pharmacoepidemiology and drug safety 2017; 26: 97–107

Published online 20 November 2016 in Wiley Online Library (wileyonlinelibrary.com) DOI: 10.1002/pds.4137

ORIGINAL REPORT

Effectiveness and safety of drugs used for stroke prevention in a

cohort of non-valvular atrial fibrillation patients from a primary care
electronic database
Maria Giner-Soriano1,2*, Albert Roso-Llorach1,2, Cristina Vedia Urgell2,3, Xavier Castells4, Dolors Capellà4,
Ignacio Ferreira-González5, Josep Maria Elorza-Ricart6, Marc Casajuana1,2, Amelia Troncoso Mariño7,
Eduard Diògene8, Bonaventura Bolíbar1,2, Concepció Violan1,2 and Rosa Morros1,2
1
Institut Universitari d’Investigació en Atenció Primària Jordi Gol (IDIAP Jordi Gol), Barcelona, Spain
2
Universitat Autònoma de Barcelona, Bellaterra (Cerdanyola del Vallès), Spain
3
Unitat de farmàcia, Servei d’Atenció Primària Barcelonès Nord i Maresme, Institut Català de la Salut, Badalona, Spain
4
Translab Research Group, Unitat de Farmacologia Clínica, Departament de Ciències Mèdiques, Facultat de Medicina, UdG, Girona,
Spain
5
Unidad de Epidemiología del Servicio de Cardiología, Hospital Vall d’Hebron y CIBER de Epidemiología y Salud Pública (CIBERESP),
Barcelona, Spain
6
System for the Improvement of Research in Primary Care (SIDIAP), Barcelona, Spain
7
Unitat de Coordinació i Estratègia del Medicament, Institut Català de la Salut, Barcelona, Spain
8
Servei de Farmacologia, Hospital Vall d’Hebron, Barcelona, Spain

ABSTRACT
Purpose The aim of this study was to assess effectiveness and safety of antithrombotics for stroke prevention in non-valvular atrial fibril-
lation in real-use conditions.
Methods We used a population-based retrospective cohort study. Information emerges from SIDIAP, a database containing anonymized
information from electronic health records from 274 primary healthcare centres of the Catalan Health Institute, Catalonia (Spain), with a ref-
erence population of 5 835 000 people.
Population includes all adults with a new diagnosis of non-valvular atrial fibrillation registered in SIDIAP from 2007 to 2012.
The main outcome of antithrombotics’ effectiveness was stroke. The main outcomes of safety were cerebral and gastrointestinal
haemorrhages. We also estimated all-cause mortality. We used multivariable Cox proportional hazard models to examine association
between antithrombotic treatment and main outcomes.
Results We included 22 205 subjects with non-valvular atrial fibrillation; 40.8% initiated on vitamin K antagonists (VKA), 33.4% on
antiplatelets and 25.8% untreated. We found stroke-risk reduction with VKA, hazard ratio (HR) 0.72 (95% confidence interval (CI),
0.58–0.91), also seen in patients with CHADS2 ≥ 2, HR 0.65 (95%CI, 0.49–0.86), and CHA2DS2-VASc ≥ 2, HR 0.66 (95%CI,
0.52–0.84). We observed a higher risk of digestive bleeding with antiplatelets, HR 1.32 (95%CI, 1.01–1.73). Both VKA and antiplatelets
were associated with reduction of all-cause mortality risk; HR 0.55 (95%CI, 0.49–0.62) and HR 0.89 (95%CI, 0.80–0.97), respectively.
Conclusions This study found a stroke-risk reduction associated with VKA and an increased risk of gastrointestinal bleeding associated
with platelet-aggregation inhibitors in comparison with untreated patients. Both antithrombotic groups showed a reduction in all-cause
mortality. Copyright © 2016 John Wiley & Sons, Ltd.

*Correspondence to: M. Giner-Soriano, Institut Universitari d’Investigació en

Atenció Primària Jordi Gol (IDIAP Jordi Gol), Gran Via de les Corts Catalanes • Conference papers:
587, àtic, 08007 Barcelona, Spain. Email: mginer@idiapjgol.info • Giner-Soriano M, Morros Pedrós R, Vedia Urgell C, Roso-Llorach A,
Castells X, Capellà D. Effectiveness and safety of antithrombotic drugs used
Prior postings and presentations in non-valvular atrial fibrillation (ESC-FA study). 12th Congress of the
• Giner-Soriano M, Vedia Urgell C, Roso-Llorach A, et al. Effectiveness, safety European Association for Clinical Pharmacology and Therapeutics.
and costs of thromboembolic prevention in patients with non-valvular atrial Madrid, 27–30 junio 2015. Best Communications Awards. Monday June
fibrillation: phase I ESC-FA protocol study and baseline characteristics of a 29th. Clinical Therapeutics, Vol. 37, Issue 8, e 12.
cohort from a primary care electronic database. BMJ Open. 2016;6(1): • Giner-Soriano M, Morros Pedrós R, Vedia Urgell C, Roso-Llorach A,
e010144. doi: 10.1136/bmjopen-2015-010144 Casajuana M, Castells X, Capellà D. Effectiveness, safety and costs of stroke
• Giner-Soriano M, Roso-Llorach A, Vedia Urgell C, et al. Drug therapy for prevention in non-valvular atrial fibrillation patients (ESC-FA study). De-
rate and rhythm control in non-valvular atrial fibrillation: a cross-sectional scription of cohorts and baseline characteristics of patients. XXVII Congreso
study with electronic health records in a primary care cohort. Clinical de la Sociedad Española de Farmacología Clínica. Sevilla, 2–4 octubre
Therapeutics 2016;1-11. doi: 10.1016/j.clinthera.2016.02.002 2014. Basic & Clinical Pharmacology & Toxicology, 115 (Suppl 3), 12–19.

Copyright © 2016 John Wiley & Sons, Ltd.

98 m. giner-soriano et al.

key words—atrial fibrillation; electronic health records; stroke; cerebral haemorrhage; gastrointestinal haemorrhage; vitamin K antagonists;
platelet-aggregation inhibitors; all-cause mortality; antithrombotic; primary health care; pharmacoepidemiology

Received 22 March 2016; Revised 4 October 2016; Accepted 25 October 2016

INTRODUCTION aspirin or other antiplatelets, or no antithrombotic

treatment.13–16
Atrial fibrillation (AF) is the most common chronic There are some studies on OAC utilization in AF
cardiac arrhythmia. Its prevalence in the general popu- patients from Europe,17–19 most of them indicating
lation in North America and Europe is 1.5–2%, and it VKA under-use, and some published in PHC in
increases with age.1–3 It represents an increasing Spain, although conducted within small sample
healthcare burden due to an ageing population.2 In sizes.5,20–24
Spain, Gómez-Doblas et al. reported a prevalence of To our knowledge, utilization of antithrombotic
4.4% in population older than 40 years attended in pri- drugs for stroke prevention in AF and their effective-
mary health care (PHC),4 and Barrios et al. described a ness and safety in real-use conditions have not been
prevalence of 6.1% in general PHC population.5 Both assessed in our setting through population studies con-
authors described a prevalence higher than 17% in ducted with electronic health records.
older than 80 years. Clua-Espuny et al. described A progressively ageing population, which in addi-
similar prevalence and an approximately 20% of tion increases incidences of AF and thromboembolic
undiagnosed cases.6 events and a necessity of assessing effectiveness and
Atrial fibrillation is associated with a variety of safety of antithrombotic agents in real-use conditions
cardiovascular conditions such as hypertension, symp- before the introduction of direct OAC in AF manage-
tomatic heart failure or stroke. In fact, it increases ment, justifies the need of conducting this study.
stroke risk by fivefold, and one in five strokes may Moreover, the most used VKA in our setting is
be attributed to AF. Mortality rate is also increased acenocoumarol, which has not been studied as much
in these patients.7 as warfarin for stroke prevention in AF, although they
Management of AF patients increasingly takes place are considered equivalents.
in PHC settings. It aims reduction of symptoms and The main objective of this study is to assess the ef-
prevention of associated complications by heart rate fectiveness of antithrombotics in real-use conditions
and rhythm control, handling of concomitant cardio- according to stroke rates and to assess the safety of
vascular disorders and stroke prevention.2,7 The drugs antithrombotics according to bleeding events rates be-
aimed at decreasing the risk of stroke and thromboem- fore the introduction of the direct OAC. All-cause
bolic events in AF are antithrombotics: oral anticoagu- mortality will be also estimated.
lant (OAC) drugs and platelet-aggregation inhibitors,
also known as antiplatelets. Current available data have
shown superiority of OAC over antiplatelets, so they METHODS
are the recommended treatment in the guidelines for
patients requiring antithrombotic therapy, even though The ESC-FA study (Effectiveness, Safety and Costs in
they have been frequently prescribed to patients Atrial Fibrillation) is a population-based retrospective
considered as non-candidate for anticoagulation.8–12 observational cohort study.
Traditionally, OAC drugs used were vitamin K The study population was all individuals older than
antagonists (VKA); acenocoumarol in Spain and 18 years with a new diagnosis of non-valvular AF reg-
warfarin in the USA and some European countries. istered in SIDIAP database (Information System for
Recently, direct OAC received marketing authoriza- the Improvement of Research in Primary Care)25–31
tion in the European Union for stroke prevention in from 2007 to 2012, who initiated antithrombotic treat-
non-valvular AF (dabigatran in 2011, rivaroxaban in ment at the cohort entry with antiplatelets or VKA or
2012 and apixaban in 2013 in Spain). remained untreated.
Apart from the clinical trials that demonstrated effi- As diagnosis and treatment start may be registered
cacy and safety of OAC in stroke prevention, their ef- in different days in ECAP™ (electronic health records
fectiveness has been assessed in some observational in PHC), we allowed a time interval of 3 months
studies, where warfarin showed lower stroke risks than from diagnosis date and start of treatment. The time

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
 real-use data on stroke prevention in atrial fibrillation 99
of cohort entry (or index date) was defined as the latest The main outcome to assess safety was major
date between diagnosis date and treatment initiation bleeding, specifically cerebral and gastrointestinal
date: In those people with a diagnosis date registered haemorrhages. The rest of the overall haemorrhages
after the start of the treatment, the index date was the included eye bleeding, genitourinary bleeding and
diagnosis date. In those who had the diagnosis regis- other bleeding. All events of interest were concur-
tered before the start of the treatment, the index date rently identified through SIDIAP (ICD10) and
was the date of treatment initiation. In the group of CMBD-AH (ICD9) databases.
non-treated, the index date was the date of diagnosis. All-cause mortality during follow-up was assessed
All patients were followed up from date of cohort through SIDIAP database. We had access to date of
entry until the first of these events occurred: end of death, but we were not able to know cause of death.
the study period, lost to follow-up or death. We obtained the information on drug exposures
from the pharmacy invoice registry. When patients
Data source stopped having antithrombotic treatment recorded in
the pharmacy invoice registry, we considered they
SIDIAP31 contains anonymized clinical information of
discontinued treatment.
all PHC Centres of the Catalan Health Institute (ICS),
All variables and outcomes of interest were analysed
the main provider of health services in Catalonia,
in the patients initiated on VKA or on antiplatelets or
Spain, managing 274 PHC practices with a reference
untreated, excluding patients receiving other anti-
population of 5 835 000 patients (80% of the Catalan
thrombotic therapies owing to small sample size (as
population). This information emerges from ECAP™,
dual therapy with VKA and antiplatelets, n = 227, or
and it includes socio-demographic characteristics,
dabigatran, n = 153).
health conditions registered as ICD10 codes, clinical
We have recently published a detailed description of
parameters, toxic habits, laboratory data and General
the ESC-FA study methodology and all baseline char-
Practitioners’ prescriptions and their corresponding
acteristics of patients.33
pharmacy invoice data.
SIDIAP may be linked with CMBD-AH (‘minimum
set of data at hospital discharge’),32 which contains di- Statistical analysis
agnoses coded with ICD9 at hospital discharge from
Descriptive statistics were used to summarize overall
all hospitals in Catalonia, to obtain the data for the
information. Categorical variables were expressed as
endpoints of the study (stroke and bleedings).
frequencies (percentage) and quantitative variables as
mean (standard deviation) or median (interquartile
Variables range).
The variables assessed at baseline were as follows: For each event, we defined time to follow-up as the
socio-demographic characteristics; stroke and bleed- time between cohort entry and the event. Patients were
ing risks by CHADS2, CHA2DS2-VASc and HAS- followed up until censored (died only for stroke and
BLED scores (HAS-BLED was calculated without bleeding events, lost to follow-up or end of observa-
‘L: labile International Normalized Ratio (INR)’ item, tion). Patients were also censored when they initiated
because INR values were missing in most patients a new antithrombotic treatment.
treated with VKA as they were new treatments, and Incidence rates of stroke and bleeding events during
in all patients from the other two groups; antiplatelets follow-up were estimated for each cohort. Incident
and untreated); comorbidities of interest before AF di- rates are presented per 1000 patient-years and their
agnosis (ICD9 and ICD10 codes, see Supporting In- corresponding 95% confidence intervals (CIs).
formation file of codes); stroke and bleeding episodes Time-to-event analysis was performed using
registered before AF diagnosis (ICD9 and ICD10 non-parametric methods like Kaplan–Meier and
codes, see Supporting Information file of codes); labo- log-rank test.
ratory data; and exposures to all antithrombotics and to Multivariate Cox proportional-hazards regression
concomitant drugs prescribed at the time of diagnosis models were fitted, adjusting for the following base-
(Anatomical Therapeutic Chemical codes). line socio-demographic characteristics and confound-
The variables assessed during follow-up were as fol- ing and predictive factors of each event: age, sex,
lows: any stroke, haemorrhages, all-cause mortality Charlson index, creatinine clearance (<30, 30–60 or
rates and exposures to all antithrombotic drugs. >60 mL/minute/1.73 m2), previous stroke, myocardial
The main outcome to assess effectiveness of infarction, peripheral artery disease, hypertension,
antithrombotic drugs during follow-up was stroke. heart failure, diabetes mellitus, hepatic impairment,

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
100 m. giner-soriano et al.
dyslipidemia or statin treatment, previous bleeding, 5724 (25.8%) untreated. These 22 205 patients were
peptic ulcer disease and co-treatments with digoxin, 72.8 (13.1) years old, and 51.4% of them were men.
β-blockers, verapamil, diltiazem, other antiarrhythmic The stroke risk measured by CHADS2 was
drugs, angiotensin converting enzyme inhibitors, an- low–moderate (0–1) in 52.6% of them. We were able
giotensin II receptor blockers, proton pump inhibitors to calculate HAS-BLED in 14 161 patients, so there
and non-steroidal anti-inflammatory drugs. were 36.2% of patients with missing values. Main
All-cause mortality was also adjusted for MEDEA baseline characteristics are shown in Table 1. During
index (deprivation index that shows the social or mate- the study period, 2207 (38.6%) patients in the
rial disadvantage accruing to a person or group in non-treated cohort initiated an antithrombotic
accordance to their city/region/country, as given in treatment, 1015 (13.7%) changed from antiplatelets
the census data in Catalonia; the higher it is, the worse to other group and 1465 (16.2%) did so from VKA
the deprivation)34 in the Cox model. to another group.
The no-treatment group was considered as the refer- All baseline characteristics for the 22 585 patients
ence in all analysis. Extended Cox models were used are shown in Tables S1 to S4 and Figure S1, and we
when the model’s proportional-hazards assumption have previously described them in a recently published
did not hold. paper.33
We conducted, as a sensitivity analysis, an
intention-to-treat analysis, in which patients from the
three groups were followed up from cohort entry up
to the first of an event occurrence, end of study period
or lost to follow-up, with no censoring when they Table 1. Main baseline socio-demographic characteristics and scores for
switched antithrombotic groups. stroke and bleeding risks of patients included, n = 22 205
We also calculated the ‘net clinical’ benefit obtained All patients included, n = 22 205
by the individuals at high risk of stroke treated with
No
VKA in comparison with the untreated individuals antithrombotic Antiplatelets VKA
and those treated with antiplatelets as follows:
Net clinical benefit = (Stroke rate off-VKA Stroke n (%)
rate on-VKA) 1.5 × (Cerebral bleeding rate on- 9057
VKA Cerebral bleeding rate off-VKA).16 5724 (25.8) 7424 (33.4) (40.8)
Stroke and bleeding rates used to this calculation
Sex (%)
were unadjusted rates. Women 46.9 50.3 48.3
Multiple imputations with 10 imputed datasets were Men 53.1 49.7 51.7
used to handle missing data. We included all the Age, years 69.6 (16.4) 74.6 (12.9) 73.4 (10.3)
(mean, SD)
potential predictive variables and survival outcome CHADS238 (%)
terms in the imputation model. Our final Cox models 0–1 61.2 52.5 47.3
were fitted using the multiply imputed datasets by ≥2 38.8 47.5 52.7
CHA2DS2-VASc39
using Rubin’s rules to combine effect estimates and (%)
standard errors to allow for the uncertainty related to 0–1 34.5 22.3 17.8
missing data. 2 34.5 19.2 19.5
3 21.4 23.8 19.5
All statistical tests were two sided at the 5% signif- ≥4 27.6 34.7 19.5
icance level. The analyses were performed using Stata HAS-BLED40,41 2921, 51.0 4813, 64.8 6427, 71.0
ver. 13 (Stata Corp., Collage Station, TX) and R ver- (n, %)
0 475, 16.2 0, 0.0 426, 6.6
sion 3.2.1 (R Foundation for Statistical Computing, 1–2 1856, 63.6 2048, 42.5 4432, 68.9
Vienna, Austria). ≥3 590, 20.2 2765, 57.4 1569, 24.4
Missing values 2803, 49.0 2611, 35.2 2630, 29.0
RESULTS CHADS2, stroke-risk score, which includes congestive heart failure, hyper-
tension, age ≥ 75 years, diabetes mellitus and stroke/TIA/TE (2 points).
From 2007 to 2012, 22 585 non-valvular AF patients CHA2DS2-VASc, stroke risk, which includes congestive heart failure,
hypertension, age ≥ 75 years (2 points), diabetes mellitus, stroke/TIA/TE
were included in ESC-FA study. We analysed out- (2 points), vascular disease, age 65–74 years and female sex. HAS-BLED,
comes of interest in 22 205 patients who were pre- hypertension (systolic blood pressure ≥ 160 mm Hg), abnormal kidney and/
scribed the most frequent antithrombotic options or or liver function, stroke, bleeding, labile INR, elderly, drugs (antiplatelets)
and/or alcohol.
remained untreated at baseline: 9057 (40.8%) patients VKA, vitamin K antagonists; SD, standard deviation; TIA, transient
initiated on VKA, 7424 (33.4%) with antiplatelets and ischemic attack; TE, thromboembolism.

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
 real-use data on stroke prevention in atrial fibrillation 101
Effectiveness assessment person-years (95%CI, 9.7–11.6). We observed a
Among 22 205 patients, we identified 648 (2.9%) higher risk of suffering digestive bleeding with
strokes. The total person-time during the follow-up antiplatelets, HR 1.32 (95%CI, 1.011.73, p = 0.041)
was 44 370.2 person-years, and stroke rate was 14.6 (Figure 2).
per 1000 patient-years (95%CI 13.5–15.8). Stroke
rates and per stroke risk in each cohort are presented in All-cause mortality
Table 2.
The Kaplan–Meier estimates of time to the event The cumulative time at risk for death was 45 112.8
stroke during follow-up are presented in Figure 1. person-years. A total of 2518 patients died (11.3%):
Adjusted Cox models for stroke and the rest of the 12.9% of the untreated cohort, 15.4% of the
outcomes are shown in Figure 2. We found a reduction antiplatelets group and 7.0% of the VKA group. Total
of stroke risk with VKA therapy, hazard ratio (HR) all-cause mortality rates and per age groups for the
0.72 (95%CI, 0.58–0.91, p = 0.005), also in patients three cohorts are shown in Table 4. We observed
with high risk of stroke; CHADS2 score ≥ 2, HR 0.65 higher rates progressively increasing with age and dif-
(95%CI, 0.49–0.86, p = 0.002) and CHA2DS2-VASc ferent between groups.
score ≥ 2, HR 0.66 (95%CI, 0.52–0.84, p = 0.001). Kaplan–Meier estimates of survival until the time of
We found a trend in increasing stroke risk with censoring are shown in Figure 3.
antiplatelets, although it was not statistically signifi- Both antiplatelets and VKA showed a significant re-
cant (adjusted HR 1.18, 95%CI, 0.96–1.46, p = 0.120). duction in the risk of all-cause mortality in the adjusted
analysis: 11% of reduction with antiplatelets, HR 0.89
(95%CI, 0.80–0.97, p = 0.013), and 45% with VKA,
Safety assessment HR 0.55 (95%CI, 0.49–0.62, p < 0.001) (Figure 2).
Incidence rates of all haemorrhages for the three co-
horts are shown in Table 3. The rate of all
haemorrhages in the whole population was 22.8 per Intention-to-treat analysis
1000 person-years (95%CI, 21.4–24.3). In the intention-to-treat analysis, we identified 892
Cerebral haemorrhage rate in overall population was (4.0%) strokes among 22 205 patients included. We
2.8 per 1000 person-years (95%CI, 2.4–3.3). No statis- found a reduction of stroke risk with VKA treatment,
tically significant differences were found in the ad- HR 0.72 (95%CI, 0.60–0.86, p < 0.001), which was
justed HR for cerebral bleeding (Figure 2), although also significant in the subgroups with CHADS2 and
we observed a positive association for VKA and cere- CHA2DS2-VASc ≥ 2.
bral bleeding, HR 1.49 (95%CI 0.87–2.57, p = 0.150). No statistically significant differences were found in
Regarding gastrointestinal haemorrhage, the event the adjusted HR for cerebral and gastrointestinal
rate in overall population was 10.6 per 1000 haemorrhages.

Table 2. Number and rates of strokes per 1000 person-years, overall and stratified by stroke-risk index, n = 22 205
Outcome No antithrombotic Antiplatelets Vitamin K antagonists

Stroke
Events 131 297 220
Event rate (95%CI) 13.3 (11.2–15.8) 20.1 (18.0–22.6) 11.1 (9.8–12.7)
Stroke in CHADS2 < 2
Events 47 86 63
Event rate (95%CI) 6.8 (5.1–9.1) 10.4 (8.4–12.8) 6.6 (5.1–8.4)
Stroke in CHADS2 ≥ 2
Events 84 211 157
Event rate (95%CI) 28.5 (23.0–35.2) 32.8 (28.6–37.5) 15.4 (13.2–18.0)
Stroke in CHA2DS2-VASc < 2
Events 13 14 18
Event rate (95%CI) 2.9 (1.7–5.1) 3.7 (2.2–6.2) 5.0 (3.2–8.0)
Stroke in CHA2DS2-VASc ≥ 2
Events 118 283 202
Event rate (95%CI) 21.7 (18.1–26.0) 25.9 (23.1–29.1) 12.5 (10.9–14.3)

CHADS2, stroke-risk score, which includes congestive heart failure, hypertension, age ≥ 75 years, diabetes mellitus and stroke/TIA/TE (2 points). CHA2DS2-
VASc, stroke risk, which includes congestive heart failure, hypertension, age ≥ 75 years (2 points), diabetes mellitus, stroke/TIA/TE (2 points), vascular dis-
ease, age 65–74 years and female sex.
CI, confidence interval; TIA, transient ischemic attack; TE, thromboembolism.

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
102 m. giner-soriano et al.

Figure 1. Kaplan–Meier estimates of time to the event ‘stroke’ during follow-up, according to the antithrombotic medication status at baseline [Color figure
can be viewed at wileyonlinelibrary.com]

Figure 2. Forest plot of adjusted hazard ratios on treatment regimen of various outcomes during follow-up, n = 22 205 [Color figure can be viewed at
wileyonlinelibrary.com]

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
 real-use data on stroke prevention in atrial fibrillation 103
Table 3. Number and rates of cerebral and gastrointestinal bleedings per 1000 person-years, stratified by high risk of stroke, n = 22 205
Outcome No antithrombotic Antiplatelets Vitamin K antagonists

All bleeding events

Events 138 321 541
Event rate (95%CI) 14.1 (11.9–16.7) 22.0 (19.7–24.5) 27.8 (25.5–30.2)
Cerebral bleeding
Events 18 40 68
Event rate (95%CI) 1.8 (1.1–2.9) 2.7 (2.0–3.6) 3.4 (2.7–4.3)
Cerebral bleeding in CHADS2 ≥ 2
Events 13 25 41
Event rate (95%CI) 4.3 (2.5–7.5) 3.8 (2.6–5.6) 3.9 (2.9–5.3)
Cerebral bleeding in CHA2DS2-VASc ≥ 2
Events 17 37 60
Event rate (95%CI) 3.1 (1.9–5.0) 3.3 (2.4–4.6) 3.6 (2.8–4.7)
Gastrointestinal bleeding
Events 85 180 206
Event rate (95%CI) 8.3 (7.0–10.7) 12.2 (10.5–14.1) 10.4 (9.0–11.9)
Gastrointestinal bleeding in CHADS2 ≥ 2
Events 37 98 115
Event rate (95%CI) 12.5 (9.0–17.2) 15.0 (12.3–18.3) 11.2 (9.3–13.4)
Gastrointestinal bleeding in CHA2DS2-VASc ≥ 2
Events 62 152 171
Event rate (95%CI) 11.4 (8.9–14.6) 13.8 (11.8–16.2) 10.5 (9.0–12.2)

CI, confidence interval.

Table 4. Number and rates of all-cause mortality per 1000 person-years per age groups, n = 22 205

Outcome No antithrombotic Antiplatelets Vitamin K antagonists

Overall all-cause mortality

Events 739 1146 633
Event rate (95%CI) 74.4 (69.2–79.9) 76.2 (72.0–80.8) 31.4 (29.1–34.0)
0–64 years old
Events 43 23 25
Event rate (95%CI) 10.5 (7.8–14.2) 7.3 (4.9–11.0) 8.7 (5.9–12.9)
65–74 years old
Events 74 77 67
Event rate (95%CI) 41.4 (32.9–52.0) 23.6 (18.9–29.5) 13.4 (10.6–17.1)
≥75 years old
Events 622 1046 541
Event rate (95%CI) 153.0 (141.4–165.5) 121.4 (114.2–128.9) 44.0 (40.5–47.9)

CI, confidence interval.

Regarding all-cause mortality, a total of 3245 antiplatelets obtained a net clinical benefit of 17.3
(14.6%) patients died during the study period; 18.3% per 1000 person-years (stroke and cerebral bleeding
of the untreated, 17.8% in the antiplatelets group and rates are shown in Tables 2 and 3).
9.6% in the VKA group. Both antiplatelets (HR 0.89, For patients with CHA2DS2-VASc ≥ 2, net clinical
95%CI 0.82–0.97, p = 0.007) and VKA (HR 0.67, benefit for VKA compared with untreated was 8.5
95%CI 0.61–0.74, p < 0.001) showed reduction in per 1000 person-years and 13.0 per 1000 person-years
the risk of all-cause mortality in the adjusted analysis. for VKA compared with antiplatelets.
Adjusted Cox models for all outcomes are shown in For patients at low or moderate risk of stroke per
Figure S2. Kaplan–Meier estimates of time to stroke CHADS2 < 2, in the VKA group compared with un-
and time to death are shown in Figures S3 and S4. treated, net clinical benefit was 2.9 per 1000
person-years, and for VKA compared with
antiplatelets, it was 2.3 per 1000 person-years.
Net clinical benefit For those patients at CHA2DS2-VASc < 2, net clin-
In the group of patients with CHADS2 ≥ 2, net clinical ical benefit for VKA compared with untreated was
benefit for VKA versus no treatment was 13.7 per 5.1 per 1000 person-years and 4.5 per 1000
1000 person-years. Patients with VKA versus person-years for antiplatelets.

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
104 m. giner-soriano et al.

Figure 3. Kaplan–Meier estimates of survival during follow-up, according to the antithrombotic medication status at baseline [Color figure can be viewed at
wileyonlinelibrary.com]

DISCUSSION The intention-to-treat analysis provided similar re-

sults to the main analysis, probably because approxi-
Our study assessed effectiveness and safety of tradi- mately 80% of patients remained on the same cohort
tional antithrombotics used for stroke prevention in during all the study period. Some differences may only
non-valvular AF under real-use conditions. We be seen after more than 2 years of follow-up in the
assessed 22 205 patients, initiated on VKA (40.8%) Kaplan–Meier estimates of time to the event stroke
or antiplatelets (33.4%) or untreated (25.8%) after (Figure 1 in the paper and Figure S4), pointing out that
AF diagnosis. the risk of suffering a stroke seems to be higher during
Untreated patients were generally younger and with the first 2 years of initiating the treatment and the
lower stroke risk (61.2% had CHADS2 index = 0–1) follow-up, as in most cardiovascular conditions. Simi-
than those initiated on VKA or antiplatelets. Never- lar benefits of anticoagulant therapy for stroke preven-
theless, we found similar proportions of patients with tion were detected in other cohort studies. Forslund
low–moderate (CHADS2 = 0–1) and with high risk of et al.35 evaluated the benefits of warfarin, aspirin or
stroke (CHADS2 ≥ 2) in both antithrombotic groups, no treatment in 41 810 AF patients from a database co-
possibly indicating that antithrombotic prescription hort. Those treated with aspirin showed increased
after AF diagnosis was not generally linked to the eval- stroke and bleeding risks. Warfarin patients had lower
uation of potential stroke risk, although unknown fac- stroke rates, different according to CHA2DS2-VASc
tors might be also involved in the decision of the score, and lower mortality rates than aspirin-treated
treatment prescription (frail patients, history of bleed- or untreated patients. The higher stroke and mortality
ing, frequent falls, living alone, etc.). rates in the aspirin group might be pointing out a con-
We found reduction of stroke risk with VKA ther- founding by indication (not treating frail patients with
apy compared with untreated. No differences in the ad- OAC but with antiplatelets), rather than therapeutic
justed HR of cerebral bleeding were found, although effects of antiplatelets.
we observed a positive association for VKA and cere- In another database study, Friberg et al.13 assessed
bral haemorrhage. We found an increase in the risk of effectiveness and safety of warfarin versus no
gastrointestinal bleeding with antiplatelets. Regarding treatment in 182 678 AF subjects. Ischemic stroke
all-cause mortality, incidence rates increased with rates increased with increasing CHA2DS2-VASc
age and were different between groups. Both VKA scores from 0% to 12% annually in patients without
and antiplatelets evidenced lower rates than no treat- warfarin and to 7% in patients with warfarin at
ment (Figure 2). baseline. Intracranial haemorrhage (ICH) occurred at

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
 real-use data on stroke prevention in atrial fibrillation 105
an annual rate of 0.6% in warfarin-treated and un- <2), indicating that patients at high risk of stroke ben-
treated patients alike, whereas bleeding of any type oc- efit from anticoagulation therapy.
curred at an annual rate of 2.3%. So the risk of Analysing the untreated cohort, we hypothesize it
ischemic stroke without OAC is higher than the risk may include two different types of patients: (i) pa-
of ICH with OAC treatment. As in our study, tients with better prognosis, probably younger with
VKA-treated patients had better effectiveness low comorbidity and low risk of stroke, which might
outcomes than untreated patients. be in a status of sinus rhythm, so they do not need
Singer et al.16 assessed stroke and ICH rates in a to be anticoagulant therapy; and (ii) patients with
cohort of 13 559 AF patients, 46.9% untreated (they worse prognosis, probably elderly with many comor-
could be on antiplatelets therapy) and 53.1% receiving bidities and high risk of stroke, who are not started
warfarin. They described an overall unadjusted rate of on OAC because they are considered frail patients
stroke and thromboembolism of 1.27% (95%CI, or they present possible contraindications such as
1.19–1.44) and an overall unadjusted rate of ICH of dementia, bleeding history, frequent falls or parti-
0.58% (95%CI, 0.51–0.68). The adjusted net clinical cular social circumstances such as living alone.8,35,37
benefit of warfarin over no-warfarin therapy was In fact, all-cause mortality rates increased with
0.68% per year, so despite the risk of ICH, age in all groups, and they were higher in the
anticoagulation reduced stroke risk. antiplatelets group than in the VKA group, and even
Similar results to ours were found by Go et al.14 higher in the untreated group than in the antithrom-
They reported a 51% lower risk of thromboembolism botic groups.
with warfarin versus no warfarin (either no antithrom- This could be pointing out a selection bias produced
botic or no aspirin). Warfarin was associated with a by confounding by indication, meaning that patients
nearly twofold increased risk of ICH, so it was effec- with worse prognosis and high stroke risk are some-
tive for preventing ischemic stroke while the absolute times treated with antiplatelets or even untreated,
increase in ICH risk was small. It also reduced the risk although OAC are indicated but they are not
of all-cause death in 31%. prescribed as patients are considered frail. These
Hylek et al.15 assessed stroke rates in a cohort of patients would have high mortality rates, which could
13 559 AF patients; 596 (4.4%) suffered a stroke explain the high all-cause mortality associated with the
during 20 months of follow-up, 32% were treated whole group of untreated. A more detailed assessment
with warfarin, 27% were treated with aspirin and of this cohort would help us to prove this hypothesis in
42% were untreated. Independent factors associated the future.
with stroke severity were being untreated, receiving
warfarin and having an INR < 2.0, age and heart
failure. Strengths and limitations
Therefore, our results showed VKA superiority in The strengths of our study are the large number of pa-
stroke prevention and mortality in comparison with tients included, representativeness for the general pop-
no therapy, without significantly increasing bleeding ulation, complete socio-demographic and health
risk, as shown in the previous studies.13–16,35 There records, long follow-up and real clinical practice data.
were no significant increases in cerebral haemorrhage To our knowledge, this is the first population-based
risk, although there was a trend for VKA in increasing study in our setting that assesses the number of pa-
this risk, as previously seen in the Forslund et al. or tients treated with traditional antithrombotics and the
Friberg et al. studies.13,14 Treatment with antiplatelets clinical results of their use in terms of stroke,
increased the risk of gastrointestinal haemorrhage, haemorrhages and mortality rates, before assessing
which is a very well-known side effect of this pharma- these clinical results including direct OAC.
cological group.9,12,35,36 Antiplatelets also showed a Some weaknesses of observational studies conducted
trend in increasing risk of stroke, although it was with electronic health records are missing
non-significant. This fact has been previously reported or incomplete information, prescriptions not linked
in similar studies.35 with diagnoses coded and possible confounders. To
We found a net clinical benefit of VKA over therapy overcome the limitation of the lack of correlation be-
with antiplatelets or no treatment in patients at high tween diagnoses and treatments, we took into account
risk of stroke (CHADS2 and CHA2DS2-VASc ≥ 2), as the antithrombotic agents initiated during the time in-
it was also found in the studies of Singer et al.16 and terval of 3 months from the date of diagnosis. To min-
Friberg et al.13 The net clinical benefit was negative imize missing information and confounders’ effects,
for patients at low–moderate risk of stroke (scores we used missing imputation techniques and fitting

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
106 m. giner-soriano et al.
Cox regression models adjusted for socio-demographic For the linkage with CMBD-AH database (or other
characteristics and for possible confounders and predic- databases), SIDIAP uses a ‘trusted third party’ in order
tive factors, respectively, as described in the Statistical to ensure confidentiality when linking both data
Analysis. sources. This third party has no access to clinical infor-
mation, only to codes and IDs.
Conclusions
This study assesses the number of patients treated with FUNDING
the antithrombotics traditionally available for stroke
prevention in non-valvular AF and the clinical results ‘ESC-FA study, study in various phases on the effec-
of their use in terms of stroke, bleeding and mortality tiveness, safety and cost of thromboembolic preven-
rates through a population-based cohort study con- tion in patients with non-valvular atrial fibrillation’
ducted with electronic health records in a PHC setting. received funding from the Ministry of Health, Social
We found reduced stroke risk with VKA, also in Policy and Equality (Spanish Government) through
patients with higher risk of stroke (CHADS2 and the 2011 Grants for Independent Clinical Research
CHA2DS2-VASc ≥ 2), in comparison with no anti- (reference EC11-251).
thrombotic treatment, with no significant increases in
the risk of cerebral and digestive haemorrhages.
We observed an increased risk of gastrointestinal KEY POINTS
haemorrhage with antiplatelets when compared with • Clinical trials and observational studies have
untreated. assessed efficacy, effectiveness and safety of oral
Both VKA and antiplatelets showed lower rates of anticoagulants for stroke prevention in non-
all-cause mortality. valvular atrial fibrillation.
We have found that anticoagulation is the best op- • Many studies on anticoagulants use for stroke
tion for stroke prevention in non-valvular AF. prevention indicate under-use of vitamin K
Other next steps in our research are to assess how antagonists.
the introduction of direct OAC dabigatran, • Vitamin K antagonists, particularly warfarin,
rivaroxaban and apixaban in the management of non- have demonstrated a higher reduction in stroke
valvular AF affects effectiveness, safety and costs of and mortality risks in atrial fibrillation patients
stroke prevention with the antithrombotics already than platelet-aggregation inhibitors or no
ascertained in the present work. antithrombotic treatment.
• We assess effectiveness and safety of drugs
traditionally used for stroke prevention in non-
CONFLICT OF INTEREST valvular atrial fibrillation before the introduction
of direct anticoagulants through a population-
The authors declare no conflict of interest. based study conducted within an automated
healthcare database that originated from primary
ETHICS STATEMENT care electronic health records.
• We have seen a reduction of stroke and mortality
The present study follows national and international risks with vitamin K antagonists, without signif-
regulations: Declaration of Helsinki Ethical Principles icantly increasing bleeding risk. Despite reduc-
for Medical Research Involving Human Subjects and ing overall mortality risk, we have found an
Good Research Practice principles and guidelines. increase in gastrointestinal haemorrhage risk
The IDIAP Jordi Gol Clinical Research Ethics with platelet-aggregation inhibitors. Vitamin K
Committee, the reference institution for research in antagonists seem to be the best antithrombotic
PHC of the ICS, approved the study protocol. option for stroke prevention in non-valvular
Regarding the data contained in the databases and atrial fibrillation.
according to Spanish legislation about confidentiality
and data protection (Ley Orgánica 15/1999 de 13 de
diciembre de Protección de Datos de Carácter
Personal), data included in SIDIAP are always ACKNOWLEDGEMENTS
anonymized. Thus, it is not necessary to ask for The authors thank Jordi Cortés for his advice in the
informed consent from the participants. statistical analysis.

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
 real-use data on stroke prevention in atrial fibrillation 107
REFERENCES Anticoagulation prescription in atrial fibrillation. Expert Opin Pharmacother
2011; 12: 1473–1479.
23. Morillas P, Pallarés V, Llisterri JL, et al. Prevalencia de fibrilación auricular y uso de
1. Camm AJ, Lip GYH, Caterina RD, et al. 2012 focused update of the ESC
fármacos antitrombóticos en el paciente hipertenso ≥ 65 años. El registro FAPRES.
Guidelines for the management of atrial fibrillation: an update of the 2010 ESC
Rev Española Cardiol 2010; 63: 943–950. doi:10.1016/S0300-8932(10)70206-2.
Guidelines for the management of atrial fibrillation. Developed with the special
24. Bertomeu-Martínez V, Morillas-Blasco P, González-Juanatey J. Empleo de
contribution of the European Heart Rhythm Association. Eur Heart J 2012;
fármacos antitrombóticos en pacientes hipertensos con fibrilación auricular
33: 2719–2749. doi:10.1093/eurheartj/ehs253.
crónica. Estudio CARDIOTENS 99. Med Clin (Barc) 2002; 118: 327–331.
2. Lip GYH, Tse HF, Lane DA. Atrial fibrillation. Lancet 2012; 379: 648–661.
25. Violán C, Foguet-Boreu Q, Hermosilla-Pérez E, et al. Comparison of the infor-
doi:10.1016/S0140-6736(11)61514-6.
mation provided by electronic health records data and a population health survey
3. Lip GYH, Brechin CM, Lane DA. The global burden of atrial fibrillation and
to estimate prevalence of selected health conditions and multimorbidity. BMC
stroke: a systematic review of the epidemiology of atrial fibrillation in regions
Public Health 2013; 13: 251. doi:10.1186/1471-2458-13-251.
outside North America and Europe. Chest 2012; 142: 1489–1498. doi:10.1378/
26. Pagès-Castellà A, Carbonell-Abella C, Avilés FF, et al. Burden of osteoporotic
chest.11-2888.
fractures in primary health care in Catalonia (Spain): a population-based study.
4. Gómez-Doblas JJ, Muñiz J, Alonso Martin J, et al. Prevalence of atrial fibrilla-
BMC Musculoskelet Disord 2012; 13: 79. doi:10.1186/1471-2474-13-79.
tion in Spain. OFRECE study results. Rev Española Cardiol (English Ed 2014;
27. Vinagre I, Mata-Cases M, Hermosilla E, et al. Control of glycemia and cardio-
67: 259–269.
vascular risk factors in patients with type 2 diabetes in primary care in Catalonia
5. Barrios V, Calderón A, Escobar C, de la Figuera M. Patients with atrial fibrilla-
(Spain). Diabetes Care 2012; 35: 774–779. doi:10.2337/dc11-1679.
tion in a primary care setting: Val-FAAP study. Rev Esp Cardiol (Engl Ed) 2012;
28. Garcia-Gil MDM, Hermosilla E, Prieto-Alhambra D, Fina F, Rosell M, Williams
65: 47–53. doi:10.1016/j.recesp.2011.08.008.
T. Construction and validation of a scoring system for the selection of high-
6. Clua-Espuny JL, Lechuga-Duran I, Bosch-Princep R, et al. Prevalence of undi-
quality data in a Spanish population primary care database (SIDIAP). Inform
agnosed atrial fibrillation and of that not being treated with anticoagulant drugs:
Prim Care 2011; 19: 135–145.
the AFABE study. Rev Española Cardiol (English Ed 2013; 66: 545–552.
29. Ramos R, Balló E, Marrugat J, et al. Validity for use in research on vascular dis-
doi:10.1016/j.rec.2013.03.003.
eases of the SIDIAP (Information System for the Development of Research in
7. Camm AJ, Kirchhof P, Lip GYH, et al. Guidelines for the management of atrial
Primary Care): the EMMA study. Rev Esp Cardiol (Engl Ed) 2012; 65: 29–37.
fibrillation: the Task Force for the Management of Atrial Fibrillation of the
doi:10.1016/j.recesp.2011.07.017.
European Society of Cardiology (ESC). Eur Heart J 2010; 31: 2369–2429.
30. Bolíbar B, Fina Avilés F, Morros R, et al. SIDIAP database: electronic clinical
doi:10.1093/eurheartj/ehq278.
records in primary care as a source of information for epidemiologic research.
8. Själander S, Själander A, Svensson PJ, Friberg L. Atrial fibrillation patients do
Med Clin (Barc) 2012; 138: 617–621. doi:10.1016/j.medcli.2012.01.020.
not benefit from acetylsalicylic acid. Europace 2014; 16: 631–638.
31. SIDIAP. Information System for the Improvement of Research in Primary Care.
doi:10.1093/europace/eut333.
SIDIAP. 2015. Available at: http://www.sidiap.org/. Accessed March 16, 2016.
9. Hansen ML, Sørensen R, Clausen MT, et al. Risk of bleeding with single, dual, or tri-
32. CatSalut. Servei Català de la Salut. Conjunt mínim bàsic de dades (CMBD). Gen-
ple therapy with warfarin, aspirin, and clopidogrel in patients with atrial fibrillation.
eral Catalunya Gencat.cat 2015. Available at: http://catsalut.gencat.cat/ca/proveidors-
Arch Intern Med 2010; 170: 1433–1441. doi:10.1001/archinternmed.2010.271.
professionals/registres-catalegs/registres/cmbd/. Accessed March 16, 2016.
10. Mant J, Hobbs FDR, Fletcher K, et al. Warfarin versus aspirin for stroke preven-
33. Giner-Soriano M, Vedia Urgell C, Roso-Llorach A, et al. Effectiveness, safety
tion in an elderly community population with atrial fibrillation (the Birmingham
and costs of thromboembolic prevention in patients with non-valvular atrial
Atrial Fibrillation Treatment of the Aged Study, BAFTA): a randomised con-
fibrillation: phase I ESC-FA protocol study and baseline characteristics of a
trolled trial. Lancet 2007; 370: 493–503. doi:10.1016/S0140-6736(07)61233-1.
cohort from a primary care electronic database. BMJ Open 2016; 6: e010144.
11. Connolly S, Pogue J, Hart R, et al. Clopidogrel plus aspirin versus oral
doi:10.1136/bmjopen-2015-010144.
anticoagulation for atrial fibrillation in the Atrial fibrillation Clopidogrel Trial
34. Domínguez-Berjón M, Borrell C, Cano-Serral G, et al. Construcción de un índice
with Irbesartan for prevention of Vascular Events (ACTIVE W): a randomised
de privación a partir de datos censales en grandes ciudades españolas (Proyecto
controlled trial. Lancet 2006; 367: 1903–1912. doi:10.1016/S0140-6736(06)
MEDEA). Gac Sanit 2008; 22: 179–187.
68845-4.
35. Forslund T, Wettermark B, Wändell P, von Euler M, Hasselström J, Hjemdahl P.
12. Connolly SJ, Pogue J, Hart RG, et al. Effect of clopidogrel added to aspirin in
Risks for stroke and bleeding with warfarin or aspirin treatment in patients with
patients with atrial fibrillation. ACTIVE-A. N Engl J Med 2009; 360:
atrial fibrillation at different CHA2DS2VASc scores: experience from the
2066–2078. doi:10.1056/NEJMoa0901301.
Stockholm region. Eur J Clin Pharmacol 2014; 70: 1477–1485. doi:10.1007/
13. Friberg L, Rosenqvist M, Lip GYH. Net clinical benefit of warfarin
s00228-014-1739-1.
in patients with atrial fibrillation: a report from the Swedish atrial
36. Staerk L, Lip GYH, Olesen JB, et al. Stroke and recurrent haemorrhage associ-
fibrillation cohort study. Circulation 2012; 125: 2298–2307. doi:10.1161/
ated with antithrombotic treatment after gastrointestinal bleeding in patients with
CIRCULATIONAHA.111.055079.
atrial fibrillation: nationwide cohort study. BMJ 2015; 351: h5876–h5876.
14. Go AS, Hylek EM, Yuchiao C, et al. Anticoagulation therapy for stroke preven-
doi:10.1136/bmj.h5876.
tion in atrial fibrillation. JAMA 2003; 290: 2685–2692.
37. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to pre-
15. Hylek EM, Go AS, Yuchiao C, et al. Effect of intensity of oral anticoagulation
vent stroke in patients who have nonvalvular atrial fibrillation. Ann Intern Med
on stroke severity and mortality in atrial fibrillation. N Engl J Med 2003; 349:
2007; 146: 857–867.
1019–1026.
38. Gage BF, Waterman a D, Shannon W, Boechler M, Rich MW, Radford MJ. Val-
16. Singer DE, Chang Y, Fang MC, et al. The net clinical benefit of warfarin
idation of clinical classification schemes for predicting stroke: results from the
anticoagulation in atrial fibrillation. Ann Intern Med 2009; 151: 297–305.
National Registry of Atrial Fibrillation. JAMA 2001; 285: 2864–2870.
doi:10.1059/0003-4819-151-5-200909010-00003.
39. Lip GYH, Nieuwlaat R, Pisters R, Lane DA, Crijns HJGM. Refining clinical risk
17. Kirchhof P, Ammentorp B, Darius H, et al. Management of atrial fibrillation in
stratification for predicting stroke and thromboembolism in atrial fibrillation
seven European countries after the publication of the 2010 ESC Guidelines on
using a novel risk factor-based approach: the Euro Heart Survey on atrial fibril-
atrial fibrillation: primary results of the PREvention oF thromboemolic events
lation. Chest 2010; 137: 263–272. doi:10.1378/chest.09-1584.
—European Registry in Atrial Fibrillation (PREFER in AF). Europace 2014;
40. Pisters R, Hohnloser SH, Connolly SJ, et al. Effect of dronedarone on clinical end
16: 6–14. doi:10.1093/europace/eut263.
points in patients with atrial fibrillation and coronary heart disease: insights from
18. Kakkar AK, Mueller I, Bassand J-P, et al. Risk profiles and antithrombotic treat-
the ATHENA trial. Europace 2014; 16: 174–181. doi:10.1093/europace/eut293.
ment of patients newly diagnosed with atrial fibrillation at risk of stroke: perspec-
41. Lip GYH, Frison L, Halperin JL, Lane DA. Comparative validation of a novel
tives from the international, observational, prospective GARFIELD registry.
risk score for predicting bleeding risk in anticoagulated patients with atrial fibril-
PLoS One 2013; 8: e63479. doi:10.1371/journal.pone.0063479.
lation: the HAS-BLED (Hypertension, Abnormal Renal/Liver Function, Stroke,
19. Scowcroft ACE, Lee S, Mant J. Thromboprophylaxis of elderly patients with AF
Bleeding History or Predisposition, Labile INR, Elderly, Drug. J Am Coll
in the UK: an analysis using the General Practice Research Database (GPRD)
Cardiol 2011; 57: 173–80. doi:10.1016/j.jacc.2010.09.024.
2000–2009. Heart 2013; 99: 127–132. doi:10.1136/heartjnl-2012-302843.
20. Lobos-Bejarano JM, Del Castillo-Rodríguez JC, Mena-González A, et al.
Características de los pacientes y abordaje terapéutico de la fibrilación auricular
en atención primaria en España: Estudio FIATE. Med Clin (Barc) 2013; 141:
279–286. doi:10.1016/j.medcli.2012.12.023. SUPPORTING INFORMATION
21. Rodríguez-Mañero M, Cordero A, Bertomeu-González V, et al. Impacto de los
nuevos criterios para el tratamiento anticoagulante de la fibrilación auricular. Additional supporting information may be found in
Rev Española Cardiol 2011; 64: 639–641. doi:10.1016/j.recesp.2011.04.005.
22. Bertomeu-González V, Alberto Cordero PM, Moreno-Arribas J, Lorenzo Fácila the online version of this article at the publisher’s
JN, Moisés Rodríguez-Mañero JC-S, González-Juanatey JR, Juan Quiles VB-M. web site.

Copyright © 2016 John Wiley & Sons, Ltd. Pharmacoepidemiology and Drug Safety, 2017; 26: 97–107
DOI: 10.1002/pds
Copyright of Pharmacoepidemiology & Drug Safety is the property of John Wiley & Sons,
Inc. and its content may not be copied or emailed to multiple sites or posted to a listserv
without the copyright holder's express written permission. However, users may print,
download, or email articles for individual use.