed DOAC: 1,806.
Randomization: 1:1, stratified (per main OPTIMAS protocol).
Setting: Multi-center UK trial; acute stroke units.
Enrollment: 2018–2024 (main OPTIMAS window; subgroup analysis reported 2025).
Mean follow-up: 90 days (primary endpoint window).
Analysis: Mixed-effects logistic regression with interaction terms between each subgroup and allocation; multivariable logistic regression for independent associations.
Primary outcome: Composite of recurrent ischemic stroke, symptomatic intracranial hemorrhage, and systemic arterial embolism at 90 days.

Population

Inclusion Criteria

Adults with acute ischemic stroke.
Confirmed atrial fibrillation (any subtype: paroxysmal, persistent, or permanent).
Eligible for DOAC therapy per clinician judgment.

Exclusion Criteria

Contraindication to DOAC therapy.
Indication for chronic anticoagulation before the qualifying stroke for a non-AF reason that would mandate immediate anticoagulation.
Active bleeding or high bleeding risk precluding randomization.
Other indication confounding the primary endpoint (per main OPTIMAS protocol).

Baseline Characteristics (overall, n = 3,619)

Mean age: 78.0 ± 9.9 years.
Sex: 45.3% women.
AF known before index stroke: 1,838 (50.8%).
AF detected after index stroke: 1,781 (49.2%).
AF subtype: persistent vs. paroxysmal distribution reported in main OPTIMAS dataset; this subgroup analysis stratifies the pre-specified subgroup of interest by AF time of diagnosis.

Interventions

Early DOAC — direct oral anticoagulant started within 4 days of the qualifying stroke.
Delayed DOAC — direct oral anticoagulant started between day 7 and day 14 of the qualifying stroke.
DOAC agent at clinician discretion (apixaban, rivaroxaban, edoxaban, or dabigatran per local practice and license).
Dose-adjusted for renal function per product label.

Outcomes

Primary Outcome (composite: recurrent ischemic stroke + symptomatic ICH + systemic arterial embolism)

For AF diagnosed before the index stroke (n = 1,838):

Early DOAC: 37/918 (4.0%) vs. Delayed DOAC: 32/920 (3.5%)
Odds ratio: 1.17 (95% CI 0.72–1.89)
Non-inferior to delayed (per main OPTIMAS prespecified margin).

For AF diagnosed after the index stroke (n = 1,781):

Early DOAC: 22/895 (2.5%) vs. Delayed DOAC: 27/886 (3.0%)
Odds ratio: 0.79 (95% CI 0.45–1.40)
P interaction = 0.312 — no heterogeneity between subgroups.

Subgroup Analyses (by AF subtype)

Persistent AF (early vs. delayed): OR 1.06 (95% CI 0.71–1.58)
Paroxysmal AF (early vs. delayed): OR 0.66 (95% CI 0.25–1.72)
P interaction = 0.377 — no heterogeneity between subtypes.

Independent association of AF subtype with primary outcome (adjusted, paroxysmal AF as reference):

Persistent AF vs. paroxysmal AF: adjusted OR 2.10 (95% CI 1.19–3.68)

AF time of diagnosis was not independently associated with primary outcome events.

Adverse Events / Safety

Per main OPTIMAS; symptomatic intracranial hemorrhage rates were comparable between early and delayed arms and by AF subgroup (specific rates per subgroup reported in main trial publication). Bleeding definitions match ISTH criteria.

Figures

*Figure 1.* CONSORT diagram (Consolidated Standards of Reporting Trials) of participant flow. DOAC indicates direct oral anticoagulant; FU, follow-up; and GP, general practitioner.

*Figure 2.* Forest plot showing primary composite outcome events according to atrial fibrillation (AF) time of diagnosis, AF subtype, and treatment allocation. DOAC indicates direct oral anticoagulant.

*Figure 3.* Kaplan-Meier curves comparing outcome events according to atrial fibrillation (AF) time of diagnosis. A , Primary outcome. B , Recurrent ischemic stroke. C , Symptomatic intracranial hemorrhage (sICH). D , All-cause mortality at 90 days.

*Figure 4.* Kaplan-Meier curves comparing outcome events at 90 days according to atrial fibrillation (AF) subtype. A , Primary outcome, ( B ) recurrent ischemic stroke, ( C ) symptomatic intracranial hemorrhage (sICH), ( D ) all-cause mortality.

Criticisms

Subgroup analysis caveat. The subgroup analysis itself is exploratory in framing, although the AF-time-of-diagnosis subgroup was pre-specified. Interaction P-values > 0.30 do not constitute formal heterogeneity testing for non-inferiority; absence of evidence ≠ evidence of absence.
90-day window. Whether the early-vs-delayed trade-off changes over longer follow-up is not addressed here.
Generalizability. UK-centric population; routine practice patterns for AF detection and DOAC initiation may differ internationally.
Persistent-AF association. Persistent AF (vs. paroxysmal) carrying a twofold risk of recurrent stroke is biologically plausible but the subgroup paper does not adjudicate AF burden or duration rigorously; the finding is hypothesis-generating rather than practice-changing on its own.

Funding

British Heart Foundation Clinical Study Project grant CS/17/6/33361. The funder had no role in study design, data collection, analysis, interpretation, or writing of the report.

What this is

OPTIMAS already established, in the main trial, that early DOAC initiation was non-inferior to delayed for the composite outcome at 90 days. This paper asks a reasonable follow-up: is that safety/efficacy trade-off the same across AF phenotypes that we see in clinic every day? The answer is essentially yes.

The paper

Authors. James Lyon, Philip S. Nash, Norin Ahmed, et al.
Title. Early Versus Delayed Anticoagulation in Acute Ischemic Stroke According to Atrial Fibrillation Subtype and Time of Diagnosis: Subgroup Analysis of the OPTIMAS Randomized Controlled Trial.
Journal. Stroke.
Year. 2026.
DOI. 10.1161/STROKEAHA.125.055037
PMCID. PMC13196862
Registration. NCT03759938

1. Shadow Audit

What the article is not saying — and on first read you might miss — is that “AF detected after stroke” patients had half the event rate of “AF known before stroke” patients (2.5–3.0% vs. 3.5–4.0%). That is a real, numerically substantial between-subgroup difference in absolute event rate that the authors attribute to underlying phenotype (“a less burdened AF”), but the adjusted comparison in this paper isn’t a clean between-subgroup test for that rate difference; the interaction test on the treatment effect is the prespecified question, and the independent association models ask the secondary question. The headline should not become “patients with AF detected after stroke do better”; that is the wrong question for this paper.

2. Inversion Engine

For the opposite conclusion to hold — that early DOAC is worse in either subgroup — you’d need a positive interaction term, i.e., the early-delayed gap would need to invert by ≥15–20% in absolute terms given the sample sizes. The 95% CIs on the ORs (0.72–1.89 and 0.45–1.40) rule out a clinically meaningful detriment to early DOAC in either subgroup; that is consistent with the main OPTIMAS finding. To flip the persistent-vs-paroxysmal finding (adjusted OR 2.10, lower CI 1.19), you’d need the AF-burden phenomenon (duration, monitoring intensity) to be a confounder the adjustment couldn’t capture.

3. Second-Order Catalyst

Stroke neurology update protocols: most centers now have an ERP that says “start the DOAC on day 3–4 unless there’s a clear ICH signal or large infarct with hemorrhagic transformation.” This subgroup analysis gives stroke program directors a clean answer to the “what if AF was discovered in-hospital vs. was known” question: nothing changes. Days 7–14 delay is no longer a defensible protocol.

4. Asymmetric Leverage

The biggest asymmetric payoff from this paper is removing delay as practice-changing evidence: if you can shave a week off the time-to-DOAC across an entire stroke service, the absolute benefit material is small (2.5-vs-3.0% is small) but applies to a huge denominator (every AF-stroke patient). Persistent vs. paroxysmal is the smaller lever here.

5. Paradigm Destroyer

What does this paper kill? The “two-week wait because we don’t yet know the AF phenotype” reflex. That reflex is now evidence-free. It also softens, without destroying, the assumption that paroxysmal and persistent AF carry equivalent recurrent stroke risk; the adjusted OR of 2.10 will be useful when adjusted risk models (CHA₂DS₂-VASc successors) get refitted.

MVP — Minimum Viable Proof

For the bedside clinician: start the DOAC by day 4 in eligible AF-stroke patients regardless of whether AF was known before the event or detected after — the 90-day safety/efficacy profile is non-inferior to delayed.

Best Combination

Pair this with main OPTIMAS (non-inferiority), the 2022 meta-analysis (AF detected after stroke has a lower recurrence rate baseline), and current AHA/ASA secondary prevention guidance. Together: don’t delay the DOAC for post-stroke AF; AF-subtype distinctions don’t change the timing decision.

Overvalue Warning

Two things to not over-read: (1) the persistent-vs-paroxysmal adjusted OR 2.10 is a secondary finding; it is hypothesis-generating, not a recommendation to escalate anticoagulation intensity beyond standard DOAC dosing in persistent AF; (2) the absence of heterogeneity tests (P interaction > 0.30) does not prove identical effects — it proves the trial was not powered to detect anything but large between-subgroup differences.

Educational use notice. The Clotbust Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.

OPTIMAS — Early vs. Delayed DOAC After Ischemic Stroke, by AF Subtype and Time of Diagnosis