Introduction
Direct oral anticoagulants (DOAC) are the treatment of choice for ischemic stroke prevention in people with atrial fibrillation (AF) [1,2]. Recent observational studies and randomized trials including ELAN (Early versus Late Initiation of Direct Oral Anticoagulants in Post-ischemic Stroke Patients with Atrial Fibrillation) support the early initiation of DOAC after an acute ischemic stroke [3-5], but did not address what the optimal antithrombotic treatment strategy up to the initiation of DOAC should be. In clinical practice, antiplatelets are commonly used until patients are started on DOAC.
However, the evidence for this approach is weak, emanating primarily from randomized controlled trials from the late 90s. The International Stroke Trial and Chinese Acute Stroke Trial showed only a small net benefit from early treatment with aspirin, which reduced the rate of recurrent ischemic stroke or death at the expense of a slightly higher risk of bleeding complications over a treatment period of a few weeks compared to no treatment [6,7]. Similar results were seen in a meta-analysis of people with stroke and AF [8]. The latest guidelines on recurrent stroke prevention in people with AF either make no mention of antiplatelet use before anticoagulant initiation (American Heart/Stroke Association) [1] or make a weak recommendation in favor of antiplatelet use based on an expert panel’s majority decision (European Stroke Organisation; ESO) [2].
With early DOAC initiation becoming increasingly more common in everyday practice, the additional benefit of a preceding short course of antiplatelet therapy has been called into question. Persisting altered platelet function after switching from antiplatelets to DOAC might even increase the risk of hemorrhagic complications, which could offset any benefit in thromboembolic protection. Notably, bridging with heparin up to the initiation of oral anticoagulants is discouraged in the latest ESO guidelines [2], as it seems to increase bleeding risk without reducing stroke recurrence [9,10].
With these considerations in mind, we wondered whether treatment with antiplatelets before DOAC initiation might be associated with any benefit or whether it might even be linked to worse outcomes compared to directly initiating DOAC without prior antiplatelet treatment. We explored these questions in a post-hoc analysis of data from the ELAN trial.
Methods
Study design and participants
This is an exploratory post-hoc analysis of ELAN trial data. The ELAN trial protocol and main results have been published elsewhere. 3,11 In short, ELAN (NCT03148457) randomized participants with acute ischemic stroke and AF to early (<48 h after minor and moderate stroke, 6-7 days after major stroke) versus late (3-4 days after minor, 6-7 days after moderate, 12-14 days after major stroke) DOAC initiation across 103 stroke units and centers in Europe, the Middle East, and Asia between October 2017 and December 2022. Antiplatelet treatment before DOAC initiation was at the discretion of the local physicians. The ELAN trial protocol allowed the inclusion of participants receiving single or short-term dual antiplatelet therapy before randomization. People with pre-existing long-term dual antiplatelet therapy were excluded from participation, and no antiplatelet treatment was allowed after DOAC initiation in any participants.
In this study, we included all evaluable randomized ELAN participants, as in the main trial analysis.3 We excluded participants with incomplete information on antiplatelet use before DOAC initiation.
The main baseline variable of interest was antiplatelet use at trial inclusion (i.e., before DOAC initiation), defined as use of aspirin, clopidogrel, or any other antiplatelet agent at any dose and in any combination. This information was captured in detail in the trial database in a standardized manner, but the indications and duration of antiplatelet treatment were not recorded. Whether antiplatelets were already used at stroke onset was also not captured and was therefore unavailable for analysis. We also used the following baseline variables, defined as in the main trial [3,11]: age, sex, stroke classification (mild, moderate, major; according to infarct size on neuroimaging based on a standardized visual rating scheme as in the main trial), National Institutes of Health Stroke Scale (NIHSS) score at trial randomization, randomized allocation to early versus late treatment, CHA2DS2-VASc (congestive heart failure, hypertension, age >75 years, diabetes, stroke or transient ischemic attack [TIA], vascular disease, age 65 to 74 years, and sex) score, detailed cardiovascular risk factors and comorbidities including history of ischemic stroke or TIA, systemic embolism, hypertension, myocardial infarction, heart failure, peripheral artery disease, diabetes, dyslipidemia, and smoking status, as well as creatinine clearance, pre-stroke modified Rankin Scale (mRS) score and stroke etiology categorized to AF-only versus AF-plus according to the presence or absence of competing stroke mechanisms other than AF as reported by local investigators.
Follow-up data included absence or presence and timing of recurrent ischemic stroke, systemic embolism, symptomatic intracranial hemorrhage (ICH), major extracranial bleeding, or vascular death within 30 days after trial inclusion, defined as described previously [3,11].
All study data were gathered by local investigators and collected in a web-based database hosted by the Clinical Trial Unit (CTU) Bern, Switzerland.
Outcomes
The primary outcome was the composite of recurrent ischemic stroke (defined as evidence of acute cerebral infarction on neuroimaging, or as a clinical diagnosis with symptoms lasting >24 hours and exclusion of other causes on neuroimaging), systemic embolism (defined as clinical or radiological evidence of abrupt arterial occlusion of an extremity or organ other than the brain in the absence of another likely mechanism), symptomatic ICH (defined as subdural, epidural, subarachnoid, or intracerebral hemorrhage leading to clinical symptoms, hospitalization, or death), major extracranial bleeding (defined as fatal, life-threatening, or clinically overt hemorrhage associated with a hemoglobin level decrease of ≥2 g/dL over a 24-hour period, transfusion of ≥2 units of packed red cells, or occurring in a critical body part), or vascular death (defined as any death due to a vascular cause, including cardiac death and death due to ischemic or hemorrhagic events) within 30 days after trial inclusion, as in the main analysis [3]. Secondary outcomes were the separate components of the primary composite outcome, all assessed at 30 days. Symptomatic ICH and major extracranial bleeding were merged into a single major bleeding outcome due to low event numbers to allow for statistical analysis. All outcomes were centrally adjudicated, as described previously [3,11].
Statistical analysis
We present baseline variables stratified by use versus non-use of antiplatelets prior to DOAC initiation using descriptive statistics, i.e., frequencies and percentages for categorical data and the median and interquartile range (IQR) for continuous data. We compared categorical variables using the χ2-test and continuous variables using the Mann-Whitney U test.
We analyzed the primary and secondary outcomes without any formal hypothesis testing using (1) Firth’s logistic regression (considering the small number of outcome events) and (2) cause-specific Cox proportional hazards regression (to account for competing risks). For the primary composite outcome, non-vascular death was the only competing risk. For the respective secondary outcomes, competing risks were non-vascular death and all other components of the composite outcome. We fitted (1) simple unadjusted models including only antiplatelet treatment before DOAC initiation (use vs. non-use; non-use as reference) as the sole independent variable without further adjusting for confounding and (2) weighted models using stabilized inverse probability of treatment weighting (SIPTW) to control for potential confounding through imbalances in all baseline data listed above. Justified by the overall low missingness rate, we used simple (single) imputation of missing values of the baseline categorical variables used for SIPTW. For this, we imputed missing data by sampling from a binary distribution with a probability derived from the prevalence observed in the available data. We report the rate of missing values for all variables. We truncated weights to the 1st and 99th percentiles of their distribution. We assessed the balance of baseline characteristics before and after weighting using the standardized mean difference (SMD). We estimated the weighted Cox models using robust standard errors. For all models, we report the model-based odds or hazard ratio (OR or HR) estimates along with 95% confidence intervals (CI) as measures of association, but refrain from presenting P values and from specifying any thresholds for “statistical significance.” We present the cumulative incidence of outcome events using the Aalen-Johansen estimator, taking into account competing risks by treating all other events as competing events.
We additionally analyzed the net clinical benefit of antiplatelet use over non-use, adopting established methodology as in prior research [12,13], with weighting of the type of events for their impact on death and disability relative to recurrent ischemic stroke or systemic embolism. We calculated the net benefit of treatment by subtracting the weighted rate of excess bleeding events attributable to antiplatelets from the rate of excess ischemic events possibly prevented by antiplatelets, according to the following formula:
where Rischemic represents the rate of recurrent ischemic stroke or systemic embolism, RICH the rate of ICH, and Rmajor extr. bleed the rate of major extracranial bleeding. Weight values were derived from the literature [13-15], with major extracranial bleeding being assigned a weight of 0.7 (weightmajor extr. bleed), while the weight for ICH varies from 1.5 to 3.1 (weightICH). We performed the net benefit analysis across the entire range of ICH weights, as in prior research [12,16]. We extracted the outcome event rates from the weighted survival model. We report the net benefit in events per 100 person-months along with 95% CI, calculated based on 1,000 bootstrap replications. For the net benefit analyses, we considered all participants but those with death as first outcome.
In sensitivity analyses, we investigated whether (1) the timing of DOAC initiation (early vs. late; according to the randomized allocation), (2) stroke etiology (AF-only vs. AF-plus), and (3) stroke classification (mild, moderate, major; according to infarct size) modify the association of antiplatelet use with the outcomes by adding appropriate interaction terms to the aforementioned weighted logistic models. For this, we report the interaction P values as a continuous measure, with smaller values indicating stronger evidence for interaction, without specifying any threshold. Additionally, we repeated all aforementioned analyses in the subgroup of participants with major stroke, which is the most pertinent for this study’s research question, after excluding participants with mild or moderate stroke. As a final sensitivity analysis, we repeated the aforementioned logistic regression and survival analyses after excluding participants with comorbidities that may have warranted antiplatelet treatment before stroke onset (i.e., those with history of previous ischemic stroke or TIA, systemic embolism, myocardial infarction, or peripheral artery disease). We used these comorbidities as a proxy for antiplatelet treatment at stroke onset and for alternative indications for antiplatelet use, since this data was not available for analysis.
Statistical analyses were performed using R version 4.3.1 (R Foundation for Statistical Computing, Vienna, Austria) based on a predefined statistical analysis plan. We conducted this study in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement for observational studies [17]. The data that support the findings of this study may be made available on reasonable request to the corresponding author and after approval by the ELAN steering committee.
Ethics
The ELAN protocol was approved by all responsible ethics committees and, if applicable, by the regulatory authorities in the countries in which the trial was conducted. The participant, next of kin or another legal representative, or an independent physician provided written informed consent before enrolment, according to country-specific requirements. ELAN was conducted in accordance with the Good Clinical Practice guidelines of the International Council for Harmonisation E6 requirements and the Declaration of Helsinki.
Results
Information on antiplatelet use before DOAC initiation was complete, so that the full ELAN dataset of 2,013 participants was available for analysis, as in the main trial.
Baseline characteristics
In total, 1,090 participants used antiplatelets before DOAC initiation, and 923 did not. In both groups, the median age was 77 years, 45.5% of participants were female, median NIHSS score at trial randomization was 3, and median CHA2DS2-VASc score was 5 (Table 1). The remaining baseline characteristics and vascular comorbidities were generally balanced between the groups except stroke size (with major stroke being more common among participants with antiplatelet treatment and minor or moderate stroke more common among those without antiplatelet treatment), randomized allocation (early DOAC initiation more common in participants without antiplatelet treatment), and history of myocardial infarction (more common in those with antiplatelet treatment). Weighting achieved good balance (absolute SMD <0.05) across all baseline characteristics (Supplementary Figure 1). Of participants treated with antiplatelets, monotherapy with aspirin was most common (88.4%), followed by monotherapy with clopidogrel (7.8%), dual therapy with aspirin and clopidogrel (2.8%), or other antiplatelet treatments (Table 2).
Primary and secondary outcome analyses
Among 2,013 participants, we observed a total of 70 primary outcome events within 30 days, consisting of 39 recurrent ischemic strokes, 8 major extracranial bleedings and 4 ICH, 13 systemic embolic events, and 21 vascular deaths. Most recurrent ischemic strokes occurred early after trial randomization at a median (IQR) of 3 (2-7) days, while major bleeding events were more evenly distributed throughout the 30-day follow-up. The number and timing of all outcomes stratified by use versus nonuse of antiplatelets before DOAC initiation are shown in Table 3. Of all 2,013 participants, 13 withdrew consent at some point during study follow-up, 1 was lost to follow-up, and 24 died from non-vascular causes. These were included in survival analyses with appropriate censoring, but were excluded from logistic regression analyses, which were therefore carried out on the remaining 1,975 participants.
In unadjusted logistic models, antiplatelet treatment before DOAC initiation was not associated with the primary composite outcome, systemic embolism, or vascular death, but showed lower odds of ischemic stroke recurrence and higher odds of major bleeding. Weighted logistic models accounting for baseline imbalances yielded virtually unchanged results (Figure 1A). Both unadjusted and weighted cause-specific Cox proportional hazards regression models showed no association of antiplatelet treatment with the primary composite outcome, systemic embolism, or vascular death. Consistent with the logistic models’ results, in both unadjusted and weighted survival models, antiplatelet treatment showed a lower hazard for ischemic stroke recurrence, but a higher hazard for major bleeding (Figure 1B). The Aalen-Johansen estimates for the primary and secondary outcomes stratified by antiplatelet treatment before DOAC initiation are presented in Figure 2.
Net clinical benefit analysis
The point estimates for the net benefit of antiplatelet use over non-use before DOAC initiation were small but consistently positive over the entire range of ICH weights used (0.57 to 0.30 weighted events possibly prevented with antiplatelet use per 100 person-months for ICH weights 1.5 to 3.1), with wide 95% CIs crossing zero (Figure 3). The detailed results of the net benefit analysis are presented in Supplementary Table 1.
Sensitivity analyses
Repeating the weighted logistic models including appropriate interaction terms showed no evidence for interaction between antiplatelet use and (1) timing of DOAC initiation, (2) stroke etiology, or (3) stroke size classification on their association with the primary composite outcome (Pinteraction=0.600, 0.302, and 0.928, respectively). There was no evidence for interaction in any of the secondary outcomes, i.e., recurrent ischemic stroke (Pinteraction= 0.728, 0.802, and 0.412, respectively), major bleeding (Pinteraction= 0.362, 0.666, and 0.969, respectively), systemic embolism (Pinteraction=0.937, 0.444, and 0.273, respectively), and vascular death (Pinteraction=0.194, 0.950, and 0.516, respectively).
Repeating all logistic regression and survival analyses in participants with major stroke (n=465) resulted in findings consistent with the whole study population analysis, with antiplatelet treatment again showing a lower risk of ischemic stroke recurrence and a higher risk of major bleeding (Supplementary Figure 2). The net benefit analysis in participants with major stroke showed consistently positive point estimates in favor of antiplatelet use across the entire range of ICH weights used, with threefold higher net benefit values compared to the whole study population analysis (1.92 to 0.81 weighted events possibly prevented per 100 person-months for ICH weights 1.5 to 3.1), again with wide 95% CIs crossing zero (Supplementary Table 2 and Supplementary Figure 3).
Finally, logistic regression and survival analyses after excluding participants with history of previous stroke or TIA, systemic embolism, myocardial infarction, or peripheral artery disease yielded consistent results with the whole study population analysis (Supplementary Figure 4).
Discussion
This exploratory post-hoc analysis of the randomized ELAN trial investigated the risks and benefits of antiplatelet treatment prior to DOAC initiation after an acute ischemic stroke in people with AF. We estimated that the risk of recurrent ischemic stroke was lower and bleeding risk higher when antiplatelets were used before DOAC initiation, but identified no signals of overall harm with antiplatelet use. However, the generally low event numbers lead to substantial imprecision in our estimates, precluding any definitive conclusions.
While common in clinical practice, bridging with antiplatelets up to DOAC initiation following an acute ischemic stroke in people with AF is not backed by strong evidence. Still, antiplatelet bridging is recommended in the latest European, United Kingdom, and Canadian guidelines for secondary stroke prevention [2,18,19], but this practice has been called into question, particularly when DOACs are started early after stroke. Despite strong research interest and a series of recent randomized controlled trials examining the timing of DOAC initiation after ischemic stroke [3,5,20], no randomized or observational studies have so far addressed the question of antiplatelet bridging before DOAC initiation.
In our study, antiplatelet use before DOAC was not associated with the primary composite outcome of recurrent ischemic stroke, systemic embolism, major bleeding, and vascular death. This held true in both logistic and time-to-event models, and even after accounting for confounders. Showing no differences in a combined vascular outcome (comprising both ischemic and hemorrhagic events) in a large cohort, our data raise no concerns about the overall effects of antiplatelet bridging and are reassuring for this practice. Moreover, we found consistent signals that the risk of recurrent ischemic stroke may be lower and the risk of bleeding higher with antiplatelets. Importantly, when considering the net clinical effect of antiplatelet use over non-use, which accounts for the different clinical importance of ischemic and hemorrhagic events rather than merely summing events up, we estimated a small net benefit with antiplatelet use. Despite substantial imprecision, these analyses again yielded no indication that antiplatelet treatment might be associated with overall harm. Of note, our main results were not modified by the timing of DOAC initiation or by stroke size. In fact, consistent findings resulted when examining the subgroup of participants with major stroke alone. This is important because DOAC initiation is delayed the longest in people with major stroke, in whom the question of antiplatelet bridging is most pertinent. Notably, our net benefit analyses in this subgroup yielded estimates of even more pronounced benefit in these patients, albeit with substantial imprecision warranting cautious interpretation.
Taken together, these findings present no concerns about the weak recommendation for antiplatelet bridging made in several guidelines and provide no justification for abandoning this practice. As a possible explanation for these findings, we posit that, in the absence of anticoagulation, even a short course of antiplatelets may reduce the early risk of recurrent ischemic stroke attributable to non-cardioembolic atherothrombotic mechanisms in particular, which often coexist with AF [21]. Although our results were not modified by stroke etiology, this hypothesis is consistent with the known modest reduction of ischemic stroke risk conferred by antiplatelets among people with AF not receiving anticoagulation [22], as best exemplified by the effects of intensive antiplatelet treatment observed in the ACTIVE-A (Atrial Fibrillation Clopidogrel Trial with Irbesartan for Prevention of Vascular Events - Aspirin arm) trial [23]. Notably, most recurrent strokes in ELAN occurred within a few days from randomization, with a pronounced early separation of the incidence curves between participants with versus without antiplatelet use—a phenomenon reminiscent of the concentrated benefit of early intensive antiplatelet therapy in the first days following a non-cardioembolic stroke that was observed in the CHANCE (Clopidogrel in High-Risk Patients with Acute Nondisabling Cerebrovascular Events), POINT (Platelet-Oriented Inhibition in New TIA and Minor Ischemic Stroke), and THALES (Acute Stroke or Transient Ischemic Attack Treated with Ticagrelor and ASA for Prevention of Stroke and Death) trials [16,24].
Our study has the following strengths. We used high-quality, prospectively collected data from a large randomized controlled trial with central outcome event adjudication. In addition, we employed several lines of statistical inquiry, all leading to consistent results, lending confidence to our findings. We acknowledge the following limitations. First, ELAN was not designed to address the research question of this post-hoc analysis of nonrandomized, unbalanced comparison groups, which should be considered purely exploratory and precludes causal inference. Second, residual confounding through unmeasured baseline imbalances might have affected our results and cannot be ruled out, despite unadjusted and weighted models (that included extensive controlling for measured confounders) yielding similar findings. Thrid, the generally low event numbers introduce imprecision in our estimates, as reflected in the wide 95% CI that cross the null effect line in several analyses, thus preventing definitive conclusions. Fourth, as the vast majority of participants with antiplatelet use had aspirin monotherapy, our results may not be generalizable to other antiplatelet treatment strategies, such as dual antiplatelet therapy. Finally, our study lacked information about the indication for antiplatelet use at trial inclusion or its duration, and had no data about antiplatelet treatment at the time of stroke onset. Therefore, we cannot rule out misclassification bias—that is, participants treated with antiplatelets before stroke, in whom antiplatelets were discontinued on admission, may have been categorized into the no-antiplatelet group. Considering the long-lasting antithrombotic effects of antiplatelets, this might have confounded or diluted our results, although sensitivity analyses excluding participants with comorbidities that may have warranted antiplatelet treatment before stroke onset or warranted treatment for indications other than bridging stroke prevention yielded highly consistent findings.
Conclusions
We identified a lower risk of stroke recurrence and higher bleeding risk with antiplatelets before DOAC initiation, but no signs of overall harm. Despite imprecision in the estimates, these findings provide no rationale to abandon the practice of starting preventive antiplatelet treatment early after an AF-associated ischemic stroke before initiating DOAC.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
