Early versus late oral anticoagulants after acute ischemic stroke in atrial fibrillation patients: a meta-analysis of randomized controlled trials

Abstract

Background

Atrial fibrillation (AF) occurs in about 20% of ischemic stroke cases. Although oral anticoagulants (OACs) reduce stroke risk in AF, the optimal timing for starting them after acute ischemic stroke remains uncertain, with varying recommendations across guidelines.

Methods

A systematic review and meta-analysis of randomized controlled trials was conducted using PubMed, Scopus, Web of Science, EMBASE, and Cochrane CENTRAL through October 2024. Studies included adults with AF and recent ischemic stroke, comparing early versus late OAC initiation. The primary outcome was recurrent ischemic stroke. Risk of bias was assessed using the Cochrane RoB2 tool.

Results

Three trials with 6522 patients were included. There were no significant differences between early and late OAC initiation in recurrent ischemic stroke (relative risk [RR] 0.82; 95% confidence interval [CI] 0.61–1.11; P = 0.19), major extracranial bleeding (RR 0.47; 95% CI 0.22–1.01; P = 0.05), intracranial hemorrhage (RR 0.93; 95% CI 0.44–1.96; P = 0.84), or mortality (RR 0.79; 95% CI 0.50–1.24; P = 0.30).

Conclusion

Early OAC initiation after ischemic stroke in AF appears as safe as later initiation, without higher bleeding or mortality risks. Given the elevated risk of early recurrence, prompt anticoagulation may be beneficial but requires further confirmation.

Atrial fibrillation (AF) is identified in at least 20% of patients with ischemic stroke, often serving as the underlying cause of the event.¹ Large randomized controlled trials (RCTs) have demonstrated that long-term use of direct oral anticoagulants (DOACs) can reduce the risk of ischemic stroke in individuals with AF by approximately two-thirds, with a relatively low incidence of intracranial hemorrhage.^2,3 However, these trials excluded patients who had experienced an acute ischemic stroke within 7 to 30 days before eligibility assessment, leaving the optimal timing for initiating anticoagulation after an acute ischemic stroke unclear.^2,3

Current guidelines regarding the initiation of DOACs following AF-related ischemic stroke are imprecise and inconsistent. In 2013, the European Heart Rhythm Association and the European Society of Cardiology proposed the “1–3–6–12 days rule”⁴ based on evidence suggesting that larger infarcts, which result in severe stroke syndromes, are more prone to hemorrhagic transformation compared to smaller ones.⁵ Meanwhile, the 2018 guidelines from the American Heart Association/American Stroke Association recommend starting DOACs 4 to 14 days after the onset of neurological symptoms in patients with ischemic stroke.⁶ Although the 2026 American Heart Association/American Stroke Association guideline suggests that early initiation of direct oral anticoagulants after AF-related ischemic stroke appears to be safe compared with delayed initiation, its efficacy in preventing early recurrent stroke remains unestablished.⁷ The lack of high-quality evidence has created uncertainty among clinicians, prompting calls for randomized interventional trials to clarify guidelines and recommendations.

In cases of acute ischemic stroke related to AF, the risk of early recurrent ischemic stroke,⁸ as well as hemorrhagic transformation,⁹ is greatest in the days immediately following the initial stroke event. Randomized and observational studies have indicated that early anticoagulation may reduce the risk of ischemic stroke without increasing the incidence of intracranial hemorrhage.^10,11 While these studies provided estimates of event rates,¹² they did not definitively establish the safety or superiority of early anticoagulation compared to delayed treatment.¹³ Therefore, in this systematic review and meta-analysis, we aimed to compare the effects of early versus late initiation of DOAC therapy in patients with AF who had experienced a stroke. We evaluated key outcomes, including recurrent stroke rates, bleeding risk, and all-cause mortality, to guide clinical guidelines and improve treatment outcomes for this high-risk population.

METHODS

This systematic review and meta-analysis was completed using the PRISMA statement¹⁴ and the Cochrane Handbook for Meta-Analyses and Systematic Reviews.¹⁵ This review was submitted to PROSPERO and published under ID CRD42024606899.

Search strategy and study selection

We searched PubMed (MEDLINE), Scopus, Web of Science, EMBASE, and the Cochrane Central Register of Controlled Trials (CENTRAL) databases until October 27, 2024. After we changed the search terms and keywords in each database, the results are displayed in Supplemental Table 1.

Table 1.

Summary characteristics of the included RCTs

Characteristic	Fischer et al. 2023	Oldgren et al. 2022	Werring et al. 2024
Design	RCT	Registry-based RCT	Phase 4 RCT
Country	Internationala	Sweden	United Kingdom
Funding source	Swiss National Science Foundation	Swedish Medical Research Council	British Heart Foundation
Recruitment duration	Nov 6, 2017–Sep 12, 2022	Apr 2, 2017–Dec 30, 2020	Jul 5, 2019–Jan 31, 2024
Centers	103	34	100
Blinding status	Open-label (assessor blinded)	Open-label (blinded endpoint study)	Open-label (blinded endpoint study)
Total participants	2032	888	3648
Early treatment definition	48 h from stroke onset in minor/moderate stroke; days 6–7 in major stroke	≤4 days from stroke onset	≤4 days from stroke onset
Late treatment definition	3–4 days minor; days 6–7 moderate; days 12–14 majorb	≥5–10 days from stroke onset	7–14 days from stroke onset
Main inclusion criteria	Adults >18 with signed consent, confirmed ischemic stroke and AF, agreement of the treating physician to prescribe OAC	Adults >18 with AF and recent ischemic stroke willing to start OAC	Adults with AF and a clinical diagnosis of acute ischemic stroke; physician uncertain of optimal OAC timing
Primary outcome	Composite: major bleeding, recurrent ischemic stroke, systemic embolism, vascular death at 30 ± 7 days after randomization	Composite: recurrent ischemic stroke, symptomatic ICH, all-cause death at 90 days of index stroke	Composite: recurrent ischemic stroke, symptomatic ICH, unclassifiable stroke syndromes,c systemic arterial embolism at 90 days after randomization
Follow-up duration	90 days	90 days	90 days

^aThe study encompassed 15 countries across Europe, the Middle East, and Asia and was conducted under the oversight of the University Hospital Bern, Switzerland.

^bInfarct size (minor, moderate, or major) was determined by site investigators using a standardized visual rating scheme, with minor infarcts defined as ≤1.5 cm, moderate as infarcts in a cortical superficial branch of major cerebral arteries, and major as larger infarcts in these regions or brain-stem/cerebellar infarcts >1.5 cm.

^cPatients in whom a clinical diagnosis of a stroke syndrome was made but who did not undergo neuroimaging for clinical reasons, such as a terminal prognosis.

AF indicates atrial fibrillation; ICH, intracerebral hemorrhage; OAC, oral anticoagulant; RCT, randomized controlled trial.

We included studies that met the following PICOS criteria: Population: adult patients, 18 years old or older, with AF and recent ischemic stroke; Intervention: early initiation of anticoagulation therapy, <4 days from stroke symptoms onset; Comparison: late initiation of anticoagulation therapy from stroke symptom onset, after >5 days from stroke symptom onset; and Outcomes: recurrent ischemic stroke, with secondary outcomes of major extracranial bleeding, systemic embolism, symptomatic intracranial hemorrhage, nonmajor bleeding, and all-cause mortality.

Studies included were RCTs. The exclusion criteria for the articles were as follows: 1) book chapters; 2) nonhuman and in vitro experiments; 3) studies with overlapping or duplicate datasets; 4) reviews, comments, letters to the editor, and guidelines; and 5) studies not published in English.

The online Covidence tool was used to conduct the review. Once duplicates were removed from the records, two authors (A.M. and M.T.) independently assessed each. During the initial full-text screening for eligibility requirements, the full texts of the records were examined by the same two authors. Disagreements were resolved by consensus and discussion with a senior author.

Data extraction

After acquiring the complete texts of the pertinent publications, we carried out a pilot extraction to precisely set up the data extraction sheet. Three sections made up the Excel (Microsoft, USA) data extraction sheet. The first section contained a summary of the features of the included studies, including the study ID, country, study design, funding source, inclusion criteria, total number of participants, definition of both the intervention and comparator, primary outcome, and follow-up period. In the second section, the participants’ baseline data were presented: age, gender, comorbidities, National Institutes of Health stroke scale score, initial treatment for stroke, medication history, and the initial international normalized ratio. The third part included outcomes data. Two reviewers (A.M. and M.T.) were responsible for data extraction. Disagreements were resolved through reaching agreement or discussions with a senior author.

Risk of bias and certainty of evidence

Using the Cochrane RoB2 method,¹⁶ two reviewers (A.M. and M.T.) independently assessed the quality of the studies. They considered five domains: measuring the outcome, selecting the reported results, deviating from the intended intervention, missing outcome data, and the risk of bias related to the randomization process. Any disagreements were discussed with a senior author.

Statistical analysis

For modeling and visualization, we used R version 2021.09, MetaInsight, and Meta-Mar; for sensitivity analysis, we used PQStat version 1.8. An odds ratio (OR) for intention to treat and a risk ratio (RR) for side effects with a 95% confidence interval (CI) were used to report dichotomous data. However, the random-effects model was used when there was significant heterogeneity. We employed the chi-square and I² tests; the chi-square test establishes whether heterogeneity exists, and the I² test establishes the degree of heterogeneity. Significant heterogeneity was defined as an I² of >50% and considerable heterogeneity as an alpha level of <0.1 for the chi-square test, according to the Cochrane Handbook (chapter 9).¹⁷

RESULTS

After the screening process, 5273 studies were screened and assessed based on their titles and abstracts; 486 articles moved to full-text screening once 3874 duplicates and 5314 irrelevant studies were excluded. Finally, we included three RCTs (Figure 1) conducted across 17 countries with 6568 patients. The studies exclusively investigated the timing of DOAC initiation in patients with ischemic stroke related to AF. The included participants’ comorbidities, baseline characteristics, and details of the RCTs are listed in Tables 1 and 2.

Figure 1.

PRISMA flow chart of the screening process.

Table 2.

Baseline characteristics of the participants

Variable	Early treatment / Late treatment
	Fischer et al. 2023 (11)	Oldgren et al. 2022 (9)	Werring et al. 2024 (12)
Participant characteristics
Sample size (N)	1006 / 1007	450 / 438	1814 / 1807
Age, mean (SD)	77 (10.4) / 77.7 (9.7)	78.4 (10.1) / 78.3 (9.7)	78.5 (9.9) / 78.5 (9.9)
Female, n (%)	459 (45.6) / 456 (45.3)	207 (46.0) / 203 (46.3)	810 (44.7) / 830 (45.9)
Medical history
TIA, n (%)	45 (4.5) / 51 (5.1)	36 (8.0) / 26 (5.9)	NA / NA
Current smoker, n (%)	103 (10.2) / 84 (8.3)	41 (9.1) / 30 (6.8)	144 (7.9) / 129 (7.1)
Ex-smoker, n (%)	147 (14.6) / 158 (15.7)	NA / NA	502 (27.7) / 517 (28.6)
Prior stroke, n (%)	128 (12.7) / 140 (13.9)	79 (17.6) / 76 (17.4)	295 (16.3) / 242 (13.4)
HTN, n (%)	690 (68.6) / 673 (66.8)	333 (74.0) / 338 (77.2)	1205 (66.4) / 1229 (68.0)
MI, n (%)	80 (8.0) / 87 (8.6)	NA / NA	162 (8.9) / 174 (9.6)
DM, n (%)	185 (18.4) / 161 (16.0)	81 (18.0) / 91 (20.8)	121 (6.7) / 127 (7.0)
AF previously known, n (%)	NA / NA	223 (49.6) / 213 (48.6)	917 (50.6) / 919 (50.9)
NIHSS score
At admission, mean (SD)	6.3 (7.4) / 6 (6.7)	6.2 (5.8) / 5.9 (5.9)	6.7 (5.9) / 6 (5.2)
At randomization, mean (SD)	3.3 (3.7) / 3.3 (3.7)	NA / NA	4.3 (3.7) / 4.3 (3.7)
Initial treatment for stroke
Thrombolysis, n (%)	391 (38.9) / 377 (37.4)	132 (29.3) / 120 (27.4)	421 (23.2) / 377 (20.9)
Thrombectomy, n (%)	207 (20.6) / 232 (23)	65 (14.4) / 56 (12.8)	131 (7.2) / 135 (7.5)
Medication history
Antiplatelet use, n (%)	518 (51.5) / 610 (60.6)	103 (22.9) / 96 (21.9)	213 (11.7) / 194 (10.7)
OAC use, n (%)	NA / NA	90 (20.1) / 85 (19.5)	582 (32.1) / 584 (32.3)
Warfarin use, n (%)	NA / NA	31 (6.9) / 34 (7.8)	61 (3.4) / 53 (2.9)
INR, mean (SD)	1.03 (0.07) / 1.03 (0.07)	2.07 (0.68) / 1.86 (0.50)	NA / NA

AF indicates atrial fibrillation; DM, diabetes mellitus; HTN, hypertension; INR, international normalized ratio; MI, myocardial infarction; NIHSS, National Institutes of Health Stroke Scale; OAC, oral anticoagulant; SD, standard deviation; TIA, transient ischemic attack.

Risk of bias and certainty of evidence

All the included RCTs demonstrated low overall risk as well as low risk of bias across all the five domains of the RoB 2 tool, with the exception of Oldgren et al, which exhibited some concerns of bias in the domain related to missing outcome data, and Werring et al, which showed some concerns of bias in the domain addressing deviations from the intended intervention (Figure 2).

Figure 2.

Quality assessment of risk of bias in the included trials. A schematic representation of risks (low = green, unclear = yellow, and high = red) for specific types of biases of each of the studies in the review.

Primary outcomes

There was no significant difference between early initiation and late initiation of DOACs in the recurrence of ischemic stroke (RR 0.82; 95% CI 0.61, 1.11; P = 0.19) (Figure 3). The pooled studies were homogenous (I²= 26%, P = 0.26).

Figure 3.

Forest plot of the primary outcome: recurrence of ischemic stroke. CI indicates confidence interval; RR, risk ratio.

Secondary outcomes

There was no significant difference between early and late initiation of DOACs in major extracranial bleeding (RR 0.47; 95% CI [0.22, 1.01]; P = 0.05) (Figure 4a), symptomatic intracranial hemorrhage (RR 0.93; 95% CI [0.44, 1.96]; P = 0.84) (Figure 4b), nonmajor bleeding (RR 1.07; 95% CI [0.79, 1.45]; P = 0.65) (Figure 4c), systemic embolism (RR 0.43; 95% CI [0.16, 1.11]; P = 0.08) (Figure 4d), and all-cause mortality (RR 0.79; 95% CI [0.50, 1.24]; P = 0.30) (Figure 4e).

Figure 4.

Forest plot of the secondary outcome: (a) major extracranial bleeding, (b) symptomatic intracranial hemorrhage, (c) nonmajor bleeding, (d) systemic embolism, (e) all-cause mortality. CI indicates confidence interval; RR, risk ratio.

The pooled studies were homogenous in major extracranial bleeding (I²= 0%, P = 0.66), symptomatic intracranial hemorrhage (I²= 0%, P = 0.94), nonmajor bleeding (I²= 0%, P = 0.43), and systemic embolism (I²= 0%, P = 0.83). However, pooled studies were heterogeneous in all-cause mortality (I²= 71%, P = 0.03). Heterogeneity was best resolved by excluding Oldgren et al (I²= 0%) (Figure 5).

Figure 5.

Sensitivity analysis of all-cause mortality.

DISCUSSION

In this systematic review and meta-analysis of three randomized controlled trials investigating the optimal timing for anticoagulation initiation following acute ischemic stroke in patients with AF, we found no significant differences in the primary outcome between early and late initiation of DOACs. Furthermore, there were no significant differences in major extracranial bleeding, symptomatic intracranial hemorrhage, nonmajor bleeding, systemic embolism, recurrent ischemic stroke, or all-cause mortality between the early and late initiation groups.

Vitamin K antagonists (VKAs), such as warfarin, are widely used anticoagulants that reduce blood clotting by inhibiting vitamin K activity. Warfarin therapy, when managed to maintain an international normalized ratio within the target range of 2.0 to 3.0, has been demonstrated to decrease the risk of stroke by approximately two-thirds in patients with AF and is more effective than antiplatelet therapy.³ VKAs are still recommended as a treatment option for AF patients with an elevated risk of thromboembolic events, as outlined in both European and American clinical guidelines.^18,19 Despite their effectiveness, only 50% to 60% of eligible patients with AF are treated with VKAs, and many of those who are prescribed VKAs receive suboptimal management.^20,21 One major barrier to optimal VKA use is the fear of bleeding complications, particularly among elderly patients.^22,23 Additionally, VKAs are associated with significant variability in patient response, have numerous food and drug interactions, and possess a narrow therapeutic index. Consequently, VKA therapy requires frequent monitoring and dose adjustments, which can be burdensome for both patients and healthcare providers, making effective management challenging in routine clinical practice.²⁴

DOACs are now recognized as the preferred alternative to warfarin for primary and secondary ischemic stroke prevention in patients with AF.^25,26 DOACs have been shown to be at least as effective as warfarin, with a favorable bleeding profile in large randomized trials. However, these trials do not guide the optimal timing for initiating DOAC therapy in the early phase after an acute ischemic stroke, as a recent stroke (within 7–30 days) was an exclusion criterion.^2,27–31

The high risk of stroke recurrence during the acute phase following an ischemic stroke supports the rationale for initiating very early stroke preventive therapy.^8,32 However, the potential benefits of early DOACs must be balanced against the risks of hemorrhagic transformation of the ischemic lesion and the development of intracerebral hemorrhage during this vulnerable period.³³ A recent systematic review of 28 trials from the Cochrane Library concluded that while early anticoagulation is associated with a reduction in recurrent ischemic strokes, this advantage is counterbalanced by a similar increase in the incidence of intracranial hemorrhages.³³

A randomized study from Korea involving 195 patients examined the initiation of treatment with either the DOAC rivaroxaban or warfarin within 5 days following a mild stroke in patients with AF. The study reported no significant differences in the primary outcome, which assessed new ischemic or hemorrhagic lesions on brain imaging after 4 weeks.³⁴ Similarly, the randomized AREST trial (Apixaban for Early Prevention of Recurrent Embolic Stroke and Hemorrhagic Transformation) evaluated early anticoagulation with apixaban, followed by later initiation of warfarin, in patients with AF who had experienced a stroke or transient ischemic attack.³⁵ Although this study had a limited sample size, it reported no cases of symptomatic intracerebral hemorrhage in the DOAC-treated patients. Meanwhile, 2.1% of patients treated with warfarin experienced symptomatic intracerebral hemorrhages.³⁵

Additionally, none of the patients in the TIMING study experienced symptomatic intracerebral hemorrhage during the 90-day follow-up period. Overall, rates of major bleeding, including intracranial hemorrhages, were very low in the first 4 weeks.¹⁰ Other observational studies conducted concurrently with or after the TIMING study have also shown significantly lower rates of symptomatic intracranial hemorrhages in patients treated early with DOACs following ischemic stroke compared to those receiving warfarin.^36–38

Our findings align with those of randomized trials investigating early initiation of DOACs, including the OPTIMAS,¹³ TIMING,¹⁰ ELAN,¹² and START trials.³⁹ Together, these results provide reassurance regarding the safety of early anticoagulation with a DOAC and challenge the current guideline-supported practice of delaying DOACs for up to 14 days after moderate-to-severe acute ischemic stroke in patients with AF. Furthermore, early DOAC initiation may offer the practical advantage of increasing the proportion of patients who begin secondary prevention treatment before hospital discharge. However, our data do not specifically demonstrate this benefit. This potential advantage warrants further investigation in future studies. However, it is important to note that DOACs are primarily eliminated via the kidneys, with renal excretion rates ranging from 80% for dabigatran to 27% for apixaban.³⁷ Therefore, caution is needed when prescribing these agents to patients with renal impairment. In individuals with mild to moderate renal dysfunction, the reduced bleeding risk associated with DOACs compared to warfarin can generally be maintained with appropriate dose adjustments. In contrast, for patients with severe renal impairment or those undergoing dialysis, the risks of DOAC therapy are not well understood. In such cases, DOACs should be avoided, and VKAs should be considered an alternative until further data from future trials clarify the safety of DOACs in these populations.³⁷

Moreover, clinical decision support systems (CDSS) have emerged as a cost-effective intervention to enhance healthcare process measures.⁴⁰ Given the findings of our study, CDSS could play an important role in guiding clinicians in making timely decisions about anticoagulation therapy. One meta-analysis highlighted the potential of CDSS to reduce the incidence of myocardial infarction and cerebral or systemic embolic events in patients with AF; however, it also noted no significant differences in rates of oral anticoagulant and antiplatelet prescribing, all-cause mortality, major bleeding, or clinically relevant nonmajor bleeding.⁴¹ CDSS can potentially support adherence to guidelines for anticoagulation initiation, which may improve patient outcomes. Future research should explore how integrating CDSS with emerging evidence on early DOAC initiation can optimize treatment strategies for patients with AF following acute ischemic stroke.

Our study has certain limitations. A key limitation is the inherent heterogeneity observed in the included studies, likely resulting from differences in study designs, patient populations, initiation time, and definition of primary composite. While the sensitivity analysis addressed this issue, heterogeneity may still influence the interpretation of the overall results. Additionally, most included studies focused on short-term outcomes, restricting our ability to conclude early versus late therapy’s long-term effectiveness and safety. Furthermore, the timing of anticoagulation initiation in clinical practice is highly individualized and influenced by patient-specific factors such as stroke severity, hemorrhagic transformation, need for surgical intervention, concomitant antithrombotic therapy, and overall bleeding risk, which could not be fully accounted for in this trial-level analysis. Lastly, excluding non-English language studies could have omitted relevant data, potentially affecting the comprehensiveness of our analysis.

Despite these limitations, our study offers critical insights into the timing of DOAC initiation for managing stroke patients with AF, providing a foundation for future research and clinical practice. While current evidence suggests no significant difference between early and late DOAC initiation in terms of recurrent stroke or major bleeding risks, the optimal timing remains unclear. Given the high risk of early recurrent strokes and the potential for hemorrhagic transformation, further investigation is needed to determine the safest and most effective timing for initiating anticoagulation therapy. Long-term studies focusing on individual patient characteristics, such as renal function and stroke severity, are essential for refining treatment strategies.

In conclusion, early initiation of DOACs is not associated with an increased risk of major bleeding or mortality compared to delayed initiation. Given the high recurrence risk of ischemic stroke during the acute phase, early anticoagulation may enhance patient outcomes, although further research is needed to confirm these benefits.

Disclosure statement/Funding

The authors report no funding or conflicts of interest.