Study reference

Study characteristics

Patient characteristics ²

Intervention (I)

Comparison / control (C) ³

Follow-up

Outcome measures and effect size ⁴

Comments

ICH

Majeed, 2010

Type of study:

Retrospective cohort

Setting:

Multicentre

Country:

Sweden, Canada

Source of funding: S.S. received consulting fees from many pharma, H.M. has received lecure fees from pharma, others have no potential conflicts of interest.

Inclusion criteria:

Patients treated with warfarin, diagnosis (radiologically confirmed) of intracranial haemorrhage (ICD-10: 1600-1629), INR>1.5

Exclusion criteria: haemorrhage caused by severe accidents with multi-trauma or haemorrhagic transformation of ischemic stroke, death <1week after bleeding

N total at baseline: 234 (first week survivors n=177)

Important prognostic factors:

median ± IQR:

total: 75 (65-80)

I: 70 (63-77)

C: 78 (70.5-72)*

Sex male, n (%):

Total: 112 (63)

I: 31 (69)

C:56 (64)

Indication for anticoagulation

Total; I; C, n(%)

AF: 102 (58); 22 (49); 79 (91)

VT: 30 (17); 0; 0

Mechanical aortic valve:19 (11); 15 (33); 4 (5)

Mechanical mitral valve:9 (5); 7 (16); 2(2)

Other 17 (10); 1(2); 2(2):

Groups comparable at baseline? Patients who did not resume warfarin were significantly older.

No resumption of warfarin

Resumption of warfarin (different time periods) (n=59, 33%)

Length of follow-up:

Median 69 weeks (IQR 19-144)

Loss-to-follow-up:

113 (48%) of the patients died during follow up, with median survival of 4.5 years.

Incomplete outcome data:

Intervention:

N (%)

Reasons (describe)

Control:

N (%)

Reasons (describe)

I vs. C

Cox proportional hazards model at different time intervals without (I) and with (C) resumption of warfarin:

Risk of recurrent intracranial haemorrhage per day

1-35 days:

0.18%vs. 0.75%

HR: 4.13

36-63 days:

0.044% vs. 0.20%

HR: 4.46

64-217 days:

0.0% vs. 0.20%

HR: ∞

>218 days:

0.0% vs. 0.0069%

HR: ∞

Risk of ischemic event per day

1-77 days:

0.068%vs. 0.0%

HR: 0.00

78-329 days:

0.039% vs. 0.0%

HR: 0.0

>330 days:

0.017% vs. 0.0035%

HR: 0.21

The modelling of risk for recurrent intracranial haemorrhage vs. ischemic stroke in patients with or without resumption of warfarin therapy is based on the population with cardiac indication for anticoagulation and/or with previous ischemic stroke and who had survived the first week without a recurrent event (n=132).

Resumption of warfarin at any given time point will increase the subsequent risk of recurrent intracranial haemorrhage and reduce the risk of a thromboembolic event. The optimal restart time must balance these 2 competing cumulative risks over the entire warfarin “treatment horizon”.

To determine the optimal restart time, we varied warfarin resumption between 1 and 50 weeks after the index intracranial bleed and calculated the total risk (before + after selected time point of resumption) of recurrence IH and of thromboembolic event through to the end of treatment. The calculation of cumulative risks used the rated displayed left (outcome). Given that it is implausible that the actual risk of intracranial bleed or thromboembolic event is 0, we have not used the observed daily risks directly, but instead we have blended the observed rates and cox model hazard ratio.

The results (figure) demonstrate how the total risk of intracranial haemorrhage and an ischemic event for the whole treatment period varies according to when warfarin is restarted. Based on this combined risk the optimal period of resumption of warfarin seems to be between 10 and 30weeks from the index intracranial haemorrhage over a survival- and treatment-horizon of 3 years.

Yung, 2012

Type of study:

Setting:

Cohort

Country:

Canada

Source of funding: none

Inclusion criteria:

Warfarin related ICH (intracerebral or subarachnoid) submitted to 13 stroke centres, >18 years, who were or were not restarted on warfarin after hospitalization.

Exclusion criteria: previous ICH of any type, other concurrent active bleeding process, pregnancy, bleeding diathesis precluding reinitiation of warfarin, recent trauma, hemorrhagic conversion from ischemic stroke, neurosurgical instrumentation, intracranial neoplasia, were palliative, or if thrombolysis was administered.

N total at baseline: 284

Intervention: 91

Control: 193

Important prognostic factors²:

age ± SD:

I: 71.8 (12.9)

C: 74.4 (11.9)

P=0.11

Sex:

I: 46 (50.5)

C: 110 (57.0)

P=0.31

Stroke type:

Intracerebral hemorrhage I: 78 (85.7)

C: 174 (90.2)

Subarachnoid hemorrhage I: 13 (14.3)

C: 19 (9.8)

P=0.27

GCS score:

I: 10.2 (3.9)

C: 7.2 (3.6)

P=0.001

Atrial fibrillation at admission:

I: 46 (50.5)

C: 89 (46.1)

P=0.49

Prosthetic valve:

I: 19 (20.9)

C: 18 (9.3)

P=0.007

Antiplatelet use:

I: 12 (13.2)

C:45 (23.3)

P=0.047

The main indications for anticoagulation included atrial fibrillation or flutter (67.3%), valve prosthesis (13.0%), and venous thromboembolic disease (10.9%).

Groups comparable at baseline? No

Warfarin was restarted (n=91) in hospital

Warfarin not restarted (n=193)

Length of follow-up: 1 year

Loss-to-follow-up:

Intervention:

N (%)

Reasons (describe)

Control:

N (%)

Reasons (describe)

Incomplete outcome data:

Intervention:

N (%)

Reasons (describe)

Control:

N (%)

Reasons (describe)

Unadjusted outcomes

I vs. C, n (%)

Death:

In hospital: 30 (33.0) vs. 98 (50.8) p=0.005

1 month: 29 (31.9) vs. 105 (54.4) p=0.001

6 months: 38 (41.8) vs. 114 (59.1) p=0.006

1 year: 44 (48) vs. 118 (61)

P=0.04

ICH expansion or recurrrence (haemorrhage recurrence of expansion confirmed on serial CT or MRI): 14 (15.4) vs. 29 (15.0)

P=0.94

Death or intracranial bleeding:

1 month: 32 (35.2) vs. 106 (54.9) p=0.002

1 year: 46 (50.5) vs. 118 (61.1) p=0.09

Death, bleeding, or thrombotic

complication at 1 year:

47 (51.60) vs. 119 (61.7) p=0.11

Multivariate analysis

Mortality (30 days):

aOR: 0.49 (0.26-0.93)

p=0.03

mortality (1 year)

aOR: 0.79 (0.43-1.43)

p=0.43

Data was collected from the registry of the Canadian stroke network (includes audit information prospectively collected on all consecutive patients with acute stroke or TIA evaluated in the ER and admitted to hospital at acute institutions; combined with data from the registered persons database.

Multivariate logistic regression model was performed including potential clinical variables with p<0.25 on univariate analysis.

Variables examined in multivariable analysis

for warfarin reinitiation included: age, gender, CNS score, mechanical valve prosthesis, hypertension, previous stroke or TIA, and antiplatelet use on admission.

Variables considered in multivariable analysis of all-cause mortality and bleeding included: age, gender, warfarin reinitiation, CNS score, presenece of intraventricular haemorrhage, systolic BP, diabetes, previous stroke or TIA, INR >3.0 and antiplatelet use.

In multivariable analysis, patients receiving concurrent antiplatelet therapy on admission were less likely to be restarted on warfarin

(adjusted odds ratio [aOR], 0.34; 95% confidence interval [CI], 0.16-0.74; P 0.007). Warfarin was more likely to be restarted in those with less severe stroke (CNS score < 7;

aOR, 2.07; 95% CI, 1.20-3.57; P 0.009) or with valve prostheses in situ (aOR 3.07; 95% CI, 1.29-7.27; P 0.011). Prior stroke or TIA, gastrointestinal bleeding, cirrhosis,

renal disease, hypertension, CHADS2 score, presentation INR (  3.0), BP, stroke location, and age were not associated with warfarin reinitiation.

Our findings suggest that patients

with mild-to-moderate stroke severity (ie, CNS score <7) without intraventricular involvement and admission INR

3.0 may be restarted on warfarin without an excess risk of hemorrhage expansion or death. Patients that are high thrombotic

risk with indications for long-term warfarin may be considered for reinitiation in-hospital, although the exact timing of

reinitiation warrants further evaluation. Conversely, patients at high risk for thrombosis with either severe stroke (CNS score >7), intraventricular hemorrhage, or supratherapeutic INR on presentation should be carefully evaluated on an individual

basis as they are more prone to poor outcomes. The routine use of neuroimaging to rule out intracranial hematoma expansion

prior to reinitiating therapy may be reasonable. Finally, all decisions

should be made in the context of each patient’s unique

clinical circumstances and comorbidities. Patients who are not

at high thrombotic risk may not derive net clinical benefit from

early reinitiation of warfarin if the potential for in-hospital

hematoma expansion or recurrent intracranial bleeding remains

high.

Gathier, 2013

Type of study:

retrospective follow-up study

Setting:

Hospital

Country: Netherlands

Source of funding: H.B. van der Worp has served as a consultant to Bristol-Meyers

Squibb.

Inclusion criteria:

ICH while treated with acenocoumarol or phenprocoumom at time of ICH, INR ≥1.1, discharge from hospital

Exclusion criteria:-

N total at baseline: 40

I1 (OAC): 12

I2 (APM): 13

Control (No ATM): 13

Important prognostic factors²:

age ± SD:

I1 (OAC): 70 (54-86)

I2 (APM): 74 (44-87)

Control (No ATM): 69 (48-85)

Reason for OAC

Atrial fibrillation:

I1 (OAC): 4 (33)

I2 (APM): 8 (62)

Control (No ATM): 6 (46)

Heart valve replacement:

I1 (OAC): 2 (17%)

I2 (APM): 0

Control (No ATM):0

Myocardial infarction

I1 (OAC): 0

I2 (APM): 2 (15)

Control (No ATM): 2 (15)

DVT/PE:

I1 (OAC): 4 (33)

I2 (APM): 0

Control (No ATM): 2 (15)

Cerebral infarction

I1 (OAC): 0

I2 (APM): 2 (15)

Control (No ATM): 1 (8)

Other

I1 (OAC): 2 (17)

I2 (APM): 1 (8)

Control (No ATM): 2 (15)

Groups comparable at baseline? no

I1 (OAC): oral anticoagulation was restarted within 2 months after ICH

I2 (APM): antiplatelet therapy was restarted within 2 months after ICH

No antithrombotic medication was started within 2 months after ICH

Length of follow-up:

Median: 3.5 years

Loss-to-follow-up:

2, 1 reason unknown, 1 declined responding to questionnaire

Incomplete outcome data:

Intervention:

N (%)

Reasons (describe)

Control:

N (%)

Reasons (describe)

Any stroke

n

I1 (OAC): 3

I2 (APM): 7

Control (No ATM): 2

Incidence per patient year:

I1 (OAC): 15.4

I2 (APM): 11.0

Control (No ATM): 5.6

Incidence ratio (when compared to no ATM): (95%CI)

I1 (OAC): 2.7 (0.5-16.3)

I2 (APM): 1.9 (0.4-9.4)

Cerebral infarction, n

I1 (OAC): 2

I2 (APM): 6

Control (No ATM): 2

Incidence per patient year:

I1 (OAC): 10.2

I2 (APM): 9.4

Control (No ATM): 5.6

Incidence ratio (when compared to no ATM): (95%CI):

I1 (OAC): 1.8 (0.3-12.9)

I2 (APM): 1.7 (0.3-8.3)

Data were obtained by reviewing medical records and mailed questionnaires

APM: antiplatelet therapy

ATM: antitrombotic medication

OAC: oral anticoagulation

The total of patient-years (PY) on OAC, antiplatelet therapy,

and no antithrombotic medication after the ICH was calculated

until the occurrence of an event or – in case no event occurred –

until the date the questionnaire was administered or the date of

death, whichever came first.

GIB

Witt, 2012

Type of study:

Retrospective cohort study

Setting: hospital

Country: US

Source of funding: CSL Behring LLC provided funding for the study

Inclusion criteria: hospitalized/emergency care encounter for GIB, outpatient purchase of warfarin and INR in 60 days before GIB, KPCO membership, no GIB during 6 months before index GIB

Exclusion criteria: -

N total at baseline: 442

Intervention: 260

Control: 182

Important prognostic factors:

age ± SD:

I: 71.8±12.0

C: 77.7±11.3

Sex:

I: 49.6% M

C: 51.1% M

Indication for anticoagulation therapy:

AF:

I: 46.2%

C: 56.6%, p=0.03

VTE:

I: 25.8%

C: 22.5%, [=0.44

Prostetic heart valve:

I: 15.4%

C: 1.1%, p<0.001

Other:

I: 12.7%

C: 19.8%, p=0.04

GIB location

Large intestine:

I: 28.1%

C: 23.6%, p=0.26

Mouth-esophagus:

I: 7.7%

C: 5.5%, p=0.37

Rectum-anus:

I: 19.6%

C: 7.1%, p<0.01

Small-intestine:

I: 2.7%

C: 3.9%, p=0.50

Stomach-duodenum:

I: 25.0%

C: 32.7%, p=0.07

Not identified:

I: 16.9%

C: 26.9%, p=0.01

Groups comparable at baseline? No, prostetic heart valve and GIB localized to rectum-anus were more common in I group

Resumption of warfarin (or continuation)

No resumption warfarin

Length of follow-up: 90 days

Loss-to-follow-up:

-

Incomplete outcome data:

-

I vs. C

Thrombotic event

(arterial: stroke (n=5), systemic embolus (n=1), venous: PR (n=3), DVT (2):

1 (0.04%) vs. 10 (5.5%), p<0.001

HR 0.05 95%-CI: 0.01-0.58

Recurrent GIB

26 (10.0%) vs. 10 (5.5%), p=0.09

Multivariable analysis:

HR 1.32 95%-CI: 0.50-3.57]

Deceased

15 (5.8%) vs. 37 (20.3%)

P<0.001

Multivariable analysis:

HR 0.31 95%-CI 0.15-0.62)

Analysis excluding deceased within 1 week: association remained

Patients who never stopped warfarin are included in the reinitiation group (n=41)

Median time to resumption of warfarin was 4 days (2-9 days)/

Multivariate analysis was corrected for:

controlled for propensity score, CDS, age, sex, indication for warfarin use, prior heart failure diagnosis, location of GIB, pre-FIB target INR, pre-GIB percentage of INR in range, reception of LMWH, length of ED/inpatient stay, acute GIB treatment

Compared with all other patients the rate of recurrent GIB was sign increased when warfarin therapy was resumed between 1 and 7 days after GIB (6.23% vs. 12.4%, p=0.03).

Most common causes of death: related to malignancy (28.8%), infection (19.2%), cardiac disease (17.3%). No recurrent GIB resulted in death

Sengupta, 2015

Type of study: prospective observational cohort

Setting:

Single setting (hospital)

Country: US

Source of funding: none

Inclusion criteria:

GIB while on systemic anticoagulation (clinically significant GIB: overt hematochezia, hematemesis, melena, guaiac-positive stools, with sign. Drop in haemoglobin).

Exclusion criteria:-

N total at baseline: 208 (excluded: 11)

Intervention: 121

Control:76

Important prognostic factors²:

Age, median (IQR), years:

I: 75 (64,82)

C:77 (66,84), p=0.32

Sex:

I: 55% M

C: 62% M, p=0.38

Indication for anticoagulation

I vs. C, p

Atrial fibrillation:

I: 71 (59%)

C: 44 (58%), p=1.00

History DVT:

I: 23 (19%)

C: 13 (17%), p=0.85

History PE:

I: 17 (14%)

C: 5 (7%), p=0.16

Prosthetic valve:

I: 18 (15%)

C: 0 (0%), p=0.000

Portal vein thrombosis: I: 2 (2%)

C: 4 (5%), p=0.21

Post surgical procedure:

I: 1 (1%) 2

C: (3%), p=0.56

Cause of GIB

Esophagitis

I: 4 (3%)

C: 2 (3%), p=1.00

Gastric ulcer

I: 8 (7%)

C: 7 (9%), p=0.58

Gastritis

I: 8 (7%)

C: 3 (4%), p=0.53

Dieulafoy's

I: 5 (4%)

C: 2 (3%), p=0.71

AVMs

I: 15 (12%)

C: 3 (4%), p=0.07

Duodenal ulcer

I: 5 (4%)

C: 6 (8%), p=0.34

Diverticulosis

I: 6 (5%)

C: 10 (13%), p=0.06

Colonic ulcer

I: 3 (2%)

C: 2 (3%), p=1.0

Hemorrhoids

I: 5 (4%)

C: 1 (1%), p=0.41

Other (ischemic colitis 3, Mallory Weiss Tear 3,GAVE 1 duodenitis 3)

I: 6 (5%)

C: 4 (5%), p=1.00

Colitis

I: 2 (2%)

C:4 (5%), p=0.21

Post polypectomy

I: 6 (5%)

C: 2 (3%), p=0.71

Radiation proctitis

I: 4 (3%)

C: 2 (3%), p=1.00

Source not identified

I: 45 (37%)

C: 27 (36%), p=0.88

Management

No blood transfusion

I: 49 (40%)

C: 24 (32%), p=0.23

No vitamin K

I: 66 (55%)

C: 46 (61%), p=0.46

No FFP

I: 74 (61%)

C: 48 (63%), p=0.88

ICU care

I: 54 (45%)

C: 37 (49%), p=0.66

Endoscopic intervention

I: 26 (21%)

C: 13 (17%), p=0.58

Groups comparable at baseline? Intervention group: more prostetic valves, prior stroke or TIA, or prior GIB. C group: more likely to have history of active malignancy.

Anticoagulation

Resumed (n=121)

Anticoagulation

Stopped (n =76)

Length of follow-up: 90 days

Loss-to-follow-up:

12%

I: 8171 (11%) person-days

C: 4295 (14%) person-days of follow-up

P=0.50

Incomplete outcome data:

Patients who died during initial hospitalization were excluded

from statistical analysis n=11 (5%)

I vs. C

Thromboembolic event

venous thromboembolism

(pulmonary embolism or deep vein thrombosis

(DVT)), arterial thromboembolism, stroke, or transient ischemic

attack.

1 (0.8%) vs, 6 (8%), p=0.003

HR multivariate analysis controlling for propensity score: 0.121,

95%-CI=0.006–0.812

Recurrent GIB

readmission to any hospital

in the 90-day follow-up period because of another episode of GIB

HR=2.17, 95% CI=0.861–6.67, P =0.10

Importantly, there were no

deaths in this group of patients readmitted with GIB.

Mortality:

HR multivariate analysis: HR=0.632, 95% CI=0.216–1.89, P =0.40

At hospital discharge, the decision to resume or discontinue anticoagulation was made by the physicians directly responsible for patient care, depending on clinician and patient preferences. We categorized patients into whether anticoagulation was resumed

or whether there was interruption of anticoagulation. Interruption of anticoagulation was defined as holding systemic anticoagulation for 72 h or more after discharge.

Analyses were adjusted

for the following factors included in the propensity score: age,

gender, Charlson comorbidity index, transfusion requirements,

and active malignancy.

After excluding the patients who restarted systemic anticoagulation

because of having a thromboembolic episode in follow-up, 15

(20%) of the original 76 patients in the anticoagulation interruption

cohort had restarted anticoagulation by the 90-day follow-up

call. In these patients, anticoagulation was restarted at a median

of 25 days from initial discharge. None of the 15 patients who

restarted anticoagulation in follow-up were readmitted to the hospital

within 90 days because of recurrent GIB.

No raw data for recurrent GIB and mortality

In summary, we found that resuming anticoagulation aft er

hospitalization for GIB was associated with a signifi cantly

decreased adjusted risk of major thromboembolic events over 90

days, without a signifi cantly increased risk of recurrent GIB. Th ese

data support the recommendation that anticoagulation should

be continued aft er an episode of GIB whenever possible.

Qureshi, 2014

Type of study:

Retrospective cohort

Setting:

Hospital

Country:

US

Source of funding: department of internal medicine of henry ford health system

Inclusion criteria:

Patients who developed major GIB while taking warfarin and then

had evidence of resolution of major GIB (defined as stability

of hemoglobin levels with <1 g decrease of hemoglobin for

48 hours)

Exclusion criteria: pts who died <72h of GIB, hospice, postoperative or

valvular AF, patients in whom primary indication for anticoagulation

was any reason other than nonvalvular AF,

warfarin was not interrupted for at least

48h

N total at baseline: 1329

Intervention: 653

Control:676

Important prognostic factors²:

age ± SD:

I: 74.8±10.7

C:75.3±10.7, p=0.43

Sex:

I: 55.7% M

C: 49.7% M, p=0.03

Groups comparable at baseline? Caucasians, patients with concomitant

upper and lower sources for GIB, diabetics, patients with

renal disease, history of coronary artery disease, and history

of falls were less likely to be restarted on warfarin (p <0.05). Overall, there were greater co-morbidity burdens in the patients who were not restarted on warfarin. Major

reasons for not restarting warfarin were physician preference

(18%) and patient’s inability to follow up with the anticoagulation

clinic (19%).

Restarting warfarin

Warfarin cessitation (or starting>6 months)

Length of follow-up: 2 years

Loss-to-follow-up:

Intervention:

N (%)

Reasons (describe)

Control:

N (%)

Reasons (describe)

Incomplete outcome data:

Intervention:

N (%)

Reasons (describe)

Control:

N (%)

Reasons (describe)

I vs. C

Recurrent major GIB (<90 days)

defined as any of the

following: (1) >2 g of hemoglobin decrease from the last known hemoglobin level warranting hospitalization, (2)

need for blood transfusion of at least 2 units, and (3) visible

bleeding by health personnel or endoscopic evidence of

stigmata of recent bleeding in the form of visible bleeding or

clot.

Adjusted Hazard Ratio (95% CI): 0.71 (0.54-0.93), p=0.01

Thromboembolism (1 year)

defined as venous thromboembolism

(pulmonary embolism and deep venous thrombosis),

arterial thromboembolism, or stroke or transient ischemic

attack.

Adjusted Hazard Ratio (95% CI): 0.71 (0.54-0.93), p=0.01

Mortality

Adjusted hazard ratio 0.66, 95% confidence

interval 0.56 to 0.81, p <0.0001

GIB was defined as a decrease in hemoglobin of 2 g/dl and/or transfusion of 2 units of packed red blood cells with at least one of the following: hematemesis, melena, hematochezia, bright red blood per rectum, blood in nasogastric aspirate, or bleeding documented during an endoscopic procedure.

Restarting warfarin was defined as prescription of warfarin with objective evidence of increase in international

normalized ratio to 2.0 with evidence of at least 2 days of discontinuation of warfarin as observed by chart review. Patients who interrupted warfarin after 1 month of restarting warfarin (54; 4.1%) were included in the group that restarted warfarin. Patients who started warfarin after 6 months of

interruption (39, 2.9%) were included in the group that did not restart warfarin (warfarin cessation group). Patients who interrupted warfarin within the first month (12; 0.9%) were included in the warfarin cessation group.

Multivariate analysis were adjusted for age, gender, race, Charlson co-morbidity index, number of blood product transfusions,

international normalized ratio on admission, and CHADS2 and HAS-BLED scores

Traumatisch hersenletsel

Albrecht, 2014

Type of study:

Retrospective analysis insurance beneficiaries

Setting:

hospital

Country:

USA

Source of funding: none reported

Inclusion criteria:medicare benificiaries, >65yrs, hospitalized for traumatic brain injury (2006-2009) who received warfarin in the month prior to the injury (for: AF).

Exclusion criteria:

N total at baseline: 10 782

I1: 1877

I2: 3934

C: 4971

Important prognostic factors²:

age ± SD:

I1: 81.0 (7.4)

I2: 80.1 (7.2)

C: 82.4 (7.3)

Sex: n male (%)

I1: 691 (37)

I2: 1323 (34)

C: 1836 (37)

Atrial fibrillation:

I1:1532 (82)
I2:3333 (85)
C: 3978 (80) p<.001

CHADs2 score >2, No. (%)

I1: 1577 (84)

I2: 3197 (81)

C: 4223 (85) p<.001

Modified HEMORR2HAGES score >3, No. (%)

I1: 1141 (61)

I2: 2093 (53)

C: 2994 (60) p<.001

Groups comparable at baseline? No, Beneficiaries differed by warfarin use categories (Table 1).

Beneficiaries who used warfarin for 7 to 12months (mean [SD]

age, 80.1 [7.2] years) were younger than those who used warfarin

for 1 to 6 months (mean [SD] age, 81.0 [7.4] years) and

thosewhodid not use warfarin (mean [SD] age, 82.4 [7.3] years;

analysis of variance P < .001). They were less likely to have a

hospital stay of 9 ormore days (15%vs 17%vs 28%; P < .001).

Beneficiaries who used warfarin for 7 to 12 months were less

likely to have Alzheimer disease and related dementias (27%

vs 35% vs 39%; P < .001).

Warfarin (or other anticoagulant) resumption

I1: for 1-6 months

I2: 7-12 months

No resumption

Length of follow-up: Mean (SD) length of follow-up after discharge from hospitalization

for TBI was 594.9 (405.6) days

Loss-to-follow-up:

-

Incomplete outcome data:

I1: 719 (38%)

I2: 1205 (31%)

C: 2358 (47%)

Control:

N (%)

Reasons (describe)

Patients transferred to skilled nursing facility (SNF) have missing values for warfarin use (other payment):

Trombotic event, n, incidence rate (per 1000)

I: n=400, 113.5 (95%CI: 102.9-125.2)

C: 562, 155.9 (143.5-169.3)

RR 0.77 [0.67-0.88]

Hemorrhagic event, n, incidence rate (per 1000):

I: 422, 119.8 (108.9-131.8)

C: 309 85.7 (76.7-95.8)

RR: 1.51 (1.29-1.78)

Hemorrhagic or ischemic stroke, n, incidence rate (per 1000):

I: 339, 96.2 (87.5-105.8)

C: 494, 137.0 (122.6-153.2)

RR: 0.83 (0.72-0.96)

The interaction between

period and lagged warfarin use was not statistically significant; therefore, the regression results are interpreted as the

effect of the lagged warfarin use variable on outcomes averaged

over the first 12 periods following discharge from hospitalization for TBI.

The final model for hemorrhagic outcomes included the following time-invariant and time-varying variables: age, sex,

race, pre-TBI hemorrhagic event, lagged warfarin use, period (continuous), other anticoagulant use in the period, length of

hospital stay (categorical), discharge to SNF, atrial fibrillation, liver disease, chronic kidney disease, ethanol abuse, malignant

neoplasm, hypertension, anemia, coagulation defect, neurological disease, and thrombotic event in the period.

The final model for thrombotic outcomes included the following time-invariant and time-varying variables: age, sex, race, pre-

TBI thrombotic event, lagged warfarin use, hemorrhagic event in the period, period (continuous), other anticoagulant use in

the period, length of hospital stay (categorical), discharge to SNF, stroke or transient ischemic attack, hypertension, diabetes mellitus, and heart failure.

Restricting our analyses to patients with atrial fibrillation did not significantly affect estimates of the effect of

warfarin receipt.

Study reference

(first author, year of publication)

Bias due to a non-representative or ill-defined sample of patients?1

(unlikely/likely/unclear)

Bias due to insufficiently long, or incomplete follow-up, or differences in follow-up between treatment groups?2

(unlikely/likely/unclear)

Bias due to ill-defined or inadequately measured outcome ?3

(unlikely/likely/unclear)

Bias due to inadequate adjustment for all important prognostic factors?4

(unlikely/likely/unclear)

ICH

Majeed, 2010

Likely (reasons for restarting therapy are not described, but likely that the factors for restarting are also prognostic factors; therefore likely to cause bias).

Unlikely

Unlikely

Unlikely

Gathier, 2013

Likely (reasons for restarting therapy are likely also prognostic factors; therefore likely to cause bias).

Unlikely

Unlikely

Likely (not all relevant prognostic factors were corrected for / available)

Yung, 2012

Likely (reasons for restarting therapy are likely also prognostic factors; therefore likely to cause bias).

Unclear (retrospective, not fully clear how long follow-up was)

Unlikely

Likely (not all relevant prognostic factors were available from registration)

GIB

Witt, 2012

Likely (reasons for restarting therapy are likely also prognostic factors; therefore likely to cause bias).

Unlikely

Unlikely

Unlikely

Sengupta, 2015

Likely (reasons for restarting therapy are likely also prognostic factors; therefore likely to cause bias).

Unlikely (relatively high loss-to-follow-up, however comparable in both groups)

Unlikely

Unlikely

Qureshi, 2014

Likely (reasons for restarting therapy are likely also prognostic factors; therefore likely to cause bias).

Unlikely

Unlikely

Likely (not all relevant prognostic factors were corrected for / available)

TBI

Albrecht, 2014

Likely (reasons for restarting therapy are likely also prognostic factors; therefore likely to cause bias).

Likely, patients transferred to skilled nursing facility (SNF) have missing values for warfarin use (other payment). Incomplete data differs for groups.

Unlikely

Unlikely