Prior Anticoagulation and Risk of Hemorrhagic Transformation in Acute Stroke: A Post Hoc Analysis of the PRODAST Study

Abstract

Background

Oral anticoagulation is highly effective in preventing ischemic events in patients with atrial fibrillation. Still, a considerable number of patients have an acute ischemic stroke or transient ischemic attack despite anticoagulation. In this study, we investigated the association of prior antithrombotic regimens with stroke severity, volume, and hemorrhagic transformation.

Methods and Results

This is a post hoc analysis of the prospective, multicenter, observational PRODAST (Prospective Record of the Use of Dabigatran in Patients With Acute Stroke or TIA) study, which was conducted in 86 stroke units in Germany between July 2015 and November 2020. In 9030 patients with atrial fibrillation who had an acute ischemic stroke or transient ischemic attack within 7 days before enrollment, we analyzed the association of anticoagulants in comparison to lack of prevalent antithrombotic treatment with clinical stroke severity, infarct size, and risk for hemorrhagic transformation. A total of 4479 patients had prior anticoagulation at the time of index event. After adjustment for confounders (arterial hypertension, diabetes, heart failure, age, and sex), patients with prior anticoagulation had less severe strokes (−2.5 National Institutes of Health Stroke Scale points [95% CI, −2.8 to −2.2]), smaller infarct sizes (−23 mL [95% CI, −44 mL to −2 mL], n=4041), and reduced odds for hemorrhagic transformation (5% versus 10%; odds ratio, 0.48 [95% CI, 0.40–0.57]) compared with patients without antithrombotic treatment. These findings were confirmed using sensitivity analyses accounting for thrombolysis and mechanical thrombectomy, as well as timing of brain imaging. Antiplatelet therapy had hardly any association with the end points compared with no antithrombotic pretreatment.

Conclusions

Prior anticoagulation was not only associated with less severe stroke and smaller infarct size but also with a reduced risk of hemorrhagic transformation compared with no antithrombotic pretreatment.

Registration

URL: https://www.clinicaltrials.gov; Unique identifier: NCT02507856.

Nonstandard Abbreviations and Acronyms

AIS

acute ischemic stroke

ARAIS

Argatroban plus r‐tPA for Acute Ischemic Stroke

BBB

blood brain barrier

DOAC

direct‐acting oral anticoagulant

ECASS

European Cooperative Acute Stroke Study

HT

hemorrhagic transformation

NIHSS

National Institutes of Health Stroke Scale

OAC

oral anticoagulation

PRODAST

Prospective Record of the Use of Dabigatran in Patients With Acute Stroke or TIA

VKA

vitamin K antagonist

Clinical Perspective.

What Is New

• This study shows that ischemic strokes occurring in patients with anticoagulation may be less severe and smaller and may have a reduced odds for hemorrhagic transformation compared with strokes in patients without prior antithrombotic treatment.

What Are the Clinical Implications?

• Contrary to the common notion, prior anticoagulation is not only safe but may actually be linked to a lower risk of early hemorrhagic transformation in acute ischemic stroke.

• This underscores the beneficial properties of oral anticoagulants and the necessity of adequate drug adherence in patients with atrial fibrillation.

• Further evidence is urgently needed to establish the clinical relevance of these findings, especially regarding the application of thrombolysis in patients with effective anticoagulation.

In ≈25% to 30% of all ischemic strokes, the underlying cause is a cardiac embolism due to atrial fibrillation (AF). However, this proportion may be even higher because a significant number of patients with cryptogenic stroke are diagnosed with AF in the long‐term course. ¹ , ² Acute ischemic stroke (AIS) due to AF is associated with larger infarct size, worse clinical outcome, and higher probability for hemorrhagic transformation (HT) as compared with stroke of other causes. ³ Oral anticoagulants can dramatically reduce the risk for recurrent stroke in patients with AF by ≈60% to 70%. ⁴ Adherence to direct‐acting oral anticoagulants (DOACs) is modestly higher compared with vitamin K antagonists (VKAs), and nonadherence is associated with adverse events. ⁵ However, even in patients undergoing effective anticoagulation, 1% to 2% of cases have an acute ischemic event each year despite this treatment based on the patient medical history. ⁶ , ⁷ A detailed characterization of strokes occurring despite anticoagulation compared with no anticoagulation is not yet available. ⁷

Use of VKAs dosed in an international normalized ratio (INR) target of 2 to 3 was associated with reduced stroke severity and mortality according to a study published in 2003. ⁸ , ⁹ Over the past years, DOACs emerged as standard therapeutics for anticoagulation in patients with nonvalvular AF. DOACs are at least as effective and safer compared with VKAs. ¹⁰ Meinel et al and Xian et al reported a lower stroke severity in patients with effective DOAC pretreatment compared with no antithrombotic pretreatment. ¹¹ , ¹² This finding is in accordance with Japanese registry studies with smaller sample sizes. ¹³ , ¹⁴ , ¹⁵ However, it is unknown whether there are differences between the various anticoagulant substances compared with no antithrombotic treatment in terms of infarct size and severity of AIS and occurrence of HT. Infarct size is a major risk factor for HT. ¹⁶ , ¹⁷ It may be hypothesized that DOACs are superior to VKAs at reducing infarct size and the risk of HT. ¹⁸

Therefore, in this study, we aimed to investigate the association between prior use of different anticoagulants and antiplatelets with stroke severity, infarct size, and the risk of HT. We used data from the PRODAST (Prospective Record of the Use of Dabigatran in Patients With Acute Stroke or TIA) study, which is the largest prospective registry study to date on early or prevalent application of anticoagulants in patients with AIS or transient ischemic attack.

METHODS

Anonymized data may be shared upon reasonable request by any qualified investigator via contacting the corresponding author.

Design and Study Population

Design and methods of the PRODAST study (ClinicalTrials.gov identifier: NCT02507856) have been previously reported. ¹⁹ , ²⁰ Briefly summarized, PRODAST is a multicenter, prospective, observational, postauthorization safety study that enrolled 10 039 patients diagnosed with AF who had experienced an AIS or transient ischemic attack within the previous 7 days. Recruitment took place at 86 stroke units in Germany from July 2015 to November 2020. Further inclusion criteria were an age of 18 years or older at the time of enrollment and diagnosis of nonvalvular AF. Patients with mechanical heart valves or valve disease that indicated a need for valve replacement within the following 3 months were excluded. Similarly, those engaged in any randomized controlled trial involving an experimental drug or medical device, women of reproductive age lacking anamnestic exclusion of pregnancy, individuals not using effective contraception, and nursing mothers were excluded. All procedures were at the discretion of the treating physicians. For this analysis, we restricted the cohort to patients with documented preceding antithrombotic monotherapy at baseline or no antithrombotic treatment. Combination treatments or parenteral antithrombotic treatments were excluded (n=1009) because the indication for these treatments is even more complex, and residual confounding becomes more likely. In total, 9030 patients were eligible for this analysis.

Data Documentation and Outcomes

The study sites were responsible for documentation of clinical and demographic data, brain imaging results (including infarct size), and evaluation of HT, as well as treatment status including information on whether anticoagulants were effective at the time of the index event. Information on acute therapies (ie, mechanical thrombectomy and intravenous thrombolysis), time of last intake of antithrombotic treatments, and timing of the considered brain imaging, was documented. For patients taking VKAs, we also considered the INR values at hospital admission. Data quality was monitored by the Center for Clinical Trials Essen with regular on‐site visits. To ensure the eventual effect of the antithrombotic therapy at the time of the event, we added a substance‐specific lag time to the last time of ingestion. This means that the end of actual treatment was postponed by 1 day for DOACs, 7 days for antiplatelets, and 5 days for VKAs as previously described. ¹⁹ For estimation of infarct volume, the ABC/2 formula (A: largest diameter of the infarction in the axial plane, B: largest diameter of the infarction at 90° to A in the axial plane, C: number of slices of the infarction multiplied by the slice thickness) was applied. ²¹ HT was graded using the European Cooperative Acute Stroke Study (ECASS) criteria ²² on baseline cerebral imaging. In a subanalysis, parenchymal hemorrhages (according to the ECASS classification of parenchymal hemorrhage 1 and 2) were considered.

Ethics

For all patients, written informed consent was provided. In the event that informed consent could not be obtained in a timely manner because of the patient's condition, the treating physician was able to decide on the study inclusion, whereupon the informed consent process was rescheduled as soon as possible. This enabled the study teams to also recruit patients with severe stroke. Ethical approval was provided by the institutional review board of the University of Duisburg‐Essen (No. 15‐6202‐BO). All procedures were conducted in accordance with German law, the 1964 Declaration of Helsinki and its later amendments, and the recommendations of the guidelines on Good Clinical Practice and Good Epidemiological Practice.

Statistical Analysis

We aimed to analyze the association of different prevalent antithrombotic agents (ie, dabigatran, apixaban, edoxaban, rivaroxaban, VKAs, or antiplatelet therapy) in comparison to no antithrombotic treatment, with clinical stroke severity, infarct size, and the risk for HT on initial brain imaging. For this purpose, linear regression analyses were applied with the National Institutes of Health Stroke Scale (NIHSS) or infarct volume (in mL) as outcome variables, respectively, and each antithrombotic agent (compared with no treatment) as predictors. Regression coefficients were then calculated. We calculated and reported CIs to assess the precision of our estimates because our goal was estimation rather than significance testing. Each model was adjusted for arterial hypertension, diabetes, heart failure, age, and sex, as these variables were identified as a sufficient adjustment set based on a directed acyclic graph (Figure S1). We also performed sensitivity analyses excluding patients with acute therapies (ie, intravenous thrombolysis or mechanical thrombectomy). In addition, we considered the time of brain imaging used for infarct size and HT evaluation, as well as INR values for patients taking VKA treatment. All analyses were conducted using SAS Enterprise Guide 7.1 (SAS Institute Inc).

RESULTS

The Table shows clinical and demographic data of the study cohort. A total of 4479 patients had prior anticoagulation at the time of the index event, with 288 receiving dabigatran, 1493 apixaban, 390 edoxaban, 900 rivaroxaban, and 1408 VKAs. A total of 1403 patients had preceding antiplatelet treatment,( and 3148 patients had no current antithrombotic therapy at the time of the index event. All patients with DOACs ie, dabigatran or factor Xa inhibitors) were taking effective anticoagulation, according to the documentation of the last drug intake. The prevalence of vascular risk factors and age in patients with different antithrombotic therapies were not different from patients without antithrombotic treatment. Median NIHSS scores ranged from 2 to 4. The median time from symptom onset until imaging used for analysis was 1 day in all groups.

Table 1.

Clinical and Demographic Characteristics

	No prior antithrombotic treatment	Prior dabigatran treatment	Prior factor Xa inhibition	Prior VKA treatment	Prior platelet inhibition
Number	3148	288	2783	1408	1403
Age, median (5th–95th percentile), y	78 (57–92)	79 (62–88)	80 (64–91)	80 (65–90)	80 (63–92)
Women	1682 (55%)	123 (43%)	1398 (50%)	587 (42%)	714 (51%)
Arterial hypertension	2446 (78%)	262 (91%)	2479 (89%)	1259 (89%)	1255 (89%)
Diabetes	696 (22%)	98 (34%)	916 (33%)	492 (35%)	455 (32%)
Congestive heart failure	274 (9%)	27 (9%)	440 (16%)	223 (16%)	205 (15%)
Renal insufficiency	75 (2%)	1 (0%)	89 (3%)	82 (6%)	73 (5%)
Liver insufficiency	28 (1%)	3 (1%)	14 (1%)	12 (1%)	15 (1%)
Dyslipidemia	966 (31%)	137 (48%)	1143 (41%)	587 (42%)	589 (42%)
Current smoker	335 (11%)	24 (8%)	174 (6%)	100 (7%)	129 (9%)
Previous smoker	689 (22%)	81 (28%)	709 (25%)	411 (29%)	339 (24%)
BMI, median (5th–95th percentile), kg/m2	26.20 (20.40–37.10)	26.50 (21.30–36.20)	26.30 (20.35–36.30)	26.60 (21.00–36.00)	26.50 (20.20–36.20)
History of ischemic stroke	282 (9%)	122 (42%)	800 (29%)	318 (23%)	319 (23%)
History of hemorrhagic stroke	13 (0%)	2 (1%)	22 (1%)	6 (0%)	10 (0%)
Potential symptomatic arterial stenosis	413 (13%)	41 (14%)	418 (15%)	180 (13%)	220 (16%)
NIHSS, median (5th–95th percentile)	4 (0–21)	3 (0–19)	3 (0–17)	2 (0–17)	4 (0–22)
Infarct volume, mL; median (5th–95th percentile) (missing %)	6.1 (0.055–16.821) (47%)	2.0 (0.023–76.908) (66%)	2.8 (0.038–11.224) (64%)	2.7 (0.040–12.150) (63%)	6.5 (0.063–14.803) (47%)
Early HT	310 (10%)	15 (5%)	131 (5%)	77 (5%)	147 (10%)
Parenchymal hematoma types 1 and 2	86 (3%)	2 (1%)	36 (1%)	22 (2%)	40 (3%)
Parenchymal hematoma type 2	26 (1%)	0 (0%)	10 (1%)	10 (1%)	16 (1%)
Time from symptom onset until brain imaging, median (5th–95th percentile), d	1 (0–9)	1 (0–9)	1 (0–8)	1 (0–8)	1 (0–9)
Thrombolysis	1004 (32%)	35 (12%)	83 (3%)	113 (8%)	411 (29%)
Mechanical thrombectomy	594 (19%)	28 (10%)	249 (9%)	132 (9%)	201 (14%)
CHA2DS2‐VASc score, median (5th–95th percentile)	5 (3–7)	6 (4–7)	6 (4–7)	6 (4–7)	6 (4–7)
Platelet count, median (5th–95th percentile), ×109/L	223 (130–367)	222 (132–371)	214 (130–357)	209 (122–348)	218 (132–375)
INR, median (5th–95th percentile)	1.04 (0.91–1.35)	1.19 (1.00–1.55)	1.18 (0.98–2.11)	2.01 (1.20–3.37)	1.04 (0.91–1.27)

BMI indicates body mass index; HT, hemorrhagic transformation; INR, international normalized ratio; NIHSS, National Institutes of Health Stroke Scale; and VKA, vitamin K agonist.

Association of Preceding Anticoagulant Use With Stroke Severity

After adjusting for confounders, compared with patients without antithrombotic treatment, those treated with DOACs had less severe strokes (−2.4 points on the NIHSS score [95% CI, −2.8 to −2.1]) (Figure 1). This was true for each DOAC agent, with minor differences and low precision of the estimates: dabigatran: −2.2 NIHSS points (95% CI, −3.0 to −1.4), apixaban: −2.8 NIHSS points (95% CI, −3.2 to −2.4), edoxaban: −2.1 NIHSS points (95% CI, −2.8 to −1.4), and rivaroxaban: −2.2 NIHSS points (95% CI, −2.6 to −1.7). Patients with prior VKA use had a comparably lower stroke severity according to the adjusted analysis: −2.7 NIHSS points (95% CI, −3.1 to −2.2). Prior antiplatelet therapy was not associated with stroke severity: −0.3 NIHSS points (95% CI, −0.8 to 0.1). Prior use of oral anticoagulants (OACs) in total (ie, DOACs+VKAs) was associated with a decrease of 2.5 NIHSS points (95% CI, −2.8 to −2.2) compared with no antithrombotic treatment. In a sensitivity analysis restricted to patients who did not receive thrombolysis (n=7384), prior anticoagulation remained associated with lower stroke severity (−1.7 NIHSS points [95% CI, −2.0 to −1.3]).

Figure 1. Adjusted estimates of differences in clinical stroke severity (measured by NIHSS) by prevalent antithrombotic treatment vs no antithrombotic treatment.

NIHSS indicates National Institutes of Health Stroke Scale, OAC,  oral anticoagulant (ie, apixaban, dabigatran, edoxaban, rivaroxaban, and VKAs); and VKA, vitamin K antagonist.

Association of Preceding Anticoagulant Use With Stroke Size

Information on infarct volumes was available in 4041 patients. Table S1 shows the baseline characteristics of patients with and without information on infarct volume. Patients with missing data on stroke volume had less severe strokes. Other baseline characteristics did not substantially differ. As depicted in Figure 2, prior use of DOACs was related to smaller infarct size on initial brain imaging according to the adjusted analysis, with a reduction of 23 mL (95% CI, −48 mL to 2 mL). The individual DOAC agents showed a similar effect size but with lower precision of estimation due to smaller strata: dabigatran: −26 mL (95% CI, −108 mL to 56 mL), apixaban: −23 mL (95% CI, −60 mL to 14 mL), edoxaban: −18 mL (95% CI, −82 mL to 45 mL), and rivaroxaban: −22 mL (95% CI, −66 mL to 22 mL). In patients with prior VKA use, infarct size was also lower: −21 mL (95% CI, −57 mL to 15 mL). Preceding antiplatelet use had no relevant association with stroke volume (−8 mL (95% CI, −40 mL to 23 mL). Considering all anticoagulants (ie, DOACs+VKAs) compared with no antithrombotic treatment was associated with a decrease in stroke volume of 23 mL (95% CI, −44 mL to −2 mL). In patients who did not receive thrombolysis, this association was also evident (−34 mL [95% CI, −60 mL to −7 mL]).

Figure 2. Adjusted estimates of differences in infarct volumes (n=4041) by prevalent antithrombotic treatment vs no antithrombotic treatment.

OAC indicates oral anticoagulant (ie, apixaban, dabigatran, edoxaban, rivaroxaban, and VKAs). VKA indicates vitamin K antagonist.

Association of Preceding Anticoagulant Use With Risk for HT

The odds for HT were more than halved in patients with prior use of DOACs compared with no antithrombotic therapy (odds ratio [OR], 0.46 [95% CI, 0.37–0.56]). This was also true for dabigatran (OR, 0.50 [95% CI, 0.30–0.86]), apixaban (OR, 0.41 [95% CI, 0.31–0.54]), edoxaban (OR, 0.60 [95% CI, 0.39–0.92]), rivaroxaban (OR, 0.46 [95% CI, 0.33–0.64]), and VKAs (OR, 0.53 [95% CI, 0.29–0.91]) but not for antiplatelet treatment (OR, 1.07 [95% CI, 0.87–1.32]) (Figure 3). Combining all OACs (ie, DOACs+VKAs) revealed an OR of 0.48 (95% CI, 0.40–0.57) compared with no antithrombotic therapy. This was also true for patients who did not receive thrombolysis (OR, 0.52 [95% CI, 0.425–0.641]). In a sensitivity analysis adjusting for NIHSS (and thus not estimating the total effect of preceding anticoagulation on HT), the aforementioned effects were only slightly weakened: total prior use of DOACs (adjusted OR [aOR], 0.60 [95% CI, 0.49–0.75]), dabigatran (aOR, 0.65 [95% CI, 0.38–1.12]), apixaban (aOR, 0.55 [95% CI, 0.41–0.73]), edoxaban (aOR, 0.76 [95% CI, 0.49–1.19]), rivaroxaban (aOR, 0.57 [95% CI, 0.41–0.80]), VKAs (aOR, 0.72 [95% CI, 0.55–0.95]), antiplatelet treatment (aOR, 1.10 [95% CI, 0.88–1.36]), and total OACs (aOR, 0.65 [95% CI, 0.54–0.78]).

Figure 3. Adjusted ORs for hemorrhagic transformation in initial brain imaging by prevalent antithrombotic treatment vs no antithrombotic treatment.

OAC indicates oral anticoagulant (ie, apixaban, dabigatran, edoxaban, rivaroxaban, and VKAs); OR, odds ratio; and VKA, vitamin K antagonist.

We also investigated the association of preceding antithrombotic treatment with parenchymal hemorrhages (according to the ECASS classification grades 1 and 2). The following effect estimates were observed: total prior use of DOACs (OR, 0.43 [95% CI, 0.29–0.63]), dabigatran (OR, 0.24 [95% CI, 0.06–0.99]), apixaban (OR, 0.30 [95% CI, 0.17–0.54]), edoxaban (OR, 0.81 [95% CI, 0.41–1.63]), rivaroxaban (OR, 0.54 [95% CI, 0.31–0.96]), VKAs (OR, 0.55 [95% CI, 0.34–0.88]), antiplatelet treatment (OR, 1.06 [95% CI, 0.72–1.54]), and all OACs (OR, 0.47 [95% CI, 0.33–0.65]).

See Figure S2 for an additional sensitivity analysis on the association of antithrombotic pretreatment with occurrence of parenchymal hemorrhages when excluding patients undergoing thrombolysis and mechanical thrombectomy. Of note, the effect sizes were largely unaltered in these analyses. Figure S2, moreover, shows an analysis regarding the timing of brain imaging that was considered for evaluation of stroke size and HT. Of note, the preventive effect of prior use of OACs was evident regardless of the timing of brain imaging as well as of acute stroke treatment.

Moreover, for patients on VKA treatment, we performed an additional subanalysis regarding INR values at hospital admission. Patients with an INR of at least 2.0 showed an OR of 0.184 for HT (95% CI, 0.045–0.744), while those with INR values <2.0 showed an OR of 0.834 (95% CI, 0.635–1.095).

DISCUSSION

The main findings of this study are that prior use of any OAC leads to less severe stroke, smaller infarct size, and a substantially reduced risk for HT >50%. No relevant differences among individual DOACs or VKA could be found while antiplatelet therapy had any association with the outcomes.

In contrast to a systematic review of 52 studies providing detailed information on a total of 12 247 patients, which found that stroke severity and infarct sizes were lower in DOAC‐treated patients compared with those taking VKAs, ⁷ we did not find a similar trend in our data. This, however, may be attributable to the fact that patients on VKA treatment were largely effectively anticoagulated based on their INR values, which was not the case in many previous studies. The overall reduction of clinical stroke severity by 2 to 3 NIHSS points in patients undergoing anticoagulation therapy is also in line with previous studies conducted worldwide. ¹¹ , ¹² , ¹³ , ¹⁴ , ¹⁵ The smaller sample size caused by missing values, however, reduced the precision of estimates regarding stroke volume. According to our results, prior use of anticoagulants is associated with a substantial decrease of ≈23 mL in imaging‐based infarct sizes on average. Hypothetically, this may be due to smaller thrombi as an effect of anticoagulation or due to a shift in other stroke causes that may lead to smaller infarcts. In accordance with these considerations, Rose et al showed that a large portion of strokes occurring despite anticoagulation may be due to reasons other than AF (ie, small‐vessel disease, arterio‐arterial embolisms, malignancies, or patent foramen ovale). ²³

Most intriguingly, we were also able to demonstrate a substantial preventive effect of all anticoagulants on the risk of HT in initial cerebral imaging. We first performed a crude analysis, as infarct size is a mediating factor in this effect. However, even after adjustment for the NIHSS, only slightly weaker effect sizes could be observed. A subanalysis regarding the presence of more severe HT in terms of parenchymal hematoma types 1 and 2 according to the ECASS criteria also demonstrated the protective effect of prior use of anticoagulants, in part even with considerably stronger effect sizes. Strikingly, in patients with prior use of VKAs, the preventive effect on HT was evident for patients with effective anticoagulation as documented by an INR of at least 2.0 at hospital admission. Notably, in our analysis, all patients with prior use of DOACs were on effective treatment at the time of the index event according to the documentation.

On a pathophysiological level, a potential explanation for this finding could be that inhibition of thrombin, whether direct or indirect, contributes to a more stable blood brain barrier (BBB) in acute stroke. HT in AIS occurs due to dysfunction of the BBB, which is, in addition to the primary cell death during AIS, further influenced by inflammation and metalloproteases. ¹⁸ Thrombin may induce BBB dysfunction by a matrix metalloprotease (MMP)–dependent mechanism. ²⁴ Therefore, anticoagulants inhibiting thrombin may be regarded as BBB stabilizing. Experimental studies have investigated the effect of anticoagulants on the risk of HT in rodent models. In a study by Bieber et al, it was shown that treatment with edoxaban led to a reduction of stroke severity in mice by reducing damage of the BBB and inflammation. ²⁵ In contrast, Pfeilschifter et al reported that prior use of VKAs dramatically increases the risk of HT in a middle cerebral artery occlusion model in mice. ²⁶ Thus, results from experimental studies on the effect of preexisting anticoagulant use on the risk of HT are conflicting. Our study provides evidence that early HT of AIS is indeed prevented by prior anticoagulant use in humans. A multicenter study by Meinel et al recently demonstrated that intravenous thrombolysis in patients currently receiving anticoagulation resulted in lower rates of intracranial hemorrhage compared with patients without anticoagulation. ²⁷ This is in line with our current findings. Of note, the number of patients who received thrombolysis despite OAC use was too small in our study to conduct a meaningful subanalysis. In the study by Meinel et al, the proportion of HT, similar to our current work, was approximately halved in patients taking OACs, which may further support the safety of thrombolysis despite anticoagulation. Furthermore, the randomized controlled trial ARAIS (Argatroban plus r‐tPA for Acute Ischemic Stroke) investigated whether adding the thrombin inhibitor argatroban to alteplase for acute ischemic stroke was safe and effective. ²⁸ Interestingly, the rates of HT were similar between both groups and not lower in the argatroban group. ²⁸ Hypothetically, BBB stabilizing effects by thrombin inhibition may thus be evident only with longer therapy as in patients with prior OAC use, such as in our study. Current guidelines are ambiguous in their recommendation regarding thrombolysis in patients with stroke with existing OAC therapy. ²⁹ Randomized controlled trials are therefore currently underway to investigate the safety and efficacy of thrombolysis in this situation. ³⁰

Of note, we decided against an analysis of end points occurring beyond those investigated in this study due to the multitude of influencing variables following hospital admission. Analyses focusing on the initiation or resumption of anticoagulation are planned or have been published elsewhere. ²⁰

The large sample size and the multicenter and prospective design are clear strengths of this study. Multivariable models and the application of directed acyclic graphs were used for addressing confounding, which was enabled by the detailed representation of clinical and demographic data. However, residual confounding may play a role in our results. Another limitation is that information on infarct size was missing in a substantial proportion of patients, leading to imprecise effect estimates in some treatment strata, while other estimations were highly robust and precise. Because smaller and multilocular infarcts may be more difficult to quantify, the analysis of infarct volumes should be interpreted with caution. However, the findings regarding the NIHSS show similar results, which further supports this interpretation. The observational nature of this study implies that the timing of brain imaging was not standardized. However, the results were consistent in corresponding and thorough sensitivity analyses. Another limitation of this study is the lack of data regarding drug monitoring of anticoagulants at admission, as well as the type of antiplatelet therapy. Due to the higher variation in drug activity levels, we have also refrained from considering preexisting parenteral antithrombotic therapies, although this limits the generalizability of the results. It remains uncertain, moreover, whether our findings also apply to patients with AF detected after stroke, ³¹ as it was not possible for our study to investigate the anticoagulant effects in this population.

In conclusion, in this large prospective registry study, we show that prior use of anticoagulants not only results in less severe stroke and smaller infarct size but also a substantially reduced risk of HT in initial imaging by >50% compared with no prior use of antithrombotics. Therefore, the results of this large real‐world study underscore the beneficial properties of OACs in AIS and the necessity of adequate drug adherence in patients with AF. Further studies are required to validate these findings, also with regard to potential therapeutic implications.

Sources of Funding

The PRODAST study was funded by an unrestricted grant from Boehringer Ingelheim. Boehringer Ingelheim was not involved in the study design, collection, analysis, interpretation of data, writing of this article, or decision to submit for publication.

Disclosures

Dr Grosse received research grants from the Lower Saxony Ministry for Science and Culture and the German Ministry of Education and Research (BMBF) as well as honoraria from Bayer and Boehringer Ingelheim. A. Hüsing received honoraria from Ipsen and Novartis. Dr Eggers received honoraria in the last 12 months from Abbvie, Stada Pharma, Bristol Myers Squibb, Daiichi Sankyo, and Bial. Dr Eschenfelder is an employee of Boehringer Ingelheim. Dr Diener received honoraria for participation in clinical trials, contribution to advisory boards or oral presentations from: Abbott, BMS, Boehringer Ingelheim, Daiichi‐Sankyo, Novo‐Nordisk, Pfizer, Portola, and WebMD Global. Dr Deiner received research grants from the German Research Council (DFG) and German Ministry of Education and Research (BMBF). He serves as editor of Neurologie up2date, Info Neurologie & Psychiatrie and Arzneimitteltherapie, as co‐editor of Cephalalgia, and is on the editorial board of Lancet Neurology and Drugs. The remaining authors have no disclosures to report.

Supporting information

Data S1

Table S1

Figures S1–S2