Original Research

Pharmacoepidemiology of Benzodiazepine and

Sedative–Hypnotic Use in a Canadian General
Population Cohort During 12 Years of Follow-up
Scott B Patten, MD, PhD1; Jeanne VA Williams MSc2; Dina H Lavorato MSc2;
Aliya Kassam, PhD3; C David Sabapathy, MD4

Objective: Benzodiazepines (BDZs) and similar sedative–hypnotics (SSHs) can have both
beneficial and adverse effects. Clinical practice guidelines indicate that the course of treatment
should usually be brief (a few weeks), but patients often take these medications for longer periods
of time. We hypothesized that treatment with antidepressants (ADs) would be associated with a
shorter duration of SSHs use as mood and anxiety disorders may underlie the symptoms usually
targeted by BDZ treatment.
Method: Our study used data from a Canadian longitudinal general health study, the National
Population Health Survey, which has collected data since 1994. Data are currently available to
2006. At each interview, all medications taken in the preceding 2 days are recorded. In our study,
we used proportional hazard models to describe patterns of initiation and discontinuation of these
medications in the general population.
Results: At each interview, the frequency of BDZ–SSH use was 2% to 3%. About 1% of the
population initiated use in each 2-year follow-up period. Contrary to expectation, taking ADs
predicted initiation of BDZ–SSHs, but not discontinuation.
Conclusions: Unexpectedly, respondents taking ADs had a higher frequency of new BDZ–SSH
use. AD use may be a marker for depression severity or comorbidity, such that the observed
results may be an artifact of confounding by these factors. Irrespective of etiology, initiation of AD
treatment does not appear to negate the risk of long-term BDZ–SSH use.
Can J Psychiatry. 2010;55(12):792–799.

Clinical Implications
· Patients undergoing AD treatment have an elevated risk of long-term BDZ–SSH use.
· The association between BDZ–SSH and AD use may be related to residual symptoms such
as sleep disturbances or anxiety symptoms or may occur because AD treatment is a
marker for more severe or more highly comorbid episodes. Tolerance, abuse or
dependence, as well as AD-induced insomnia or agitation may also contribute.
· It may be valuable to incorporate BDZ–SSH discontinuation strategies proactively as a
component of treatment when these are used in combination with ADs.

Limitations
· The National Population Health Survey (NPHS) is an epidemiologic data source and cannot
discern the details of therapeutic decisions. Survey data cannot determine the
appropriateness of specific medication regimens.
· The measure of medication use in the NPHS covers only the 2 days preceding an interview
such that episodic use may sometimes appear to represent new use or discontinued use.
· The association of long-term BDZ use with AD treatment does not provide strong evidence
of a causal connection—AD treatment is better regarded as an indicator of long-term risk.

Key Words: epidemiology, longitudinal studies, sedative–hypnotic medications, mood

disorders, anxiety disorders, population studies

792 W La Revue canadienne de psychiatrie, vol 55, no 12, décembre 2010

 Pharmacoepidemiology of Benzodiazepine and Sedative–Hypnotic Use in a Canadian General Population Cohort During 12 Years of Follow-up

B enzodiazepine and SSHs have sedative, hypnotic, anx-

iolytic, muscle relaxant, and anticonvulsant properties.
In addition to their therapeutic properties, BDZ–SSHs are
objectives of our study included estimation of the incidence
and pattern of new use and the frequency and pattern of dis-
continuation of BDZ–SSH medications in relation to MDEs
associated with adverse effects, including cognitive prob- and the use of ADs.
lems,1,2 an increased risk of falls,3–6 an increased frequency of
motor vehicle accidents,7–9 and dependence and abuse. 10,11 Method
For the most part, the pharmacoepidemiologic literature is The NPHS is a longitudinal study based on a nationally repre-
restricted to estimates of the prevalence and cross-sectional sentative sample of 17 276 respondents selected from the
correlates of BDZ–SSH use.12–14 For example, the ESEMeD Canadian household population by Statistics Canada in 1994.
found that anxiety and depressive disorders, female sex, older The NPHS longitudinal cohort has been interviewed every
age, help seeking for emotional problems, low education, and 2 years since that time. The NPHS methodology is described
pain predicted BDZ use in 6 European countries.15 Similarly, in previous publications.20 The estimated response rate to
in Canada, cross-sectional correlates of BDZ use include 2006 is 77.0%.20 This response rate does not include NPHS
female sex, (higher) age, low income, low education, and the respondents who left the sampling frame (died or were insti-
presence of mood or anxiety disorders.13 Few longitudinal tutionalized) during follow-up as they were considered to
studies have examined BDZ use in Canada. have been successfully followed. The 1994 baseline inter-
views were conducted in person (75%), but 99% of follow-up
Zopiclone (a cyclopyrrolone derivative) and zaleplon (a
interviews were conducted over the phone.
pyrazolopyrimidine) are non-BDZ–SSHs that share pharma-
cological properties, adverse effects, and risks with The NPHS interview includes items covering demographics,
BDZ.10,16,17 However, these medications are used more spe- health status, health care use, and health determinants.
cifically for insomnia and less frequently for anxiety. Both Respondents were also asked to retrieve containers (boxes,
zopiclone and zaleplon are listed under Schedule IV by the bottles, tubes, or packages) of medications taken in the 2 days
International Narcotics Control Board.18 However, neither preceding an interview. Specific medications were recorded
zopiclone nor zaleplon are classified as controlled drugs in using ATC codes. These ATC codes were used in the current
Canada, whereas BDZs are. analysis to identify respondents who were taking a
Treatment with BDZ–SSHs is often initiated during treatment BDZ–SSH and to identify the use of ADs. We examined the
of depressive disorders, particularly when symptoms of anxi- incidence of new use of BDZ–SSH as the frequency of
ety are present.19 Such use is often intended to be brief, with reported use at subsequent cycles after not reporting use at an
discontinuation of the BDZ–SSH being intended when a initial cycle. A particular concern is the pattern of persistent
coprescribed AD begins to work. Some courses of BDZ–SSH BDZ–SSH use, as this may be more problematic than
treatment may become prolonged if there are residual symp- short-term use. An analysis was therefore performed requir-
toms of depression or comorbidities that lead to longer-term ing 2 consecutive cycles of new use. As the patterns of initia-
treatment. Some ADs, may cause insomnia or agitation as an tion and discontinuation may differ depending on the type of
adverse effect and this may also lead to an increased or more medication, additional analyses were performed in which
prolonged coprescribing of BDZ–SSHs. Tolerance and abuse zopiclone and zaleplon were analyzed separately from BDZ.
may also lead to prolonged BDZ–SSH exposure. The As the 2-day record of use covers a brief period of time, we
also examined (for comparative purposes) the incidence of
use of sleeping pills in the preceding month, as documented
by responses to a different item: “In the past month, did you
Abbreviations used in this article
take sleeping pills?”
AD antidepressant
ATC Anatomic Therapeutic Classification To examine trends in the frequency of BDZ–SSH use, we ini-
BDZ benzodiazepine
tially tabulated unadjusted frequencies by year of interview.
We also tabulated the types of medications used and the fre-
CIDI-SFMD Composite International Diagnostic
Interview—Short Form for Major Depression
quency of use of more than one BDZ–SSH at the same time.
To examine the impact of ADs and MDEs on initiation and
DSM Diagnostic and Statistical Manual of Mental
discontinuation of BDZ–SSHs, we used discrete (grouped)
Disorders
time proportional hazard models to estimate HRs. These
ESEMeD European Study of the Epidemiology of Mental models included both age and sex to generate estimates of the
Disorders
effect of ADs and MDE on subsequent BDZ–SSHs adjusted
HR hazard ratio for these nonmodifiable determinants of use. We also devel-
MDE major depressive episode oped models incorporating additional potential determinants
NPHS National Population on Health Study to assess possible confounding effects. As we were interested
SSH similar sedative–hypnotic
in the effects of AD use on subsequent BDZ–SSH use, AD
use was incorporated into these models as a time-varying

The Canadian Journal of Psychiatry, Vol 55, No 12, December 2010 W 793
Original Research

covariate. Where applicable, other covariates were also with no clear trend suggesting an increasing or decreasing
allowed to vary over time. These included self-reported pro- frequency of use (Figure 1). A detailed description of the dis-
fessionally diagnosed painful conditions (migraine, injury, or position of the sample is provided in Figure 2. When the sam-
back pain), employment status, education level, alcohol con- ple was restricted to 7780 respondents with completed
sumption, smoking, and income. Employment status was interviews at each of the 7 NPHS cycles from 1994 to 2006
dichotomized into working, compared with not working, sta- (this includes some respondents with partial responses), 90%
tus. Education was categorized according to whether a reported no BDZ–SSH use at any cycle. A small proportion
respondent had more or less than secondary level education. (less than 0.5%) reported use at each of the 7 NPHS cycles
Excessive alcohol consumption was classified using (Table 1). At each cycle, more than 90% of the respondents
self-report items in which consumption of 5 or more drinks on who reported taking a BDZ–SSH were taking only one such
a single occasion was recorded. Smoking was categorized by medication. The most commonly used medications were
distinguishing between current smokers and nonsmokers, lorazepam and zopiclone. In 2006, 50.6% (95% CI 44.3% to
with the latter category including both former and 56.8%) of respondents taking a BDZ–SSH were taking
never-smokers. In the analysis concerned with initiation of lorazepam and 18.9% (95% CI 14.2% to 23.5%) were taking
use, we also included a variable that was intended to represent zopiclone.
a general propensity to access health services as such propen-
sity could plausibly be a determinant of exposure to Among 15 254 potentially eligible respondents, 14 825 were
BDZ–SSH. For this purpose, we used counts of past-year vis- not taking BDZ–SSHs at the baseline interview and were
its to general practitioners or family physicians, categorizing considered eligible for analyses concerned with BDZ–SSH
these counts as 1, 2, or 3 or more visits. Terms representing initiation. At the first follow-up visit 2 years later, 1.3%
interactions between the predictive variables and time-period (95% CI 1.1% to 1.5%) reported using BDZ–SSHs, 0.9%
indicator variables were used to evaluate the proportional haz- (95% CI 0.6% to 1.2%) of men and 1.7% (95% CI 1.3% to
ards assumption. Statistical significance was evaluated using 2.1%) of women. This frequency of use increased with age:
likelihood ratio tests for the set of relevant interaction terms. from 0.6% (95% CI 0.4% to 0.9%) in the group aged 12 to 45
Other P values reported in the paper derive from Wald tests. years to 2.1% (95% CI 1.5% to 2.7%) in the group aged 46 to
65 years and 3.4% (95% CI 2.2% to 4.5%) in the group aged
The NPHS interview included the CIDI-SFMD,21 which 66 years and older category. The onset of new long-term use
assesses past-year MDE. The CIDI-SFMD is scored with a as characterized by new use at 2 consecutive cycles was
predictive probability algorithm based on the number of lower, varying between 0.3% and 0.5% at each cycle. The
symptom-based criteria fulfilled during a 2-week period in the frequency of new past-month sleeping pill use ranged
year preceding the interview. The instrument was scored at between 2.1% and 3.6% across the cycles. The pattern of
the 90% predictive probability level, indicating endorsement effect of age (older > younger) and sex (female > male) on the
of 5 symptoms. frequency of new use was comparable in analyses that exam-
ined consecutive cycle new use and new self-reported sleep-
The approach to scoring of the CIDI-SFMD is broadly consis- ing pill use. Also, inclusion or exclusion of zopiclone or
tent with the DSM-IV A criterion requirement for MDE.22 zaleplon in the outcome category did not change this pattern.
However, the CIDI-SFMD was originally validated against
DSM-IIIR criteria.21 The scoring algorithm stipulates that at No violations of the proportional hazard assumption were
least one of these symptoms must be depressed mood or loss found. In unadjusted analyses, AD use was strongly predic-
of interest or pleasure, also broadly consistent with DSM-IV. tive of initiation of BDZ–SSHs (unadjusted HR 3.8, 95% CI
2.9 to 4.9). In a proportional hazard model including MDE,
The NPHS used a stratified multistage sampling procedure. age, sex, and interaction terms, a sex by MDE interaction was
Replicate sampling weights provided by Statistics Canada identified (HR for the interaction term 0.3, 95% CI 0.1 to 0.7,
and an associated bootstrap procedure for variance estimation P = 0.006). In men, the age-adjusted HR for ADs was 4.6
were used to account for design effects. All analyses were (95% CI 1.8 to 11.6) and for MDE was 8.4 (95% CI 3.6 to
conducted using Stata23 at the Prairie Regional Data Centre on 19.7), whereas in women these HRs were 6.8 (95% CI 4.4 to
the University of Calgary Campus. Our study was approved 10.5) and 2.2 (95% CI 1.3 to 3.9), respectively. Table 2 pres-
by the University of Calgary Conjoint Health Research Ethics ents a model incorporating a broader set of covariates. As
Board. expected, age and sex were associated with initiation of
BDZ–SSH, as were pain complaints, smoking, and having 3
or more primary care visits in the past year. These adjust-
Results ments attenuated the associations of MDE and AD use with
Among the 17 276 respondents in the NPHS longitudinal file, BDZ–SSHs, but the associations remained evident and were
15 254 were considered eligible for this analysis because they statistically significant. There were no significant interac-
were aged 13 years and older at the baseline interview. Con- tions involving sex, but there was an interaction between
sistent with prior reports,13 the frequency of prevalent MDE and age, with the effect of MDE being smaller in peo-
BDZ–SSH use was between 2% and 3% at each NPHS cycle, ple aged 66 years and older. The model presented in Table 2

794 W La Revue canadienne de psychiatrie, vol 55, no 12, décembre 2010

 Pharmacoepidemiology of Benzodiazepine and Sedative–Hypnotic Use in a Canadian General Population Cohort During 12 Years of Follow-up

Figure 1 Frequency of BDZ–SSH use by cycle in the NPHSa

3.5

Frequency of BDH–SSH use, %

2.5

1.5

0.5

0
1994 1996 1998 2000 2002 2004 2006

a
Error bars represent 95% CIs deriving from a bootstrap procedure incorporating 500 replicate sampling
weights

Figure 2 Sample attrition for eligible respondents, 1994 to 2006

The Canadian Journal of Psychiatry, Vol 55, No 12, December 2010 W 795
Original Research

Figure 3 Proportion of BDZ–SSH-treated respondents in 1994 reporting use at

subsequent cyclesa

70

60
Continued use, %
50

40

30

20

10

0
1996 1998 2000 2002 2004 2006
NPHA cycle
NPHS cycle

a
Error bars represent 95% CIs deriving from a bootstrap procedure incorporating 500 replicate sampling
weights

does not include covariates that were not significantly associ- BDZ–SSH use. To assess whether this resulted in a distortion
ated with new BDZ–SSH use. Inclusion of these additional of the epidemiologic results, we modelled the determinants
variables in more complex models did not lead to substantial of new self-reported sleeping pill use, incorporating various
changes in the HRs. covariate adjustments. The results were similar to those
reported above. For example, the HR for AD use, adjusted for
With application of the persistent use definition (new use on age and sex was 2.7 (95% CI 2.1 to 3.5) and that for MDE was
2 consecutive cycles), the associations for ADs and MDE per- 2.1 (95% CI 1.7 to 2.8).
sisted with covariate adjustments. In men, after adjustment for
age, general practitioner visits, and pain (the only covariates The next part of the analysis focused on discontinuation of
continuing to attain significance in this analysis) the HR for BDZ–SSHs. The analysis of discontinuation included
AD use was 3.6 (95% CI 1.9 to 6.6) and for MDE was 5.7 429 respondents who reported taking a BDZ–SSH at the
(95% CI 3.2 to 10.1). Inclusion of additional covariates did baseline interview. The frequency with which these respon-
not substantially change the HR estimates. In women, the dents continued their BDZ–SSHs during the next 2 years was
adjusted HRs were 5.4 (95% CI 3.4 to 8.4) for ADs and 1.9 54.5% (95% CI 47.8% to 61.1%). However, among those
(95% CI 1.1 to 3.3) for MDE. Examining new BDZ use who also reported use in 1996, only 44.5% (95% CI 35.5% to
(excluding zopiclone and zaleplon), the results were broadly 53.6%) continued use during the subsequent 2-year cycle.
comparable: the age adjusted HR for AD use in men was 3.5 Figure 3 shows the proportion taking a BDZ–SSH at each
(95% CI 1.6 to 7.5), whereas MDE was 3.3 (95% CI 1.8 to cycle among people who reported use of BDZ–SSH at the
6.1). In women, the age adjusted HR for AD use was 2.5 initial interview in 1994. Although these estimates are impre-
(95% CI 1.7 to 3.7) and for MDE was 1.4 (95% CI 0.9 to 2.2), cise, the probability of discontinuation appears to decline
the latter effect did not achieve statistical significance (P = with increasing duration of use. A similar pattern was
0.16). Using zopiclone or zaleplon use as the outcome vari- observed in the analysis of self-reported sleeping pill use.
able (excluding BDZs), the age and sex adjusted HRs in this Sample size limitations precluded an analysis of the
analysis were 2.6 (95% CI 1.6 to 4.4) for AD use and 2.6 non-BDZ medications.
(95% CI 1.6 to 4.4) for MDE.
ADs are likely candidates for increasing the rate of
A concern is the possibility that the past 2-day approach to BDZ–SSH discontinuation because of their effectiveness in
measurement would result in a lack of specificity for new use the treatment of mood and anxiety disorders. A proportional
as sporadic users may be detected if they happen to take a hazard model, including an indicator variable for AD use,
medication in the 2 days preceding the interview. As noted MDE as a time-varying characteristic, sex, and age was used
above, the frequency of self-reported past-month sleeping pill to evaluate this possibility. In this model, AD use was not
use was about twice as high as that for ATC coded (past 2-day) associated with discontinuation (HR 0.8, 95% CI 0.4 to 1.5),

796 W La Revue canadienne de psychiatrie, vol 55, no 12, décembre 2010

 Pharmacoepidemiology of Benzodiazepine and Sedative–Hypnotic Use in a Canadian General Population Cohort During 12 Years of Follow-up

Table 1 Proportion taking BDZ–SSHs, by number of Discussion

NPHS cycles with reported usea
The results presented in our study extend an existing, largely
Number of cycles cross-sectional, literature about the pharmacoepidemiology
taking BDZ–SSHs Frequency (%) 95% CI (%)
of BDZ–SSH use. Prevalent use of BDZ–SSH arises as a
0 91.6 90.9–92.4 steady-state outcome of new use, discontinuation of use, and
1 4.2 3.7–4.8 associated mortality in the population. The longitudinal data
available in the NPHS cannot conclusively delineate the bal-
2 1.6 1.3–1.9
ance between these factors because the data collected do not
3 0.9 0.7–1.1
reconstruct a full record of medication use in the 2 years
4 0.6 0.4–0.8 between interviews. However, the NPHS does provide a gen-
5 0.4 0.3–0.6 eral perspective on the pattern of use. In the general house-
6 0.3 0.2–0.5 hold population not taking a BDZ–SSH, 1.0% to 1.3% were
7 0.3 0.1–0.4 found to be taking one of these medications 2 years later.
Among those taking an AD, about one-half were no longer
a
Among people providing responses at each interview cycle (n = 7780) taking them 2 years later. Among the 2% to 3% of the popula-
tion taking BDZ–SSH, a sizable proportion (about one-third)
appear to have initiated use within the preceding 2 years and a
Table 2 Proportional hazard model predicting sizable proportion (about one-half) appear to discontinue
initiation of BDZ–SSH use with adjustment for over the next 2 years. A smaller proportion have a pattern of
covariates including primary care visits highly persistent use, but these can be expected to accumulate
HR 95% CI to a great extent in the pool of prevalent BDZ–SSH users.
MDE 1.9 1.4–2.6
BDZ–SSHs are frequently used in conjunction with ADs for
AD use 2.2 1.6–3.0 symptom management early in the course of treatment for
Female 1.6 1.3–2.0 depressive disorders. ADs and BDZ–SSHs may also be used
Age 46 to 65 years 2.4 2.0–3.0 in combination for the management of anxiety disorders.
Age >66 years 5.2 4.1–6.6 Clinical expectation holds that the risk of long-term
BDZ–SSH use should be diminished if patients receive effec-
Pain complaints 1.7 1.4–2.1
tive treatment for underlying mood or anxiety disorders.
Back problems 1.3 1.1–1.7
Unexpectedly, the analysis found that respondents taking
Currently smoking 1.3 1.0–1.5a ADs had a higher risk of BDZ use, even after adjustment for
Primary care visit MDE. It is conceivable that AD side effects may contribute to
1 1.1 0.8–1.6 this pattern, as some ADs may trigger insomnia, agitation, or
2 1.3 1.0–1.9b anxiety as adverse effects. However, it is also possible that
respondents receiving treatment both with ADs and
³3 1.7 1.3–2.3
BDZ–SSH have more severe illness and that residual symp-
MDE by age >66 years interaction 0.2 0.1–0.6 toms (such as insomnia or anxiety) are more likely to persist
a
P = 0.02
after partial resolution of MDE. Another possibility is that
b
P = 0.09
even when patients respond to an AD the BDZ–SSH may not
be discontinued, either because of rebound symptoms or
because of hesitancy on the part of physicians to alter a suc-
cessful combination. Finally, ADs may sometimes be used in
an attempt to treat insomnia, which may subsequently lead to
hypnotic exposure. Such effects cannot be disentangled
using the NPHS data, but it is clear that patients taking ADs
nor was MDE (HR 0.8, 95% CI 0.5 to 1.5). Sex was also not are at an elevated risk of persistent BDZ–SSH use. In other
associated with probability of discontinuation (HR 1.1, words, ADs are an indicator of risk for subsequent
95% CI 0.7 to 1.8). HRs comparing the group aged 46 to 65 BDZ–SSH use. The NPHS has additional limitations. The
measure of MDE is a brief predictive measure that is not
years (HR 1.0, 95% CI 0.5 to 1.9) and the group aged 66 years
expected to be as accurate as the full version of the CIDI
and older (HR 1.4, 95% CI 0.7 to 2.5) to the baseline (45 years
instrument. Inaccurate measurement often diminishes the
and younger) age group did not suggest an effect of age on the strength of association observed in epidemiologic studies.24
discontinuation pattern. Nearly identical results were seen
when past-month sleeping pill use was evaluated using the Assessment of BDZ–SSH use during the 2 days preceding an
same approach. In our study, the HR for AD use was 0.8 (95% interview is a sensitive indicator of continuous use, but is not
CI 0.4 to 1.6). specific for continuous use. Some patients with sporadic use

The Canadian Journal of Psychiatry, Vol 55, No 12, December 2010 W 797
Original Research

will also be detected. Infrequent sporadic use would be less dependence, or to an offsetting of therapeutic benefits by
likely to influence the analysis than more frequent use. For AD-related adverse effects.
example, a person taking BDZ–SSH on 10% of days would
have an 81% chance of being classified as a nonuser at the Acknowledgements
baseline assessment and to therefore be considered eligible Our study was supported by a research grant from Servier
for inclusion in the initiation of use analysis. The same Canada. Servier Canada is involved in the research and
respondent has a 19% chance of being classified as a development of medication in the field of depression, creating a
BDZ–SSH user at a follow-up interview. If these events are potential conflict of interest in relation to the current project. In
addition to the grant that funded this work, Dr Patten has received
independent, the respondent would have a 15% chance of
consulting fees from Servier Canada. Servier funded the study
being misclassified as a new user. However, this would fall to without placing any restrictions on the analysis or interpretation
3% in the part of the analysis requiring 2 consecutive cycles of the data. Servier did not seek or obtain any control over
with use. While such misclassification could bias the fre- publication of the results. The analysis plan, and the plan to
quency estimates upward, it would be more likely to bias the publish the results, was formulated prior to the results being
HRs toward the null value of one through a process of dilu- known. A copy of the proposal is available from the authors upon
request. Dr Patten is a Senior Health Scholar with the Alberta
tion. The comparability of results for past-month sleeping pill
Heritage Foundation for Medical Research and Research. This
use helps to confirm that such misclassification was not a analysis was based on data collected by Statistics Canada in the
major source of bias. NPHS. However, the analyses and interpretations presented do
not reflect the views of Statistics Canada. Dr Patten received a
The NPHS does not include an assessment of anxiety disor- speaking honorarium from Lundbeck Canada and was also paid a
fee by Lundbeck for reviewing a grant submitted to that company
ders, such that the role of these disorders in determining BDZ for funding (2009). He is also a paid member of a data safety
use cannot be clarified. An additional comorbidity that is not monitoring board for an isotretinoin trial (Cipher
assessed in the NPHS is BDZ dependence. A lack of detailed Pharmaceuticals).
clinical information about these comorbid conditions means
that the results reported here should not be interpreted as an References
assessment of the clinical appropriateness of BDZ–SSH use.
For example, some patients with long-term use may achieve 1. Bramness JG, Skurtveit S, Morland J. Clinical impairment of
positive outcomes, whereas others may encounter negative benzodiazepines—relation between benzodiazepine concentrations and
impairment in apprehended drivers. Drug Alcohol Depend.
outcomes from short-term use. The estimates presented here 2002;68(2):131–141.
provide a description of the overall epidemiologic pattern 2. Hogan DB, Maxwell CJ, Fung TS, et al. Prevalence and potential consequences
of benzodiazepine use in senior citizens: results from the Canadian Study of
rather than an assessment of the appropriateness of any partic- Health and Aging. Can J Clin Pharmacol. 2003;10(2):72–77.
ular pattern of use. 3. Wagner AK, Zhang F, Soumerai SB, et al. Benzodiazepine use and hip fractures
in the elderly: who is at greatest risk? Arch Inter Med.
2004;164(14):1567–1572.
The proportion of respondents taking BDZ–SSHs who no 4. Ray WA, Griffin MR, Downey W. Benzodiazepines of long and short
elimination half-life and the risk of hip fracture. JAMA.
longer reported doing so in subsequent cycles declines with 1989;262(23):3303–3307.
the duration of use (Figure 3). This pattern is consistent with, 5. Neutel CI, Hirdes JP, Maxwell CJ, et al. New evidence on benzodiazepine use
but does not provide conclusive evidence for, the emergence and falls: the time factor. Age Ageing. 1996;25(4):273–278.
6. Maxwell CJ, Neutel CI, Hirdes JP. A prospective study of falls after
of tolerance to BDZ–SSHs. The proportion of respondents benzodiazepine use: a comparison of new and repeat use. Pharmacoepidemiol
who reported taking these medications throughout follow-up Drug Saf. 1997;6(1):27–35.
7. Neutel CI. Risk of traffic accident injury after a prescription for a
was small, suggesting that continuous multi-year use of these benzodiazepine. Ann Epidemiol. 1995;5(3):239–244.
medications is uncommon. However, it may be noted that this 8. Hemmelgarn B, Suissa S, Huang A, et al. Benzodiazepine use and the risk of
motor vehicle crash in the elderly. JAMA. 1997;278(1):27–31.
part of the analysis excluded many elderly respondents who 9. Barbone F, McMahon AD, Davey PG, et al. Association of road-traffic
were removed from the sampling frame owing to death or accidents with benzodiazepine use. Lancet. 1998;352(9137):1331–1336.
institutionalization during the 12 years of follow-up. Highly 10. Sikdar S, Ruben SM. Zopiclone abuse among polydrug users. Addiction.
1996;91(2):285–286.
persistent use may be more common in this older age group. 11. O’Connor K, Belanger L, Marchand A, et al. Psychological distress and
adaptational problems associated with discontinuation of benzodiazepines.
Addict Behav. 1999;24(4):537–541.
ADs are more decisive and effective treatments for mood and 12. Kassam A, Carter B, Patten SB. Sedative hypnotic use in Alberta. Can J
some anxiety disorders than BDZ–SSH. Expectation holds Psychiatry. 2006;51(5):287–294.
13. Kassam A, Patten SB. Hypnotic use in a population-based sample of over
that with the onset of activity of an AD, discontinuation of a thirty-five thousand interviewed Canadians. Popul Health Metr. 2006;4:15.
BDZ–SSH should usually be feasible owing to the expected 14. Rawson NS, D’Arcy C. Sedative-hypnotic drug use in Canada. Health Rep.
1991;3(1):33–57.
improvement in sleep and other symptoms. However, the
15. Demyttenaere K, Bonnewyn A, Bruffaerts R, et al. Clinical factors influencing
epidemiologic data presented here indicate that AD-treated the prescription of antidepressants and benzodiazepines: results from the
patients have an elevated risk of persistent BDZ–SSH use. It European study of the epidemiology of mental disorders (ESEMeD). J Affect
Disord. 2008;110(1–2):84–93.
will be important to determine whether this is due to problems 16. Rush CR, Frey JM, Griffiths RR. Zaleplon and triazolam in humans: acute
in service delivery (for example, failure to follow through behavioral effects and abuse potential. Psychopharmacology.
1999;145(1):39–51.
with plans to discontinue the medications), to residual symp- 17. Sikdar S, Ayonrinde O, Sampson E. Physical dependence on zopiclone. BMJ.
toms, to tolerance or rebound effects, BDZ abuse, or 1998;317(7151):146.

798 W La Revue canadienne de psychiatrie, vol 55, no 12, décembre 2010

 Pharmacoepidemiology of Benzodiazepine and Sedative–Hypnotic Use in a Canadian General Population Cohort During 12 Years of Follow-up

18. International Narcotics Control Board. Green list annex to the annual statistical
report on psychotropic substances (form P) [Internet]. 23rd ed. Vienna (AT):
International Narcotics Control Board; 2003 [cited 2010 Jul 25]. Available from: Manuscript received March 2010, revised, and accepted July 2010.
1
http://www.incb.org/pdf/e/list/green.pdf. Professor, Department of Community Health Sciences, University of
19. Dunlop BW, Davis PG. Combination treatment with benzodiazepines and SSRIs Calgary, Calgary, Alberta; Member, Hotckhiss Brain Institute, Calgary,
for comorbid anxiety and depression. Prim Care Companion J Clin Psychiatry. Alberta.
2
2008;10(222):228. Research Associate, Department of Community Health Sciences,
20. Statistics Canada. Statistics Canada National Population Health Survey University of Calgary, Calgary, Alberta.
3
Household Component Cycle 1 to 7 (1994/1995 to 2006/2007) longitudinal Postdoctoral Fellow, Mental Health Commission of Canada and
documentation. Ottawa (ON): Statistics Canada; 2009. University of Calgary, Calgary, Alberta.
4
21. Kessler RC, Andrews G, Mroczek D, et al. The World Health Organization Community Medicine Resident, Department of Community Health
Composite International Diagnostic Interview Short-Form (CIDI-SF). Int J Sciences, University of Calgary, Calgary, Alberta; Clinical Fellow,
Methods Psychiatr Res. 1998;7:171–185. Alberta Innovates, Health Solutions, Calgary, Alberta.
22. American Psychiatric Association. Diagnostic and statistical manual of mental Address for correspondence: Dr SB Patten, Department of Community
disorders. 4th ed. Text revision. Washington (DC): APA; 2000. Health Sciences, University of Calgary, 3rd Floor, TRW, 3280 Hospital
23. StataCorp. Stata [computer program]. Version 9.0. College Station (TX): Stata Drive NW, Calgary, AB T2N 4Z6; [email protected]
Corporation; 2005.
24. Kleinbaum DG, Kupper LL, Morgenstern H. Information bias. Epidemiologic
research. New York (NY): Van Nostrand Reinhold; 1982. p 220–241.

Résumé : Pharmaco-épidémiologie de l’utilisation des benzodiazépines et

des sédatifs-hypnotiques dans une cohorte de la population générale
canadienne durant 12 ans de suivi
Objectif : Les benzodiazépines (BDZ) et les sédatifs-hypnotiques semblables (SHS) peuvent avoir
des effets tant bénéfiques qu’indésirables. Les guides de pratique clinique indiquent que le cours du
traitement devrait habituellement être bref (quelques semaines), mais les patients prennent souvent
ces médicaments pendant de plus longues périodes. Nous avons émis l’hypothèse que le traitement
aux antidépresseurs (AD) serait associé à une utilisation de SHS de plus courte durée car les
troubles de l’humeur et anxieux peuvent être sous-jacents aux symptômes habituellement ciblés par
le traitement aux BDZ.
Méthode : Notre étude a utilisé les données d’une étude longitudinale canadienne sur la santé
générale, l’Enquête nationale sur la santé de la population (ENSP), qui recueille des données depuis
1994. Les données présentement disponibles sont celles de 2006. À chaque entrevue, tous les
médicaments pris dans les 2 jours précédents ont été enregistrés. Dans notre étude, nous avons
utilisé des modèles à risques proportionnels pour décrire les tendances de l’initiation et de la
cessation de ces médicaments dans la population générale.
Résultats : À chaque entrevue, la fréquence d’utilisation des BDZ–SHS était de 2 % à 3 %. Environ
1 % de la population initiait l’utilisation à chaque période de suivi de 2 ans. Contrairement à nos
prévisions, le fait de prendre des AD prédisait l’initiation des BDZ–SHS, mais pas la cessation.
Conclusions : Étonnamment, les répondants prenant des AD avaient une fréquence plus élevée
d’utilisation nouvelle des BDZ–SHS. L’utilisation d’AD peut être un marqueur de la gravité ou de la
comorbidité de la dépression, de sorte que les résultats observés peuvent être un artéfact de
confusion par ces facteurs. Sans égard à l’étiologie, l’initiation du traitement aux AD ne semble pas
annuler le risque d’une utilisation à long terme des BDZ–SHS.

The Canadian Journal of Psychiatry, Vol 55, No 12, December 2010 W 799