Psychiatry and Clinical Neurosciences 2014; 68: 255–262
12117
Regular Article
Benzodiazepine use and risk of stroke: A retrospective
population-based cohort study
Wei-Shih Huang, MD,1,4 Chih-Hsin Muo, MSc,3 Shih-Ni Chang,
Chon-Haw Tsai, MD1,4 and Chia-Hung Kao, MD2,4*
Departments of 1Neurology, 2Nuclear Medicine and PET Center, China Medical University Hospital, 3Management Office
for Health Data, 4Graduate Institute of Clinical Medical Science, School of Medicine, College of Medicine, China Medical
University, Taichung, and 5Department of Health Promotion and Health Education, National Taiwan Normal University,
Taipei, Taiwan
Aim: The aim of this study was to investigate the
possible association between benzodiazepine (BZD)
use and risk of incident stroke by utilizing data from
2000 to 2003 from the National Health Insurance
system of Taiwan.
Methods: Study subjects consisted of 38 671 patients
with new BZD use and 38 663 people without BZD
use who were frequency-matched for age, sex and
baseline comorbidity with BZD users. All subjects
had no history of stroke. Each study patient’s case was
followed until a new diagnosis of stroke was made or
until the patient was censored by loss to follow up,
death, or termination of insurance. The study lasted
until the end of 2009. A Cox proportional hazards
regression model was used to estimate the incidences
and hazard ratios (HR) of stroke.
TROKE IS THE second leading cause of death
S globally1 and the leading cause of disability in
adults.2 Identification of potential risk factors is one
of the most important methods for preventing the
recurrence of stroke.
For the general population, benzodiazepines
(BZD) are one of the most frequently prescribed
classes of drugs. Among older adults, BZD use ranges
*Correspondence: Chia-Hung Kao, MD, Graduate Institute of
Clinical Medical Science, School of Medicine, College of Medicine,
China Medical University, No. 2, Yuh-Der Road, Taichung 404,
Taiwan. Email: d10040@mail.cmuh.org.tw
Received 12 June 2013; revised 12 September 2013; accepted 18
September 2013.
© 2013 The Authors
Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
doi:10.1111/pcn.
MSc,3 Yen-Jung Chang, PhD,3,5
Results: The HR of hemorrhagic stroke was signifi-
cantly lower in the BZD group when compared with
the non-BZD group. For patients aged 20–39 years,
the HR of ischemic stroke was significantly higher
in the BZD group when compared with the non-BZD
group. Compared to the non-BZD group, patients
with a lower annual dosage (<1 g) or duration (<30
days) of BZD use had a lower risk of stroke in the elder
group (P < 0.0001) and patients with a higher annual
dosage (≥ 4 g) or duration (≥ 95 days) of BZD use had
a higher risk of stroke in all age groups (P < 0.0001).
Conclusions: Our findings may suggest neuroprotect-
ion under lower-dosage BZD use and neurotoxicity
under higher-dosage BZD use.
Key words: benzodiazepine, cohort study, stroke.
from 10% to 42% worldwide.3–6 Cheng et al.7 found
that the rate of BZD use among Taiwanese older
adults was even higher: the 1-year-period prevalence
was approximately 43%. The relation between BZD
use and subsequent risk of stroke remains controver-
sial. Some studies have indicated that BZD use may
improve acute stroke outcomes through GABAergic
stimulation,8,9 whereas other studies have indicated
that there are greater risks of stroke in BZD users
because of possible GABAA-mediated neurochemical
mechanisms.10–13 Because BZD are so commonly pre-
scribed, even a small risk accompanying their use
could have important clinical implications and also
spark the interest of the public. A large population-
based study has the potential to clarify the relation
255
256 W.-S. Huang et al.
between BZD and stroke. We therefore used the
National Health Insurance (NHI) database of Taiwan
to investigate the relation between exposure to
BZD and the occurrence of stroke in the Taiwanese
population.
METHODS
Data sources
This study was based on reimbursement data from
the Taiwanese NHI system, which contained all
medical claims from the system. The NHI system has
provided affordable health care for all of the island’s
residents since 1996. From its inception, the NHI
system has covered health care for more than 96% of
the total Taiwanese population and maintains con-
tractual relations with more than 97% of all clinics
and hospitals in Taiwan. As of 2007, more than 99%
of the population was enrolled in the insurance
program. For research and administrative purposes,
the National Health Research Institute (NHRI) from
the Department of Health built several randomly
selected claim databases representative of the entire
population (generally referred to as NHIRD).
The data we used were a sub-data set from the
NHIRD containing a record of all reimbursement
claims from 1996 to 2009 for 1 000 000 randomly
selected patients representative of the population
(approximately 5% of all insurants). Details on this
population-based cohort database have previously
been published.14 Diagnoses were coded according to
the ICD-9-CM.
We confirm that all data were de-identified and
analyzed anonymously. In addition, this study was
also approved by the Ethics Review Board at China
Medical University (CMU-REC-101-012).
Study patients
Patients who newly underwent treatment for at least
2 months between the years 2000 and 2003 were
defined as the study group (BZD group). We excluded
patients with histories of stroke before the index date,
as well as those younger than 20 years. ‘Index date’ in
this study refers to the first date of BZD prescription.
A total of 38 671 patients were extracted from the
database to form the BZD group.
For each BZD group patient identified, approxi-
mately one insurant with no record of BZD treatment
© 2013 The Authors
Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
Psychiatry and Clinical Neurosciences 2014; 68: 255–262
and no history of stroke was randomly selected from
the same index month and year. They were then
frequency-matched to each BZD case according to age
(5-year intervals), sex and baseline comorbidity,
including coronary artery disease (CAD), diabetes,
hypertension and hyperlipidemia. These insurants
were designated the non-BZD comparison cohort
(non-BZD group).
Based on the ICD-9-CM classification, we searched
for baseline comorbid conditions in study patients,
including CAD (ICD-9-CM: 410–414), hypertension
(ICD-9-CM: 401–405), diabetes mellitus (ICD-9-CM:
250), and hyperlipidemia (ICD-9-CM: 272). We also
considered the ischemic stroke prevention medica-
tion, including aspirin, clopidogrel and warfarin.
Those medications were used before the beginning of
the study.
Each patient’s case was followed until the follow-
ing outcomes: a new diagnosis of stroke (ICD-9-CM
430–438; ischemic stroke, ICD-9-CM 433–438;
hemorrhagic stroke, ICD-9-CM 430–432); censor-
ship for loss to follow-up, death, or latest withdrawal
from the NHI system; or until the end of the
follow-up period on 31 December 2009 (whichever
came first).
Annual duration and dosage levels for BZD in the
exposure cohort were stratified by 25th percentile:
<0.3, 0.3–0.9, 1.0–3.9 and ≥ 4.0 g for annual dosage
and <15, 15–29, 30–94 and ≥ 95 days for annual
duration.
Statistical analysis
The χ2-test was used to compare distributions of sex,
age, and comorbidities between BZD and non-BZD
groups. The incidence densities of stroke were calcu-
lated for each cohort. The BZD-to-non-BZD hazard
ratios (HR) of stroke with 95% confidence interval
(CI) for each variable (including overall risk of stroke
and stroke-subtype-specific risk) were estimated using
univariate and multivariate Cox proportional hazards
regression models. For future projections, we esti-
mated the incidences and HR of stroke according to
comorbidity. The association between annual dosage/
duration for BZD use and stroke was also assessed. All
analyses were performed using SAS 9.1 (SAS Institute,
Cary, NC, USA). A significance level of 0.05 was con-
sidered statistically significant. In stratified analyses,
the significance level was set at 0.0045 for avoiding the
type I error.
Psychiatry and Clinical Neurosciences 2014; 68: 255–262 Benzodiazepine use and risk of stroke 257

0.94–1.09) and adjusted HR of 0.97 (95%CI = 0.90–

RESULTS
A total of 77 334 patients ≧20 years of age, including
38 663 patients with no BZD exposure and 38 671
patients with BZD exposure were included in our
analysis (Table 1). Female patients accounted for the
majority of the BZD group (55.4%). More than half of
the BZD group (55.3%) was between 20 and 49 years
of age. The BZD group had the proportionally highest
percentages of hypertension (29.5%), and then hyper-
lipidemia (15.1%), CAD (12.6%), and diabetes mel-
litus (9.54%). Compared with the non-BZD group,
the BZD group likely received aspirin (8.19% vs
7.26%, P < 0.0001) (Table 1).
Table 2 shows overall and age-specific incidence
densities and HR for stroke according to sex and
BZD-use status. Overall, the incidence of stroke was
not statistically lower in the non-BZD group (5.44 per
1000 person-years) than in the BZD group (5.60 per
1000 person-years) with a crude HR of 1.01 (95%CI =
1.05) after adjusting for aspirin use. Furthermore,
stratified analysis by age showed that the risk of
stroke increased significantly for BZD patients aged
20–39 years (adjusted HR = 1.99, 95%CI =1.32–
3.01), but decreased for BZD patients aged 60–69
years (adjusted HR = 0.74, 95%CI =0.64–0.85)
(Table 2). HR stratified by sex showed the same trend
in both men and women. In particular, for men aged
60–69 years, the BZD group had a significantly
decreased risk of stroke compared with the non-BZD
group (HR = 0.70, 95%CI =0.59–0.84).
Table 3 shows the association of BZD use with risks
of different stroke subtypes according to age. The HR
of hemorrhagic stroke was lower in the BZD group
when compared with the non-BZD group (HR = 0.77,
95%CI = 0.64–0.92). For patients aged 20–39 years,
the HR of developing ischemic stroke was 2.53-fold
higher for the BZD group (95%CI = 1.48–4.35).
Conversely, for patients in the BZD group aged 40

Table 1. Baseline characteristics between BZD group and non-BZD group in 2000–2003

BZD

No Yes
N = 38 663 N = 38 671
Variables n % n % P-value
Sex
Female 21 438 55.5 21 439 55.4
Male 17 225 44.6 17 232 44.6
Age, years
20–39 11 918 30.8 11 921 30.8
40–49 9 486 24.5 9 486 24.5
50–59 7 072 18.3 7 072 18.3
60–69 5 673 14.7 5 673 14.7
≥ 70 4 514 11.7 4 519 11.7
Means (SD) 48.9 (15.8) 49.1 (15.7)
Comorbidity
CAD 4 850 12.5 4 854 12.6
Diabetes 3 687 9.54 3 691 9.54
Hypertension 11 399 29.5 11 404 29.5
Hyperlipidemia 5 844 15.1 5 849 15.1
Medication
Aspirin 2 806 7.26 3 168 8.19 <0.0001
Clopidogrel 45 0.12 61 0.16 0.12
Warfarin 143 0.37 173 0.45 0.09
χ2-test and t-test.
BZD, benzodiazepine; CAD, coronary artery disease.

© 2013 The Authors

Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
258 W.-S. Huang et al. Psychiatry and Clinical Neurosciences 2014; 68: 255–262

Table 2. Comparisons of incidence density of stroke between BZD group and non-BZD group by characteristics

BZD

No Yes HR (95%CI)
Variables Cases Person-years Rate† Cases Person-years Rate† Crude Adjusted
All
20–39 31 73 996 0.42 86 96 731 0.89 2.03 (1.34–3.06)* 1.99 (1.32–3.01)
40–49 113 54 397 2.08 179 77 062 2.32 1.05 (0.83–1.33) 1.03 (0.81–1.31)
50–59 190 36 680 5.18 295 55 906 5.28 0.98 (0.82–1.18) 0.97 (0.81–1.17)
60–69 356 26 179 13.60 474 43 266 10.96 0.75 (0.65–0.86)* 0.74 (0.64–0.85)*
≥ 70 446 17 584 25.36 661 29 546 22.37 0.86 (0.76–0.97) 0.85 (0.75–0.96)
Overall 1136 208 837 5.44 1695 302 511 5.60 1.01 (0.94–1.09) 0.97 (0.90–1.05)
Women
20–39 8 40 984 0.20 28 56 085 0.50 2.33 (1.06–5.13) 2.30 (1.05–5.06)
40–49 39 30 702 1.27 74 46 418 1.59 1.17 (0.79–1.73) 1.17 (0.79–1.73)
50–59 66 19 990 3.30 117 32 538 3.60 1.04 (0.77–1.42) 1.03 (0.76–1.40)
60–69 129 13 140 9.82 209 23 284 8.98 0.84 (0.67–1.05) 0.83 (0.66–1.04)
≥ 70 196 7 689 25.49 291 13 524 21.52 0.81 (0.68–0.98) 0.81 (0.67–0.97)
Overall 438 112 506 3.89 719 171 849 4.18 1.05 (0.93–1.18) 1.01 (0.90–1.14)
Men
20–39 23 33 011 0.70 58 40 645 1.43 2.00 (1.23–3.24)* 1.95 (1.20–3.17)
40–49 74 23 695 3.12 105 30 644 3.43 1.05 (0.78–1.41) 1.02 (0.76–1.38)
50–59 124 16 690 7.43 178 23 368 7.62 0.99 (0.79–1.25) 0.99 (0.78–1.24)
60–69 227 13 039 17.41 265 19 982 13.26 0.72 (0.60–0.86)* 0.70 (0.59–0.84)*
≥ 70 250 9 895 25.26 370 16 023 23.09 0.90 (0.77–1.06) 0.89 (0.78–1.04)
Overall 698 96 331 7.25 976 130 662 7.47 1.02 (0.92–1.12) 0.97 (0.88–1.07)
*P < 0.005.
†
Per 1000 person-year.
BZD, benzodiazepine; CI, confidence interval; HR, hazard ratio, compared to non-BZD group.
Adjusted for aspirin use.

years or older, the hemorrhagic stroke (HR = 0.70,

95%CI =0.64–0.92) was lower than in the non-BZD
group (Table 3).
Young people without comorbidities in the BZD
group in particular had a 2.80-fold higher risk of
stroke than those in the non-BZD group (95%CI =
1.67–4.71). Overall, BZD patients with CAD or hyper-
tension had lower risks of stroke than patients in the
non-BZD group, and these trends carried through to
older adult patients (Table 4).
The association between the annual dosage/
duration levels for BZD and stroke is presented in
Table 5. Compared to the non-BZD group, subjects
who annually received under the median dosage
(<1 g) or duration (<30 days) had a significantly lower
risk of stroke. Conversely, those who received over the
first quartile dosage or duration had a significantly
higher risk. There was the same trend in those aged 40
years or older.
DISCUSSION
Compared to the non-BZD group, patients with BZD
use in our study had a lower risk of hemorrhagic
stroke, especially in subjects older than 40 years
(P < 0.005, Table 3). Conversely, there was a higher
risk of ischemic stroke in patients aged 20–39 years
who used BZD as compared with the non-BZD group
(P < 0.005, Table 3). Patients with a lower annual
dosage (<1 g) or duration (<30 days) of BZD use had
a significantly lower risk of stroke in the elder group
and patients with a higher annual dosage (≥4 g) or
duration (≥95 days) of BZD use had a significantly
higher risk of stroke in all age groups (P < 0.0001,
Table 5).
To our knowledge, this study was the first
population-based investigation of this group of
38 671 Taiwanese patients who use BZD. For the
comparison group, we randomly frequency-matched

© 2013 The Authors

Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
Psychiatry and Clinical Neurosciences 2014; 68: 255–262 Benzodiazepine use and risk of stroke 259

Table 3. Comparisons of incidence density of stroke between BZD group and non-BZD group among stroke subtype

BZD

No Yes HR (95%CI)
Age group Cases Rate† Cases Rate† Crude Adjusted
Ischemic stroke
20–39 years-old 17 0.23 60 0.62 2.60 (1.52–4.46)* 2.53 (1.48–4.35)*
≥ 40 years-old 900 6.67 1382 6.72 0.98 (0.90–1.07) 0.95 (0.87–1.03)
Overall 917 4.39 1442 4.77 1.07 (0.98–1.16) 1.02 (0.94–1.11)
Hemorrhagic stroke
20–39 years-old 14 0.19 26 0.27 1.34 (0.70–2.57) 1.34 (0.70–2.56)
≥ 40 years-old 205 1.52 227 1.10 0.71 (0.59–0.86)* 0.70 (0.58–0.85)*
Overall 219 1.05 253 0.84 0.78 (0.65–0.94) 0.77 (0.64–0.92)*
*P < 0.005.
†
Per 1000 person-year.
BZD, benzodiazepine; CI, confidence interval; HR, hazard ratio.
Adjusted for aspirin use.

each patient with BZD use to a person from the cohort
without BZD use according to age, sex, and index
calendar year.
Two points should be mentioned before further
discussion. First, BZD, a class of psychoactive drugs
for medical problems, such as anxiety, insomnia, and
convulsion, are recognized as one of the most widely
prescribed medications worldwide.15 Second, physi-
cians in Taiwan prescribe BZD for anxiety and insom-
nia, which is similar to global practice. Our subjects

Table 4. Comparisons of incidence density of stroke between BZD group and non-BZD group among comorbidity
Overall 20–39 years old ≥40 years old

Rate† / Case Rate†/ Case Rate†/ Case

Variable Non-BZD BZD HR (95%CI) Non-BZD BZD HR (95%CI) Non-BZD BZD HR (95%CI)
‡
Non-comorbidity 1.84/256 2.23/415 1.16 (0.99–1.35) 0.27/18 0.80/69 2.80 (1.67–4.71)* 3.30/238 3.49/346 1.00 (0.85–1.18)
CAD‡
No 4.32/820 4.52/1204 0.99 (0.90–1.08) 0.38/28 0.85/81 2.12 (1.38–3.25)* 6.78/792 6.53/1123 0.91 (0.83–1.00)*
Yes 16.47/316 13.70/491 0.80 (0.70–0.93)* 2.63/3 2.56/5 0.76 (0.48–3.23) 17.35/313 14.34/486 0.80 (0.69–0.92)*
Diabetes‡
No 4.49/866 4.70/1298 0.99 (0.91–1.08) 0.39/28 0.89/84 2.15 (1.40–3.30)* 6.98/838 6.70/1214 0.91 (0.83–1.00)
Yes 16.70/270 15.01/397 0.85 (0.72–0.99) 2.17/3 1.06/2 0.49 (0.08–2.95) 18.06/267 16.08/395 0.84 (0.72–0.98)
Hypertension‡
No 2.28/362 2.75/599 1.13 (1.00–1.29) 0.28/20 0.84/77 2.78 (1.70–4.55)* 3.88/342 4.16/522 1.01 (0.88–1.15)
Yes 15.53/774 12.91/1096 0.80 (0.73–0.88)* 3.53/11 1.89/9 0.53 (0.22–1.28) 16.33/763 13.56/1087 0.80 (0.73–0.88)*
Hyperlipidemia‡
No 4.81/875 4.88/1256 0.96 (0.88–1.05) 0.38/27 0.84/78 2.11 (1.36–3.27)* 7.65/848 7.15/1178 0.89 (0.82–0.97)
Yes 9.75/261 9.75/439 0.93 (0.80–1.08) 1.46/4 2.03/8 1.29 (0.39–4.29) 10.69/257 10.49/431 0.91 (0.78–1.06)
Medication use
No 4.67/918 4.87/1358 1.02 (0.93–1.11) 0.38/28 0.85/80 2.12 (1.37–3.26)* 7.19/890 6.91/1278 0.93 (0.85–1.01)
Yes 17.87/218 14.31/337 0.79 (0.67–0.94) 2.47/3 2.16/6 0.83 (0.21–3.31) 19.57/215 15.93/331 0.81 (0.68–0.96)

*P < 0.005.
†
Per 1000 person-year.
‡Adjusted for aspirin use.

BZD, benzodiazepine; CI, confidence interval; HR, hazard ratio.

Medication included aspirin, clopidogrel and warfarin.

© 2013 The Authors

Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
260 W.-S. Huang et al. Psychiatry and Clinical Neurosciences 2014; 68: 255–262

Table 5. Comparisons of incidence density of stroke among dosage and duration of benzodiazepine

Overall 20–39 years old ≥40 years old

† † †
Variable Cases Rate HR (95%CI) Cases Rate HR (95%CI) Cases Rate HR (95%CI)
None 1136 5.44 1.00 (reference) 31 0.42 1.00 (reference) 1105 8.19 1.00 (reference)
Annual dosage, g
<0.3 167 2.06 0.37 (0.31–0.43)*** 16 0.57 1.26 (0.69–2.30) 151 2.87 0.34 (0.29–0.40)***
0.3–0.9 289 3.90 0.68 (0.59–0.77)*** 11 0.47 1.04 (0.52–2.08) 278 5.50 0.64 (0.56–0.73)***
1.0–3.9 445 5.94 1.02 (0.91–1.20) 24 1.08 2.43 (1.42–4.14)** 421 7.97 0.92 (0.82–1.03)
≥ 4.0 794 10.96 1.86 (1.70–2.04)*** 35 1.55 3.49 (2.15–5.67)*** 759 15.25 1.73 (1.58–1.90)***
Annual duration, days
<15 87 1.04 0.18 (0.15–0.23)*** 8 0.26 0.58 (0.27–1.26) 79 1.47 0.17 (0.14–0.22)***
15–29 241 3.42 0.60 (0.52–0.69)*** 16 0.68 1.55 (0.84–2.83) 225 4.77 0.56 (0.48–0.65)***
30–94 458 5.76 0.99 (0.89–1.10) 24 1.00 2.23 (1.31–3.81)** 434 7.83 0.90 (0.81–1.01)
≥ 95 909 13.27 2.22 (2.04–2.43)*** 38 2.01 4.57 (2.84–7.35)*** 871 17.54 1.98 (1.81–2.16)***
*P < 0.05, **P < 0.001, ***P < 0.0001.
†
Per 1000 person-year.
CI, confidence interval; HR, hazard ratio.
Adjusted for aspirin use.
Annual dosage and duration stratified by quartile.

were retrieved from a random sample of 187 413
people enrolled in Taiwan’s NHI program.16
BZD enhance the effects of the neurotransmitter
gamma-aminobutyric acid (GABA), which results in
sedative, hypnotic, anxiolytic, anticonvulsant, muscle
relaxant, and amnesic functions. GABA is the main
inhibitory neurotransmitter in the central nervous
system and acts by reducing depolarization-induced
and ischemia-induced glutamate release.17,18 GABA
can trigger hyperpolarization of neurons through
anion channels (GABAA) and presynaptic G-protein-
coupled receptors (GABAB).19 This hyperpolarization
counteracts the depolarization, which is the initiating
event in the biochemical ischemic cascade.20 Enhanc-
ing the activity of the GABA-ergic system may be
neuroprotective in acute stroke and result in a signifi-
cantly better outcome21 because of its inhibiting
effects on presynaptic glutamate release, the activities
of N-methyl-D–aspartate (NMDA) and non-NMDA
receptors, voltage-gated ion channels, and perhaps
peri-infarct depolarizations.22,23 GABA-ergic activa-
tion reduces the metabolic respiratory rate, preserv-
ing glucose and reducing acidosis, and facilitates
local cerebral blood flow.24 GABA receptor agonists
can induce hypothermia, which is also regarded as a
neuroprotective condition for acute stroke.25,26 To test
these theories, a multicenter, randomized, stratified,
double-blind, placebo-controlled clinical human
trial was conducted. This 2006 trial reported that
diazepam treatment in acute stroke resulted in lower
early recurrence rates of stroke in infarct patients, but
the results were not statistically significant.9 Our
study showed a lower risk of hemorrhagic stroke in
any aged patients with BZD use (P < 0.001), espe-
cially in patients older than 40 years (P < 0.0001,
Table 3). We also found that patients who annually
received lower dosage (<1 g) or duration (<30 days)
BZD had a significantly lower risk of stroke in the
elder group (P < 0.0001, Table 5). These results
support the above theories.
Conversely, some studies have indicated that BZD
use may possibly increase the risk of stroke. Animal
studies revealed that both blood flow and glucose use
decreased following administration of the GABA
agonist in the conscious rats10 and blockade of GABAA
receptors in rats led to a significant increase in regional
cerebral blood flow (rCBF) in the focal ischemic
area.11 An experimental challenge trial with the seda-
tive midazolam, a non-selective BZD that potentiates
the activity of GABA at the GABAA receptor, has shown
temporary re-emergence of many elements of a clini-
cal syndrome that had been precipitated in the acute
state after a focal brain infarct but had undergone
considerable remission in the days to years after.12
Kozma et al. found that their BZD control group had
significantly greater odds of stroke-related events
when compared with patients using risperidone (odds
ratio of 1.96; P = 0.001) and patients using all atypical
antipsychotics (odds ratio of 2.05, P < 0.001).13 In
another placebo-controlled study of risperidone, the

© 2013 The Authors

Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
Psychiatry and Clinical Neurosciences 2014; 68: 255–262
rate of stroke was twice as common in the risperidone-
treated group (4%) as in the placebo group (2%).27
Similarly, in placebo-controlled trials of olanzapine,
the rate of stroke was 1.3% in the olanzapine-treated
group and 0.4% in the placebo group.28 Patients recov-
ering from stroke who use drugs from the BZD class
have been reported to have a poorer recovery of sen-
sorimotor29 and upper-extremity30 functions. Gold-
stein31 documented the risks accompanying the use of
diazepam and certain other medications by people
who have had a stroke and suggests that BZD should
generally be avoided in certain stroke patients.32 Some
patients who recovered from neurological deficits
demonstrated a transient reemergence of syndromes
after administration of anesthesia.33–35 Our study
showed that BZD use would significantly increase risk
of ischemic stroke in younger patients aged 20–39
years and patients who annually received higher
dosage (≥4 g) or duration (≥95 days) BZD had a
significantly higher risk of stroke in all age groups
(P < 0.0001, Table 5). These results support the above
theories.
Because of the population-based design of this
study, we were able to achieve a high level of inherent
representativeness in this study. Furthermore, the data
on BZD use and stroke diagnoses used in this study
were extremely reliable. However, this study has limi-
tations. First, detailed information, such as smoking
habits, alcohol consumption, body mass index, socio-
economic status, and family history, were not avail-
able from the NHIRD. All of these variables are major
risk factors for stroke and could plausibly be associ-
ated with BZD prescription. However, because the
NHIRD covers nearly the entire population of Taiwan
with a universal reimbursement policy, it is unlikely
that these variables influenced the decisions of physi-
cians to prescribe BZD. Second, evidence derived from
a cohort study is generally of lower quality than that
derived from randomized control trials. The cohort
study design is subject to many biases related to con-
founding adjustment. Despite our meticulous study
design with adequate controls over confounding
factors, bias may have existed in unmeasured or
unknown confounders. Third, diagnoses from the
NHIRD claims were made primarily for administra-
tive billing purposes and were not verified scientifi-
cally. We were unable to contact patients directly on
their BZD use because entries were anonymous and
identified only by number. Finally, our data did
not include prescriptions for BZD that occurred
prior to 1996. Therefore, cumulative dosage of BZD
Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
Benzodiazepine use and risk of stroke 261
may have been underestimated, weakening observed
associations.
Conclusions
Compared to the non-BZD group, this retrospective
population-based cohort study showed a signifi-
cantly lower risk of hemorrhagic stroke in all age
patients, especially in subjects older than 40 years
(P < 0.0001) and a significantly higher risk of isch-
emic stroke in patients aged 20–39 years (P < 0.0001)
with BZD use. There was also a BZD-dose–response
phenomenon. Compared to the non-BZD group,
patients with a lower annual dosage (<1 g) or dura-
tion (<30 days) of BZD use had a significantly lower
risk of stroke in the elder group and patients with a
higher annual dosage (≥ 4 g) or duration (≥ 95 days)
of BZD use had a significantly higher risk of stroke in
all age groups. This may suggest neuroprotection
under lower-dosage BZD use and neurotoxicity under
higher-dosage BZD use. The mechanisms of these
effects, whether due to neuroendocrine responses,
stress hormone levels or sympathetic tone, remain a
question for further research.
ACKNOWLEDGMENTS
This work was supported by ‘Aim for the Top Univer-
sity Plan’ of the National Chiao Tung University
and Ministry of Education, Taiwan, R.O.C., National
Science Council (NSC 99–2314-B-039-016-MY2),
the Taiwan Department of Health Clinical Trial and
Research Center of Excellence (DOH102-TD-B-111-
004), the Taiwan Department of Health Cancer
Research Center of Excellence (DOH102-TD-C-111-
005), and the China Medical University Hospital
(DMR-96-010). The authors declare that they have no
conflict of interest or financial interest invested in this
work, either collectively or individually.
REFERENCES
1. Di Carlo A. Human and economic burden of stroke. Age.
Ageing 2009; 38: 4–5.
2. Leys D. Atherothrombosis: A major health burden.
Cerebrovasc. Dis. 2001; 11 (Suppl 2): 1–4.
3. Madhusoodanan S, Bogunovic OJ. Safety of benzodiaz-
epines in the geriatric population. Expert Opin. Drug Saf.
2004; 3: 485–493.
4. Hogan DB, Maxwell CJ, Fung TS, Ebly EM. Prevalence and
potential consequences of benzodiazepine use in senior
© 2013 The Authors
262 W.-S. Huang et al.
citizens: Results from the Canadian study of health and
aging. Can. J. Clin. Pharmacol. 2003; 10: 72–77.
5. Zandstra SM, Furer JW, van de Lisdonk EH et al. Different
study criteria affect the prevalence of benzodiazepine use.
Soc. Psychiatry Psychiatr. Epidemiol. 2002; 37: 139–144.
6. Alvarenga JM, Loyola Filho AI, Firmo JO, Lima-Costa MF,
Uchoa E. Prevalence and sociodemographic characteristics
associated with benzodiazepines use among community
dwelling older adults: The Bambui Health and Aging
Study (BHAS). Rev. Bras. Psiquiatr. 2008; 30: 7–11.
7. Cheng JS, Huang WF, Lin KM, Shih YT. Characteristics
associated with benzodiazepine usage in elderly outpa-
tients in Taiwan. Int. J. Geriatr. Psychiatry 2008; 23: 618–
624.
8. Lodder J, Heuts-van Raak L, Kessels F. GABAergic stimula-
tion by benzodiazepines at stroke onset may ameliorate
functional outcome in cardioembolic stroke patients.
Cerebrovasc. Dis. 1996; 6: 118.
9. Lodder J, van Raak L, Hilton A, Hardy E, Kessels A. Diaz-
epam to improve acute stroke outcome: Results of the
early gaba-ergic activation study in stroke trial. A random-
ized double-blind placebo-controlled trial. Cerebrovasc.
Dis. 2006; 21: 120–127.
10. Kelly PT, McCulloch J. The effects of the GABAergic
agonist muscimol upon the relationship between local
cerebral blood flow and glucose utilization. Brain Res.
1983; 258: 338–342.
11. Chi OZ, Hunter C, Liu X, Chi Y, Weiss HR. Effects of
GABA(A) receptor blockade on regional cerebral blood
flow and blood-brain barrier disruption in focal cerebral
ischemia. J. Neurol. Sci. 2011; 301: 66–70.
12. Lazar RM, Fitzsimmons BF, Marshall RS et al. Reemergence
of stroke deficits with midazolam challenge. Stroke 2002;
33: 283–285.
13. Kozma C, Engelhart L, Long S et al. No evidence for rela-
tive stroke risk in elderly dementia patients treated with
risperidone versus other antipsychotics. Proceedings of the
Annual Meeting of the American Psychiatric Association;
2004;May 1–6, New York.
14. Kao CH, Sun LM, Su KP et al. Benzodiazepine use possibly
increases cancer risk: A population-based retrospective
cohort study in Taiwan. J. Clin. Psychiatry 2012; 73: e555–
e560.
15. Ballenger JC. Benzodiazepines. In: Schatzberg AF,
Nemeroff CB (eds). Textbook of Psychopharmacology. The
American Psychiatric Press, Arlington, 1995; 215–225.
16. Fang SY, Chen CY, Chang IS, Wu EC, Chang CM, Lin KM.
Predictors of the incidence and discontinuation of long-
term use of benzodiazepines: A population-based study.
Drug Alcohol Depend. 2009; 104: 140–146.
17. Nelson RM, Green AR, Lambert DG, Hainsworth AH. On
the regulation of ischemia-induced glutamate efflux from
rat cortex by GABA; in vitro studies with GABA,
clomethiazole and pentobarbitone. Br. J. Pharmacol. 2000;
130: 1124–1130.
© 2013 The Authors
Psychiatry and Clinical Neurosciences © 2013 Japanese Society of Psychiatry and Neurology
Psychiatry and Clinical Neurosciences 2014; 68: 255–262
18. Vaishnav A, Lutsep HL. GABA agonist: Clomethiazole.
Curr. Med. Res. Opin. 2002; 18 (Suppl 2): s5–s8.
19. Wilby MJ, Hutchinson PJ. The pharmacology of
chlormethiazole: A potential neuroprotective agent. CNS
Drug Rev. 2004; 10: 281–294.
20. Tuttolomondo A, Di Sciacca R, Di Raimondo D
et al. Neuron protection as a therapeutic target in acute
ischemic stroke. Curr. Top. Med. Chem. 2009; 9: 1317–
1334.
21. Wahlgren NG, Bornhov S, Sharma A et al. The
Clomethiazole Acute Stroke Study (class): efficacy results
in a subgroup of 545 patients with total anterior circula-
tion syndrome. J. Stroke Cerebrovasc. Dis. 1999; 8: 231–
239.
22. Tecoma ES, Choi DWGAB. A-ergic neocortical neurons are
resistant to NMDA receptor-mediated injury. Neurology
1989; 39: 676–682.
23. Hossmann KA. Periinfarct depolarisations. Cerebrovasc.
Brain Metab. Rev. 1996; 8: 195–208.
24. Edvinsson L, Krause DN. Pharmacological characteriza-
tion of GABA receptors mediating vasodilation of verebral
arteries in vitro. Brain Res. 1979; 173: 89–97.
25. Klassman L. Therapeutic hypothermia in acute stroke. J.
Neurosci. Nurs. 2011; 43: 94–103.
26. Visser SA, Pozarek S, Martinsson S, Forsberg T, Ross SB,
Gabrielsson J. Rapid and long-lasting tolerance to
clomethiazole-induced hypothermia in the rat. Eur.
J. Pharmacol. 2005; 512: 139–151.
27. Wooltorton E. Risperidone (risperdal): increased rate of
cerebrovascular events in dementia trials. CMAJ 2002;
167: 1269–1270.
28. Wooltorton E. Olanzapine (zyprexa): Increased incidence
of cerebrovascular events in dementia trials. CMAJ 2004;
170: 1395.
29. Goldstein LB, Matchar DB, Morgenlander JC, Davis JN.
Drugs influence the recovery of function after stroke.
Stroke 1990; 21: 179.
30. Goldstein LB. Common drugs may influence motor recov-
ery after stroke. Neurology 1994; 44 (Suppl 2): A289.
31. Goldstein LB. Potential effects of common drugs on stroke
recovery. Arch. Neurol. 1998; 55: 454–456.
32. Goldstein LB. Influence of common drugs and related
factors on stroke outcome. Curr. Opin. Neurol. 1997; 10:
52–57.
33. Lazar RM, Marshall RS, Pile-Spellman J et al. Anterior
translocation of language in patients with left cerebral
arteriovenous malformation. Neurology 1997; 49: 802–
808.
34. Warach S, Gaa J, Siewert B, Wielopolski P, Edelman RR.
Acute human stroke studied by whole brain echo planar
diffusion-weighted magnetic resonance imaging. Ann.
Neurol. 1995; 37: 231–241.
35. Fazekas F, Fazekas G, Schmidt R, Kapeller P, Offenbacher
H. Magnetic resonance imaging correlates of transient
cerebral ischemic attacks. Stroke 1996; 27: 607–611.