Antipsychotic Drug Use and Mortality in Older Adults with Dementia
Background: Antipsychotic drugs are widely used to manage be-
havioral and psychological symptoms in dementia despite concerns
about their safety.
Objective: To examine the association between treatment with
antipsychotics (both conventional and atypical) and all-cause mor-
tality.
Design: Population-based, retrospective cohort study.
Setting: Ontario, Canada.
Patients: Older adults with dementia who were followed between
1 April 1997 and 31 March 2003.
Measurements: The risk for death was determined at 30, 60, 120,
and 180 days after the initial dispensing of antipsychotic medica-
tion. Two pairwise comparisons were made: atypical versus no
antipsychotic use and conventional versus atypical antipsychotic
use. Groups were stratified by place of residence (community or
long-term care). Propensity score matching was used to adjust for
differences in baseline health status.
Results: A total of 27 259 matched pairs were identified. New use
of atypical antipsychotics was associated with a statistically signifi-
cant increase in the risk for death at 30 days compared with
nonuse in both the community-dwelling cohort (adjusted hazard
ratio, 1.31 [95% CI, 1.02 to 1.70]; absolute risk difference, 0.2
percentage point) and the long-term care cohort (adjusted hazard
ratio, 1.55 [CI, 1.15 to 2.07]; absolute risk difference, 1.2 percent-
age points). Excess risk seemed to persist to 180 days, but unequal
rates of censoring over time may have affected these results. Rel-
ative to atypical antipsychotic use, conventional antipsychotic use
was associated with a higher risk for death at all time points.
Sensitivity analysis revealed that unmeasured confounders that in-
crease the risk for death could diminish or eliminate the observed
associations.
Limitations: Information on causes of death was not available. Many
patients did not continue their initial treatments after 1 month of
therapy. Unmeasured confounders could affect associations.
Conclusions: Atypical antipsychotic use is associated with an in-
creased risk for death compared with nonuse among older adults
with dementia. The risk for death may be greater with conventional
antipsychotics than with atypical antipsychotics.
Sudeep S. Gill, MD, MSc; Susan E. Bronskill, PhD; Sharon-Lise T. Normand, PhD; Geoffrey M. Anderson, MD, PhD; Kathy Sykora, MSc; Kelvin Lam, MSc; Chaim M. Bell, MD, PhD; Philip E. Lee, MD; Hadas D. Fischer, MD; Nathan Herrmann, MD; Jerry H. Gurwitz, MD; and Paula A. Rochon, MD, MPH. Antipsychotic Drug Use and Mortality in Older Adults with Dementia. Ann Intern Med. 2007;146:775-786.
Various challenging behavioral and psychological symp-
toms commonly develop in older adults with demen-
tia and predispose them and their caregivers to poor out-
comes (1). Nonpharmacologic strategies are recommended
as first-line management for these symptoms (2), but they
may be difficult to implement in clinical practice (3). For
many reasons, antipsychotic medications are routinely pre-
scribed in this setting (4, 5). Conventional antipsychotics,
such as haloperidol, have been available since the 1950s.
Meta-analyses of clinical trials evaluating conventional
antipsychotics to treat agitation in dementia show that
these agents have modest efficacy and important adverse
effects compared with placebo (6, 7). In the past decade,
use of newer “atypical†antipsychotics has been rapidly in-
creasing in clinical practice because these agents were
thought to produce fewer adverse effects than conventional
agents (2). A Canadian study found that the prevalence of
antipsychotic use in older adults increased from 2.2% in
1993 to 3.0% at the end of 2002. In that study, atypical
antipsychotics, which were unavailable in 1993, accounted
for 82.5% of all antipsychotics dispensed in 2002 (8).
Short-term randomized, controlled trials (RCTs) have
studied the role of atypical antipsychotics in the manage-
ment of behavioral and psychological symptoms of demen-
tia (2, 9). An RCT involving 421 outpatients with Alzhei-
mer disease and psychosis, aggression, or agitation concluded
that the adverse effects of these newer drugs offset their
advantages (10). As a result, improvements in behavioral
symptoms with antipsychotic drug treatment do not nec-
essarily lead to improvements in overall quality of life for
patients or their caregivers (11).
____
Context
Recent reports suggest that antipsychotics are associated
with increased risk for death in patients with dementia.
Contribution
This large, population-based study from Canada assessed
the risk for death after dispensation of antipsychotics in
older adults with dementia. New use of antipsychotics
compared with nonuse was associated with increased risk
for death at 30 days. Conventional agents were associated
with higher risks than were atypical agents.
Caution
Sensitivity analyses showed that unmeasured confounders
might diminish or erase observed associations.
Implication
Both conventional and atypical antipsychotics may be
associated with an increased risk for death in elderly
persons with dementia.
—The Editors
____
In April 2005, the U.S. Food and Drug Administra-
tion (FDA) issued a public health advisory that the use of
atypical antipsychotics to treat elderly patients with de-
mentia was associated with an increased risk for death
compared with placebo (12). In June 2005, Health Canada
issued a similar warning and additional data (13). These
warnings stem from reviews of RCTs that involve the atyp-
ical agents risperidone, olanzapine, quetiapine, and aripi-
prazole. The mortality rate was approximately 1.6 to 1.7
times higher than with placebo and was greater with anti-
psychotics than with placebo in 15 of the 17 trials reviewed
by the U.S. FDA (12). The warnings extend to all cur-
rently available atypical antipsychotics. Other publications
have provided support for these warnings and have raised
further safety concerns about older conventional antipsy-
chotics (14–16).
Important questions remain unanswered. Although
RCTs provide the best evidence of treatment efficacy and
harm, the individual RCTs in this case had low event rates.
Reliable estimates of the mortality risk were generated only
when data were combined by meta-analysis (14). Further-
more, these RCTs were generally short in duration and
could not provide information about the long-term effect
of antipsychotics on mortality (14, 17). Finally, these trials
provide estimates of harm primarily for atypical antipsy-
chotics. Relatively few data are available on harms associ-
ated with older conventional antipsychotics. Studies suggest
that important differences may exist in the safety profiles of
conventional and atypical agents (15, 16, 18, 19).
Using population-based data, we sought to determine
the risk for all-cause mortality in older adults with demen-
tia who received atypical antipsychotics, conventional anti-
psychotics, or no antipsychotic. Because important baseline
differences exist among these groups, we used propensity
score matching to improve their comparability. We also
evaluated the effect of duration of treatment with antipsy-
chotics on the risk for death.
METHODS
Data Sources
Ontario is Canada’s most populous province. During
our study, Ontario had a population of approximately 12
million people, of whom 1.4 million were 65 years of age
or older. A universally funded health program covers nearly
all physician services, medications, and hospital services for
patients 65 years of age or older in Ontario. Information
from 4 administrative health care databases was linked to
develop the study cohort: pharmacy records from the On-
tario Drug Benefit program, hospitalization records from
the Canadian Institute for Health Information Discharge
Abstract Database, physician billing information for in-
patient and outpatient services from the Ontario Health
Insurance Plan, and basic demographic information and
vital statistics from the Registered Persons Database. We
used encrypted unique identifiers that are common among
databases to link anonymous information on demographic
characteristics and health services utilization for patients in
our study. Little basic information on patients is missing in
these databases. For example, the coding accuracy of drug
claims in the Ontario Drug Benefit program database is
excellent, with an error rate of only 0.7% (20).
The study was approved by the ethics review board of
Sunnybrook and Women’s College Health Sciences Cen-
tre, Toronto, Ontario, Canada.
Dementia Cohort
We identified a cohort of all Ontario residents 66
years of age or older with a diagnosis of dementia (in the
Ontario Health Insurance Plan or Discharge Abstract Data-
base) between 1 April 1997 and 31 March 2002. To focus
on antipsychotic drug treatment for behavioral and psy-
chological symptoms of dementia, we excluded patients
who had evidence of other psychotic disorders (such as
schizophrenia) or were receiving palliative care services. To
reduce the potential for selection bias, we studied only new
users of antipsychotics and excluded those who had re-
ceived antipsychotics in the year before cohort entry (21).
Exposure to Antipsychotics
We identified new use of antipsychotics if any agent
available through the Ontario Drug Benefit program was
dispensed after cohort entry. Cohort entry (that is, the
index date) was defined as the date of the first dispensed
antipsychotic drug. Available atypical drugs included olan-
zapine, quetiapine, and risperidone, and available conven-
tional drugs included chlorpromazine, flupenthixol, flu-
phenazine, haloperidol, loxapine, pericyazine, perphenazine,
pimozide, thioridazine, and trifluoperazine. Clozapine was
rarely used in Ontario during the study period, and we
therefore excluded patients who were receiving this medi-
cation. Other atypical antipsychotics (such as aripiprazole
and ziprasidone) are not licensed for use in Canada. We
decided that exposure to an antipsychotic was discontinued
(and we censored follow-up) if the patient did not refill his
or her antipsychotic prescription within an interval com-
posed of the days of drug supply plus a grace period of
20%. For example, we censored follow-up for a patient
who did not refill his or her 60-day antipsychotic prescrip-
tion within 72 days. We also censored follow-up for pa-
tients who switched from atypical to conventional antipsy-
chotics (or vice versa). However, we continued follow-up
for patients who switched from 1 atypical antipsychotic to
another, because data suggest no statistically significant dif-
ference in the risk for death associated with individual
drugs in this class (13, 14, 16). We applied the same rules
to conventional antipsychotics.
All-Cause Mortality
The primary outcome was all-cause mortality, as re-
corded in the Registered Persons Database (for patients
who were not hospitalized at the time of death) or the
Discharge Abstract Database (for patients who died while
hospitalized). To assess the influence of the duration of
antipsychotic exposure on the outcome, we evaluated the
risk for death at 30, 60, 120, and 180 days after the initial
dispensing of antipsychotic medication.
Cohort Matching
We stratified the dementia cohort to support separate
analyses among persons living in the community and those
residing in long-term care at cohort entry. Studies have
demonstrated that rates of antipsychotic prescribing are
substantial among older adults newly admitted to long-
term care facilities (4). Furthermore, long-term care resi-
dents typically carry a greater burden of comorbid disease
and are more vulnerable to adverse drug events than are
their counterparts in the community (22, 23).
Our first objective was to determine the risk for death
among older adults with dementia who received atypical
antipsychotics compared with those who were not exposed
to any antipsychotic. Because antipsychotic use was not
randomly assigned in the study cohorts, we addressed po-
tential confounding and selection biases by developing a
propensity score for antipsychotic use. We then applied
this score to match users of atypical antipsychotics with
nonusers in the dementia cohort. The rationale and meth-
ods underlying the use of a propensity score for a proposed
causal exposure variable are described elsewhere (24). Re-
cent studies provide guidance on the selection of variables
to include in the propensity score (25, 26). We developed
a logistic regression model by using 42 covariates describ-
ing patient characteristics. Tables 1 and 2 list many of the
characteristics included in the propensity score.
After a structured and iterative assessment of the bal-
ance of measured covariates between atypical antipsychotic
users and nonusers, we applied additional forced matching
to a variable indicating the presence of a hospitalization in
the 90 days before cohort entry. This variable was thought
to be an important predictor of new antipsychotic use and
was not well balanced between groups in the propensity
score. Using the resulting predicted probabilities as pro-
pensity scores, we matched each atypical antipsychotic user
with a nonuser by utilizing a caliper width (that is, interval
for successful match) of 0.6 of the SD of the log odds of
the propensity score. This method has been demonstrated
to remove approximately 90% of the bias from measured
confounders (27). These steps were completed separately
for each stratum (community-dwelling older adults com-
pared with long-term care residents), resulting in 2 groups
of matched pairs.
Our second objective was to determine the risk for
death with conventional antipsychotic use compared with
atypical antipsychotic use. We repeated the propensity
score–matching process to create matched cohorts of new
users of conventional and atypical antipsychotics. We strat-
ified analyses to examine older adults residing in the com-
munity and in long-term care separately. We developed 4
models, 1 for each of 2 drug use comparisons (atypical
antipsychotic use vs. nonuse and conventional vs. atypical
antipsychotic use) within each of 2 residential settings
(community and long-term care).
We used a “greedy matching†algorithm (available at
www2.sas.com/proceedings/sugi26/p214-26.pdf) to match
new users of atypical antipsychotics to nonusers and new
users of conventional antipsychotics to new users of atypi-
cal antipsychotics. Atypical antipsychotic users were eligi-
ble for both comparisons. We matched only patients with
an adequate match as defined earlier and sacrificed com-
pleteness of matching for accuracy of matching.
Statistical Analysis
To examine the effects of antipsychotic use on mortal-
ity, we first examined absolute risk differences in cumula-
tive mortality rates. We estimated 95% CIs for the abso-
lute risk differences by using bootstrap methods with
matched pairs repeated 1000 times. We then conducted
survival analysis by using Cox proportional hazards models
on our 4 matched cohorts. To address possible residual
confounding after the matching process, we performed tra-
ditional risk adjustment with covariates that strongly influ-
ence mortality (28). The covariates in our models were age,
sex, and dichotomous variables for the individual medical
conditions included in the Charlson comorbidity index
(acute myocardial infarction, congestive heart failure, peri-
pheral vascular disease, cerebrovascular disease, chronic
pulmonary disease, connective tissue disease, ulcer disease,
mild liver disease, diabetes [with and without end-organ
damage], hemiplegia or paraplegia, moderate or severe re-
nal disease, primary cancer, moderate or severe liver dis-
ease, and metastatic cancer) (29). Analyses involved strati-
fied Cox proportional hazards models using the STRATA
option of the PROC PHREG procedure in SAS, version
9.1 (SAS Institute, Inc., Cary, North Carolina), to account
for the matched pair design. We ran separate models for
the 30-, 60-, 120-, and 180-day end points. We confirmed
the proportional hazards assumption for all models by us-
ing an interaction term between the independent variable
and time. Because the P value for the term was greater than
0.05 for all comparisons, we concluded that the assump-
tion of proportional hazards was satisfied.
Although propensity score matching helps to balance
groups on measured covariates, our study is observational,
and unknown or unmeasured confounders may influence
the results. As other investigators have done (30, 31), we
conducted sensitivity analyses to investigate the potential
effects of unmeasured confounders on the observed associ-
ations. Our sensitivity analyses investigated the effects of a
hypothetical binary confounder on the observed hazard ra-
tios for 2 comparisons in community-dwelling patients:
the 30-day outcome for atypical antipsychotic use com-
pared with no antipsychotic use, and the 30-day outcome
for conventional antipsychotic use compared with atypical
antipsychotic use. We varied the prevalence of the hypo-
thetical unmeasured confounder in the 2 comparison
groups, as well as the relative hazard of death associated
with this unmeasured confounder.
Role of the Funding Source
The Canadian Institutes for Health Research funded
the study. The funding source had no role in the design,
conduct, or reporting of the study or in the decision to
submit the manuscript for publication.
RESULTS
Matching Process and Patient Characteristics
For the comparison of atypical antipsychotic use with
nonuse, we identified 9100 matched pairs of community-
dwelling older adults and 4036 matched pairs of long-term
care residents. For the comparison of conventional with
atypical antipsychotic use, we identified 6888 matched
pairs of community-dwelling older adults and 7235
matched pairs of long-term care residents. Thus, the study
included data on 27 259 propensity score–matched pairs.
The matching process produced excellent balance for co-
variates in the propensity score (Tables 1 and 2). The
standardized differences for these comparisons were nearly
all less than 10%. No value was missing for demographic
data or data used to construct the covariates (all analysis
variables were mandatory fields in our administrative data-
bases). In the community-dwelling cohort, atypical anti-
psychotic users started treatment with risperidone (75.2%),
olanzapine (19.6%), and quetiapine (5.2%) and conven-
tional antipsychotic users started treatment with haloperi-
dol (60.2%), loxapine (17.9%), thioridazine (10.3%),
chlorpromazine (5.8%), and perphenazine (3.5%). The
breakdown of antipsychotic use was similar in the long-
term care cohort.
In all comparisons, matches were limited by the smaller
of the 2 comparison groups. In the comparison of atypical
antipsychotic use with nonuse, the matched pairs of commu-
nity-dwelling older adults and long-term care residents repre-
sented 96% and 48% of eligible atypical antipsychotic users,
respectively. The lower match rate in the long-term care co-
hort reflects the smaller pool of nonusers available in this set-
ting, because recently institutionalized people with dementia
are frequently exposed to antipsychotics (4). In the compari-
son of conventional with atypical antipsychotic use, the
matched pairs of community-dwelling older adults and long-
term care residents represented 95% and 92% of eligible con-
ventional antipsychotic users, respectively. To support the ap-
propriateness of excluding patients for whom we could not
find adequate matches, we examined the characteristics of un-
matched atypical antipsychotic users. Compared with
matched patients, unmatched patients were younger, were
more likely to be male, were more likely to have had recent
hospitalizations with delirium and investigations (such as
computed tomography of the head), and had more physician
and hospital visits. Those differences in the demographic pro-
file and health services utilization of unmatched and matched
atypical antipsychotic users were all statistically significant.
Atypical Antipsychotic Use versus Nonuse
Figure 1 plots cumulative mortality for atypical anti-
psychotic use and nonuse. Table 3 shows crude event rates
and absolute risk differences for mortality at 30 and 180
days. The number of patients at risk decreased over time in
an unequal manner. We were concerned that this might
represent informative censoring. We therefore examined
the characteristics of the remaining patients at 60 days and
found that they stayed well balanced despite the unequal
rates of censoring. New use of atypical antipsychotic med-
ications was associated with a statistically significant in-
crease in the risk for death at 30 days compared with non-
use in both the community-dwelling cohort (adjusted
hazard ratio, 1.31 [95% CI, 1.02 to 1.70]; absolute risk
difference, 0.2 percentage point) and the long-term care
cohort (adjusted hazard ratio, 1.55 [CI, 1.15 to 2.07]; ab-
solute risk difference, 1.2 percentage points). The risk for
death associated with atypical antipsychotic use persisted to
180 days in both cohorts (Tables 3 and 4). Although the
hazard ratios for mortality in the long-term care cohort
were similar to those in the community-dwelling cohort,
cumulative mortality was almost twice as high among long-
term care residents (Figure 1).
Conventional versus Atypical Antipsychotic Use
As shown in Figure 2, use of conventional antipsy-
chotics was associated with an even greater risk for death
than that observed with atypical antipsychotic use. This
risk was evident at 30 days: The adjusted hazard ratio was
1.55 (CI, 1.19 to 2.02) for the community-dwelling cohort
and 1.26 (CI, 1.04 to 1.53) for the long-term care cohort
(adjusted risk difference for both groups, 1.1 percentage
points). The risk again persisted to 180 days (adjusted haz-
ard ratio, 1.23 [CI, 1.00 to 1.50]; absolute risk difference,
2.6 percentage points and 1.27 [CI, 1.09 to 1.48]; absolute
risk difference, 2.2 percentage points, respectively) (Tables
3 and 5).
Sensitivity Analyses
Sensitivity analyses for 30-day outcomes with atypical
antipsychotic use compared with nonuse found that an
unmeasured confounder that increased mortality could sta-
tistically significantly diminish the observed relationship
between use of atypical antipsychotics and death (Table 6).
For example, if the unmeasured confounder was associated
with a moderate increase in mortality (bivariate hazard ra-
tio, 1.5), the association between atypical antipsychotic use
and death was no longer statistically significant regardless
of the prevalence of the unmeasured confounder in the 2
groups. On the other hand, an unmeasured confounder
would need to be more strongly associated with mortality
to reduce the link between conventional antipsychotic use
and death at 30 days in the comparison of conventional
and atypical antipsychotic use (Table 6). For example, the
hazard ratio for death with conventional antipsychotic use
loses statistical significance only when the unmeasured
confounder has a bivariate hazard ratio of at least 2.0 and is
very unevenly distributed (that is, 30% prevalence among
atypical antipsychotic users and 60% prevalence among
conventional antipsychotic users).
DISCUSSION
Our study provides further evidence that use of atyp-
ical antipsychotics is associated with a small but significant
increase in mortality among older adults with dementia. In
addition, the risk for death associated with antipsychotics is
apparent after as little as 1 month of use and may persist
for 6 months. Finally, these data provide independent con-
firmation of reports that use of conventional antipsychotics
confers an even greater risk for death than does use of
atypical antipsychotics (15, 18).
Our population-based, matched cohort study includes
5 times as many atypical antipsychotic users and nonusers
as that in a meta-analysis of RCTs (14), includes many
conventional antipsychotic users (a group for whom few
data from RCTs exist), and provides follow-up data to 6
months. Our results complement and extend the findings
of studies that examined the relationship between anti-
psychotic exposure and death in different patient samples
(14–16, 18, 19, 32–37). The meta-analysis by Schneider
and colleagues (14) includes 15 RCTs that evaluated sev-
eral atypical antipsychotics in the short-term management
of neuropsychiatric symptoms of dementia. Our results are
consistent with that meta-analysis and provide further in-
sight in a nontrial population with longer follow-up and
greater exposure to conventional antipsychotics. Wang and
colleagues (15) used U.S. administrative data to compare
the risk for death among new users of conventional and
atypical antipsychotics. They used various sophisticated
observational study techniques to address potential con-
founding factors. Similar to our results, Wang and col-
leagues (15) found that conventional antipsychotics were
associated with a statistically significantly higher risk for
death at all intervals up to 180 days after treatment initia-
tion. In the first 180 days of use, the absolute event rates
were dramatic: 17.9% of patients who began using conven-
tional antipsychotics and 14.6% of patients who began
using atypical antipsychotics died (15).
The potential causes of death associated with antipsy-
chotic use merit consideration. We did not have informa-
tion on the proximate causes of death for all patients (38,
39). Nonetheless, several plausible mechanisms can be pro-
posed. First, antipsychotics may prolong the QT interval,
predisposing patients to arrhythmias and sudden cardiac
death (40–46). Second, sedation and accelerated cognitive
decline brought on by exposure to antipsychotics may in-
crease the risk for aspiration syndromes and choking (47–
49). Aspiration pneumonia is an important cause of death
among people with dementia (38). Third, several studies
have found a link between atypical antipsychotic use and
venous thromboembolism (50, 51); therefore, pulmonary
embolism may be an underrecognized cause of sudden
death in these patients. Fourth, a risk for cerebrovascular
events may be associated with antipsychotic use, although
this risk has been questioned (52, 53). Finally, antipsychot-
ics may contribute to events that are not initially recog-
nized as the first step in a sequence that promotes prema-
ture death, such as falls leading to hip fractures (54).
Although details are limited, deaths in the RCTs seem to
have been primarily related to cardiac arrhythmias and as-
piration pneumonia (12). A review of the RCTs that eval-
uated olanzapine supports these proposed mechanisms of
harm (16).
These results have important implications for clinical
practice. First, conventional antipsychotics seem to be as-
sociated with a higher risk for death than are atypical anti-
psychotics (15, 18). Second, the estimated mortality rate
among study participants was high, especially in the long-
term care setting (as one might expect among vulnerable
older adults with dementia) (55). Despite these high mor-
tality rates, we can still detect excess mortality associated
with exposure to antipsychotics. In the U.S. FDA and
Health Canada reviews (12, 13), the risk for death seemed
to be a class effect with all atypical antipsychotics studied,
and 2 meta-analyses confirmed these findings (14, 16).
Thus, switching between individual atypical antipsychotics
in an attempt to modify the risk for death cannot be rec-
ommended. Finally, and perhaps most complex, the role of
atypical antipsychotics in the management of behavioral
and psychological symptoms of dementia must be carefully
reviewed (56, 57). Rabins and Lyketsos (17) suggest an
approach that limits the use of these drugs to situations in
which “there is an identifiable risk of harm to the patient
or others, when the distress caused by symptoms is signif-
icant, or when alternate therapies have failed and symptom
relief would be beneficial.†Because the risk for death asso-
ciated with antipsychotics develops quickly and may persist
for up to 6 months, clinicians must reevaluate benefits and
risks frequently and consider discontinuation of treatment
when appropriate. Studies have shown that some patients
receiving antipsychotics can be successfully weaned from
these medications when monitored closely (58–61). Anti-
psychotic therapies should not be initiated if effective non-
drug treatments are available or if symptoms are unlikely to
respond to antipsychotic treatment (for example, repetitive
vocalizations or wandering). Nonpharmacologic treatments
for neuropsychiatric symptoms of dementia have been re-
viewed (62). Clinical trials involving behavior management
and caregiver education have shown benefits in both com-
munity-dwelling persons (63–65) and long-term care resi-
dents (66) and may help to minimize antipsychotic use.
Efforts are needed to facilitate the implementation of these
effective interventions into clinical practice (3).
Our study has important limitations. First, we used
administrative data and observational study techniques.
Our risk estimates are relatively small and may therefore be
confounded by other variables, and our ability to control
for differences in the cohorts was limited to variables with
available data. The sensitivity analyses highlight these lim-
itations. Nevertheless, our results are consistent with find-
ings from other research, including a meta-analysis of
RCTs (14). Second, we did not examine the risk for death
posed by individual antipsychotic drugs. However, regula-
tory warnings and 2 meta-analyses have shown that the
increased risk for death is consistent among the individual
drugs in this class (12–14, 16). Third, we could not exam-
ine the causes of death. Fourth, we did not examine dose–
response relationships, given the complexity of our study
design and changes in dosages over time. Of note, no sta-
tistically significant relationship between antipsychotic
dose and risk for death was found in the meta-analysis of
olanzapine RCTs (16). Fifth, we could not match all po-
tentially eligible patients, which may limit the generaliz-
ability of our findings. Finally, we restricted our study to
older adults with dementia. The safety of antipsychotics
when used for other indications (for example, schizophre-
nia and delirium) requires further evaluation (36, 37, 67).
In conclusion, we show that older adults with demen-
tia who are exposed to atypical antipsychotics have a small
but significant increase in overall mortality that is evident
as early as 1 month after initiation of treatment, and this
risk may persist for 6 months. Use of conventional anti-
psychotics seems to confer an even greater risk for death
than does atypical antipsychotic use. These findings high-
light the need to carefully balance potential risks and ben-
efits when considering antipsychotic treatment for older
adults with dementia and emphasize the need to limit use
of these drugs to situations in which nonpharmacologic
measures have provided an inadequate response.
From Queen’s University, Kingston, Ontario, Canada; Institute for
Clinical Evaluative Sciences and University of Toronto, Toronto, On-
tario, Canada; Harvard Medical School and Harvard School of Public
Health, Boston, Massachusetts; University of British Columbia, Vancou-
ver, British Columbia, Canada; and Meyers Primary Care Institute of the
University of Massachusetts Medical School, Fallon Clinic Foundation,
and Fallon Community Health Plan, Worcester, Massachusetts.
Grant Support: By a Canadian Institutes for Health Research operating
grant (no. 53124) and a Chronic Disease New Emerging Team program
grant (no. 54010). The New Emerging Team program receives joint
sponsorship from the Canadian Diabetes Association, the Kidney Foun-
dation of Canada, the Heart and Stroke Foundation of Canada and the
Canadian Institutes for Health Research Institutes of Nutrition, Metab-
olism & Diabetes and Circulatory & Respiratory Health. Dr. Gill is
supported by an Ontario Ministry of Health and Long-Term Care Ca-
reer Scientist Award. Dr. Bronskill is supported by a New Investigator
Award through the New Emerging Team program. Dr. Anderson is
supported by a Chair in Health Management Strategies from the Uni-
versity of Toronto. Dr. Rochon is supported by a Canadian Institutes for
Health Research Investigator Award. Dr. Lee is supported by a fellow-
ship grant from Eli Lilly.
Potential Financial Conflicts of Interest: Consultancies: P.E. Lee (Pfizer
Canada, Janssen-Ortho), N. Herrmann (Janssen, Novartis, Pfizer Inc.);
Honoraria: P.E. Lee (Pfizer Inc., Janssen-Ortho, Novartis), N. Herrmann
(Janssen, Eli Lilly, Novartis, AstraZeneca); Expert testimony: N. Herr-
mann (Janssen); Grants received: P.E. Lee (Eli Lilly), N. Herrmann
(Janssen).
Requests for Single Reprints: Sudeep S. Gill, MD, MSc, St. Mary’s of
the Lake Hospital, 340 Union Street, Kingston, Ontario K7L 5A2,
Canada.
Current author addresses and author contributions are available at www
.annals.org.
References
1. McKeith I, Cummings J. Behavioural changes and psychological symptoms in
dementia disorders. Lancet Neurol. 2005;4:735-42. [PMID: 16239180]
2. Sink KM, Holden KF, Yaffe K. Pharmacological treatment of neuropsychiat-
ric symptoms of dementia: a review of the evidence. JAMA. 2005;293:596-608.
[PMID: 15687315]
3. Covinsky KE, Johnston CB. Envisioning better approaches for dementia care
[Editorial]. Ann Intern Med. 2006;145:780-1. [PMID: 17116923]
4. Bronskill SE, Anderson GM, Sykora K, Wodchis WP, Gill S, Shulman KI,
et al. Neuroleptic drug therapy in older adults newly admitted to nursing homes:
incidence, dose, and specialist contact. J Am Geriatr Soc. 2004;52:749-55.
[PMID: 15086656]
5. Liperoti R, Mor V, Lapane KL, Pedone C, Gambassi G, Bernabei R. The use
of atypical antipsychotics in nursing homes. J Clin Psychiatry. 2003;64:1106-12.
[PMID: 14628988]
6. Schneider LS, Pollock VE, Lyness SA. A metaanalysis of controlled trials of
neuroleptic treatment in dementia. J Am Geriatr Soc. 1990;38:553-63. [PMID:
1970586]
7. Lanctôt KL, Best TS, Mittmann N, Liu BA, Oh PI, Einarson TR, et al.
Efficacy and safety of neuroleptics in behavioral disorders associated with demen-
tia. J Clin Psychiatry. 1998;59:550-61; quiz 562-3. [PMID: 9818639]
8. Rapoport M, Mamdani M, Shulman KI, Herrmann N, Rochon PA. Anti-
psychotic use in the elderly: shifting trends and increasing costs. Int J Geriatr
Psychiatry. 2005;20:749-53. [PMID: 16035128]
9. Lee PE, Gill SS, Freedman M, Bronskill SE, Hillmer MP, Rochon PA.
Atypical antipsychotic drugs in the treatment of behavioural and psychological
symptoms of dementia: systematic review. BMJ. 2004;329:75. [PMID:
15194601]
10. Schneider LS, Tariot PN, Dagerman KS, Davis SM, Hsiao JK, Ismail MS,
et al. Effectiveness of atypical antipsychotic drugs in patients with Alzheimer’s
disease. N Engl J Med. 2006;355:1525-38. [PMID: 17035647]
11. Ballard CG, Margallo-Lana ML. The relationship between antipsychotic
treatment and quality of life for patients with dementia living in residential and
nursing home care facilities. J Clin Psychiatry. 2004;65 Suppl 11:23-8. [PMID:
15264968]
12. U.S. Food and Drug Administration. FDA issues public health advisory for
antipsychotic drugs used for treatment of behavioral disorders in elderly patients.
FDA Talk Paper T05-13. Rockville, MD: U.S. Food and Drug Administration;
11 April 2005. Accessed at www.fda.gov/bbs/topics/ANSWERS/2005
/ANS01350.html on 9 March 2007.
13. Health Canada. Health Canada advises consumers about important safety
information on atypical antipsychotic drugs and dementia. Advisory 2005-63.
Ottawa: Health Canada; 15 June 2005. Accessed at www.hc-sc.gc.ca/ahc-asc/me-
dia/advisories-avis/2005/2005_63_e.html on 9 March 2007.
14. Schneider LS, Dagerman KS, Insel P. Risk of death with atypical antipsy-
chotic drug treatment for dementia: meta-analysis of randomized placebo-
controlled trials. JAMA. 2005;294:1934-43. [PMID: 16234500]
15. Wang PS, Schneeweiss S, Avorn J, Fischer MA, Mogun H, Solomon DH,
et al. Risk of death in elderly users of conventional vs. atypical antipsychotic
medications. N Engl J Med. 2005;353:2335-41. [PMID: 16319382]
16. Kryzhanovskaya LA, Jeste DV, Young CA, Polzer JP, Roddy TE, Jansen JF,
et al. A review of treatment-emergent adverse events during olanzapine clinical
trials in elderly patients with dementia. J Clin Psychiatry. 2006;67:933-45.
[PMID: 16848653]
17. Rabins PV, Lyketsos CG. Antipsychotic drugs in dementia: what should be
made of the risks? [Editorial]. JAMA. 2005;294:1963-5. [PMID: 16234504]
18. Nasrallah HA, White T, Nasrallah AT. Lower mortality in geriatric patients
receiving risperidone and olanzapine versus haloperidol: preliminary analysis of
retrospective data. Am J Geriatr Psychiatry. 2004;12:437-9. [PMID: 15249282]
19. Barnett MJ, Perry PJ, Alexander B, Kaboli PJ. Risk of mortality associated
with antipsychotic and other neuropsychiatric drugs in pneumonia patients.
J Clin Psychopharmacol. 2006;26:182-7. [PMID: 16633149]
20. Levy AR, O’Brien BJ, Sellors C, Grootendorst P, Willison D. Coding
accuracy of administrative drug claims in the Ontario Drug Benefit database. Can
J Clin Pharmacol. 2003;10:67-71. [PMID: 12879144]
21. Ray WA. Evaluating medication effects outside of clinical trials: new-user
designs. Am J Epidemiol. 2003;158:915-20. [PMID: 14585769]
22. Gurwitz JH, Field TS, Judge J, Rochon P, Harrold LR, Cadoret C, et al.
The incidence of adverse drug events in two large academic long-term care facil-
ities. Am J Med. 2005;118:251-8. [PMID: 15745723]
23. Fahey T, Montgomery AA, Barnes J, Protheroe J. Quality of care for elderly
residents in nursing homes and elderly people living at home: controlled obser-
vational study. BMJ. 2003;326:580. [PMID: 12637404]
24. D’Agostino RB Jr. Propensity score methods for bias reduction in the com-
parison of a treatment to a non-randomized control group. Stat Med. 1998;17:
2265-81. [PMID: 9802183]
25. Brookhart MA, Schneeweiss S, Rothman KJ, Glynn RJ, Avorn J, Stürmer
T. Variable selection for propensity score models. Am J Epidemiol. 2006;163:
1149-56. [PMID: 16624967]
26. Austin PC, Grootendorst P, Anderson GM. A comparison of the ability of
different propensity score models to balance measured variables between treated
and untreated subjects: a Monte Carlo study. Stat Med. 2007;26:734-53.
[PMID: 16708349]
27. Rosenbaum PR, Rubin DB. Constructing a control group using multivariate
matched sampling methods that incorporate the propensity score. Am Stat 1985;
39:33-8.
28. D’Agostino RB Jr. Propensity score methods for bias reduction in the com-
parison of a treatment to a nonrandomized control group. In: D’Agostino RB Jr,
ed. Tutorials in Biostatistics. Volume 1: Statistical Methods in Clinical Studies.
New York: J Wiley; 2004.
29. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of
classifying prognostic comorbidity in longitudinal studies: development and val-
idation. J Chronic Dis. 1987;40:373-83. [PMID: 3558716]
30. Kern LM, Powe NR, Levine MA, Fitzpatrick AL, Harris TB, Robbins J,
et al. Association between screening for osteoporosis and the incidence of hip
fracture. Ann Intern Med. 2005;142:173-81. [PMID: 15684205]
31. Schneeweiss S. Sensitivity analysis and external adjustment for unmeasured
confounders in epidemiologic database studies of therapeutics. Pharmacoepide-
miol Drug Saf. 2006;15:291-303. [PMID: 16447304]
32. Suh GH, Shah A. Effect of antipsychotics on mortality in elderly patients
with dementia: a 1-year prospective study in a nursing home. Int Psychogeriatr.
2005;17:429-41. [PMID: 16252375]
33. Scarmeas N, Brandt J, Albert M, Hadjigeorgiou G, Papadimitriou A,
Dubois B, et al. Delusions and hallucinations are associated with worse outcome
in Alzheimer disease. Arch Neurol. 2005;62:1601-8. [PMID: 16216946]
34. Trifiro` G, Verhamme KM, Ziere G, Caputi AP, Ch Stricker BH, Sturken-
boom MC. All-cause mortality associated with atypical and typical antipsychotics
in demented outpatients. Pharmacoepidemiol Drug Saf. 2006. [PMID:
17036366]
35. Livingston G, Walker AE, Katona CL, Cooper C. Antipsychotics and cog-
nitive decline in Alzheimer’s disease: the LASER-Alzheimer’s disease longitudinal
study. J Neurol Neurosurg Psychiatry. 2007;78:25-9. [PMID: 16801350]
36. Joukamaa M, Heliövaara M, Knekt P, Aromaa A, Raitasalo R, Lehtinen V.
Schizophrenia, neuroleptic medication and mortality. Br J Psychiatry. 2006;188:
122-7. [PMID: 16449697]
37. Tiihonen J, Walhbeck K, Lönnqvist J, Klaukka T, Ioannidis JP, Volavka J,
et al. Effectiveness of antipsychotic treatments in a nationwide cohort of patients
in community care after first hospitalisation due to schizophrenia and schizoaf-
fective disorder: observational follow-up study. BMJ. 2006;333:224. [PMID:
16825203]
38. Keene J, Hope T, Fairburn CG, Jacoby R. Death and dementia. Int J
Geriatr Psychiatry. 2001;16:969-74. [PMID: 11607941]
39. Ray WA. Observational studies of drugs and mortality. N Engl J Med.
2005;353:2319-21. [PMID: 16319379]
40. Ray WA, Meredith S, Thapa PB, Meador KG, Hall K, Murray KT. Anti-
psychotics and the risk of sudden cardiac death. Arch Gen Psychiatry. 2001;58:
1161-7. [PMID: 11735845]
41. Hennessy S, Bilker WB, Knauss JS, Margolis DJ, Kimmel SE, Reynolds
RF, et al. Cardiac arrest and ventricular arrhythmia in patients taking antipsy-
chotic drugs: cohort study using administrative data. BMJ. 2002;325:1070.
[PMID: 12424166]
42. Straus SM, Bleumink GS, Dieleman JP, van der Lei J, ’t Jong GW, Kingma
JH, et al. Antipsychotics and the risk of sudden cardiac death. Arch Intern Med.
2004;164:1293-7. [PMID: 15226162]
43. Liperoti R, Gambassi G, Lapane KL, Chiang C, Pedone C, Mor V, et al.
Conventional and atypical antipsychotics and the risk of hospitalization for ven-
tricular arrhythmias or cardiac arrest. Arch Intern Med. 2005;165:696-701.
[PMID: 15795349]
44. Straus SM, Sturkenboom MC, Bleumink GS, Dieleman JP, van der Lei J,
de Graeff PA, et al. Non-cardiac QTc-prolonging drugs and the risk of sudden
cardiac death. Eur Heart J. 2005;26:2007-12. [PMID: 15888497]
45. Ray WA, Meador KG. Antipsychotics and sudden death: is thioridazine the
only bad actor? [Editorial]. Br J Psychiatry. 2002;180:483-4. [PMID: 12042224]
46. Hennessy S, Bilker WB, Knauss JS, Kimmel SE, Margolis DJ, Morrison
MF, et al. Comparative cardiac safety of low-dose thioridazine and low-dose
haloperidol. Br J Clin Pharmacol. 2004;58:81-7. [PMID: 15206997]
47. Warner J. Risk of choking in mental illness. Lancet. 2004;363:674. [PMID:
15001323]
48. Berzlanovich AM, Muhm M, Sim E, Bauer G. Foreign body asphyxia-
tion—an autopsy study. Am J Med. 1999;107:351-5. [PMID: 10527037]
49. Ruschena D, Mullen PE, Palmer S, Burgess P, Cordner SM, Drummer
OH, et al. Choking deaths: the role of antipsychotic medication. Br J Psychiatry.
2003;183:446-50. [PMID: 14594921]
50. Zornberg GL, Jick H. Antipsychotic drug use and risk of first-time idiopathic
venous thromboembolism: a case-control study. Lancet. 2000;356:1219-23.
[PMID: 11072939]
51. Liperoti R, Pedone C, Lapane KL, Mor V, Bernabei R, Gambassi G.
Venous thromboembolism among elderly patients treated with atypical and con-
ventional antipsychotic agents. Arch Intern Med. 2005;165:2677-82. [PMID:
16344428]
52. Gill SS, Rochon PA, Herrmann N, Lee PE, Sykora K, Gunraj N, et al.
Atypical antipsychotic drugs and risk of ischaemic stroke: population based ret-
rospective cohort study. BMJ. 2005;330:445. [PMID: 15668211]
53. Liperoti R, Gambassi G, Lapane KL, Chiang C, Pedone C, Mor V, et al.
Cerebrovascular events among elderly nursing home patients treated with con-
ventional or atypical antipsychotics. J Clin Psychiatry. 2005;66:1090-6. [PMID:
16187764]
54. Normand SL, Sykora K, Li P, Mamdani M, Rochon PA, Anderson GM.
Readers guide to critical appraisal of cohort studies: 3. Analytical strategies to
reduce confounding. BMJ. 2005;330:1021-3. [PMID: 15860831]
55. Tschanz JT, Corcoran C, Skoog I, Khachaturian AS, Herrick J, Hayden
KM, et al. Dementia: the leading predictor of death in a defined elderly popula-
tion: the Cache County Study. Neurology. 2004;62:1156-62. [PMID:
15079016]
56. Ballard C, Cream J. Drugs used to relieve behavioral symptoms in people
with dementia or an unacceptable chemical cosh? Argument. Int Psychogeriatr.
2005;17:4-12; discussion 22-9. [PMID: 15945588]
57. Shah A, Suh GH. A case for judicious use of risperidone and olanzapine
in behavioral and psychological symptoms of dementia (BPSD). Favour.
Int Psychogeriatr. 2005;17:12-22. [PMID: 15945589]
58. Ray WA, Taylor JA, Meador KG, Lichtenstein MJ, Griffin MR, Fought R,
et al. Reducing antipsychotic drug use in nursing homes. A controlled trial of
provider education. Arch Intern Med. 1993;153:713-21. [PMID: 8447709]
59. Thapa PB, Meador KG, Gideon P, Fought RL, Ray WA. Effects of anti-
psychotic withdrawal in elderly nursing home residents. J Am Geriatr Soc. 1994;
42:280-6. [PMID: 7907098]
60. Meador KG, Taylor JA, Thapa PB, Fought RL, Ray WA. Predictors of
antipsychotic withdrawal or dose reduction in a randomized controlled trial of
provider education. J Am Geriatr Soc. 1997;45:207-10. [PMID: 9033521]
61. van Reekum R, Clarke D, Conn D, Herrmann N, Eryavec G, Cohen T,
et al. A randomized, placebo-controlled trial of the discontinuation of long-term
antipsychotics in dementia. Int Psychogeriatr. 2002;14:197-210. [PMID:
12243210]
62. Livingston G, Johnston K, Katona C, Paton J, Lyketsos CG; Old Age Task
Force of the World Federation of Biological Psychiatry. Systematic review of
psychological approaches to the management of neuropsychiatric symptoms of
dementia. Am J Psychiatry. 2005;162:1996-2021. [PMID: 16263837]
63. Callahan CM, Boustani MA, Unverzagt FW, Austrom MG, Damush TM,
Perkins AJ, et al. Effectiveness of collaborative care for older adults with Alzhei-
mer disease in primary care: a randomized controlled trial. JAMA. 2006;295:
2148-57. [PMID: 16684985]
64. Vickrey BG, Mittman BS, Connor KI, Pearson ML, Della Penna RD,
Ganiats TG, et al. The effect of a disease management intervention on quality
and outcomes of dementia care: a randomized, controlled trial. Ann Intern Med.
2006;145:713-26. [PMID: 17116916]
65. Belle SH, Burgio L, Burns R, Coon D, Czaja SJ, Gallagher-Thompson D,
et al. Enhancing the quality of life of dementia caregivers from different ethnic or
racial groups: a randomized, controlled trial. Ann Intern Med. 2006;145:727-38.
[PMID: 17116917]
66. Fossey J, Ballard C, Juszczak E, James I, Alder N, Jacoby R, et al. Effect of
enhanced psychosocial care on antipsychotic use in nursing home residents with
severe dementia: cluster randomised trial. BMJ. 2006;332:756-61. [PMID:
16543297]
67. Gill SS, Seitz D, Rochon PA. Atypical antipsychotic drugs, dementia, and
risk of death [Letter]. JAMA. 2006;295:495-6 [PMID: 16449609]
Current Author Addresses: Dr. Gill: St. Mary’s of the Lake Hospital,
340 Union Street, Kingston, Ontario K7L 5A2, Canada.
Drs. Bronskill, Anderson, Bell, Fischer, and Rochon; Ms. Sykora; and
Mr. Lam: Institute for Clinical Evaluative Sciences, G Wing, 2075 Bay-
view Avenue, Toronto, Ontario M4N 3M5, Canada.
Dr. Normand: Department of Health Care Policy, Harvard Medical
School, 180 Longwood Avenue, Boston, MA 02115-5899.
Dr. Lee: St. Paul’s Hospital, 1081 Burrard Street, 9th Floor, Providence
Building, Vancouver, British Columbia V6T 2B5, Canada.
Dr. Herrmann: Sunnybrook Health Sciences Centre, 2075 Bayview
Avenue, Toronto, Ontario M4N 3M5, Canada.
Dr. Gurwitz: Meyers Primary Care Institute, 630 Plantation Street,
Worcester, MA 01605.
Author Contributions: Conception and design: S.S. Gill, S.E. Bronskill,
S.-L.T. Normand, G.M. Anderson, C.M. Bell, P.E. Lee, N. Herrmann,
J.H. Gurwitz, P.A. Rochon.
Analysis and interpretation of the data: S.S. Gill, S.E. Bronskill, S.-L.T.
Normand, G.M. Anderson, K. Sykora, K. Lam, C.M. Bell, P.E. Lee,
H.D. Fischer, N. Herrmann, J.H. Gurwitz, P.A. Rochon.
Drafting of the article: S.S. Gill, S.E. Bronskill.
Critical revision of the article for important intellectual content: S.S.
Gill, S.E. Bronskill, G.M. Anderson, K. Sykora, K. Lam, C.M. Bell, P.E.
Lee, H.D. Fischer, N. Herrmann, J.H. Gurwitz, P.A. Rochon.
Final approval of the article: S.S. Gill, C.M. Bell, N. Herrmann, J.H.
Gurwitz, P.A. Rochon.
Statistical expertise: S.S. Gill, S.E. Bronskill, S.-L.T. Normand, K.
Sykora, K. Lam.
Administrative, technical, or logistic support: K. Sykora, K. Lam, H.D.
Fischer.
