Etable 1 Search Strategy in Pubmed

eTable 1 Search Strategy in PubMed

Step / Search Term / Results
1 / "Atrial Fibrillation"[Mesh] OR atrial fibrillat* OR atrial flutter* / 55,215
2 / "Anticoagulants"[Mesh] OR anticoagulants / 192,277
3 / "apixaban" [Supplementary Concept] OR Apixaban OR Eliquis / 1,204
4 / "dabigatran" [Supplementary Concept] OR Dabigatran OR Pradaxa / 2,414
5 / "edoxaban" [Supplementary Concept] OR edoxaban OR Lixiana / 362
6 / "rivaroxaban" [Supplementary Concept] OR Rivaroxaban OR Xarelto / 1,958
7 / "Warfarin"[Mesh] OR warfarin / 22,465
8 / "Platelet Aggregation Inhibitors"[Mesh] OR Antiplatelet* / 37,183
9 / "Aspirin"[Mesh] OR Aspirin OR acetylsalicylic acid / 58,105
10 / "clopidogrel" [Supplementary Concept] OR Clopidogrel OR Plavix / 10,070
11 / "Ticlopidine"[Mesh] OR ticlopidine OR Ticlid / 8,871
12 / #1 AND (OR/ #2 - #11) / 9,043
13 / Limit to human subjects, #12 NOT "Animals"[Mesh] NOT "Humans"[Mesh] / 9,009
14 / Limit publication type, NOT NOT (((((((((((("Case-Control Studies"[MeSH]) OR "Case Reports"[Publication Type]) OR “Correspondence as Topic”[MeSH]) OR "Comment"[Publication Type]) OR "Editorial"[Publication Type]) OR "Guideline"[Publication Type]) OR "Practice Guideline"[Publication Type]) OR "Letter"[Publication Type]) OR "Clinical Trials, Phase I as Topic”[MeSH]) OR "Cross-Sectional Studies”[MeSH]))) NOT ((("randomized controlled trial "[All Fields]) OR "randomized controlled trial"[Publication Type]) OR "Cohort Studies"[Mesh])) / 4,895
15 / Limit to English language / 3,986

eTable 2. Trial-specific definitions of major bleeding

Study (Author, year) / Major bleeding
Fontaine et al, 2014 / ISTH criteria
Quintanilla et al, 2014 / Transfusion of ≥2 units of packed red blood cells + hemoglobin drop of ≥2 gr/dl.
Hernandez et al, 2014 / Intracranial haemorrhage, hemoperitoneum, inpatient or emergency department stays for gastrointestinal, hematuria or NOS haemorrhage.
Graham et al, 2014 / Fatal bleeding, hospitalized bleeding requiring transfusion or hospitalization with bleeding into critical organ.
Larsen et al, 2014 / Not reported
Larsen et al, 2013 / Bleeding from or into an organ or sudden hemoglobin drop leading to hospital stay.
Ho et al, 2012 / ISTH criteria
Laliberte et al, 2014 / Not reported
RE-LY, 2009 / ISTH criteria + bleeding events requiring inotropic agents or surgery
PETRO, 2007 / ISTH criteria + bleeding events requiring surgery
NCT 01136408, 2007 / ISTH criteria
ROCKET AF, 2011 / ISTH criteria + bleeding events associated with permanent disability
J-ROCKET AF, 2012 / ISTH criteria
ARISTOTLE, 2011 / ISTH criteria
ARISTOTLE J, 2011 / ISTH criteria
ENGAGE AF-TIMI 48, 2013 / ISTH criteria
Yamashita et al, 2012 / Modified ISTH criteria (blood transfusion four units or more)
Weitz et al, 2012 / ISTH criteria
Chung et al, 2012 / Modified ISTH criteria (blood transfusion ≥800ml packed red blood cells or whole blood)
Liu et al, 2013 / ISTH criteria
BAFTA, 2007 / Fatal haemorrhage, need transfusion or surgery, intracranial haemorrhage
WASPO, 2007 / Intracranial haemorrhage, hemoglobin drop of 2 g/dL, need for blood transfusion
ACTIVE W, 2006 / Requiring transfusion of two or more units of whole blood or red cells
AFASAK2, 1998 / Fatal, potentially life-threatening bleeding, requiring surgery
AFASAK, 1989 / Not reported
CAFA, 1991 / A fall in hemoglobin level of 2 g/dL (1.24 mmol/L) or more, or requiring transfusion, or bleeding into a sensitive location such as pericardium or retina
JAST, 2006 / Fatal bleeding, bleeding requiring hospitalisation, transfusion or fall in haemoglobin of 4g/dL or more
AVERROES, 2011 / ISTH criteria
ACTIVE, 2009 / Requiring transfusion of two or more units of whole blood or red cells or severe bleeding including fatal bleeding, drop in haemoglobin level of 5g/dL or more, hypotension requiring inotropic agents, intraocular bleeding leading to substantial loss of vision, requiring surgery, intracranial haemorrhage, or blood transfusion four units or more
PATAF, 1999 / Requiring hospital admission and blood transfusion or causing fall in haemoglobin of 2g/dL or more
SPAF, 1991 / Any bleeding that involved the central nervous system, management requiring hospitalisation with transfusion and/or surgery, or permanent residual impairment

International Society on Thrombosis and Haemostasis (ISTH) criteria:

1. Fatal bleeding and/or 2. Symptomatic bleeding in a critical area or organ, such as intracranial, intraspinal, intraocular, retroperitoneal, intraarticular or pericardial, or intramuscular with compartment syndrome and/or 3. Bleeding causing a fall in hemoglobin level of 2 g/dL (1.24 mmol/L) or more, or leading to transfusion of two or more units of whole blood or red cells.

eTable 3. Characteristics of excluded studies

No. / Study (Author, year) / Reason for Exclusion
1 / Abe et al, 20151 / FDA Adverse Event Reporting System; single-arm study.
2 / Annweiler et al, 20152 / Cognitive impairment is not an outcome of interest.
3 / Arihiro et al, 20153 / Secondary prevention of stroke is not an outcome of interest.
4 / Ashburner et al, 20154 / Single-arm study.
5 / Barcellona et al, 20155 / Patient preference is not an outcome of interest.
6 / Beyer-Westendorf et al, 20156 / Single-arm study.
7 / Beyer-Westendorf et al, 20157 / Single-arm study.
8 / Bo et al, 20158 / Utilisation study.
9 / Fareau et al, 20159 / Quality of life is not an outcome of interest.
10 / A. Gorst-Rasmussen et al, 201510 / Patient adherence is not an outcome of interest.
11 / Granziera et al, 201511 / Single-arm study.
12 / Haim et al, 201512 / Utilisation study.
13 / Laliberte et al, 201513 / Hospitalisation day is not an outcome of interest.
14 / Lip et al, 201514 / Single-arm study.
15 / Palacio et al, 201515 / Patient preference is not an outcome of interest.
16 / Sauer et al, 201516 / Patient adherence is not an outcome of interest.
17 / Shore et al, 201517 / Single-arm study.
18 / Smythe et al, 201518 / Index event for patient identification = outcome of interest (bleeding).
19 / Tamayo et al, 201519 / Single-arm study.
20 / Tamayo et al, 201520 / Single-arm study.
21 / Tung et al, 201521 / Single-arm study.
22 / Xian et al, 201522 / Index event for patient identification = outcome of interest (stroke).
23 / Alonso et al, 201423 / Index event for patient identification = outcome of interest (bleeding).
24 / Armbruster et al, 201424 / Utilisation study.
25 / Aslan et al, 201425 / Case-control study.
26 / Azoulay et al, 201426 / Case-control study.
27 / Beyer-Westendorf et al, 201427 / Duplicate (conference abstract); single-arm study.
28 / Beyer-Westendorf et al, 201428 / Single-arm study.
29 / Boettger et al, 201429 / Patient adherence is not an outcome of interest.
30 / Labaf et al, 201430 / Case-control study.
31 / Mao et al, 201431 / Unable to retrieve article.
32 / Sokolova et al, 201432 / Unable to retrieve conference abstract.
33 / Halvorsen et al, 201333 / Duplicate in another journal.
34 / Ng et al, 201334 / Contained details of patient characteristics only; AVERROES study.
35 / Azoulay et al, 201235 / Case-control study.
36 / Hart et al, 200836 / Original study not target population of interest.
37 / Currie et al, 200637 / Case-control study.
38 / Aronow et al, 199938 / Unable to retrieve article.

eTable 4a. Risk of Bias Assessment of Randomised, Controlled Trials

Study (Author, year) / Random sequence generation / Allocation concealment / Blinding of participants and personnel / Blinding of outcome assessors / Incomplete data / Selective reporting / Overall
RE-LY, 2009 / Low (computer generated sequence) / Low (interactive voice response system) / High (open-label) / Low (blinded evaluation of outcomes) / High (1.5 times drop-out rate in dabigatran 150 group) / Low (no missing outcomes) / High
PETRO, 2007 / Unclear (inadequate description of random sequence generation) / High (open-label for randomization) / High (open to identity of dabigatran or warfarin) / Unclear (no details on blinding) / High (2 times drop-out rate in dabigatran 150 group) / Low (no missing outcomes) / High
NCT 01136408, 2007 / Unclear (inadequate description of random sequence generation) / Unclear (unclear concealment) / High (open-label) / Unclear (no details on blinding) / High (1.5 times drop-out rate in dabigatran 150 group) / Unclear / High
ROCKET AF, 2011 / Low (computer generated sequence) / Low (interactive voice response system) / Low (double-blind) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / Low
J-ROCKET AF, 2012 / Unclear (inadequate description of random sequence generation) / Unclear (unclear concealment) / Low (double-blind) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / Unclear
ARISTOTLE, 2011 / Low (computer generated sequence) / Low (interactive voice response system) / Low (double-blind) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / Low
ARISTOTLE J, 2011 / Low (sequential treatment assignment with balancing for prognostic factors) / Unclear (unclear concealment) / High (open-label warfarin) / Low (blinded evaluation of outcomes) / High (1.7 times drop-out rate in warfarin group) / Low (no missing outcomes) / High
ENGAGE AF-TIMI 48, 2013 / Low (computer generated sequence) / Low (interactive voice response system) / Low (double-blind) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / Low
Yamashita et al, 2012 / Low (biased coin method) / Unclear (unclear concealment) / High (open-label warfarin) / Low (blinded evaluation of outcomes) / High (2 times drop-out rate in edoxaban group) / Low (no missing outcomes) / High
Weitz et al, 2012 / Low (generated by independent biostatistician) / Low (interactive automated telephone system) / High (open to identity of edoxaban or warfarin) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / High
Chung et al, 2012 / Low (randomisation by independent group) / Low (allocation by independent group) / High (open to identity of edoxaban or warfarin) / Low (blinded evaluation of outcomes) / High (2 times drop-out rate in edoxaban group) / Low (no missing outcomes) / High
Liu et al, 2013 / Unclear (inadequate description of random sequence generation) / Unclear (unclear concealment) / Unclear (no details on blinding) / Unclear (no details on blinding) / Unclear (no details on drop-outs) / Low (no missing outcomes) / Unclear
BAFTA, 2007 / Low (randomisation by independent group) / Low (allocation by GPs telephoning randomisation service) / High (open-label) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / High
WASPO, 2007 / Low (computer generated sequence) / Low (sealed enveloped in numbered sequences) / High (open-label) / Unclear (no details on blinding) / High (2 times drop-out rate in warfarin group) / Low (no missing outcomes) / High
ACTIVE W, 2006 / Low (computer generated sequence) / Low (interactive voice response system) / High (open-label) / Low (blinded evaluation of outcomes) / High (premature termination) / High (missing outcomes) / High
AFASAK2, 1998 / Low (computer generated sequence) / Low (adequate concealment) / High (open-label) / Low (blinded evaluation of outcomes) / High (premature termination) / High (missing outcomes) / High
AFASAK, 1989 / Low (computer generated sequence) / Low (sealed packages) / High (open-label for warfarin) / High (no blinding) / High (2 times drop-out rate in warfarin group) / Low (no missing outcomes) / High
SPAF2, 1994 / Low (computer generated sequence) / Low (adequate concealment) / High (open-label) / High (no blinding) / Low (low drop-out) / Low (no missing outcomes) / High
CAFA, 1991 / Low (predetermined random order) / Low (sequential use of medication packages) / Low (double-blind) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / Low
JAST, 2006 / Low (predetermined random sequence) / Unclear (unclear concealment) / High (open-label) / Low (blinded evaluation of outcomes) / High (premature termination) / High (missing outcomes) / High
LASAF, 1998 / Unclear (inadequate description of random sequence generation) / Unclear (unclear concealment) / High (open-label) / Unclear (no details on blinding) / Unclear
(insufficient details) / High (missing outcomes) / High
AVERROES, 2011 / Low (computer generated sequence) / Low (interactive voice response system) / Low (double-blind) / Low (blinded evaluation of outcomes) / High (premature termination) / Low (no missing outcomes) / High
ACTIVE, 2009 / Low (computer generated sequence) / Low (interactive voice response system) / Low (double-blind) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / Low (no missing outcomes) / Low
PATAF, 1999 / Low (central randomisation) / Low (telephone) / High (single-blinded) / Low (blinded evaluation of outcomes) / Low (balanced drop-out) / High (bleeding outcomes not reported for stratum 1 vs 2) / High
SPAF, 1991 / Low (computer generated sequence) / Low (adequate concealment) / High (open-label for warfarin) / High (no blinding) / Low (balanced drop-out) / Low (no missing outcomes) / High

eTable 4b. Risk of Bias Assessment of Non-Randomised Studies

Study (Author, year) / Bias due to confounding / Bias in selection of participants into the study / Bias in measurement of interventions / Bias due to departures from intended interventions / Bias due to missing data / Bias in measurement of outcomes / Bias in selection of reported results / Overall
Hsien-Yen Chang et al, 2015 / Moderate
(propensity score adjustment) / Low (new user design) / Low (claims database) / Moderate (control for switching to other interventions not reported) / Moderate (outcome data reasonably complete) / Moderate (outcomes objectively measured) / Moderate (all outcomes reported) / Moderate
Abraham et al, 2015 / Moderate
(propensity score adjustment) / Low (new user design) / Low (claims database) / Moderate (control for switching to other interventions not reported) / Moderate (outcome data reasonably complete) / Moderate (outcomes objectively measured) / Moderate (all outcomes reported) / Moderate
Fontaine et al, 2014 / Serious (no adjustment) / Serious (no new user design; possible that selection based on intervention) / Moderate (manual chart review) / Moderate (control for switching to other interventions not reported) / Serious (missing information on confounders) / Moderate (outcomes objectively measured) / Moderate (all outcomes reported) / High
Ellis et al, 2014 / NA / NA / NA / NA / NA / NA / NA / Unpublished study