Download New Evidence on Benzodiazepine Use and Falls: The Time Factor

Age and Ageing 1996.25:273-278
New Evidence on Benzodiazepine Use and
Falls: The Time Factor
C. INEKE NEUTEL, JOHN P. HIRDES, COLLEEN J. MAXWELL,
SCOTT B. PATTEN
Summary
The objectives of this prospective study were to calculate incidence rates for fall-related hospitalization, to
compare the effect of risk factors among benzodiazepine (BZD) users and unexposed controls, and to examine
variations in risks according to length of time following a BZD prescription. Data were derived from
Saskatchewan Health linked data bases, leading to information on 468 hospitalizations for injury due to falls
among a study population of 321 422.
Incidence rates per 10000 within 28 days of the prescription fill date were 26.2, 12.1 and 9.0 for BZD
sedative users, BZD tranquillizer users and for unexposed controls, respectively. Incidence rates increased with
age, and were higher for women than for men. Results from multivariate logistic regression models also showed
a greater risk of falling for BZD users but the odds ratio was higher for men than for women. A history of
treatment for alcohol abuse was a very strong risk factor for falls among both men (odds ratio, 10.7) and women
(odds ratio, 4.3).
The highest risk of serious injury due to falls was within 15 days of filling the prescription, with an odds ratio
of 3.6 for BZD sedatives and 2.6 for BZD tranquillizers. Risk decreased with further increase of time after the
BZD fill date. For the individual BZD, flurazepam and triazolam showed the highest increase in risk with odds
ratios of 3.4 and 2.7, respectively, while oxazepam, lorazepam and diazepam showed odds ratios of 2.2, 2.0 and
1.8 (all odds ratios mentioned are statistically significant at p < 0.05).
Introduction
Injuries due to falls represent an important public
health concern for all ages, but particularly for elderly
people. Falls are the second most important cause of
injury-related mortality for all age groups, but the
leading cause for those aged over 75. Falls also
represent the most important injury-related reason for
hospital admissions for all ages [1—4]. Identification of
modifiable risk factors [5, 6] could lead to the
development of preventive measures that may increase
quality of life, reduce premature mortality and lower
health care costs.
Risk factors for falls include advanced age, problems
in motor control, and a variety of chronic and acute
illnesses [7]. Increasing evidence suggests that benzodiazepine (BZD) use represents an important iatrogenic
risk factor for older adults in both community and
institutional settings [8-13]. The use of tranquillizers
and sedatives increases substantially among older age
groups, so that more than 10% of men and 20% of
women over 65 report using tranquillizers and/or
sleeping pills [14, 15]. Clinically, non-pharmacological
treatments are available for insomnia (e.g. sleep hygiene
measures) and anxiety disorders (e.g. behavioural
therapy). An understanding of the risks associated
with BZD treatment is essential for informed riskbenefit analysis of the various treatment options.
The associations between individual BZDs and risk
of falls or fractures have been the subject of several
studies in the past few years. BZDs with long
elimination half-life were found to be associated with
a higher risk for falls than BZDs with a short
elimination half-life [16, 17]. Another study found
that the shorter-acting BZDs have a similar increased
risk for hip fracture (e.g. 3.5 for temazepam use [12]).
Also, several studies have demonstrated an increased
risk of falling in older persons who are receiving
multiple medications [18-20].
However, the published literature has some important limitations. Many retrospective studies are based
on subject recall [21]. Subjects who fall may be more
likely to remember drug exposures, since they are
motivated to seek an explanation for the fall. It is also
often difficult to establish the temporal relationship
between BZD exposure and falls. The strength of the
present study is that it is designed to address both of
these limitations. It has a controlled cohort study
design and allows for prospective analysis of fall-related
injury after BZD use. In addition, the study permits an
C. I. NEUTEL ET AL.
274
evaluation of the effect of time lapse between a BZD
exposure in a previously unexposed person and the risk
of falls. The objective of this study is to evaluate the
association of a first BZD prescription with subsequent
hospitalization for injury due to falls.
Methods
The Saskatchewan Health Databases were developed for the
province's health care delivery and payment claim systems,
but they have also become valuable for population research.
The databases used for the present study are the Health
Insurance Registration file, the Prescription Drug Claim and
the Hospital In-Patient Database [22].
The study design is a cohort or prospective study in which
the exposed cohort, people who received a prescription for
BZD, are compared with the unexposed cohort, a sample
selected from the Saskatchewan population. The population
included only adults over 20 years old without a BZD
prescription in the prior 6 months, from among those eligible
for coverage under the Saskatchewan Health Plan (over 95%
of the Saskatchewan population). The exposed population
represents persons who filled a prescription for a BZD
sedative (triazolam or flurazepam) or a BZD tranquillizer
(oxazepam, lorazepam or diazepam) between 1979 and 1986.
These five BZD covered by far the majority of the BZD use
(over 90%) in Saskatchewan at this time. The data collection
for this study was discontinued after 1986 because changes in
the administration of the drug plan resulted in the drug data
being temporarily incomplete. Because this study relates an
individual prescription to a potential adverse event, changes
in prescribing patterns in more recent years should not affect
the conclusions of this study.
To obtain an unexposed group comparable with the BZD
exposed population, several criteria were used in the selection
of control subjects. Specifically, the population of the
Registration File was stratified by sex and 10-year age
groups and two control subjects were randomly selected
from the appropriate stratum for each triazolam user, so that
the age/sex distribution of the control subjects was similar to
that of the triazolam group. In order to control for seasonal
factors, members of the control group were provided with an
artificial prescription date with the same seasonal distribution
as the prescription date of members of the BZD users' groups.
These dates were used as the start of the follow-up period (up
to 60 days) to track hospital admissions for injury due to falls.
As with the exposed group, only control subjects who had not
received a prescription for one of the study BZDs in the last 6
months before the artificial prescription dates were selected.
Further details regarding the sample methodology used in the
present study have been provided in an earlier publication
[23]. Table I provides the BZD and comparison group sizes
and the number of falls occurring in those groups.
The data collected for each individual included age, sex, use
of concomitant drugs (other tranquillizers, sedatives, narcotic
analgesics, antipsychotics, antidepressants and anticonvulsants)
over the 30 days prior to the prescription date, history of
treatment for alcohol/drug abuse over the preceding 12 months
and the presence or absence of social assistance at the time of the
BZD prescription date. The history of treatment for drug or
alcohol abuse was determined by previous hospitalization for
treatment of those conditions. The outcome variable of injuries
due to falls is based on hospital records, using ICD-9 codes
E880 to E888 [24]. Within the context of this study, all
mentions of falls, injuries due to falls, or hospitalization, refer to
hospitalization for injuries due to falls. The diagnosis used is the
discharge diagnosis, but the date which determines whether an
admission is within the follow-up period of 60 days is the
hospital admission date. In case of multiple falls in the followup period, only the first hospitalization is included in the
analysis.
Analysis: Age-specific or age-adjusted incidence rates were
calculated for hospital admissions for falls within 4 weeks of
filling a BZD prescription. This allows for a comparison of the
BZD sedative and BZD tranquillizer users with the
unexposed population for men and women of different age
groups as well as the calculation of relative risks adjusted for
the age distribution of the comparison groups. The exposed
and control subjects were categorized by 10-year age groups.
The standard population needed for the age adjustment
calculation was obtained by summation of the populations of
the three comparison groups: BZD sedatives, BZD tranquillizers, and unexposed controls. Multiple logistic regression
models, including all relevant risk factors available in the
linked Saskatchewan Health Databases (e.g. age, other drug
use, receipt of social assistance, alcohol abuse), were
calculated separately for men and women aged 60 years and
older to examine the independent effects associated with BZD
sedative and tranquillizer use. All anlayses were performed
using SAS-version 6 (SAS Institute Inc., Cary NC).
Results
Table II presents age-adjusted and age-specific rates
Table I. Study population size, and number of hospital admissions for falls within 4 weeks after the first BZD prescription, by age
and sex
BZD sedatives
Total
Men
Women
Age (years)
20-59
60-79
80+
BZD tranquillizers
Control group
No. of
subjects
No. of
falls
No. of
subjects
No. of
falls
No. of
subjects
No. of
falls
77184
31161
46 023
202
178
131
146 684
53 938
92 746
64
114
97 554
38910
58 644
88
22
66
41876
28 033
7275
36
73
93
100960
38 874
6850
28
29
121
45713
39 544
12 297
43
41
71
4
NEW EVIDENCE ON BENZODIAZEPINE USE AND FALLS
275
Table II. Age-adjusted and age-specific rates per 10 000, unadjusted relative risks* and risk differences'1 for falls within 4 weeks
after the first BZD prescription, by age and sex
Unadjusted 1isk estimates
Rates of hospitalization for falls
10000
Total
Men
Women
Age (years)
20-59
60-79
80+
BZD
BZD
sedatives
tranquillizers
26.2
22.8
28.5
12.1
11.9
12.3
per
Unexposed
controls
BZD sedative group
vs. controls
BZD tranquillizer group
vs. controls
Relative
risk
Risk
difference
Relative
risk
17.2
17.1
1.3
2.1
1.1
3.1
6.2
1.0
7.1
1.6
2.6
4.9
6.2
11.3
2.9
4.0
2.5
9.0
5.7
17.2
8.6
5.7
0.8
10.8
7.8
25.5
127.8
16.9
86.1
10.7
33.3
2.4
3.8
14.8
94.5
Risk
difference
52.8
* The relative risk for this table was calculated by dividing the exposed rate by the unexposed rate.
The relative risk difference was calculated by subtracting the unexposed rate from the exposed rate.
for falls for BZD sedative users, BZD tranquillizer
users and the control group. Hospitalization rates are
highest for those exposed to BZD sedatives, followed
by users of BZD tranquillizers and lowest for the
control group. Among the three age groups, individuals
80 years and over show a much higher rate of
hospitalization due to falls than the other two groups.
Table II also presents the unadjusted relative risks and
risk differences for the comparison groups. The BZD
sedative exposed men have a risk of hospitalization due
to falls 4.0 times that of unexposed men, while similarly
exposed women have a risk of 2.5 times that of the
unexposed women. Among the three age groups, the
relative risks are highest for the under-60-year age
group, with risk estimates of 10.8 and 7.1 for those
exposed to BZD sedatives and tranquillizers, respectively. Conversely, the older subjects (aged 80 years and
older) have the highest risk difference at 94.5 per 10000
and 52.8 per 10 000 for those exposed to BZD sedatives
and tranquillizers, respectively.
Table III shows adjusted odds ratios for risk of
falls by BZD use and other variables for the population
60 years and older, stratified by sex. The odds ratios
for BZD sedative use in this table are consistent
with the relative risks reported in Table II. Specifically, compared with unexposed controls, male BZD
sedative users had approximately a four-fold increase in
risk while female BZD sedative users were approximately twice as likely to experience a fall-related
hospitalization. For both men and women, the
adjusted odds ratios associated with BZD tranquillizer
use were somewhat lower than those observed for BZD
sedative use. Use of antipsychotics and other sedatives
was associated with a statistically significantly increased
Table III. Adjusted odds ratiosf for hospitalization for falls within 4 weeks of filling a first BZD prescription among men and
women aged 60 years and over
Odds ratio (95% CL)
Variables
Men (n* = 110)
Women (n1 = 261)
BZD sedatives
BZD tranquillizers
Age (10-year intervals)
Antipsychotics
Other sedatives
Anticonvulsants
Antidepressants
History of alcohol/drug abuse
Social assistance recipient
4.0* (2.4-6.6)
2.5* (1.4-4.3)
2.6* (2.0-3.4)
1.5(0.8-2.9)
1.1 (0.6-1.9)
1.6(0.6-4.5)
0.4(0.1-1.2)
10.7* (5.4-21.0)
0.8 (0.3-2.3)
2.3* (1.7-3.2)
1.6* (1.2-2.3)
2.9* (2.4-3.4)
2.7* (2.0-3.8)
1.6* (1.2-2.1)
1.1 (0.6-2.3)
1.3 (0.9-1.9)
4.3* (1.3-13.6)
1.2(0.8-1.9)
f Obtained from multiple logistic regression models, adjusted for all variables listed above.
In this and subsequent tables: • Statistically significant at p < 0.05; *n = number of falls.
C. I. NEUTEL ET AL.
276
Table IV. Adjusted odds ratiosf for hospitalization for falls by
time after filling a first BZD prescription among subjects aged
60 years and over
Odds ratio (95% CL)
Time after
BZD (days)
n
BZD sedatives
BZD tranquillizers
<15
15-28
29-60
219
152
312
3.6* (2.5-5.2)
2.3* (1.6-3.4)
1.4* (1.1-1.9)
2.6* (1.7-3.7)
1.4(0.9-2.1)
1.1 (0.9-1.5)
f Obtained from multiple logistic regression models, also
adjusted for age, sex and concomitant use of other sedatives.
risk of hospitalization for falls for women, but not for
men. A history of treatment for alcohol and drug abuse
was strongly associated with risk of falls for both men
and women, with a particularly high odds ratio for men
(10.7).
Table IV presents adjusted odds ratios for risk of
hospitalization for falls by length of time after first
prescription. The odds ratios are highest for the first 2
weeks after the prescription was filled and then decrease
with increasing length of follow-up for both BZD
sedatives and tranquillizers.
Table V presents adjusted odds ratios for fall-related
hospitalizations associated with the individual BZD.
The highest odds ratios were found for flurazepam (3.4)
and for triazolam (2.7). The three BZDs categorized as
tranquillizers—diazepam, lorazepam, and oxazepam—
had odds ratios ranging from 1.8 to 2.2.
Discussion
The present study is among the first to report incidence
rates for fall-related hospitalizations after a first
prescription for a BZD. Both relative risks and odds
ratios showed a strong association between BZD
exposure and subsequent falls (Tables II and III).
This association is particularly strong in the period
Table V. Adjusted odds ratiosf for hospitalization for falls
within 4 weeks after a first BZD prescription among subjects
aged 60 years and over
Odds ratios (95% CL)
(n = 371)
Triazolam
Flurazepam
Diazepam
Lorazepam
Oxazepam
2.7* (2.0-3.6)
3.4* (2.5-4.7)
1.8* (1.3-2.5)
2.0* (1.3-3.1)
2.2* (1.4-3.4)
f Obtained from multiple logistic regression model, also
adjusted for age, sex and concomitant use of other sedatives.
shortly after the receipt of the BZD prescription (Table
IV) and is not accounted for by the various potential
confounding variables examined here.
The present findings report new information not
discussed elsewhere in the literature, but there are some
shortcomings to be noted. Exposure to BZD is based on
filling a prescription at a pharmacy. Some of the BZD
group may not have consumed the BZD while a few of
the control group may have obtained BZD from
aquaintances. However, the rate of misclassification is
highly unlikely to differ depending on whether falls
occurred. Biases based on non-differential misclassification would tend to decrease the observed differences
between the comparison groups, leading to a more
conservative estimate of effect (i.e. a lower relative risk
or odds ratio). Another limitation relates to the paucity
of information on each patient, including details on the
indications for BZD use and measures of the patient's
physical and mental health status. The data presented
in Table III address comorbidity to some extent, but
alternative explanations may not be accounted for
entirely. Conditions that lead to BZD use might also
increase the risk of hospitalization due to falls.
Comorbidity is a problem that increases with age and
might lead to a greater propensity to falls.
The present study offers several advantages for the
study of BZD use and risk of falls. The data used in this
analysis have the advantage of a reliable outcome
measure, as detection rate for hospitalization due to
falls is very high, probably complete, and the specificity
for diagnoses is also high. That is, it is unlikely that
coders would favour falls as the diagnosis for BZD
users over non-users, because the coders could not be
aware of BZD use. Also, the prospective nature of the
data allows for the establishment of a temporal order
between BZD use and falls. The higher odds ratio
associated with the shorter follow-up period after BZD
exposure may reflect the greater likelihood of current
BZD use. Use of BZD is likely to decrease over time,
with or without adverse side effects [25] and this may
explain a weaker association with falls over time.
Another explanation for the reduced odds ratio with
longer follow-up could be an increase in tolerance of the
effects of BZD.
The present results are consistent with evidence
reported in the literature examining a variety of
outcomes for falls. The strength of the association in
this study is somewhat higher than that of other studies
[8-11], but this could be explained by the closer time
relation between BZD use and the falls in this study.
This is especially likely in the light of the data of Table
IV. Table V shows the risk for individual BZD which is
somewhat at variance with two other studies which
found that long-acting BZD (e.g. diazepam, flurazepam) had a higher risk while the shorter-acting BZD
(e.g. triazolam, lorazepam, oxazepam) were associated
with little or no risk [12, 16, 17]. The present study
shows that flurazepam and triazolam have the higher
risk, although the three other BZD studied also have a
statistically significant increase in risk.
NEW EVIDENCE ON BENZODIAZEPINE USE AND FALLS
Some of the covariates also show some interesting
results. The strong association for history of treatment
for alcohol abuse shown in this study is consistent with
some other publications [26], while other studies failed
to find an association between alcohol use and fallrelated injuries [27, 28]. The present study suggests
that there may be a long-term risk for falls related to
drug and alcohol abuse, because not all those with a
history of abuse would continue to have that problem.
The odds ratios shown here are sufficiently high to
warrant further research.
An evaluation of the number of falls attributable to
BZD (and thus, the impact of BZD use) was made by
subtracting the incidence of falls in BZD non-users
from that in BZD users. Table II presents the risk
differences between the exposed and unexposed
comparison groups. Given the assumption of a causal
relation between BZD use and falls, these data suggest
that for those 80 years of age and older exposed to BZD
sedatives, 94.5 falls per 10 000 exposed could have been
prevented by not giving the BZD sedatives. Similarly,
52.8 per 10 000 could have been prevented in this age
group exposed to BZD tranquillizers.
Because of the numbers of falls that could potentially
be prevented in a population that is very vulnerable to
major disability because of falls, it is important to
continue to study the elderly population in order to
elucidate the causal mechanisms at work. Intervention
studies examining alternatives to BZD use are of the
highest priority. Falls have repeatedly been shown to
represent a serious public health problem and BZDs are
consistently implicated as major risk factors. Intervention studies can advance our understanding of the
mechanism linking BZD use with falling and suggest
effective ways of dealing with this problem.
Acknowledgement
This study is based in part on data provided by the
Saskatchewan Department of Health. The interpretation
and conclusions contained herein do not necessarily represent those of the Government of Saskatchewan or the
Saskatchewan Department of Health.
References
1. Tait H, Brazeau-Meahan K. Injuries and seniors: the
Canadian context. Ottawa: Statistics Canada, 1993.
2. Baker SP, Harvey AH. Fall injuries in the elderly. Clin
Geriatr Med 1985;1:5O1-12.
3. Sattin RW, Lambert HDA, Devito CA, et al. The
incidence of fall injury events among the elderly in a
denned population. Am J Epidemiol 1990;131:1028-37.
4. Herings RMC. Geneesmiddelen als determinant van
ongevallen. University of Utrecht and Ministerie van
VWS, 1994.
5. Hornbrook MC, Stevens VS, Wingfield DJ, Hollis JF,
Greenlick MR, Ory MG. Preventing falls among
community-dwelling older persons: results from a
randomized trial Gerontologist 1994;34:16-23.
6. Overstall PW. Falls. Rev Clin Gerontol 1992;2:31-8.
277
7. Horak FB, Shupert CL, Mirka A. Components of
postural dyscontrol in the elderly: a review. Neurobiol
Aging 1989;10:727-38.
8. Ray WA, Griffin MR, Schafmer W, Baugh DK, Melton
CJ. Psychotropic drug use and the risk of hip fracture. N
EnglJMed 1987;316:363-9.
9. Tinetti ME, Speechley M, Ginter SF. Risk factors for
falls among elderly persons living in the community. N
EnglJMed 1988;319:1701-7.
10. Gryfe CI, Amies A, Ashley MJ. A longitudinal study of
falls in an elderly population: I. Incidence and morbidity.
Age and Ageing 1977;6:201-10.
11. Ryynanen OP, Kivela SK, Honkanen R, Laippala P,
Saano V. Medication and chronic diseases as risk factors
for falling injuries in the elderly. Scand J Soc Med
1993;4:264-71.
12. dimming RG, Klineberg RJ. Psychotropics, thiazide
diuretics and hip fractures in the elderly. Med J Aust
1993;158:414-17.
13. Sorock GS, Shimkin EE. Benzodiazepine sedatives and
the risk of falling in a community-dwelling elderly
cohort. Arch Int Med 1988;148:2441-4.
14. Quinn K, Baker M, Evans B. Who uses prescription drugs?
Results from a population-wide study in Saskatchewan.
Saskatchewan Health 1993;90.
15. Laurier C, Dumas J, Gregoire J-P, Duval L. L'Utilisations des tranquillants, sedatifs et somniferes: analyse des
donnees de I'enquete Sante Quebec. Sante Quebec, 1990.
16. Ray WA, Griffin MR, Downey, W. Benzodiazepines of
long and short elimination half-life and the risk of hip
fracture. JAMA 1989;262:3303-7.
17. Cummings SR, Nevitt MC, Browner WS, et al. Risk
factors for hip fracture in white women. N Engl J Med
1995;332:767-73.
18. Blake AJ, Morgan K, Bendall MJ, et al. Falls by elderly
people at home: prevalence and associated factors. Age
Ageing 1988;17:365-72.
19. Campbell AJ, Borrie MJ, Spears GF. Risk factors for falls
in a community-based prospective study of people 70
years and older. J Gerontol 1989;44:112-17.
20. Prudham D, Grimley Evans J. Factors associated with
falls in the elderly: a community study. Age Ageing
1981,10:141-6.
21. Campbell AJ. Drug treatment as a cause of falls in old age:
a review of the offending agents. Drugs Aging
1991;l:289-302.
22. Malcolm E, Downey W, Strand LM, McNutt M, West
R. Saskatchewan Health linkable data bases and
pharmacoepidemiology. PMS 1993;6:175-264.
23. Neutel CI, Downey W, Senft D. Medical events after a
prescription for a benzodiazepine. Pharmacoepidemiol
Drug Safety 1995;4:63-73.
24. World Health Organization. Manual of the international
statistical classification of diseases, injuries, and causes of
death. 9th edn. Geneva, 1977.
25. Weintraub M. Compliance in the elderly. Clin Geriatr
Merfl990;6:445-52.
26. Heather N. Alcohol, accidents, and aggression. Br MedJ
1994;308:1254.
27. Nelson DE, Sattin RW, Langolis JA, DeVito CA,
Stevens JA. Alcohol as a risk factor for fall injury
events among the elderly persons in the community.
J Am Geriatr Soc 1992;40:658-61.
28. Adams WL, Magruder-Habib K, Trued S, Broome HL.
Alcohol abuse in elderly emergency department patients.
J Am Geriatr Soc 1992;+0:1236-40.
C. I. NEUTEL ET AL.
Authors' addresses
C. I. Neutel
Risk Management & Regulatory Affairs,
Drugs Directorate,
Health Canada,
Ottawa, Ontario K1A 0L2, Canada
J. P. Hirdes,
Department of Health Studies and Gerontology,
University of Waterloo and Grand River Hospital,
3570 King Street East,
Kitchener,
Ontario N2A 2W1, Canada
C. J. Maxwell
Clinical Epidemiology Unit,
SCOHS-Elisabeth Bruyere Pavilion,
43 Bruyere Street,
Ottawa, Ontario, KIN 5C8,
Canada
S. B. Patten
Department of Psychiatry,
Calgary General Hospital,
Calgary, AJberta, T2N 4N1
Canada
Received in revised form 23 November 1995