2022

Cerebrovascular disease
J Neurol Neurosurg Psychiatry: first published as 10.1136/jnnp-2021-327236 on 11 October 2021. Downloaded from http://jnnp.bmj.com/ on March 1, 2024 at Chulalongkorn University.
Original research
Early versus late start of direct oral anticoagulants

after acute ischaemic stroke linked to atrial
fibrillation: an observational study and individual
patient data pooled analysis
Gian Marco De Marchis ‍ ‍,1 David J. Seiffge ‍ ‍,1,2 Sabine Schaedelin,3
Duncan Wilson,4 Valeria Caso,5 Monica Acciarresi,6 Georgios Tsivgoulis ‍ ‍,7,8
Masatoshi Koga ‍ ‍,9 Sohei Yoshimura,9 Kazunori Toyoda,9 Manuel Cappellari,10
Bruno Bonetti,10 Kosmas Macha,11 Bernd Kallmünzer,11 Carlo W. Cereda,12
Philippe Lyrer,1 Leo H. Bonati,1 Maurizio Paciaroni,13 Stefan T. Engelter,1,14
David J. Werring ‍ ‍4
► Additional supplemental ABSTRACT trials, however, patients with AIS were excluded
material is published online Objective The optimal timing to start direct oral for at least 7 days up to 6 months after AIS, given
only. To view, please visit
the journal online (http://dx. anticoagulants (DOACs) after an acute ischaemic stroke the fear of bleeding into the territory of acute cere-
doi.org/10.1 136/jnnp-2021- (AIS) related to atrial fibrillation (AF) remains unclear. bral ischaemia. A recent individual patient data
327236). We aimed to compare early (≤5 days of AIS) versus late analysis of seven prospective observational studies
(>5 days of AIS) DOAC-start. found that the advantage of DOAC over VKA is
For numbered affiliations see preserved in the early phase after AIS.2 It remains
Methods This is an individual patient data pooled
Protected by copyright.
end of article.
analysis of eight prospective European and Japanese unclear how early DOAC can safely be started after
Correspondence to cohort studies. We included patients with AIS related to AIS. Four randomised controlled clinical trials are
Dr Gian Marco De Marchis, non-valvular AF where a DOAC was started within 30 comparing an early to a later DOAC- start after
Neurology and Stroke Center, days. Primary endpoints were 30-day rates of recurrent AIS associated with AF (ELAN (NCT03148457),
University Hospital Basel, AIS and ICH. OPTIMAS (EudraCT, 2018-0 03 859–38), TIMING
Basel, BS, Switzerland; gian. (NCT0291348) and START (NCT03021928)),
Results A total of 2550 patients were included. DOACs
demarchis@usb.ch
were started early in 1362 (53%) patients, late in 1188 but results are not available yet.3 In this individual
The work as been presented as (47%). During 212 patient-years, 37 patients had a patient data pooled analysis of prospective obser-
a poster at the European Stroke recurrent AIS (1.5%), 16 (43%) before a DOAC was vational studies, we compared the 30-day rates of
Organisation Conference 2019, started; 6 patients (0.2%) had an ICH, all after DOAC- recurrent AIS, ICH and mortality between patients
Milano, Italy.
start. In the early DOAC-start group, 23 patients (1.7%) in whom DOACs were started early (≤5 days)
Received 25 May 2021 suffered from a recurrent AIS, while 2 patients (0.1%) versus late (>5 days) after an AIS or transient isch-
Accepted 27 September 2021 had an ICH. In the late DOAC-start group, 14 patients aemic attack (TIA) associated with AF.
Published Online First 11 (1.2%) suffered from a recurrent AIS; 4 patients (0.3%)
October 2021
suffered from ICH. In the propensity score-adjusted METHODS
comparison of late versus early DOAC-start groups, there Selection of the study centres
was no statistically significant difference in the hazard of The selection of the study centres has been described
recurrent AIS (aHR=1.2, 95% CI 0.5 to 2.9, p=0.69), ICH in detail previously.2 Briefly, we contacted the prin-
(aHR=6.0, 95% CI 0.6 to 56.3, p=0.12) or any stroke. cipal investigators of peer- reviewed prospective,
Conclusions Our results do not corroborate concerns observational studies published in English based
that an early DOAC-start might excessively increase the on real-life cohorts in which DOACs were admin-
risk of ICH. The sevenfold higher risk of recurrent AIS istered within 3 months after the index stroke.
than ICH suggests that an early DOAC-start might be The characteristics of the contributing centres are
reasonable, supporting enrolment into randomised trials detailed in table 1. There is no overlap between the
comparing an early versus late DOAC-start. cohort studies studied, that is, all included patients
are unique.
© Author(s) (or their

employer(s)) 2022. No INTRODUCTION Inclusion and exclusion criteria
commercial re-use. See rights Among patients with atrial fibrillation (AF), direct We included patients with (1) AIS (defined as a focal
and permissions. Published oral anticoagulants (DOAC) are at least as effec- neurological deficit with acute onset and presence
by BMJ. tive as vitamin K antagonists (VKAs) in preventing of a corresponding lesion on diffusion- weighted
To cite: De Marchis GM, acute ischaemic strokes (AISs). The main advantage MRI (DWI) or, if no MRI was acquired, signs of
Seiffge DJ, Schaedelin S, et al. of DOAC over VKA is the lower rate of intracra- early ischaemic injury on CT) or TIA (defined as an
J Neurol Neurosurg Psychiatry nial haemorrhage (ICH), as shown in four pivotal acute onset focal neurological deficit of presumed
2022;93:119–125. randomised controlled clinical trials.1 In these ischaemic origin without a corresponding lesion on
De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2022;93:119–125. doi:10.1136/jnnp-2021-327236 119
Table 1 Single-centre and multicentre studies participating in the individual patient data analysis
Patients contributed to final
Study period cohort, n (%) Maximum follow-up period*
Single centre
NOACISP (Novel Oral Anticoagulants In Stroke Patients, Basel/ 2012–2017 376 (14.7) Up to 3.8 years
Switzerland)16
Verona (Italy)22 2013–2015 225 (8.8) 3 months
Erlangen (Germany)23 2011–2013 201 (7.9) Up to 1 year
Lugano (Switzerland) 2014–2018 32 (1.3) 3 months
Multicentre
RAF-NOAC (29 centres in Europe/Asia)18 2014–2016 768 (30) 3 months
SAMURAI-NVAF (18 centres in Japan)24 2011–2014 439 (17) Up to 3.5 years
CROMIS-2 (80 centres in the UK and one in the Netherlands)25 2011–2015 436 (17) Up to 5.4 years
RAF (29 centres in Europe/Asia)26 2012–2014 73 (2.9) 3 months
*Minimum follow-up period of 3 months for all studies. Because of rounding, percent do not add up to 100%.
DWI or, if no MRI was acquired, lasting less than 24 hours); (2) Endpoint
diagnosis of non-valvular AF, either known prior to the index The coprimary endpoints were the time to occurrence of recur-
event or detected after the event; (3) oral anticoagulation with rent AIS, ICH and any stroke (AIS or ICH).
DOAC, either continued (for those already on anticoagulation
on admission), started or resumed within 30 days after the index Statistical analysis
event; and (4) prospective follow-up for at least 30 days after the We compared demographic and clinical baseline characteristics
index event for the presence or absence of recurrent AIS, ICH among patients in the early versus late DOAC-start using the
and death (of any cause). Patients who died prior to the first Pearson χ2 test for categorical variables and the Mann-Whitney
planned follow-up, that is, within the first 3 months following U test for continuous variables. Our main goal was to compare
the index event, were included. the event rates between the early versus late DOAC-start groups.
We excluded patients with (1) mechanical heart valves, (2) Traditional Cox analysis would, however, be biased since the
rheumatic or severe mitral valve stenosis, (3) oral anticoag- timepoint of the endpoint could have influenced the group allo-
ulation with VKA, (4) ICH after the index event but prior to cation. For instance, a recurrent AIS at day 3—before DOAC-
DOAC-start, and (5) oral anticoagulation started later than 30 start—may have prompted the treating physicians to start DOAC
days after the index event or with missing information on the immediately after the recurrence, with a DOAC-start between
initiation date of oral anticoagulants. day 3 and day 5. Hence, to compare the time-to-events between
the early and late DOAC- start groups, we considered only
patients who were event-free up to 5 days of the index event
Definition of early and late DOAC-start (n=2535)—that is, a landmark analysis. Sensitivity analysis
Early DOAC-start was defined as ≤5 days from the index with landmark at 3 and 7 days were performed. We assessed
event, late DOAC-start as >5 days. The rationale for this cut- the association between the timepoint of DOAC-start (early vs
off is statistical rather than biological, as 5 days represented the late) and the endpoints using a mixed effects Cox proportional
median interval of DOAC- start, leaving us with two groups hazards regression model to compute HRs with 95% CIs. Land-
balanced in size as the published individual patient data study mark analysis was performed for Cox analyses only. Propensity
comparing DOAC to VKA after a recent AIS.2 score weighting was used to adjust for confounders (ie, baseline
As a sensitivity analysis, we assessed whether adherence to variables with p<0.05 in univariate analysis and intravenous
the guideline of the European Society of Cardiology (ESC) on thrombolysis) to compute adjusted HRs (aHR). Only the first
the timing of DOAC-start by was associated with the endpoint outcome event is used and patients who suffered a competing
rate, as recently performed by Eun et al.4 The definition of the event (eg, death following ICH) were censored at the day of
guideline-adherent group was as follows: National Institutes of their first event. To take competing risk into account, we ran a
Health Stroke Severity Scale (NIHSS) score of <8 and DOAC- Fine-Grey model.
start of ≥3 days of the index event; or NIHSS score of ≥8 and Concerning the timepoint of death, 22 patients had died
≤15 and DOAC-start of ≥6 days; or (3) NIHSS score of within 90 days of the qualifying event, but for 15 of these, the
≥16 and DOAC-start of≥12 days. Earlier DOAC-starts were exact day of death was unknown. For these 15 patients, we used
defined as non-guideline-adherent.4 multiple imputations to estimate the time of death. To this end,
for each patient, a random point between the time of anticoag-
ulation (ie, the last point the patient was known to be alive) and
Follow-up the point the patient was known to be dead was imputed. We did
We included only studies with a planned follow-up of at least 30 100 imputations. For the plots, a random imputation was used.
days after index event for recurrent AIS (defined as new neuro- The models are fitted on each imputation. Thereafter, we used
logical symptoms and evidence for ischaemic stroke on CT or Rubin’s rule to pool the estimates.
MRI); ICH defined as new neurological symptoms associated We assessed treatment effects in the following predefined
with the detection of ICH on CT or MRI as defined within the subgroups: (1) patients with minor stroke or TIA (defined
criteria of the International Society on Thrombosis and Haemo- as an NIHSS score of ≤3),6 (2) patients with severe stroke
stasis5; and all-cause mortality (including fatal AIS or ICH). (defined as an NIHSS score of >15),7 (3) elderly patients
120 De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2022;93:119–125. doi:10.1136/jnnp-2021-327236
Table 2 Baseline demographic and clinical characteristics

Late*
Early* anticoagulation anticoagulation
(n=1362) (n=1188) P value
" "!"
"#"!% Age (years), median 78.0 (71–84) 77.0 (70–83) 0.06
"!" "%"'! (IQR)
" "
&$" !" " Women 637 (46.8) 567 (47.7) 0.66
AIS as qualifying 1283 (94.1) 1172 (98.7) <0.001
event

Baseline antiplatelet 353 of 1035 (34.1) 420 of 1074 (39.1) <0.001
agents
Figure 1 Patient flow. DOAC, direct oral anticoagulant. Baseline vitamin K 222 of 1345 (16.5) 150 of 1181 (12.7) <0.001
antagonist
Baseline DOAC 94 of 1345 (7.0) 31 of 1181 (15.5) <0.001
(defined as aged ≥80 years vs <80 years), (4) patients with Baseline statins 172 of 688 (25.0) 102 of 358 (28.5) 0.26
impaired renal function (defined as creatinine clearance History of 10 of 1000 (1.0) 8 of 667 (1.2) 0.84
intracranial
of <60 mL/min/1.73 m2 using the CKD-EPI equation8), and
haemorrhage
(5) patients treated with acute recanalisation therapies for
Diabetes mellitus 298 of 1360 (21.9) 275 of 1185 (23.2) 0.47
the index stroke (intravenous thrombolysis and/or endovas-
Hypertension 1026 of 1360 (75.4) 899 of 1182 (76.0) 0.78
cular therapy). Thereby, each binary variable indicating a
Dyslipidaemia 421 of 1063 (39.6) 301 of 717 (42.0) 0.33
subgroup was included in a separate model and the interac-
Impaired renal 453 of 1231 (36.8) 260 of 1044 (24.9) <0.001
tion term between the covariate and the treatment was esti-
function†
mated. A significant interaction (ie, p<0.05) term indicates
Current smoking 176 of 1313 (13.4) 260 of 1156 (22.5) <0.001
that the estimated difference between treatments differs
NIHSS score on 4.0 (1–8) 6.0 (3–12) <0.001
between the subgroups. The estimated HRs were presented admission
as forest plots. For the subgroup analyses—given the small CHA2DS2-Vasc 5.0 (4–6) 5.0 (4–6) 0.02
endpoint numbers—we used the composite of recurrent AIS,
HAS-BLED 3.0 (2–3) 3.0 (2–4) 0.001
ICH and death within 30 days. An α-level of 0.05 was used
Intravenous 324 of 1350 (24.0) 281 of 1180 (23.8) 0.95
to determine statistical significance. The statistical analyses thrombolysis
were carried out using R. Endovascular 43 of 1194 (3.6) 53 of 1128 (4.7) 0.22
treatment
Categorical variables are given in number of patients having the characteristic/total
patients available for analysis and (%).
Data availability statement Continuous variables are displayed as median and IQR.
Deidentified participant data will be shared on written *Early: ≤5 days of the index event; late: between 6 and 30 days of the index event.
request by qualifying investigators who provide a research †Impaired renal function defined as creatinine clearance of <60 mL/min/1.73 m2.
protocol. AIS, acute ischaemic stroke; DOAC, direct oral anticoagulant; NIHSS, National
Institutes of Health Stroke Severity Scale.
RESULTS Endpoint analysis for the whole cohort

Baseline data and demographics Between the index event and 30 days follow-up, 37 patients
The pooled individual patient data cohort comprised 5421 had a recurrent AIS (1.5%), 57% of which happened under
patients, 2871 of which were excluded; flow of patients is DOAC (21/37) and 41% within the first 5 days of the index
showed on figure 1. Five patients were excluded because ICH event (15/37). Six patients had an ICH (0.2%), and 15
occurred prior to any DOAC-start; two of these five patients patients died within the first 30 days (0.6%). The composite
suffered from ICH following acute recanalisation therapy. The 30-day rate of any stroke was 1.7% (43/2550).
final cohort included 2550 patients (see table 2). Overall, the
median age was 77 years (IQR 70–84) and 1204 patients were Endpoint analysis for the early DOAC-start group (DOAC
women (47%). Stroke severity was milder in the early DOAC- started within 5 days of the index event)
start group (NIHSS score of 4 (IQR 1–8) vs 6 (IQR 3–12), In the early DOAC-start group, within 5 days of the index
p<0.001). The index event was AIS in 2455 patients (95.5%), a event, 8 patients (0.6%) suffered from a recurrent AIS before
TIA in 95 (4.5%). Prior to the index event, 773 (37%) patients DOAC was even started, 15 patients (1.1%) suffered from
were on antiplatelet therapy, 497 patients (19%) were on oral a recurrent AIS after DOAC-start; only 2 patients (0.1%)
anticoagulants. After the index event, DOACs were started suffered from ICH, all after DOAC- start. The composite
early (≤5 days of the index event) in 1362 (53%) patients, late 30-day rate of any stroke was 1.8% (25/1362).
(6–30 days of the index event) in 1188 (47%). There were
76 patients who suffered the index stroke under oral antico- Endpoint analysis for the late DOAC-start group (DOAC
agulation and where anticoagulation was continued without started between 6 and 30 days of the index event)
pausing (5.6% of the early DOAC-start group). NIHSS score In the late DOAC-start group, within 30 days of the index
on admission was missing in 160 patients (6.3%), renal func- event, eight patients (0.7%) suffered from a recurrent AIS
tion in 276 patients (10.8%) and information on intravenous before DOAC was started; six patients (0.5%) suffered from
thrombolysis in 22 patients (0.9%). a recurrent AIS after DOAC- start; four patients (0.3%)
Figure 3 Subgroup analyses. The endpoint is the composite of recurrent
acute ischaemic stroke, intracranial haemorrhage and any-cause mortality
within 30 days of the index event.
Subgroup analyses
The rate of the composite endpoint (recurrent AIS, ICH and
death within 30 days) did not differ significantly across subgroups
(figure 3, no significant interaction).
DISCUSSION
In this individual patient data pooled analysis of eight inter-
national, prospective cohort studies including 2550 patients
Figure 2 (A–C) Thirty-day outcomes and landmark analyses at day 5. (A) treated with DOACs within 30 days of an AF-related AIS or
Recurrent AIS, (B) ICH and (C) any stroke. AIS, acute ischaemic stroke; ICH, TIA, we found that (1) the 30-day rate of recurrent AIS (1.5%)
intracranial haemorrhage; AC, anticoagulation. was more than sevenfold higher than the rate of ICH (0.2%);
(2) 40% of recurrent AIS occurred early—within 5 days of
the index event—and 43% of recurrent AIS occurred before
any DOAC-start, being thus potentially preventable; (3) both
suffered from ICH, all after DOAC- start. The composite
early (≤5 days) and late (>5 days) as well as ESC guideline
30-day rate of any stroke was 1.5% (18/1188)
adherent versus non- adherent DOAC- starts were similar
in terms of recurrent AIS, ICH and any stroke at 30- day
Endpoint comparison late versus early DOAC-start groups and follow-up. Our results do not corroborate the concern that
by adherence to the ESC guidelines early anticoagulation with DOACs might increase the risk of
In the propensity score weighted comparison of late versus ICH compared with a delayed strategy. Given the seven times
early DOAC- groups, no statistically significant difference higher risk of recurrent AIS—with almost half occurring in
in the HRs was observed for the endpoint of recurrent AIS the first 5 days and before any DOAC- start—our findings
(aHR=1.2, 95% CI 0.5 to 2.9, p=0.69), ICH (aHR=6.0, suggest that early DOAC-start after AF-related AIS or TIA
95% CI 0.6 to 56.4, p=0.12) and any stroke (aHR=1.44, might be reasonable and justifies enrolment of patients into
95% CI 0.66 to 3.33, p=0.33) (see figure 2A–C). A Fine-Grey ongoing randomised controlled trials.
model to take competing risk into account did not change The risk of a recurrent AIS is front-loaded—it peaks imme-
these HRs. The propensity score weights were estimated diately after the index AIS/TIA and drops thereafter: in our
using the following variables: NIHSS score at onset, type of study, over the first 5 days, the risk of AIS was 0.12% per day (15
event (AIS vs TIA), intravenous thrombolysis, any anticoag- patients with recurrent AIS/2550 patients/5 days), whereas—
ulation at baseline, antiplatelets at baseline, impaired renal over the following 25 days—the risk of AIS was 0.03% per
function, current smoking status, CHA2DS2-Vasc and HAS- day (22 patients with recurrent AIS/2550 patients/25 days).
BLED scores. Sensitivity analyses with Kaplan-Meier curves These event rates are considerably higher than those reported
for the landmark analyses at days 3 and 7 are presented in in the subgroup analyses of the four pivotal randomised clin-
online supplemental figures 1 and 2, respectively. ical trials comparing DOACs to VKAs for AF, in which the
A sensitivity analysis by adherence to the ESC Guidelines subgroups of patients with a remote history of cerebrovas-
revealed that DOAC-start was guideline-adherent in 1445 cular events had a rate of recurrent AF-related AIS between
patients (57%) and non-adherent in 1105 patients (43%). 0.0048% per day in the RE-LY trial (dabigatran group, exclu-
There was no significant difference in the propensity score sion of patients up to 6 months after AIS) and 0.0064% per
adjusted rates for ischaemic stroke (aHR 1.15, 95% CI 0.58 day in ROCKET-AF (rivaroxaban group, exclusion of patients
to 2.31, p=0.69), ICH (aHR 1.28, 95% CI 0.22 to 7.56, up to 3 months after AIS), that is, around 10-fold lower than
p=0.79) and any-stroke (aHR 1.17, 95% CI 0.61 to 2.23, ours.9 10
p=0.64); the Kaplan- Meier curves are presented in the The rate of a recurrent AIS observed in our study was
online supplemental figure 3. considerably lower than reported in earlier studies (0.3% and
1.1% per day).11 12 One potential reason for this difference is on this important question, four randomised clinical trials
the choice of anticoagulant; in the earlier studies—published are comparing different timepoints of DOAC-start: ELAN
between 1984 and 1993—the anticoagulant of choice was (NCT03148457, Switzerland/International), OPTIMAS
often heparin; in the year 2000, however, the Heparin in (EudraCT, 2018-003859-38; UK), TIMING (NCT02961348,
Acute Embolic Stroke Trial (HAEST) found no evidence Sweden) and START (NCT03021928, USA).3 The event
that low-molecular-weight heparin, started within 30 hours rates we report here might be informative in determining
from stroke onset, was superior to aspirin at 14 days, adding the sample size needed in the ongoing trials to demonstrate
to the body of evidence that heparin—both fractioned and a difference in efficacy and safety between early and late
unfractioned—does not protect effectively against recurrent DOAC treatment.
AIS.13 14 Also, the cited studies looked only at the first 14 days Strengths of our study include the international multi-
following the index AIS, a time frame with a higher recur- centre design, the inclusion of a current, large cohort of
rence rate than over a longer follow-up period as in our study patients—all treated with DOACs. The 30- day follow- up
(30 days). focusses on the critical first few days following an AIS, where
The lower AIS-rate in our study is not likely to result from the complication risk is highest. The main limitation is that
the exclusion of VKA-treated patients, as a recent analysis this study is not randomised, and residual confounding
comparing DOAC to VKA within 3 months of an AF-related cannot be ruled out. Propensity scores can only take account
AIS showed similarly low rates of recurrent AIS between the for known confounders, not for unknown or unmeasured
DOAC and VKA groups.2 ones, such as lesion size, location and haemorrhagic trans-
The low 30-day rate of ICH is likely due to the exclu- formation. Also, propensity scores can compensate only in
sion of VKA-treated patients, as DOACs are associated with
part for the low event rate—the risk of overfitting cannot be
a lower ICH-rate than VKA.2 15 The low 30-day rate of ICH
ruled out. The immortal time bias also merits discussion: in
aligns well with four studies on patients with AF-related AIS
the early DOAC-start group, eight patients suffered from a
with clinical follow-up of ≥3 months, in which the annual-
recurrent AIS within the first 5 days but prior to any DOAC-
ised rate of ICH was between 1.2 and 9.5 times lower than
start; here, the early DOAC-start is likely to have been a
the annualised rate of recurrent AIS.16–19The study with
reaction triggered by the recurrent AIS itself; these patients
the highest annualised ICH- rate was RAF- NOAC (6.4%
would bias the 30-day recurrent AIS rate (artificially higher
per year), but most ICH occurred beyond 30 days from
AIS onset, making the association between ICH and early in the early DOAC-start or lower in the late DOAC-start
group). We addressed this issue with the landmark analysis.
DOAC-start uncertain.
In our study, seven patients suffered from ICH between The landmark analysis considers only patients who were
the index AIS/TIA and before DOAC- start, two of them event-free up to 5 days following the index event, in order to
immediately after acute endovascular treatment. In these estimate whether the event rates after day five differ between
seven patients, DOAC-start was performed late (>5 days). the early and late DOAC-start groups. We acknowledge that
We excluded these patients because their ICH could not be the selection of the time point of 5 days to discriminate
attributed to the DOAC-start following the index event. Also, between early and late DOAC initiation may be considered
including these patients would have resulted in a falsely higher as arbitrary; we tried to compensate this through the analysis
30-day ICH rate in the late DOAC-start group. ‘Primary’ on adherence to the ESC-guidelines. The amount of inter-
haemorrhagic transformation (ie, independent of anticoagu- action analyses is problematic given the limited number of
lation or recanalisation therapies) is a known complication events – the results are exploratory and shall be viewed with
of large AIS, including those due to AF. In comparison, in caution. Finally, as in many observational study comparing
the aspirin arm of the HAEST trial, symptomatic ICH was treatments, the possibility of confounding by indication
reported in four patients (1.8%) at 14 days of their AF-related cannot be ruled out.
AIS. In conclusion, our results—based on individual assess-
Because no routine imaging was performed on follow-up, ments used to select patients to start early versus late—do not
reported events are confined to symptomatic ICH. Asymp- corroborate concerns that early anticoagulation with DOACs
tomatic haemorrhagic transformation—although more may increase the risk of ICH. The sevenfold higher risk of
frequent—does not seem to portend symptomatic ICH after recurrent AIS than ICH—with almost half of these occurring
DOAC are started: in a recent prospective study on 60 patients before DOAC- start—suggests that early DOAC- start after
with an acute AF-related AIS, baseline MRI showed asymp- AF-related AIS might be reasonable. This supports enrolment
tomatic petechial haemorrhagic transformation in 25 (42%) into ongoing randomised trials comparing an early versus late
of patients; rivaroxaban was started in all of the 60 patients, DOAC anticoagulation after AIS linked to AF.
after an average of 3 days, and none developed symptomatic
haemorrhagic transformation by day 7.20 Author affiliations
1
The optimal time to start DOAC remains controversial. Neurology and Stroke Center, University Hospital Basel & University of Basel, Basel,
Our data suggest that an early DOAC-start (≤5 days) might Switzerland
2
Neurology and Stroke Center, Inselspital, University Hospital Bern, Bern, Switzerland
be reasonable, given that 43% of recurrent AIS were poten- 3
Clinical Trial Unit, University Hospital of Basel & University of Basel, Basel,
tially preventable, and that the early DOAC- start group Switzerland
did not show the feared excess in ICH. Similar findings 4
Stroke Research Center, Department of Brain Repair and Rehabilitation, UCL
were observed in the recent, comprehensive analysis of the Institute of Neurology and The National Hospital for Neurology and Neurosurgery,
SAMURAI-NVAF registry.21 Current guidelines are inconsis- London, UK
5
Stroke Unit, Santa Maria Misericordia Hospital, Perugia, Italy
tent on when to start oral anticoagulation after AF-related 6
Department of Neurology, San Giovanni Battista Hospital of Foligno, Foligno,
AIS, and do not distinguish between use of DOAC and VKA, Umbria, Italy
that is, they do not take into account the lower ICH-rates 7
Second Department of Neurology, ’Attikon’ Hospital, University of Athens, School of
observed with DOAC compared with VKA. To shed light Medicine, Athens, Greece

8
Department of Neurology, University of Tennessee Health Science Center, Memphis, were approved by the ethics committee in the National Cerebral and Cardiovascular
Tennessee, USA Center (M23-18-3 and M29-077).
9
Department of Cerebrovascular Medicine, National Cerebral and Cardiovascular Provenance and peer review Not commissioned; externally peer reviewed.
Center, Osaka, Japan
10
Department of Neuroscience, Azienda Ospedaliera Universitaria Integrata Verona, Data availability statement Data are available upon reasonable request.
Verona, Italy The data used during this study are available from the corresponding author on
11
Department of Neurology, University of Erlangen-Nuremberg, Erlangen, Germany reasonable request.
12
Neurocenter of Southern Switzerland, Ospedale Regionale di Lugano, Lugano, Supplemental material This content has been supplied by the author(s). It
Switzerland has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have
13
Department of Neurology & Stroke Unit, San Giuseppe Hospital IRCSS Multimedica, been peer-reviewed. Any opinions or recommendations discussed are solely those
Milano, Italy of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and
14
Department of Neurology & Neurorehabilitation, University Center for Medicine of responsibility arising from any reliance placed on the content. Where the content
Aging and Rehabilitation Basel, Felix Platter Hospital, Basel, Switzerland includes any translated material, BMJ does not warrant the accuracy and reliability
of the translations (including but not limited to local regulations, clinical guidelines,
Twitter Gian Marco De Marchis @gmdemarchis terminology, drug names and drug dosages), and is not responsible for any error
and/or omissions arising from translation and adaptation or otherwise.
Acknowledgements We acknowledge the following colleagues (from the RAF
Study): Giancarlo Agnelli, Fabio Bandini, Shadi Yaghi, Karen L Furie, Prasanna Tadi, ORCID iDs
Cecilia Becattini, Marialuisa Zedde, Azmil H Abdul-Rahim, Kennedy R Lees, Andrea Gian Marco De Marchis http://orcid.org/0000-0002-0342-9780
Alberti, Michele Venti, Cataldo D’Amore, Maria Giulia Mosconi, Ludovica Anna Cimini, David J. Seiffge http://orcid.org/0000-0003-3890-3849
Riccardo Altavilla, Jessica Fusaro, Paolo Bovi, Monica Carletti, Alberto Rigatelli, Georgios Tsivgoulis http://orcid.org/0000-0002-0640-3797
Jukka Putaala, Liisa Tomppo, Turgut Tatlisumak, Simona Marcheselli, Alessandro Masatoshi Koga http://orcid.org/0000-0002-6758-4026
Pezzini, Loris Poli, Alessandro Padovani, Luca Masotti, Vieri Vannucchi, Sung-Il Sohn, David J. Werring http://orcid.org/0000-0003-2074-1861
Gianni Lorenzini, Rossana Tassi, Francesca Guideri, Maurizio Acampa, Giuseppe
Martini, George Ntaios, George Athanasakis, Konstantinos Makaritsis, Efstathia
Karagkiozi, Konstantinos Vadikolias, Chrissoula Liantinioti, Maria Chondrogianni, REFERENCES
Nicola Mumoli, Domenico Consoli, Franco Galati, Simona Sacco, Antonio Carolei, 1 Ruff CT, Giugliano RP, Braunwald E, et al. Comparison of the efficacy and safety of
Cindy Tiseo, Francesco Corea, Walter Ageno, Marta Bellesini, Giovanna Colombo, new oral anticoagulants with warfarin in patients with atrial fibrillation: a meta-
Giorgio Silvestrelli, Alfonso Ciccone, Alessia Lanari, Umberto Scoditti, Licia Denti, analysis of randomised trials. Lancet 2014;383:955–62.
Michelangelo Mancuso, Miriam Maccarrone, Leonardo Ulivi, Giovanni Orlandi, Nicola 2 Seiffge DJ, Paciaroni M, Wilson D, et al. Direct oral anticoagulants versus vitamin K
Giannini, Gino Gialdini, Tiziana Tassinari, Maria Luisa De Lodovici, Giorgio Bono, antagonists after recent ischemic stroke in patients with atrial fibrillation. Ann Neurol
Christina Rueckert, Antonio Baldi, Sebastiano D’Anna, Danilo Toni, Federica Letteri, 2019;85:823–34.
Martina Giuntini, Enrico Maria Lotti, Yuriy Flomin, Alessio Pieroni, Odysseas Kargiotis, 3 Seiffge DJ, Werring DJ, Paciaroni M, et al. Timing of anticoagulation after
Theodore Karapanayiotides, Serena Monaco, Mario Maimone Baronello, Laszló recent ischaemic stroke in patients with atrial fibrillation. Lancet Neurol
Csiba, Lilla Szabó, Alberto Chiti, Elisa Giorli, Massimo Del Sette, Davide Imberti, 2019;18:117–26.
Dorjan Zabzuni, Boris Doronin, Vera Volodina, Patrik Michel, Peter Vanacker, Kristian 4 Eun M-Y, Kim J-Y, Hwang Y-H, et al. Initiation of Guideline-Matched oral
Barlinn, Lars-Peder Pallesen, Jessica Barlinn, Dirk Deleu, Gayane Melikyan, Faisal anticoagulant in atrial Fibrillation-R elated stroke. J Stroke 2021;23:113–23.
Ibrahim, Naveed Akhtar and Vanessa Gourbali. 5 Schulman S, Kearon C. Subcommittee on control of anticoagulation of the S,
Contributors GMDM, DJS, MP amd DJW conceived and designed the study. standardization Committee of the International Society on T, haemostasis. Definition
GMDM, DJS, SS, DW, VC, MA, GT, MK, SY, KT, MC, BB, KM, BK, CWC, PL, LHB, MP, SE of major bleeding in clinical investigations of antihemostatic medicinal products in
and DJW acquired and analysed the data. GMDM, DJS, SS, DW, VC, MA, GT, MK, SY, non-surgical patients. J Thromb Haemost 2005;3:692–4.
KT, MC, BB, KM, BK, CWC, PL, LHB, MP, SE and DJW drafted a significant portion of 6 Fischer U, Baumgartner A, Arnold M, et al. What is a minor stroke? Stroke
the manuscript or figures. 2010;41:661–6.
7 Mazya MV, Lees KR, Collas D, et al. Iv thrombolysis in very severe and
Funding The authors have not declared a specific grant for this research from any severe ischemic stroke: results from the SITS-ISTR registry. Neurology
funding agency in the public, commercial or not-for-profit sectors. 2015;85:2098–106.
Competing interests The following companies manufacture drugs involved in 8 Levey AS, Stevens LA, Schmid CH, et al. A new equation to estimate glomerular
this study: Bayer (BY, rivaroxaban), Boehringer Ingelheim (BI, dabigatran), Pfizer/ filtration rate. Ann Intern Med 2009;150:604–12.
Bristol Meyer Squibb (PB, apixaban) and Daiichi Sankyo (DS, edoxaban). GMDM: 9 Diener H-C, Connolly SJ, Ezekowitz MD, et al. Dabigatran compared with
scientific advisory boards and travel honoraria: BY; speaker honoraria: PB. DJS: warfarin in patients with atrial fibrillation and previous transient ischaemic
scientific advisory boards: BY and PB. GT: scientific advisory boards: BY, DS and BI. attack or stroke: a subgroup analysis of the RE-LY trial. Lancet Neurol
MK: speaker honoraria from DS, BY and PB. KT: speaker honoraria from DS, BY, BI 2010;9:1157–63.
and PB. KM: speaker honoraria from BY, travel grant from PB. PL: scientific advisory 10 Hankey GJ, Patel MR, Stevens SR, et al. Rivaroxaban compared with warfarin
boards: BY, DS and BI; funding for travel or speaker honoraria: BY and BI; research in patients with atrial fibrillation and previous stroke or transient ischaemic
funding: BI. LHB: consultancy or advisory board fees or speaker’s honoraria from attack: a subgroup analysis of rocket AF. Lancet Neurol 2012;11:315–22.
BY and PB. MP: honoraria as a member of the speaker bureau of BI, BY and PB. SE: 11 Hart RG, Coull BM, Hart D. Early recurrent embolism associated with
funding for travel or speaker honoraria: BY, BI, PB and DS; scientific advisory boards: nonvalvular atrial fibrillation: a retrospective study. Stroke 1983;14:688–93.
BY, BI and PB; educational grant from PB; research grant from DS. DJW: speaking 12 Paciaroni M, Agnelli G, Micheli S, et al. Efficacy and safety of anticoagulant treatment
honoraria: BY. in acute cardioembolic stroke: a meta-analysis of randomized controlled trials. Stroke
2007;38:423–30.
Patient consent for publication Not applicable.
13 Berge E, Abdelnoor M, Nakstad PH, et al. Low molecular-weight heparin versus
Ethics approval This study involves human participants and was approved by aspirin in patients with acute ischaemic stroke and atrial fibrillation: a double-blind
an ethics committees of participating centres. The NOACISP Long Term registry randomised study. The Lancet 2000;355:1205–10.
and the current analysis of pooled individual patient data were approved by the 14 Anon. The International stroke trial (IST): a randomised trial of aspirin,
ethics committee in Basel, Switzerland (EKNZ 2014–027). Patients provided written subcutaneous heparin, both, or neither among 19435 patients with acute
consent for participation in NOACISP Long Term. The requirement for additional ischaemic stroke. International stroke trial Collaborative group. Lancet
local ethical approval differed among participating centres and was acquired by 1997;349:1569–81.
the local principal investigator as well as written informed consent from the patient 15 Xian Y, Xu H, O’Brien EC, et al. Clinical effectiveness of direct oral
if necessary. For the other study centres, we invited all the corresponding authors anticoagulants vs warfarin in older patients with atrial fibrillation and ischemic
of all the relevant studies in 2017. Patients from the Neurocentro della Svizzera stroke: findings from the patient-c entered research into outcomes stroke
Italiana, Lugano, Switzerland, were included based on the decision of the local ethics patients prefer and effectiveness research (prosper) study. JAMA Neurol
committee (2017–02041 CE 3298); accordingly, patients were informed about 2019;76:1192–202.
the anonymisation and use of their routinely collected data for research purposes. 16 Seiffge DJ, Traenka C, Polymeris A, et al. Early start of DOAC after ischemic
Patients who denied use of their data were excluded from the analysis. CROMIS-2 stroke: risk of intracranial hemorrhage and recurrent events. Neurology
was approved by the National Research Ethics Committee, London Queen Square. 2016;87:1856–62.
Patients with capacity gave informed written consent. When patients could not 17 Arihiro S, Todo K, Koga M, et al. Three-Month risk-benefit profile of anticoagulation
consent, we obtained written consent from a proxy as defined by relevant local after stroke with atrial fibrillation: the SAMURAI-Nonvalvular atrial fibrillation (NVAF)
legislation. The SAMURAI-NVAF registry and the collaboration with the joint initiative study. Int J Stroke 2016;11:565–74.

18 Paciaroni M, Agnelli G, Falocci N, et al. Early recurrence and major bleeding 22 Cappellari M, Carletti M, Danese A, et al. Early introduction of direct oral
in patients with acute ischemic stroke and atrial fibrillation treated with anticoagulants in cardioembolic stroke patients with non-valvular atrial fibrillation. J
Non-Vitamin-K oral anticoagulants (RAF-N OACs) study. J Am Heart Assoc Thromb Thrombolysis 2016;42:393–8.
2017;6:e007034. 23 Macha K, Volbers B, Bobinger T, et al. Early initiation of anticoagulation with direct
19 Wilson D, Ambler G, Banerjee G, et al. Early versus late anticoagulation for ischaemic oral anticoagulants in patients after transient ischemic attack or ischemic stroke. J
stroke associated with atrial fibrillation: multicentre cohort study. J Neurol Neurosurg Stroke Cerebrovasc Dis 2016;25:2317–21.
Psychiatry 2019;90:320–5. 24 Toyoda K, Arihiro S, Todo K, et al. Trends in oral anticoagulant choice for acute stroke
20 Gioia LC, Kate M, Sivakumar L, et al. Early rivaroxaban use after patients with nonvalvular atrial fibrillation in Japan: the SAMURAI-NVAF study. Int J
cardioembolic stroke may not result in hemorrhagic transformation: a Stroke 2015;10:836–42.
prospective magnetic resonance imaging study. Stroke 2016;47:10.1161/ 25 Charidimou A, Wilson D, Shakeshaft C, et al. The clinical relevance of microbleeds in
STROKEAHA.116.013491:1917–9. stroke study (CROMIS-2): rationale, design, and methods. Int J Stroke 2015;10 Suppl
21 Mizoguchi T, Tanaka K, Toyoda K, et al. Early initiation of direct oral A100:155–61.
anticoagulants after onset of stroke and short- and long-term outcomes 26 Paciaroni M, Agnelli G, Falocci N, et al. Early recurrence and cerebral bleeding in
of patients with nonvalvular atrial fibrillation. Stroke 2020;51:10.1161/ patients with acute ischemic stroke and atrial fibrillation: effect of anticoagulation
STROKEAHA.119.028118:883–91. and its timing: the Raf study. Stroke 2015;46:2175–82.

BMJ Publishing Group Limited (BMJ) disclaims all liability and responsibility arising from any reliance
Supplemental material placed on this supplemental material which has been supplied by the author(s) J Neurol Neurosurg Psychiatry
Landmark analysis at day 3

AIS
1.00
0.99
Proportion without event
0.98
early AC
late AC
0.97
0.96
0.95
0 5 10 15 20 25 30
Time since qualifying event (in days)
ICH
1.00
0.99
0.98
early AC
late AC
0.97
0.96
0.95
0 5 10 15 20 25 30
any stroke
1.00
0.99
0.98
early AC
late AC
0.97
0.96
0.95
0 5 10 15 20 25 30
De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2021;0:1–6. doi: 10.1136/jnnp-2021-327236

Landmark analysis at day 7

AIS
1.00
0.99
0.98
early AC
late AC
0.97
0.96
0.95
0 5 10 15 20 25 30
ICH
1.00
0.99
0.98
early AC
late AC
0.97
0.96
0.95
0 5 10 15 20 25 30
any stroke
1.00
0.99
0.98
early AC
late AC
0.97
0.96
0.95
0 5 10 15 20 25 30

ESC Guideline adherent versus non-adherent

1.00
0.99 AIS
AIS
0.98
NOT Guideline−adherent
Guideline−adherent
0.97
0.96
0.95
0 5 10 15 20 25 30
ICH
1.00
0.99
0.98
0.97
0.96
0.95
0 5 10 15 20 25 30
any stroke
1.00
0.99
0.98
0.97
0.96
0.95
0 5 10 15 20 25 30

2022

Uploaded by

Copyright:

Available Formats

2022

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

2022

Uploaded by

Copyright:

Available Formats

Cerebrovascular disease

Early versus late start of direct oral anticoagulants

© Author(s) (or their

RESULTS Endpoint analysis for the whole cohort

De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2022;93:119–125. doi:10.1136/jnnp-2021-327236 123

124 De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2022;93:119–125. doi:10.1136/jnnp-2021-327236

De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2022;93:119–125. doi:10.1136/jnnp-2021-327236 125

Landmark analysis at day 3

Time since qualifying event (in days)

Time since qualifying event (in days)

Time since qualifying event (in days)

De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2021;0:1–6. doi: 10.1136/jnnp-2021-327236

Landmark analysis at day 7

Time since qualifying event (in days)

Time since qualifying event (in days)

Time since qualifying event (in days)

De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2021;0:1–6. doi: 10.1136/jnnp-2021-327236

ESC Guideline adherent versus non-adherent

Time since qualifying event (in days)

Time since qualifying event (in days)

Time since qualifying event (in days)

De Marchis GM, et al. J Neurol Neurosurg Psychiatry 2021;0:1–6. doi: 10.1136/jnnp-2021-327236

You might also like