Original Paper

Neuroepidemiology 2018;51:166–176 Received: June 7, 2018

Accepted: June 8, 2018
DOI: 10.1159/000490741 Published online: August 28, 2018

Risk of Hemorrhagic Stroke in Patients

Exposed to Nonsteroidal Anti-Inflammatory
Drugs: A Meta-Analysis of Observational Studies
Md. Mohaimenul Islam a, b Tahmina Nasrin Poly a, b Bruno Andreas Walther d
     

Hsuan-Chia Yang b Ming-Chin Lin a, e Yu-Chuan Li a–c, f

     

a Graduate
Institute of Biomedical Informatics, College of Medicine Science and Technology, Taipei Medical
University, Taipei, Taiwan; b International Center for Health Information Technology (ICHIT), Taipei Medical
 

University, Taipei, Taiwan; c Chairman, Department of Dermatology, Wan Fang Hospital, Taipei, Taiwan;
 

d Department of Biological Sciences, National Sun Yat-sen University, Kaohsiung City, Taiwan; e Department of
   

Neurosurgery, Shuang Ho Hospital, Taipei Medical University, New Taipei City, Taipei, Taiwan; f TMU Research Center
 

of Cancer Translational Medicine, Taipei, Taiwan

Keywords between the use of NSAIDs and hemorrhagic stroke. The

Nonsteroidal anti-inflammatory drugs · Hemorrhagic
stroke · Observational studies · Meta-analysis
Abstract
Background and Aim: Nonsteroidal anti-inflammatory
drugs (NSAIDs) are one of the most common pain relief
medications, but the risk of hemorrhagic stroke in patients
taking these medications is unclear. In this study, our aim
was to systematically review, synthesize, and critique the
epidemiological studies that evaluate the association be-
tween NSAIDs and hemorrhagic stroke risk. We therefore
assessed the current state of knowledge, filling the gaps in
our existing concern, and make a recommendation for fu-
ture research. Methods: We searched for articles in PubMed,
EMBASE, Scopus, and Web of Science between January 1,
1990, and July 30, 2017, which reported on the association
search was limited to studies published in English. The qual-
ity of the included studies was assessed in accordance with
the Cochrane guidelines and the Newcastle-Ottawa criteria.
Summary risk ratios (RRs) with 95% CI were pooled using a
random-effects model. Subgroup and sensitivity analyses
were also conducted. Results: We selected 15 out of the 785
unique abstracts for full-text review using our selection cri-
teria, and 13 out of these 15 studies met all of our inclusion
criteria. The overall pooled RR of hemorrhagic stroke was
1.332 (95% CI 1.105–1.605, p = 0.003) for the random effect
model. In the subgroup analysis, a significant risk was ob-
served among meloxicam, diclofenac, and indomethacin
users (RR 1.48; 95% CI 1.149–1.912, RR 1.392; 95% CI 1.107–
1.751, and RR 1.363; 95% CI 1.088–1.706). In addition, a
greater risk was found in studies from Asia (RR 1.490, 95%
CI 1.226–1.811) followed by Europe (RR 1.393, 95% CI 1.104–
1.757) and Australia (RR 1.361, 95% CI 0.755–2.452). Conclu-
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université

© 2018 S. Karger AG, Basel Yu-Chuan (Jack) Li, MD, PhD, FACHI, FACMI
Prof. College of Medical science and Technology
Department of Biomedical Informatics, Taipei Medical University
E-Mail karger@karger.com
250-Wuxing Street, Xinyi District, Taipei 11031 (Taiwan)
www.karger.com/ned
Downloaded by:

E-Mail jaak88 @ gmail.com; jack @ tmu.edu.tw

sion: Our results indicated that the use of NSAIDs is signifi-
cantly associated with a higher risk of developing hemor-
rhagic stroke. These results should be interpreted with
caution because they may be confounded owing to the ob-
servational design of the individual studies. Nevertheless,
we recommend that NSAIDs should be used judiciously,
and their efficacy and safety should be monitored proac-
tively. © 2018 S. Karger AG, Basel
Introduction
Stroke is the second most common cause of mortal-
ity worldwide [1, 2], and hemorrhagic stroke is respon-
sible for about 40% of all the deaths [3]. High blood
pressure is the most important risk factor for stroke [4],
and several established risk factors are associated with
stroke such as high cholesterol, smoking, physical inac-
tivity, and diabetes [5]. The use of certain drugs may
increase the risk of hemorrhagic stroke, such as oral
contraceptives that increase the risk of subarachnoid
hemorrhage [6]; however, their use is not associat-
ed with an increased risk of intracerebral hemorrhage
[7].
Nonsteroidal anti-inflammatory drugs (NSAIDs) are
the commonly and widely used medications for analgesic
and inflammatory treatment [8]. Despite their benefits of
pain relief, concern about the overall safety of NSAIDs
has been increasing because NSAIDs have been associ-
ated with an increased risk of cardiovascular disease,
cerebrovascular disease, upper gastrointestinal bleeding,
and perforation [9–11]. However, the impact of NSAIDs
on the risk of stroke, especially hemorrhagic stroke, has
received little attention. A few epidemiological studies in-
vestigated the association between them, while the under-
lying biological mechanisms that may increase the risk of
hemorrhagic stroke after the use of NSAIDs remain large-
ly unknown.
NSAIDs inhibit both COX-1 and COX-2; however,
the inhibition of platelet COX by NSAIDs leads to plate-
let aggregation, vasoconstriction, and a reduction in
thrombotic events. This pharmacological mechanism
might therefore increase the risk of hemorrhagic stroke.
Due to the lack of a firm biological explanation and the
mixed results from epidemiological studies, we there-
fore conducted an updated meta-analysis of observa-
tional studies to evaluate the possible association be-
tween the use of NSAIDs and the risk of hemorrhagic
stroke.
Hemorrhagic Stroke Risk and NSAIDs
Use
Methods
This meta-analysis was conducted in accordance with the Pre-
ferred Reporting Items for Systematic and Meta-analysis [12].
Database
The search strategy was developed in consultation with our
2 pharmacists (M.M.I. and T.N.P.) who are experts in drug and
disease association research [13–15]. PubMed, EMBASE, Scopus,
and Web of Science were searched for publications published be-
tween January 1, 1990, and July 30, 2017, reporting on the associa-
tion between NSAIDs use and hemorrhagic stroke. We used the
following text words as search terms: “NSAIDs,” “Diclofenac,”
“Ketorolac,” “Indomethacin,” “Ibuprofen,” “Naproxen,” “Piroxi-
cam,” “Meloxicam,” “Sulindac,” “Mefanamic acid,” “Celecoxib,”
“Rafecoxib,” and “Hemorrhagic stroke.” We also searched the ref-
erence lists of all included full-text articles to identify any studies
missed in the initial search. In this process, Google Scholar was
used to find academic articles citing eligible articles. However, we
did not consider any unpublished study or references that existed
only as abstract. Finally, we compiled all references, and Endno-
teX7 (Thomson Reuters) was used to manage and remove the du-
plicates.
Eligibility Criteria
In this stage, our 4 authors from International Center for
Health Information Technology screened all the titles and ab-
stracts, and each author reviewed all of them independently. At
this initial stage, the following criteria were selected to allow the
inclusion of any relevant study. We considered only studies that
were published in English and original research using any obser-
vational design (e.g., case-control or cohort), which reported on
NSAIDs use and the risk of hemorrhagic stroke. We excluded all
editorials, short communications, or case studies.
After exclusion of irrelevant and articles with insufficient infor-
mation, 2 of our authors (M.M.I. and T.N.P.) from the same lab
independently reviewed all included full-text articles. They also
checked for any duplication of selected studies. We finally includ-
ed studies if they (1) were published in English (2) were reported
patients treated with NSAIDs and increased the risk of hemor-
rhagic stroke (3) were provided OR/Hazard ratio (HR) with 95%
CI or sufficient information to calculate the risk ratio (RR) with
95% CI. We excluded articles if they were not observational and
the outcome of interest was not hemorrhagic stroke. All studies
meeting inclusion criteria at this stage were reviewed by 2 authors
(M.M.I. and T.N.P.) to ensure the appropriateness of inclusion in
the final analysis stroke. If any disagreements arose between these
2 authors, it was resolved by consensus with our main supervisor
(Y.-C.L.).
Data Extraction
In this stage, the remaining 13 studies were then extracted,
and 2 authors (M.M.I. and T.N.P.) collected all the relevant data
from each of these 13 studies. These 2 authors garnered the fol-
lowing data: study information (e.g., author, year of publication),
study characteristics (age and gender of participants, location,
duration of data collection), condition information (i.e., data
source, condition definition, and total number of participants),
and OR or HR.
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université
Neuroepidemiology 2018;51:166–176 167
DOI: 10.1159/000490741
Downloaded by:
 Primary and Subgroups Analysis
We evaluated the association between NSAIDs and hemor-
rhagic stroke risk as a primary analysis. We also performed sub-
group analyses by the type of study design (cohort and case-con-
trol), study region (Asia, Australia, Europe, and North America),
methodological quality (low-quality vs high quality, see details be-
low), and individual NSAIDs (Diclofenac, Indomethacin, Ketoro-
lac, Meloxicam, etc.).
Quality Assessment
We used Newcastle-Ottawa Scale to assess the methodological
quality of the included observational studies [16]. Included studies
in our meta-analysis were evaluated with 3 categories: selection
(4 stars) and comparability (2 stars) of study groups, and assess-
ment of the outcome of interest (3 stars). The star rating system was
used to indicate the quality, with a score from 0 to 9. However,
0–6 stars defined as low quality and 7–9 stars as high quality.
Statistical Analyses
In this section, we collected the ORs and the HRs with 95% CIs
for NSAIDs use and hemorrhagic stroke risk from included each
observational studies. It is well established when the disease inci-
dence was low in the cohort studies, we could regard observed ORs
as close approximations of RR. In addition, we combined them with
HRs, resulting in a common estimate of RR. However, RR >1 indi-
cate an increased risk of the hemorrhagic stroke, and RR <1 indi-
cate a decreased risk of the hemorrhagic stroke. We evaluated sta-
tistical significance using the 95% CIs. If the 95% CI did not include
the neutral value of 1, we considered the risk statistically significant.
We combined estimates from both case-control and cohort studies
and used the most adjusted estimate available in each study because
we considered adjusted estimates to be the most valid.
In our meta-analysis, we used a random effect model in our fi-
nal analysis because we assumed that it will reduce heterogeneity
among the studies. The comprehensive meta-analysis package
(Version: 3) was used to make the pooled estimate, forest plot and
subgroup analysis. The meta-analysis of proportion uses the bi-
nominal distribution for analysis. We quantified heterogeneity us-
ing the I2 statistic, and its significance was determined based on the
accompanying p value in the Cochran Q test. However, an I2 value
of 0% indicates no observed heterogeneity, and increasing values
represent greater amounts of heterogeneity; values of 25, 50, and
75% indicate low, moderate, and high levels of heterogeneity re-
spectively. Q Values arising from the random effects models were
also used to quantify heterogeneity.
Results
Study Screening
We retrieved 785 unique titles and abstracts, of which
770 articles were excluded based on our predefined study
inclusion criteria described in the Method section of our
study. We reviewed the full text of the selected 15 articles, and
also assessed their reference lists for relevant articles. Based
on the review, 13 studies [17–29] met all of our study inclu-
sion criteria. Figure 1 summarized our selection process.
168 Neuroepidemiology 2018;51:166–176
DOI: 10.1159/000490741
Study Characteristics
A qualitative synthesis of the 13 studies meeting all of
our inclusion criteria is summarized in Table 1. Eight of
these studies were case-control design studies [18–24, 28],
and 5 were cohort studies [18, 19, 24, 26, 27]. The date of
the publication of the 13 studies was 1999 [20] to 2017 [17].
Five studies were conducted in Europe [20, 24, 26, 27], 4 in
Asia [17, 21–23], 3 in Australia [20, 25, 28], and 1 in North
America [29]. The number of participants ranged from 134
to 4,354, and the total number of participants was 1,839,462.
All studies confirmed hemorrhagic stroke patients by us-
ing medical records. Included studies tried to reduce con-
founding factors by estimating adjusted risk with various
confounding factors. However, all included studies consid-
ered hemorrhagic stroke as either subarachnoid hemor-
rhage or intracerebral hemorrhage. Subarachnoid hemor-
rhage was diagnosed based on clinical symptoms via either
a brain image (CT, MRI) or evidence of xanthochromia on
a lumbar puncture. Additionally, intracerebral hemor-
rhage was diagnosed based on clinical symptoms and the
detection of blood in the brain parenchyma. Ventricles by
CT or MRI were also considered to be able to identify in-
tracerebral hemorrhage.
Methodological Quality of the Studies
The methodological quality of the 13 studies was as-
sessed based on Newcastle-Ottawa Scale. The method-
ological quality score ranged from 6 to 9; the mean was
7.69 (Table 2).
NSAIDs Use and Hemorrhagic Stroke Risk
Among the 13 studies, NSAIDs use was significantly
associated with an increased risk of hemorrhagic stroke
when compared with non-use. The overall pooled in-
crease of hemorrhagic stroke in patients with NSAIDs use
was RR 1.332 (95% CI 1.105–1.605) with a significant het-
erogeneity for the random effect model (Fig. 2). The over-
all pooled for intracerebral hemorrhagic stroke was RR
1.16 (95% CI 0.96–1.403, I2 = 0, Q = 0.206), and subarach-
noid hemorrhage stroke was RR 1.13 (95% CI 0.737–
1.733, I2 = 0, Q = 0.117).
Subgroup Analysis
We also performed sub-group analyses to evaluate the
influence of the study design, region, methodological
quality of included studies, and different classes of
NSAIDs medication where these characteristics could be
the possible source of heterogeneity (Table 3).
Eight case-control and 6 cohort studies provided the
risk estimation of hemorrhagic stroke patients with
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université
Islam/Poly/Walther/Yang/Lin/Li
Downloaded by:
 Color version available online
Records identified through Additional records identified

Identification
database searching through other sources
(n = 784) (n = 1)

Records after duplicates removed

(n = 423)

Screening
Records screened Records excluded
(n = 210) (n = 213)

Full-text articles Full-text articles excluded (n = 2)

assessed for eligibility Full-text duplication = 1
Eligibility

(n = 15) Ineligible study design = 1

Studies included in
qualitative synthesis
(n = 13)
Included

Studies included in
Fig. 1. Diagram of study selection, adopted quantitative synthesis
from Preferred Reporting Items for Sys- (meta-analysis)
(n = 13)
tematic and Meta-analysis (PRISMA) flow
diagram.

NSAIDs therapy. The pooled RR of case-control studies
was 1.090 (95% CI 0.998–1.192, p = 0.055) with signifi-
cant heterogeneity (I2 = 0000; p = 0.868; Q = 3.175). The
pooled RR of cohort studies was 1.592 (95% CI 1.179–
2.148, p = 0.002) in the random effect model and signifi-
cant heterogeneity was detected (I2 = 89.676; p = 0.002;
Q = 38.745).
Eleven high-quality methodological studies separately
assessed the impact of NSAIDs therapy on the risk of
hemorrhagic stroke. There was an increased risk of hem-
orrhagic stroke among NSAIDs users compared to non-
users in the random effect model with RR 1.219 (95% CI
0.991–1.498) and significant heterogeneity was detected
(I2 = 90.435; p < 0.0001; Q = 104.552). Three low quality
methodological studies evaluated the impact of NSAIDs
on hemorrhagic stroke risk. The overall pooled RR were
2.182 (95% CI 1.772–2.686) in the random effect model
and significant heterogeneity was detected (I2 = 59.143;
p = 0.005; Q = 0.945).
Five studies from Europe evaluated the risk of hemor-
rhagic stroke with NSAIDs use. The overall pooled RR was
1.393 (95% CI 1.104–1.757, p = 0.005) in the random effect
model, and significant heterogeneity was detected (I2 =
56.984; p = 0.054; Q = 9.302). The hemorrhagic stroke in
patients with NSAIDs therapy was significantly higher in
the 4 studies from Asia (RR 1.490; 95% CI 1.226–1.811, p <
0.0001) in the random effect model with significant het-
erogeneity detected (I2 = 36.835; p = 0.191; Q = 4.749).
When we compared the 3 studies from Australia, we also
found an increased risk (RR 1.361; 95% CI 0.755–2.452,
p = 0.305) in the random effect model with significant het-
erogeneity (I2 = 93.812; p < 0.000; Q = 32.32).
The hemorrhagic stroke risk in patients using diclof-
enac was (RR 1.392; 95% CI 1.107–1.751) in the ran-
dom effect model. Additionally, in the random effect
model, the pooled RR for naproxen, Indomethacin,
Piroxicam, Ketorolac, meloxicam users were RR 1.486
(95% CI 1.014–2.178), RR 1.363 (95% CI 1.088–1.706),
RR 1.220 (95% CI 0.711–2.095), RR 2.722 (95% CI
1.047–7.078), RR 1.482 (95% CI 1.149–1.912) respec-
tively.
Sensitivity Analysis
To assess whether a single study had a substantial in-
fluence on the main results, we excluded each study con-
secutively and then evaluated its effect on the summary
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université

Hemorrhagic Stroke Risk and NSAIDs Neuroepidemiology 2018;51:166–176 169

Use DOI: 10.1159/000490741
Downloaded by:
Table 1. Summary of the thirteen observational studies regarding NSAIDs use and hemorrhagic stroke risk

Author Publication Country Study Study HSP Adjustment

year design duration number

Hsu et al. [17] 2017 Taiwan Cohort 2000–2012 255 1, 2, 16, 17, 22, 23, 30
Bak et al. [18] 2003 Denmark CC 1994–1999 867 1, 2, 8, 15
Johnsen et al. [19] 2003 Denmark CC 1991–1999 912 1, 2, 5, 14, 15, 16, 17, 18, 19, 22
Thrift et al. [20] 1999 Australia CC 1990–1992 331 2, 16, 17, 20, 21, 22
Choi et al. [21] 2008 South Korea CC 2002–2004 940 1, 2, 14, 15, 16, 17, 20, 21, 22, 23, 24, 25
Chuang et al. [22] 2015 Taiwan CC 2005–2010 258 1, 2, 8, 10, 12, 19, 17, 18, 20, 24, 25, 26, 27, 28
Chang et al. [23] 2010 Taiwan CC 2005–2006 9456 3, 5, 6, 7, 8, 9, 10, 16, 17, 24, 27, 28, 29, 30, 31, 32, 33
Lapi et al. [24] 2016 Italy CC 2002–2012 578 1, 2, 3, 16, 17, 21, 22, 28, 31
Caughey et al. [25] 2011 Australia Cohort 2001–2008 350 N/A
Fosbøl et al. [26] 2014 Denmark Cohort 1997–2005 N/A 1, 2
Haag et al. [27] 2008 Netherland Cohort 1991–2004 134 1, 2, 17, 21, 24, 29
Mangoni et al. [28] 2010 Australia CC 2002–2006 1391 1, 2, 15, 16, 17, 24, 25, 31
Roumie et al. [29] 2008 USA Cohort 1999–2004 4354 1, 2, 3, 8, 9, 10, 12, 14, 21, 28

1, age; 2, sex; 3, antihypertensive; 4, statins; 5, insulin; 6, sulfonylurea; 7, thiazolidinedione; 8, beta-blockers; 9, calcium channel blockers; 10, diure-
tics; 11, non-aspirin antiplatelet; 12, anti-diabetics; 13, antiarrhythmics and antianginal drugs; 14, anticoagulant; 15, salicylic acid; 16, hypertension; 17,
diabetes mellitus; 18, chronic bronchitis; 19, antihypertensive; 20, cardiovascular disease; 21, smoking status; 22, alcoholism; 23, hyperlipidemia; 24,
arthritis; 25, respiratory tract infection; 26, gout; 27, ACE inhibitors; 28, statin; 29, atrial fibrillation; 30, congestive heart failure; 31, chronic renal disease;
32, chronic lung disease; 33, peptic ulcer disease.
CC, case-control; HSP, hemorrhagic stroke patients; NSAIDs, nonsteroidal anti-inflammatory drugs.

estimates and heterogeneity of the main analysis. As Publication Bias

drugs act differently in each human body [30], we per-
formed sensitivity analysis to evaluate gender effects. The
pooled RR in male patients from 4 studies was 1.130 (95%
CI 0.778–1.642, p = 0.521). The overall pooled RR in fe-
male patients from 3 studies was 1.271 (95% CI 0.954–
1.694, p = 0.101). No particular association was observed
in the anti-platelets use prior to hemorrhagic stroke.
However, Thrift et al. [20] analyzed adjusted ORs of pri-
mary intracerebral hemorrhage for use of aspirin on the
basis of different dose regimens. Five studies evaluated
the risk of hemorrhagic stroke with NSAIDs use ≥30 days
(RR 1.304, 95% CI 1.05–1.61, I2 = 60.25, Q = 10.07) [18,
19, 22–24]. Two studies estimated NSAIDs according to
plasma half-life <12 and >12 h and the overall pooled RR
1.24 (95% CI 1.009–1.547, I2 = 0.00, Q = 0.248) and RR
0.85 (95% CI 0.512–1.431, I2 = 0.00, Q = 0.08) respec-
tively [18, 24]. Eight studies had ≤5 years’ follow-up pe-
riod and the risk of hemorrhagic stroke with NSAIDs was
RR 1.17 (95% CI 0.911–1.527, I2 = 92.816, Q = 0.97.436).
The meta-analysis of the observational studies re-
vealed several types of bias. If a large number of stud-
ies  was included, the visual interpretation and results
of the test of asymmetry of the funnel plot of the pub-
lication are reliable [31, 32]. The funnel plot in Figure
3 indicates the existence of some publication bias.
The   Egger’s regression test of the funnel asymmetry
showed a slightly insignificant publication bias (p val-
ue = 0.07).
Discussion
Principle Findings
We conducted an updated meta-analysis to evaluate
the association between NSAIDs and the risk of hemor-
rhagic stroke. The findings from our updated meta-
analysis indicate that NSAIDs significantly increased
the risk of hemorrhagic stroke. Our results are different
from previously published meta-analysis [33]. A robust
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université

170 Neuroepidemiology 2018;51:166–176 Islam/Poly/Walther/Yang/Lin/Li

DOI: 10.1159/000490741
Downloaded by:
Table 2. Methodological quality assessment of the observational studies using the NOS

Case-control Study Selection Comparability Exposure Total

definition representativeness selection definition control for ascertainment same method of non- (0–9)
adequate of the cases of controls of controls important of exposure ascertainment for response
factor or cases and controls rate
additional
factor

Bak et al. [18], * * * * ** * * 8

2003
Johnsen et al. [19], * * * ** * * * 8
2003
Thrift et al. [20], * * * * ** * * * 9
1999
Choi et al. [21], * * * * * * * * 8
2008
Chuang et al. [22], * * * * ** * * 8
2015
Chang et al. [23], * * * ** * * * 8
2010
Lapi et al. [24], * * * * * * * * 8
2016
Mangoni et al. [28], * * * * * * * 7
2010

Case-control Study Selection Comparability Exposure Total

selection of representativeness ascertainment outcome comparability assessment follow-up adequacy (0–9)
non-exposed of the cohort of exposure of interest of cohorts on of outcome long enough of follow-up
cohort the basis of the for outcomes of cohorts
design or to occur
analysis

Hsu et al. [17], * * * * ** * * * 9

2017
Caughey et al. [25], * * * * * * 6
2011
Fosbøl et al. [26], * * * * * * 6
2014
Haag et al. [27], * * * * * * * * 8
2008
Roumie et al. [29], * * * * * * * 7
2008

A “star (*)” system of the Newcastle-Ottawa Scale (NOS) has been developed for the methodological quality assessment: each study can be awarded a maximum of one
star for each numbered item within the selection and exposure categories, while a maximum of 2 stars can be given for the comparability category.

analysis was conducted consisting of primary and sub- Biological Plausibility

group analysis. However, these findings should be in-
terpreted with caution due to comprehensive statisti-
cal  analysis and significant clinical heterogeneity
among the included studies. We retrieved and analyzed
data from previously published observational studies;
therefore, our findings cannot clarify any possible caus-
al relationships and may have several confounding fac-
tors.
The mechanism of the association between NSAIDs use
and the risk of hemorrhagic stroke remains inconclusive.
However, there are several possible explanations of the as-
sociation. First, NSAIDs promote rapid accumulation of
blood within the brain parenchyma that is linked to the dis-
ruption of the normal anatomy and increased pressure in
the lumen. Second, the long-term use of NSAIDs facilitates
arterial aneurysms rapture, and other vascular anomalies.
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université

Hemorrhagic Stroke Risk and NSAIDs Neuroepidemiology 2018;51:166–176 171

Use DOI: 10.1159/000490741
Downloaded by:
 Statistics for each study
Risk Lower Upper Relative
Study name ratio limit limit Z-value p-value Risk ratio and 95% CI weight
Bak (2003) 1.200 0.900 1.600 1.242 0.214 8.41
Johnsen (2003) 1.130 0.809 1.578 0.717 0.473 7.86
Thrift (1999) 1.030 0.521 2.036 0.085 0.932 4.37
Choi (2007) 1.120 0.765 1.640 0.583 0.560 7.30
Chuang (2015) 1.380 0.792 2.405 1.137 0.256 5.44
Chang (2010) 1.440 0.871 2.381 1.421 0.155 5.96
Lapi (2015) 1.150 0.818 1.616 0.805 0.421 7.78
Caughey (2011) 2.190 1.736 2.763 6.612 0.000 9.04
Fosbol (2014) 2.150 1.351 3.422 3.229 0.001 6.36
Haag (2008) 1.770 1.295 2.419 3.582 0.000 8.12
Mangoni (2010) 1.040 0.931 1.162 0.694 0.488 10.13
Hsu (2017) 1.660 1.601 1.721 27.487 0.000 10.47
Roumie (2008) 0.790 0.611 1.022 -1.796 0.072 8.76
1.332 1.106 1.605 3.015 0.003

0.1 0.2 0.5 1 2 5 10

Decreased risk Increased risk

Fig. 2. Risk ratio (RR) with 95% CI from the studies of risk of hemorrhagic stroke in patients who received NSAIDs.

Table 3. Subgroup analysis of the association between NSAIDs and hemorrhagic stroke risk

Variables Number Pooled estimates Test of heterogeneity Method

of studies
RR (95% Cl) p value Q value p value I 2, % RE

All studies 13 1.332 (1.105–1.605) 0.003 114.773 <0001 89.545 RE

Study design
Case control 8 1.090 (0.998–1.192) 0.055 3.175 0.868 0.000 RE
Cohort 5 1.592 (1.179–2.148) 0.002 38.745 0.000 89.676 RE
Region
Europe 5 1.393 (1.104–1.757) 0.005 9.302 0.054 56.996 RE
Asia 4 1.490 (1.226–1.811) <0001 4.749 0.191 36.835 RE
Australia 3 1.361 (0.755–2.452) 0.305 32.32 <0.0001 93.812 RE
North America 1 0.79 (0.611–1.022) 0.072 N/A N/A N/A RE
Methodology quality
High 11 1.219 (0.991–1.498) 0.061 104.552 <0001 90.435 RE
Low 2 2.182 (1.772–2.686) <0.0001 0.005 0.945 0.000 RE
COX-1 inhibitors
Ibuprofen 9 1.165 (0.969–1.401) 0.105 19.580 0.012 59.143 RE
Diclofenac 9 1.392 (1.107–1.751) 0.005 35.406 0.000 77.405 RE
Naproxen 8 1.486 (1.014–2.178) 0.042 36.843 0.000 81.001 RE
Indomethacin 5 1.362 (1.088–1.706) 0.007 3.354 0.500 0.000 RE
Piroxicam 5 1.220 (0.711–2.095) 0.470 6.592 0.159 33.317 RE
Meloxicam 5 1.482 (1.149–1.912) 0.002 7.343 0.119 45.523 RE
Ketorolac 3 2.722 (1.047–7.078) 0.040 3.381 0.184 40.842 RE
Sulindac 2 1.094 (0.770–1.555) 0.615 0.725 0.395 0.000 RE
Mefanamic acid 2 1.013 (0.624–1.644) 0.959 1.216 0.270 17.755 RE
COX-2 inhibitors
Celecoxib 5 1.170 (0.918–1.491) 0.204 11.691 0.020 65.786 RE
Rafecoxib 4 1.692 (1.072–2.670) 0.024 14.061 0.003 78.664 RE

RE, random effect model; NSAIDs, nonsteroidal anti-inflammatory drugs; N/A, not applicable.
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université

172 Neuroepidemiology 2018;51:166–176 Islam/Poly/Walther/Yang/Lin/Li

DOI: 10.1159/000490741
Downloaded by:

0
0.1
Standard error
0.2
0.3
0.4
–2.0
Fig. 3. Funnel plot shows the association
between NSAIDs use and hemorrhagic
stroke risk.
Third, NSAIDs are responsible for the imbalance of various
prostaglandins, thromboxane, prostacyclin, and their ac-
tion on vascular function, platelet aggregation, and smooth
muscle proliferation. COX-1 affects the biosynthesis of
thromboxane, which is a potent vasoconstrictor and platelet
agonist. Inhibition of thromboxane A2 synthesis by NSAIDs
triggers the risk of bleeding complications, and might in-
crease the risk of hemorrhagic stroke. Finally, renal prosta-
glandins (prostaglandins E2 and prostaglandin I2) have in-
hibitory effects on sodium reabsorption at the loop of Hen-
le and on water reabsorption at the collecting tubules. The
inhibition of COX-2 enzyme by NSAIDs is linked to intra-
vascular volume expansion. Renal impairment is associated
with a greater neurological deficit following hemorrhagic
stroke [34]. Mild stages of renal disease and estimated glom-
erulus filtration rate less than 60 mL/min/1.73 m2 increase
the risk of hemorrhagic stroke in the general population
[35]. Hypertension could be induced by the combination of
salt/fluid retention and muscle vasoconstriction, which
could contribute hemorrhagic stroke. As there is no con-
crete evidence between the association and because the in-
tracerebral hemorrhage and subarachnoid hemorrhage are
quite different and affect different age groups, more studies
are needed to explore the possible association.
Public Health Implication and Clinical Practice
Stroke is the third and fourth leading cause of death in
the United States and the United Kingdom respectively.
Hemorrhagic stroke occurs due to a blood vessel bursting
within or on the surface of the brain. According to the
national clinical guideline for stroke, 10–15% of hemor-
rhagic patients die before reaching the hospital [36].
Women are prone to have more strokes with increasing
age [37], while men are at a higher risk of having it at a
Hemorrhagic Stroke Risk and NSAIDs
Use
Funnel plot of standard error by log risk ratio
–1.5 –1.0 –0.5 0 0.5 1.0 1.5 2.0
Log risk ratio
younger age. Patients experience a greater range of dis-
abilities than for any other condition [38]. It is estimated
that more than half of stroke survivors face visual prob-
lems [39], a third experience some level of aphasia [40,
41], over 3 quarter experience arm weakness, and almost
3 quarter experience leg weakness [42].
NSAIDs have been used widely over the counter in
pharmacies, and it is usually taken without the advice of
physicians/ pharmacists because often people consider
them safe medication. Patients with a history of chronic
pain typically headache and lower pain admit to having
taken regular NSAIDs for long periods of time [43]. Ad-
ditionally, the increasing rate of aging populations and
growing use of self-medication are responsible for in-
creasing OTC NSAIDs use. However, a significant amount
of epidemiological studies reported that the use of NSAIDs
is associated with an increased risk of chronic arterial fi-
brillation [44], gastrointestinal tract bleeding [45], and
myocardial infraction [46], which can lead to death. Our
meta-analysis raises the possibility of an increased risk of
hemorrhagic stroke among NSAIDs users. Therefore, we
advocate that it is the time to carefully reevaluate the out-
come of this drugs class. Our findings are thus of major
importance to public health because the impact of NSAIDs
on hemorrhagic stroke has only received little attention.
In recent days, among the general population, NSAIDs are
being prescribed on a large scale and more frequently be-
cause they have been found to be effective treatment for
obtaining relief from pain, fever, and inflammation. But
the finding of our study showed that the risk of ill effects
as a result of NSAIDs use increased by 32% in users and
this is likely to generate a potential health hazard in the
general public. Hence, any patients who are taking NSAIDs
should discuss immediately with physicians if they experi-
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université
Neuroepidemiology 2018;51:166–176 173
DOI: 10.1159/000490741
Downloaded by:
ence adverse outcomes regarding hemorrhagic stroke
such as sudden severe headache, vision changes, loss of
balance, loss of speech, or difficulty understanding speech.
Moreover, clinical guidelines on the use of NSAIDs are so
much warranted for a safer and efficient treatment.
Strengths and Limitations
Our study has several strengths. First, multiple high-
quality and methodologically rigorous observational
studies were included. Therefore, it is the most compre-
hensive meta-analysis so far that has evaluated the pos-
sible association of the use of NSAIDs and stroke risk.
Second, we examined the association in greater detail,
that is, we stratified our results by study design, method-
ological quality, individual NSAIDs use, and so on.
Third, all included studies had a higher number of par-
ticipants, hence the findings about stroke risk patients
with NSAIDs is more reliable. Finally, our study summa-
rized findings from 4 continents, which make our study
more robust.
Our study also has some important limitations. First,
there was substantial heterogeneity across the study in the
main analysis of stroke risk. However, we could explain our
findings based on the various study designs, regional ef-
fects, various study methods, and individual NSAIDs use.
We could not ignore heterogeneity and study results vari-
ability although we included a higher number of studies.
But, we used the random effect model in our analysis, which
could reduce heterogeneity among the included studies.
Second, we used the Egger test but it is considered to lack
power due to the reduction of potential bias among studies.
Third, we could not provide risk information about African
and South-American region due to lack of data. Fourth,
most of the studies included in this meta-analysis only in-
cluded older populations. Therefore, the results might not
be generalizable to the general population, especially to
younger patients with a different baseline cardiovascular
risk. Finally, we could not be categorized according to age
because each study cauterized age into different groups.
However, level of dosage has not been analyzed due to lack
of information in the included studies. Only one study, Bak
et al. [18] estimated the effect of different types of NSAIDs
dose (<1 DDD and ≥1 DDD) and the risk of intracerebral
hemorrahage.
Recommendations and Future Directions
Since the use of NSAIDs has been increasing, and our
results show an increased risk of hemorrhagic stroke
among the NSAIDs users, physicians should take into con-
sideration this potential risk, and effective assessment is
174 Neuroepidemiology 2018;51:166–176
DOI: 10.1159/000490741
essential if stroke symptoms are suspected. To ensure safe
and effective treatment, it is therefore indispensable to
monitor treatment outcomes and patient’s condition in a
timely basis. The decision to prescribe NSAIDs should be
made in the light of the individual patient’s benefits rather
than the risks. In addition, observational studies cannot
clarify whether the observed epidemiologic outcome is the
causal effect or the results of some unmeasured confound-
ing variables. Therefore, more randomized control studies
are warranted to reduce the confounding factors. Further-
more, experimental animal and biological models are also
needed to identify a possible biological mechanism that
links the use of NSAIDs and the risk of hemorrhagic stroke.
Conclusions
The present updated meta-analysis was designed to
evaluate the impact of NSAIDs use on hemorrhagic stroke
risk. On the basis of our findings, we suggest that the use
of NSAIDs is associated with a higher risk of hemorrhag-
ic stroke. Physicians might preferentially avoid prescrib-
ing NSAIDs to patients with a risk of hemorrhagic stroke
and monitor the outcome; pharmacists should follow the
guidelines and assist patients with the appropriate dosage
and administration to avoid unfavorable side effects. Ad-
ditionally, healthcare policy makers should promote im-
proved quality and safety of NSAIDs medication through
patient’s education, physicians training, and assessment
of disease condition.
Acknowledgment
This work was financially supported by the “TMU Research
Center of Cancer Translational Medicine” from The Featured Ar-
eas Research Center Program within the framework of the Higher
Education Sprout Project by the Ministry of Education (MOE) in
Taiwan.
Ethics Statement
This study has been approved by our Institution’s Ethics Com-
mittee and has, therefore, been performed in accordance with the
ethical standards laid down in the 1964 Declaration of Helsinki.
Disclosure Statement
The authors declare that there are no conflicts of interest to
disclose.
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université
Islam/Poly/Walther/Yang/Lin/Li
Downloaded by:
 References
 1 Feigin VL, Krishnamurthi RV, Parmar P,
Norrving B, Mensah GA, Bennett DA, Bark-
er-Collo S, Moran AE, Sacco RL, Truelsen T,
et al: Update on the global burden of ischemic
and hemorrhagic stroke in 1990–2013: the
GBD 2013 study. Neuroepidemiology 2015;
45:161–176.
  2 Bonzini M, Ferrario MM, Bertù L, Bono G,
Vidale S, Veronesi G, Chambless L, Cesana
GC: Temporal trends in ischemic and hemor-
rhagic strokes in Northern Italy: results from
the Cardiovascular Monitoring Unit in
Northern Italy population-based register,
1998–2004. Neuroepidemiology 2012; 39: 35–
42.
  3 Feigin VL, Forouzanfar MH, Krishnamurthi
R, Mensah GA, Connor M, Bennett DA, Mo-
ran AE, Sacco RL, Anderson L, Truelsen T, et
al: Global and regional burden of stroke dur-
ing 1990-2010: findings from the Global Bur-
den of Disease Study 2010. Lancet 2014; 383:
245–255.
  4 Go AS, et al; American Heart Association Sta-
tistics Committee and Stroke Statistics Sub-
committee: Heart disease and stroke statis-
tics–2013 update: a report from the American
Heart Association. Circulation 2013; 127:e6–
e245.
  5 Lee KR, Bae JH, Hwang IC, Kim KK, Suh HS,
Ko KD: Effect of influenza vaccination on risk
of stroke: a systematic review and meta-anal-
ysis. Neuroepidemiology 2017;48:103–110.
  6 Algra AM, Klijn CJ, Helmerhorst FM, Algra
A, Rinkel GJ: Female risk factors for sub-
arachnoid hemorrhage: a systematic review.
Neurology 2012;79:1230–1236.
  7 Plu-Bureau G, Hugon-Rodin J, Maitrot-Man-
telet L, Canonico M: Hormonal contracep-
tives and arterial disease: an epidemiological
update. Best Pract Res Clin Endocrinol Metab
2013;27:35–45.
  8 Manthripragada AD, Schernhammer ES, Qiu
J, Friis S, Wermuth L, Olsen JH, Ritz B: Non-
steroidal anti-inflammatory drug use and the
risk of Parkinson’s disease. Neuroepidemiol-
ogy 2011;36:155–161.
  9 Group ADAPT Research: Cardiovascular and
cerebrovascular events in the randomized,
controlled Alzheimer’s Disease Anti-Inflam-
matory Prevention Trial (ADAPT). PLoS
Clin Trials 2006;1:e33.
10 Laporte JR, Ibanez L, Vidal X, Vendrell L, Le-
one R: Upper gastrointestinal bleeding associ-
ated with the use of NSAIDs: newer versus
older agents. Drug Saf 2004;27:411–420.
11 Trelle S, Reichenbach S, Wandel S, Hildeb-
rand P, Tschannen B, Villiger PM, Egger M,
Jüni P: Cardiovascular safety of non-steroidal
anti-inflammatory drugs: network meta-
analysis. BMJ 2011;342:c7086.
12 Liberati A, Altman DG, Tetzlaff J, Mulrow C,
Gøtzsche PC, Ioannidis JP, Clarke M, De-
vereaux PJ, Kleijnen J, Moher D: The PRIS-
MA statement for reporting systematic re-
views and meta-analyses of studies that evalu-
Hemorrhagic Stroke Risk and NSAIDs
Use
ate health care interventions: explanation and
elaboration. PLoS Med 2009;6:e1000100.
13 Islam MM, Yang HC, Nguyen PA, Poly TN,
Huang CW, Kekade S, Khalfan AM, Debnath
T, Li YJ, Abdul SS: Exploring association be-
tween statin use and breast cancer risk: an up-
dated meta-analysis. Arch Gynecol Obstet
2017;296:1043–1053.
14 Islam MM, Iqbal U, Walther B, Atique S,
Dubey NK, Nguyen PA, Poly TN, Masud JH,
Li YJ, Shabbir SA: Benzodiazepine use and
risk of dementia in the elderly population: a
systematic review and meta-analysis. Neuro-
epidemiology 2016;47:181–191.
15 Yang HC, Nguyen PAA, Islam M, Huang
CW, Poly TN, Iqbal U, Li YJ: Gout drugs use
and risk of cancer: a case-control study. Joint
Bone Spine 2018, Epub ahead of print.
16 Stang A: Critical evaluation of the Newcastle-
Ottawa scale for the assessment of the quality
of nonrandomized studies in meta-analyses.
Eur J Epidemiol 2010;25:603–605.
17 Hsu PS, Lin HH, Li CR, Chung WS: Increased
risk of stroke in patients with osteoarthritis: a
population-based cohort study. Osteoarthri-
tis Cartilage 2017;25:1026–1031.
18 Bak S, Andersen M, Tsiropoulos I, García Ro-
dríguez LA, Hallas J, Christensen K, Gaist D:
Risk of stroke associated with nonsteroidal
anti-inflammatory drugs: a nested case-con-
trol study. Stroke 2003;34:379–386.
19 Johnsen SP, Pedersen L, Friis S, Blot WJ,
McLaughlin JK, Olsen JH, Sørensen HT:
Nonaspirin nonsteroidal anti-inflammatory
drugs and risk of hospitalization for intrace-
rebral hemorrhage: a population-based case-
control study. Stroke 2003;34:387–391.
20 Thrift AG, McNeil JJ, Forbes A, Donnan GA:
Risk of primary intracerebral haemorrhage
associated with aspirin and non-steroidal an-
ti-inflammatory drugs: case-control study.
BMJ 1999;318:759–764.
21 Choi NK, Park BJ, Jeong SW, Yu KH, Yoon
BW: Nonaspirin nonsteroidal anti-inflamma-
tory drugs and hemorrhagic stroke risk: the
Acute Brain Bleeding Analysis study. Stroke
2008;39:845–849.
22 Chuang SY, Yu Y, Sheu WH, Tsai YT, Liu X,
Hsiung CA, Tsai HJ: Association of short-
term use of nonsteroidal anti-inflammatory
drugs with stroke in patients with hyperten-
sion. Stroke 2015;46:996–1003.
23 Chang CH, Shau WY, Kuo CW, Chen ST, Lai
MS: Increased risk of stroke associated with
nonsteroidal anti-inflammatory drugs: a na-
tionwide case-crossover study. Stroke 2010;
41:1884–1890.
24 Lapi F, Piccinni C, Simonetti M, Levi M, Lora
Aprile P, Cricelli I, Cricelli C, Fanelli A: Non-
steroidal anti-inflammatory drugs and risk of
cerebrovascular events in patients with osteo-
arthritis: a nested case-control study. Intern
Emerg Med 2016;11:49–59.
25 Caughey GE, Roughead EE, Pratt N, Killer G,
Gilbert AL: Stroke risk and NSAIDs: an Aus-
Neuroepidemiology 2018;51:166–176
DOI: 10.1159/000490741
tralian population-based study. Med J Aust
2011;195:525–529.
26 Fosbøl EL, Olsen AM, Olesen JB, Andersson
C, Kober L, Torp-Pedersen C, Gislason GH:
Use of nonsteroidal anti-inflammatory drugs
among healthy people and specific cerebro-
vascular safety. Intern J Stroke 2014; 9: 943–
945.
27 Haag MD, Bos MJ, Hofman A, Koudstaal PJ,
Breteler MM, Stricker BH: Cyclooxygenase
selectivity of nonsteroidal anti-inflammatory
drugs and risk of stroke. Arch Intern Med
2008;168:1219–1224.
28 Mangoni AA, Woodman RJ, Gilbert AL,
Knights KM: Use of non-steroidal anti-in-
flammatory drugs and risk of ischemic and
hemorrhagic stroke in the Australian veteran
community. Pharmacoepidemiol Drug Saf
2010;19:490–498.
29 Roumie CL, Mitchel EF Jr, Kaltenbach L, Ar-
bogast PG, Gideon P, Griffin MR: Nonaspirin
NSAIDs, cyclooxygenase 2 inhibitors, and the
risk for stroke. Stroke 2008;39:2037–2045.
30 Islam MM, Iqbal U, Walther BA, Nguyen PA,
Li YJ, Dubey NK, Poly TN, Masud JHB,
Atique S, Syed-Abdul S: Gender-based per-
sonalized pharmacotherapy: a systematic re-
view. Arch Gynecol Obstet 2017; 295: 1305–
1317.
31 Higgins JP, Green S: Cochrane Handbook for
Systematic Reviews of Interventions. Chich-
ester, John Wiley & Sons, 2011.
32 Poly TN, Islam MM, Walther BA, Yang HC,
Nguyen PA, Huang CW, Shabbir SA, Li YJ:
Exploring the association between statin use
and the risk of Parkinson’s disease: a meta-
analysis of observational studies. Neuroepi-
demiology 2017;49:142–151.
33 Ungprasert P, Matteson EL, Thongprayoon
C: Nonaspirin nonsteroidal anti-inflammato-
ry drugs and risk of hemorrhagic stroke: a sys-
tematic review and meta-analysis of observa-
tional studies. Stroke 2016;47:356–364.
34 Lee JG, Lee KB, Jang IM, Roh H, Ahn MY,
Woo HY, Hwang HW: Low glomerular filtra-
tion rate increases hemorrhagic transforma-
tion in acute ischemic stroke. Cerebrovasc Dis
2013;35:53–59.
35 Ovbiagele B, Schwamm LH, Smith EE, Grau-
Sepulveda MV, Saver JL, Bhatt DL, Hernan-
dez AF, Peterson ED, Fonarow GC: Hospital-
ized hemorrhagic stroke patients with renal
insufficiency: clinical characteristics, care
patterns, and outcomes. J Stroke Cerebrovasc
Dis 2014;23:2265–2273.
36 Party ISW: National clinical guideline for
stroke, London: Royal College of Physicians,
2012.
37 Wang Y, Rudd AG, Wolfe CD: Age and ethnic
disparities in incidence of stroke over time:
the South London Stroke Register. Stroke
2013;44:3298–3304.
38 Adamson J, Beswick A, Ebrahim S: Is stroke
the most common cause of disability? J Stroke
Cerebrovasc Dis 2004;13:171–177.
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université
175
Downloaded by:
39 Rowe FJ, Walker M, Rockliffe J, Pollock A,
Noonan C, Howard C, Glendinning R, Currie
J: Care provision and unmet need for post
stroke visual impairment. Final report 2013.
www.stroke.org.uk/sites/default/files/final_
report_unmet_need_2013.pdf?
40 Dickey L, Kagan A, Lindsay MP, Fang J, Row-
land A, Black S: Incidence and profile of inpa-
tient stroke-induced aphasia in Ontario, Can-
ada. Arch Phys Med Rehabil 2010; 91: 196–
202.
41 Engelter ST, Gostynski M, Papa S, Frei M,
Born C, Ajdacic-Gross V, Gutzwiller F, Lyrer
PA: Epidemiology of aphasia attributable to
first ischemic stroke: incidence, severity, flu-
176 Neuroepidemiology 2018;51:166–176
DOI: 10.1159/000490741
ency, etiology, and thrombolysis. Stroke 2006;
37:1379–1384.
42 Lawrence ES, Coshall C, Dundas R, Stewart J,
Rudd AG, Howard R, Wolfe CD: Estimates of
the prevalence of acute stroke impairments
and disability in a multiethnic population.
Stroke 2001;32:1279–1284.
43 Yaxley J, Litfin T: Non-steroidal anti-inflam-
matories and the development of analgesic
nephropathy: a systematic review. Ren Fail
2016;38:1328–1334.
44 De Caterina R, Ruigómez A, Rodríguez LA:
Long-term use of anti-inflammatory drugs
and risk of atrial fibrillation. Arch Intern Med
2010;170:1450–1455.
45 Bhala N, Emberson J, Merhi A, Abramson
S, Arber N, Baron J, Bombardier C, Cannon
C, Farkouh M, FitzGerald G: Vascular and
Upper Gastrointestinal Effects of Non-Ste-
roidal Anti-Inflammatory Drugs: Meta-
Analyses of Individual Participant Data
from Randomised Trials. Lancet 2013; 382:
769–779.
46 Latimer N, Lord J, Grant RL, O’Mahony R,
Dickson J, Conaghan PG: Cost effective-
ness of COX 2 selective inhibitors and tra-
ditional NSAIDs alone or in combina-
tion with a proton pump inhibitor for
­people with osteoarthritis. BMJ 2009; 339:
b2538.
134.157.146.115 - 8/29/2018 8:14:25 AM
Bibliothèque de Sorbonne Université
Islam/Poly/Walther/Yang/Lin/Li
Downloaded by: