Oral anticoagulation in patients with atrial fibrillation and acute ischaemic stroke: design and baseline data of the prospective multicentre Berlin Atrial Fibrillation Registry Free

Abstract

Introduction

Atrial fibrillation (AF) is the most frequent cardiac arrhythmia worldwide and its prevalence increases with age. Atrial fibrillation is associated with approximately every fifth ischaemic stroke and patients with AF-related strokes have a high recurrence rate.¹ Long-term oral anticoagulation (OAC) using the vitamin K antagonist (VKA) warfarin reduces the risk of recurrent stroke in AF patients by about two-thirds compared to placebo.² However, due to feared bleeding complications and multiple drug interactions, VKAs are widely underused in routine practice and less than half of all patients still take a prescribed warfarin 2 years after ischaemic stroke.³ According to the results of randomized multicentre trials⁴, national and international guidelines now recommend OAC using a VKA or a non-vitamin K-dependent oral anticoagulant (NOAC) for stroke prevention in patients with AF and prior ischaemic stroke.^1,5,6 Compared to VKAs, NOACs have a fixed daily dosage, a lower rate of drug–drug and drug–food interactions and avoid the need for serial coagulation measurements.⁷ Ongoing AF registries and retrospective (claims) data analyses demonstrate an increasing use of NOACs in AF patients with or without stroke.^8–10

Nevertheless, prospectively collected data on OAC use in the acute and post-acute phase of ischaemic stroke after the approval of the NOACs are rare. While ongoing multicentre registries like RASUNOA prime (NCT02533960) or PRODAST (NCT02507856) do not have a long-term follow-up after acute stroke in AF patients, large AF registries like ORBIT-AF II (NCT 01701817), GARFIELD (NCT01090362), or EORP-AF¹¹ do not explicitly focus on acute stroke patients. Furthermore, all mentioned registries are unable to provide a representative assessment of clinical practice in a certain area due to a selection of study sites and patients.

To provide a representative evaluation of OAC use for (secondary) stroke prevention in AF patients in a defined region, we have established the multicentre prospective Berlin Atrial Fibrillation Registry, including all Departments of Neurology in the area of Berlin, Germany running a stroke unit. All study sites together treat 80–90% of all stroke patients in the wider Berlin area with about 4 million inhabitants.¹² Herewith we report the design of the Berlin Atrial Fibrillation Registry, the baseline characteristics of registry patients as well as factors associated with OAC use at the time of stroke/transient ischaemic attack (TIA) or at hospital discharge after the index stroke.

Methods

Ethical conduct

The Charité - Universitätsmedizin Berlin is the sponsor of the investigator-initiated prospective multicentre Berlin Atrial Fibrillation Registry (NCT02306824). The ongoing hospital-based registry received primary approval from the Ethical Committee of the Charité - Universitätsmedizin Berlin, Germany (EA2/033/14). Registry patients provided written informed consent. All study procedures are carried out following the Declaration of Helsinki.

Study design

All 14 Departments of Neurology in the area of Berlin, Germany conducting a stroke unit participated in the Berlin Atrial Fibrillation Registry (see Supplementary material online, Table SI for details), serving for a source population of more than 3 million inhabitants. Study sites were initiated in 2014 and 2015 by members of the study coordinating office located at the Center for Stroke Research Berlin, Charité - Universitätsmedizin Berlin (see Supplementary material online, Table SII). Inclusion and exclusion criteria are shown in Table 1. Conformation of stroke or TIA was based on brain imaging criteria and clinical criteria, following the WHO definition.¹³ No recommendations were given to the study patients or treating physicians at the 16 participating stroke units (Charité operates three stroke units). The predefined primary outcome measure is patient-reported persistence to OAC (either NOAC or VKA) in stroke patients with AF at 12 months after the index stroke. Detailed information on predefined secondary outcome measures can be found in Supplementary material online, Table SIII. Serious adverse events (SAEs) like recurrent ischaemic stroke or TIA, intracerebral haemorrhage, subarachnoid haemorrhage, major bleeding, myocardial infarction, peripheral arterial embolism, and death of any cause are evaluated by an independent Critical Event Committee (CEC). The ongoing central follow-up is conducted by the study coordinating office and will provide detailed information on stroke recurrence rate and major bleeding rate in patients prescribed an OAC within days after acute ischaemic stroke. Besides the frequency of OAC prescription at 3, 12, 24, 48, 60, and 72 months after stroke, patient self-reported persistence and quality of adherence to OAC as well as self-reported quality of life will be addressed (Supplementary material online, Table SIII).

Data assessment and monitoring

Members of the local study sites registered all patients using the electronic Case Report Form (eCRF) in Research Electronic Data Capture (REDCap™) after enrolment (see Supplementary material online for details). Collected information on in-hospital stay after the index stroke included patients’ demographics, past medical history, AF type, stroke type, individual stroke treatment, and stroke severity [assessed by the National Institutes of Health Stroke Scale (NIHSS) score¹⁴ on admission and the modified Rankin Scale (mRS) score at discharge¹⁵], antithrombotic medication on admission as well as at discharge, level of education, clinical diagnosis of depression, cognitive impairment, and the presence of SAEs. The HAS-BLED score was assessed according to guidelines.¹ Classification of measured and previously documented international normalized ratio (INR) values as ‘labile INR’ was at the discretion of the treating physicians. On-site monitoring was conducted by the study central office and started after enrolment of the first five patients. Statistical analysis was performed within a scientific cooperation with the Institute for Clinical Epidemiology and Biometry at the University of Würzburg, Germany.

Statistical methods

The sample size estimation can be found in the Supplementary material online. Baseline patient characteristics are presented as absolute and relative frequencies or median and quartiles. Univariable odds ratios (OR) and respective confidence intervals (CI) are calculated using binary logistic regression analysis. Parameters with P-values <0.1 were transferred into multivariable binary logistic regression analysis and variable selection was conducted using backward selection algorithm. Multicollinearity was examined using the variance inflation factor. As sensitivity analysis, we used generalized linear modelling to estimate OR and CI adjusting for study site. We used the SPSS function ‘GENLIN’ to model a generalized estimation equation. We included the study site as clustering variable, i.e. data across study sites are still assumed to be independent, whereas data within study sites are assumed to be correlated. By using a logit link function, the generalized estimation equation estimates a binary logistic regression model that was extended to a second level (namely the study site level, which complements the patient level). Statistical analysis was performed using SPSS (Version 24, SPSS Inc., Chicago, IL, USA). Two-sided P-values <0.05 were considered statistically significant.

Results

From December 2014 to November 2016, 1080 hospitalized patients with suspected acute ischaemic stroke and known AF before the index stroke or a first episode of AF during the in-hospital stay were enrolled on a consecutive basis. Inclusion criteria (Table 1) were fulfilled in 1050 patients; 29 patients were classified as stroke mimics at hospital discharge, one patient had a long-term indication for OAC besides AF and two patients withdrew consent within the in-hospital stay. The baseline characteristics of the analysed 1048 study patients are presented in Table 2. Patients’ median age was 77 years [interquartile range (IQR) 72–83], 503 (48%) patients were female and 762 (73%) had known AF before enrolment. Of those, 350 (33%) had documented AF during the in-hospital stay after the index stroke. Overall, 254 (24%) had a TIA as index event. The median NIHSS score on admission was 2 points [IQR 1–5], 157 (15.0%) of all patients received intravenous thrombolysis and 81 (7.7%) patients underwent endovascular thrombectomy. The median mRS at hospital discharge was 2 [IQR 1–4], including 309 (29.5%) patients without any stroke-related neurological deficit and 272 (26.1%) patients who were bed-ridden after stroke (mRS score 4–5). Overall, 6 (0.6%) registry patients died in hospital.

Impact factors on oral anticoagulation use at the time of the index stroke

Antithrombotic medication on hospital admission due to the index stroke is depicted in Figure 1 according to AF status pre-admission and pre-stroke CHA₂DS₂-VASc score. While 286 (27%) study patients had a first episode of AF after hospital admission, 762 (73%) of all study patients had known AF. Of those patients, 5 (0.5%) had a pre-stroke CHA₂DS₂-VASc score of 0. In the remaining 757 patients with a given indication for OAC, 470 (62.1%) received OAC [in detail 229 (30.3%) VKA and 241 (31.8%) NOAC, 12 (1.6%) received heparin in therapeutic dosage, 164 (21.7%) had an antiplatelet and 111 (14.7%) had no antithrombotic medication (Figure 2, column A)]. The median INR of those 229 registry patients taking a VKA on hospital admission because of the index stroke was 2.0 [IQR 1.7–2.4] (range 1.0–5.6). Of those, 105 (45.9%) patients had an INR within therapeutic range (2–3) on admission. In 275 registry patients, no anticoagulation was prescribed immediately before hospital admission despite given indication for anticoagulation according to present guideline recommendations. In 264 (96%) patients, the treating physicians recognized the given indication for anticoagulation before the index stroke. In 115 (41.8%) out of these 275 patients, no anticoagulation was ever established. In 149 (54.2%) patients, previous OAC was stopped before the index stroke. A reason for stopping anticoagulation was not provided in 104 (69.8%) out of 149 patients. Besides patients’ choice to stop anticoagulation (n = 34; 22.8%), a pervious intracranial bleed (n = 9; 6.0%), a previous (non-intracranial) major bleeding (n = 27; 18.1%), or planned surgery (n = 22; 14.8%) were most often mentioned reasons.

Figure 1

Antithrombotic medication on admission according to AF status and CHA₂DS₂-VASc score before admission for acute ischaemic stroke in 1048 patients with AF. AF, atrial fibrillation; TIA, transient ischaemic attack.

Open in new tabDownload slide

Figure 2

Antithrombotic medication on admission as well as at hospital discharge after the index stroke in 757 patients with known AF and CHA₂DS₂-VASc ≥1 before the index stroke. AF, atrial fibrillation; NOAC, non-vitamin K-dependent oral anticoagulant; VKA, vitamin K antagonist.

Open in new tabDownload slide

Multivariable analysis (Table 3) revealed significant associations of prior stroke or TIA [OR 1.97 (95% CI 1.40–2.78)], diabetes [OR 1.64 (95% CI 1.17–2.28)], and age [65–74 years: OR 2.00 (95% CI 1.13–3.54), 75–84 years: OR 2.81 (95% CI 1.65–4.78), ≥85 years: OR 2.13 (95% CI 1.28–4.17)] with OAC prescription at the time of stroke/TIA. Atrial fibrillation type (paroxysmal vs. non-paroxysmal), sex, or known cardiovascular risk factors like hypertension, heart failure, vascular disease, or renal impairment [glomerular filtration rate (GFR) <90 mL/min] were not associated with OAC prescription at the time of stroke/TIA. The variance inflation factor values were <1.3 for all variables considered in the multiple logistic regression model.

Impact factors on oral anticoagulation use at hospital discharge after the index stroke

Antithrombotic medication in 1042 surviving registry patients was distributed at hospital discharge as follows: 847 (81.3%) received OAC, 26 (2.5%) received heparin in therapeutic dosage, 117 (11.2%) had antiplatelet, and 52 (5.0%) patients had no antithrombotic medication. Prior ischaemic stroke or TIA [OR 1.64 (95% CI 1.08–2.48)], OAC on admission [OR 1.77 (95% CI 1.16–2.70)], and an index TIA [OR 2.07 (95% CI 1.18–3.65) in comparison with an index stroke] were associated with OAC prescription at discharge. A severe neurological handicap at discharge [mRS score 4–5; OR 0.35 (95% CI 0.21–0.57)] and conducted carotid endarterectomy [OR 0.18 (95% CI 0.06–0.62)] were inversely associated with OAC prescription at discharge. Patients’ age, sex, AF type (first episode vs. paroxysmal vs. non-paroxysmal vs. unknown), or known cardiovascular risk factors like hypertension, heart failure, diabetes, vascular disease, or renal impairment (GFR <90 mL/min) had no significant impact on OAC prescription at discharge. These results were consistent with those gained from a sensitivity analysis adjusting for enrolling study sites (Table 4).

Antithrombotic medication on hospital discharge is shown in column B of Figure 2 for all 757 study patients with known AF and CHA₂DS₂-VASc ≥1 before the index stroke. While the proportion of patients prescribed a NOAC more than doubled compared to antithrombotic medication on admission, the proportion of patients prescribed a VKA or an antiplatelet was about halved. At discharge, less than 6% of all patients with known AF had no antithrombotic medication at all.

Impact factors on non-vitamin K-dependent oral anticoagulant prescription at hospital discharge after the index stroke

Of the 847 patients prescribed OAC at hospital discharge, 137 (16.2%) received a VKA and 710 (83.8%) received a NOAC (Table 5). An index TIA [OR 0.56 (95% CI 0.34–0.94) in comparison with an index stroke], co-existing heart failure [OR 0.49 (95% CI 0.26–0.93)], endovascular thrombectomy on admission [OR 12.9 (95% CI 1.59–104)], and medication on hospital admission were associated with NOAC prescription at discharge. Compared to patients with no antithrombotic medication on admission, patients prescribed a NOAC at the time of stroke/TIA had a significantly higher chance of NOAC prescription at discharge [OR 15.6 (95% CI 1.97–122)], while patients prescribed a VKA at the time of stroke/TIA had a significantly lower chance of NOAC prescription at discharge [OR 0.04 (95% CI 0.02–0.09)]. Patients’ age, sex, AF type, or known risk factors including prior stroke or TIA had no significant impact on prescription of NOAC on hospital discharge. These results were consistent with those gained from a sensitivity analysis adjusting for enrolling study sites (Table 5).

Discussion

The Berlin Atrial Fibrillation Registry is an investigator-initiated prospective multicentre cohort study gathering comprehensive data on (secondary) stroke prevention in AF patients in an unselected urban population in Germany. The baseline data of the registry indicate that about 60% of all patients with known AF and an at least moderate stroke risk were anticoagulated at the time of the index stroke (Figure 1). At hospital discharge after ischaemic stroke or TIA, more than 80% of all registry patients were anticoagulated and about 80% of those were prescribed a NOAC (Tables 4and5). These percentages are slightly higher than a retrospective analysis of quality assurance data in the federal state of Hesse, Germany in 2015 or in a retrospective cohort study of Danish AF patients in 2004–2007.^9,10

While prescription of OAC at the time of stroke or TIA in registry patients with known AF was associated with prior stroke or TIA, diabetes, and old age (Table 3), there was no association with other items of the CHA₂DS₂-VASc score like heart failure, hypertension, vascular disease, or sex. In comparison, a retrospective study from Korea reported an association of OAC at the time of stroke with prior stroke but also with younger age, higher CHADS₂ score and heart failure according univariate analyis.¹⁶ Of those with VKA on admission (n = 229), 105 (46%) had an INR within therapeutic range. These findings indicate that the quality of anticoagulation in our patient cohort is higher than in studies reported previously.^17,18 However, we cannot comment on the quality of OAC in registry patients prescribed a VKA within weeks before the index stroke/TIA, as previous INR values were not assessed. Moreover, we have no detailed information on the intake of a prescribed NOAC before the index stroke/TIA, as coagulation tests assessing thrombin-clotting time or the present anti-Xa activity was not part of clinical routine on admission. In 275 of registry patients, no anticoagulation was prescribed immediately before hospital admission. In 54% of these patients, anticoagulation was initiated but stopped before the index stroke or TIA. A specific reason for stopping anticoagulation was not provided in the vast majority of these patients.

At hospital discharge (median 6 days after the index stroke), patients with prior stroke or TIA and patients with an index TIA (using an index ischaemic stroke as reference) had higher probability of prescription of OAC. This may be due to the fact that the assumed intracranial bleeding risk in TIA patients is lower compared to patients with ischaemic stroke.⁷ Moreover, patients with severe neurological handicap or carotid endarterectomy had a lower probability of OAC prescription (Table 4). Previous studies in German patients with a medical history of stroke before the approval of the NOACs showed an inverse relationship between patients’ age and OAC prescription.¹⁹ However, we did not observe a statistical significant association between the a priori defined age categories and OAC prescription in patients with acute stroke in our dataset which might also be caused by the limited statistical power. We assume, that the observed differences correspond to the increased awareness regarding medical stroke prevention after the index stroke. Furthermore, general practitioners mainly prescribed medical stroke prevention before the index stroke and a neurologist prescribed medical stroke prevention at hospital discharge. The observed 80% rate of OAC at hospital discharge is higher compared to smaller single-centre cohort studies, reporting an OAC rate of 55–70% in the NOAC era or 38–58% in the pre-NOAC era.^20–23

Underpinned by the baseline data of the Berlin Atrial Fibrillation Registry, prescription of a NOAC after stroke seems to be standard nowadays as also found in other studies.^8,9 In registry patients with known AF before enrolment the proportion of patients prescribed a NOAC more than doubled at hospital discharge compared to baseline (Figure 2). Interestingly, patients with intake of the VKA phenprocoumon at the time of the index stroke were discharged on phenprocoumon in about half of all cases (Figure 2, Supplementary material online, Table SIV). In addition, the probability to prescribe a VKA (and not a NOAC) was higher in patients with heart failure. Once again, patients’ age or AF type had no impact on NOAC prescription. The magnitude of the observed impact factors on OAC (Table 4) or NOAC prescription (Table 5) was consistent after adjusting for the recruiting study sites.