Anda di halaman 1dari 10


Risk of Death With Atypical Antipsychotic

Drug Treatment for Dementia
Meta-analysis of Randomized Placebo-Controlled Trials
Lon S. Schneider, MD, MS Context Atypical antipsychotic medications are widely used to treat delusions, ag-
Karen S. Dagerman, MS gression, and agitation in people with Alzheimer disease and other dementia; how-
Philip Insel, MS ever, concerns have arisen about the increased risk for cerebrovascular adverse events,
rapid cognitive decline, and mortality with their use.

MAJORITY OF ELDERLY PA - Objective To assess the evidence for increased mortality from atypical antipsy-
tients with dementia de- chotic drug treatment for people with dementia.
velop aggression, delusions, Data Sources MEDLINE (1966 to April 2005), the Cochrane Controlled Trials Reg-
and other neuropsychiatric ister (2005, Issue 1), meetings presentations (1997-2004), and information from the
symptoms during their illness course. sponsors were searched using the terms for atypical antipsychotic drugs (aripiprazole,
Antipsychotic medications are com- clozapine, olanzapine, quetiapine, risperidone, and ziprasidone), dementia, Alzhei-
monly used to treat these behaviors, mer disease, and clinical trial.
along with psychosocial and environ- Study Selection Published and unpublished randomized placebo-controlled, parallel-
mental interventions. They have been group clinical trials of atypical antipsychotic drugs marketed in the United States to
the mainstay of psychopharmacologi- treat patients with Alzheimer disease or dementia were selected by consensus of the
cal treatment for this purpose during authors.
the last several decades despite their Data Extraction Trials, baseline characteristics, outcomes, all-cause dropouts, and
clear overuse in the 1980s and federal deaths were extracted by one reviewer; treatment exposure was obtained or esti-
regulations implemented in the early mated. Data were checked by a second reviewer.
1990s requiring their oversight and Data Synthesis Fifteen trials (9 unpublished), generally 10 to 12 weeks in dura-
monitoring in nursing homes.1 tion, including 16 contrasts of atypical antipsychotic drugs with placebo met criteria
During the last decade, the newer (aripiprazole [n=3], olanzapine [n=5], quetiapine [n=3], risperidone [n=5]). A total
atypical antipsychotic drugs (ie, ris- of 3353 patients were randomized to study drug and 1757 were randomized to pla-
peridone, olanzapine, quetiapine, and cebo. Outcomes were assessed using standard methods (with random- or fixed-
aripiprazole, in order of introduction) effects models) to calculate odds ratios (ORs) and risk differences based on patients
randomized and relative risks based on total exposure to treatment. There were no
have largely replaced the older conven- differences in dropouts. Death occurred more often among patients randomized to
tional or first-generation antipsy- drugs (118 [3.5%] vs 40 [2.3%]. The OR by meta-analysis was 1.54; 95% confi-
chotic drugs (eg, haloperidol and thio- dence interval [CI], 1.06-2.23; P=.02; and risk difference was 0.01; 95% CI, 0.004-
ridazine) and have been considered 0.02; P=.01). Sensitivity analyses did not show evidence for differential risks for in-
preferred treatments for these behav- dividual drugs, severity, sample selection, or diagnosis.
ioral disturbances associated with de- Conclusions Atypical antipsychotic drugs may be associated with a small increased
mentia.2,3 Reasons for this preference risk for death compared with placebo. This risk should be considered within the con-
include emerging clinical trials evi- text of medical need for the drugs, efficacy evidence, medical comorbidity, and the
dence,4-8 perceived relative safety ad- efficacy and safety of alternatives. Individual patient analyses modeling survival and
vantages compared with older antipsy- causes of death are needed.
chotic drugs and other medications, the JAMA. 2005;294:1934-1943
opinions of expert clinicians, and ex-
Author Affiliations are listed at the end of this article. Sciences, Keck School of Medicine, University of South-
For editorial comment see p 1963. Corresponding Author: Lon S. Schneider, MD, MS, ern California, 1510 San Pablo St, HCC 600, Los An-
Department of Psychiatry and the Behavioral geles, CA 90033 (

1934 JAMA, October 19, 2005Vol 294, No. 15 (Reprinted) 2005 American Medical Association. All rights reserved.

Downloaded From: on 02/25/2013


pectations of efficacy.2,3 There is little Drug Administration (FDA) followed METHODS

clinical trials evidence supporting the with additional warnings of increased Search Strategy, Trials Selection,
efficacy of other classes of psycho- cerebrovascular adverse events to the and Data Retrieval
tropic medication, such as benzodi- US prescribing information for ris- MEDLINE (1966 to April 2005) and the
azepines, anticonvulsants, and anti- peridone in April 2003 (http://www Cochrane Controlled Trials Register
depressants, for the treatment of, olanzapine in January (2005, Issue 1)29 were searched, using
aggression, psychotic symptoms, or agi- 2004 (, and the terms aripiprazole, clozapine,
tation in patients with dementia.9 aripiprazole in February 2005 (http: olanzapine, quetiapine, risperidone, and
The perceived relative safety advan- // There is limited ziprasidone (atypical antipsychotic
tages of atypical drugs compared with public access to these data, however, be- drugs marketed in the United States),
conventional antipsychotic drugs or cause most of the trials have not been dementia, Alzheimer disease, and clini-
other medications include lesser car- published, cerebrovascular adverse cal trial. Conference programs, ab-
diovascular adverse effects (eg, ortho- events were not reported in some trials, stract books, proceedings, Web post-
static hypotension and repolarization and adverse events occurring less than ings from available conferences,
delays), less sedation, postural insta- 5% of the time are often not reported. proceedings, abstracts, poster presen-
bility, falls, movement disorders, and Deaths during antipsychotic clinical tations, and slides from geriatric medi-
thermodysregulation. The few direct trials may be consequent to an adverse cine, psychiatric, neurological, and geri-
comparison trials, however, are inad- event caused by the drugs, and atric psychiatric professional society
equate to address this.5,10-12 Moreover, because they are classified by federal meetings since 1999 were hand-
both conventional antipsychotic use regulation as serious adverse searched. Pharmaceutical manufactur-
and the presence of psychotic symp- events 25 are generally included in ers were queried and information was
toms have been associated with more clinical trials reports. At a medical requested as needed.
rapid cognitive decline in patients with conference in 2002, the FDA reported Trials were included in the study
dementia.13-18 on deaths occurring in a sample of analyses if they met the following cri-
Recently, there has been concern 1452 patients with dementia from teria: parallel group, double-blinded,
about an increased risk for cerebrovas- placebo-controlled trials of atypical placebo-controlled with random as-
cular adverse events (mainly stroke and antipsychotic drugs (approximately signment to an orally administered an-
transient ischemic episodes, and some 32% received placebo, 59% atypical tipsychotic or placebo; patients had Alz-
instances of loss of consciousness or drugs, and the rest mostly received heimer disease, vascular dementia,
syncope), metabolic syndrome, and haloperidol). There was a high rate of mixed dementia, or a primary demen-
other adverse events that may be spe- deaths in the placebo group (164.7 per tia; and numbers of patients random-
cifically caused by certain atypical 1000 patient-years) and higher rates of ized, dropouts, and deaths were ob-
drugs.19-22 The evidence for cerebrovas- 242.5 and 276.3 per 1000 patient- tainable. Additional information had to
cular adverse events was found through years for atypical and conventional be available with respect to sample se-
pooled analyses of mainly nursing home drugs,26 respectively, implying RRs of lection criteria, location of patients, ran-
clinical trials of risperidone and olanza- 1.47 and 1.68 for atypical and conven- domization, double-blinding, trials du-
pine for people with dementia per- tional drugs, respectively, compared rations, medication dosage ranges and
formed by regulators who had access with placebo. On April 11, 2005, the formulations, and outcomes. Trials did
to unpublished data.20,23 These analy- FDA issued a health advisory warning not need to be published or peer-
ses assessing incidences during clini- of an increased risk for death with reviewed and could be reported in
cal trials all show increased relative risks atypical drugs in persons with demen- manuscripts, technical trials reports,
(RRs) or odds ratios (ORs) ranging from tia but did not provide data.27 Despite posters, letters, or slide formats. (Some
2 to 4, albeit with some not reaching the FDA warning and a lack of con- sources presented incomplete informa-
nominally statistically significant P val- trolled trials proving efficacy, consult- tion and additional information was ob-
ues. ants found atypical antipsychotic tained though other data presenta-
Health Canada advised Canadian drugs beneficial in calming agitated or tions or from sponsors).
physicians in late 2002 to reassess aggressive elderly patients and noted Information extracted included de-
the risks and benefits . . . in elderly that there were no good alternatives.28 sign, selection criteria (dementia diag-
patients with dementia . . . [and to] In light of these events and the ex- noses and presence of psychosis of de-
counsel their patients/caregivers to panding evidence base, we conducted mentia30), medication doses, locations,
immediately report signs and symp- independent individual study meta- trials durations, age, sex, baseline cog-
toms of potential cerebrovascular analyses of atypical antipsychotic drug nitive scores, numbers randomized, and
adverse events so that diagnosis can trials to assess the evidence for mortal- the outcomes, all-cause dropouts, and
be made and treatment options con- ity associated with their use in elderly deaths occurring during the double-
sidered. . . . 24 The US Food and patients with dementia. blind trial period or within 30 days of
2005 American Medical Association. All rights reserved. (Reprinted) JAMA, October 19, 2005Vol 294, No. 15 1935

Downloaded From: on 02/25/2013


discontinuing double-blind treat- of at least 50% were taken as indica- were obtained, which contained infor-
ment. Because there were few dose- tors of heterogeneity of outcomes. mation on placebo-controlled trials, and
ranging trials, sparse outcomes, and to We compared the following sub- 13 with unique information on 10 trials
avoid multiple comparisons with the groups as sensitivity analyses: whether were retained.
same placebo group, we aggregated dos- or not sample selection required that From other sources, including medi-
age groups within each trial to make one patients had to have psychosis of de- cal letters and direct communications
contrast with the placebo group. Data mentia,30 outpatient vs nursing home from pharmaceutical companies and
were abstracted by one investigator status, cognitive severity (mean base- news articles in medical journals, 18 pla-
(K.S.D.) and checked by another in- line Mini-Mental State Examination cebo-controlled, randomized con-
vestigator (P.I.). Any discrepant data score per trial 10 or 10), or by drug trolled trials were identified. Three ris-
were rereviewed by the investigators to used. Differences between 2 or more peridone trials were not included
ensure that accurate data were obtained. subgroups were investigated by sub- because of unavailability of data, includ-
After the trials to be included were tracting the sum of the heterogeneity ing one 4-week-long nursing home trial
identified, we extracted or calculated 2 statistics of the subgroups from the in Belgium (RIS-BEL-14, n=39), one 12-
total drug and placebo exposure dura- overall 2 statistic and comparing the week-long multicenter nursing home
tion (patient-years of treatment) after result with a 2 distribution with a df trial terminated early (RIS-INT-83,
noticing that this information could be of 1 less than the number of n=18), and one 12-week-long outpa-
obtained from several sources, includ- subgroups.35 tient trial in Germany, which used het-
ing a presentation by the FDA, 23 a As an additional analysis, rates for erogeneous, patients with organic psy-
manuscript under review,31 a letter to deaths were expressed as the number chosis syndrome (RIS-GER-16, n=815)
health care professionals from a phar- of events divided by total duration of (A. Greenspan, Janssen Pharmaceuti-
maceutical company,32 and any infor- exposure to drug or placebo in patient- cals Inc, written communication, De-
mation that could be estimated from an- years and RRs were calculated for each cember 7, 2004). No other placebo-
other article.33 drug and summarized by meta- controlled trials of atypical drugs in
analysis using a random-effects model. patients with dementia were identified.
Statistical Analyses
The dropouts or deaths and the num- RESULTS Trials and Patient Characteristics
ber randomized into each drug and pla- Search Flow Fifteen trials fulfilled criteria and
cebo for each trial were statistically The search strategy yielded 352 were included in the review
combined using the DerSimonian and MEDLINE citations and 118 Coch- (TABLE).4-8,11,12,33,37-51 These trials in-
Laird random-effects model for drop- rane Controlled Trials Register cita- cluded 3 aripiprazole trials, 2 in nurs-
outs and the Mantel-Haenszel fixed- tions. A total of 27 MEDLINE and 24 ing homes and 1 with outpatients (10-
effects model for deaths. Effects were Cochrane Controlled Trials Register week durations); 5 olanzapine trials, 2
expressed as ORs and absolute risk dif- citations were retrieved as likely in nursing homes, 3 with outpatients,
ferences with their 95% confidence in- placebo-controlled trials, from which and 1 with a risperidone comparison (6-
tervals (CIs) and P values using Re- 5 and 6 citations, respectively, were to 26-week durations); 5 risperidone
view Manager version 4.2.7 software retained as fulfilling search criteria trials, including the outpatient trial
(The Cochrane Collaboration, Ox- (FIGURE 1). One placebo-controlled above with an olanzapine compari-
ford, England). Effects were calcu- trial of olanzapine (n = 16 patients)36 son, 4 in nursing homes, 1 with a halo-
lated for each drug-placebo contrast, as was not included because the only peridol comparison (8- to 12-week du-
meta-analytic summaries for each drug, available documentation was an rations); and 3 quetiapine nursing
and for all atypical drugs combined. A abstract with inadequate informa- homes trials, 1 with a haloperidol com-
funnel plot in which sample size was tion. The 6 trials from the Cochrane parison and another with rivastig-
plotted against the log OR of the out- results included the 5 trials from mine (10- to 26-week durations). One
come was used to evaluate potential re- MEDLINE; 5 were primary trial trial is counted both as a risperidone
trieval bias and to compare the pub- articles, 1 was a review with informa- trial and an olanzapine trial. Thus, 11
lished trials with the nonpublished tion about an olanzapine trial not trials were performed in nursing homes
trials. contained elsewhere.37 One recently and 4 with outpatients. Eight trials al-
2 Tests and the I2 statistic derived published trial of quetiapine was lowed dosage adjustment, 2 trials were
from the 2 values were used to test het- included but was not identified in fixed-dose, and 5 trials were dose-
erogeneity among the contrasts. I2 ap- the literature search because it had ranging with 2 or 3 fixed doses of study
proximates the proportion of total varia- not been published when the search drug.
tion in the effect size estimates that is was performed.33 Overall, 3353 patients were random-
due to heterogeneity rather than sam- Twenty-four posters and slide pre- ized to drug (603 were randomized to
pling error.34 An error P.20 and I2 sentations from medical conferences aripiprazole, 1184 to olanzapine, 391
1936 JAMA, October 19, 2005Vol 294, No. 15 (Reprinted) 2005 American Medical Association. All rights reserved.

Downloaded From: on 02/25/2013


to quetiapine, and 1175 to risperi- extent of cognitive impairment ranged (95% CI, 0.004-0.02; P=.01). There was
done) and 1757 were randomized to from mild to severe with 13 trials hav- no significant heterogeneity among the
placebo. In 2 trials, one comparing ing mean Mini-Mental State Examina- outcomes (for OR: 152 =8.45, P=.90, and
quetiapine and the other comparing ris- tion scores of 11.3 (range of means per for risk difference: 152 =13.63, P=.55;
peridone with haloperidol and pla- trial, 5.4-21.5) on a 30-point scale. I2 =0% for both analyses; FIGURE 2). A
cebo, 293 were randomized to funnel plot graphing log ORs against
haloperidol. Meta-analysis Outcomes sample size did not show evidence of
Overall, 87% of all patients had Alz- We found 118 deaths in the atypical an- selection bias with symmetry around
heimer disease; the weighted mean (SD) tipsychotic drug groups and 40 in the the mean overall effect, or asymmetry
age per trial was 81.2 (7.8) years; and placebo groups, a simple pooled inci- between published and unpublished
70% were women. Nine trials allowed dence of 3.5% and 2.3% per trial, re- trials (data not shown).
patients with Alzheimer disease only spectively. The overall OR by meta- The risk differences for death in pa-
and comprised 53% of the patients; 6 analysis for death in patients treated with tients treated with aripiprazole vs pla-
allowed patients to have various de- antipsychotic drugs compared with pla- cebo were 0.01 (95% CI, 0.01 to 0.03;
mentia diagnoses and included 73% of cebo was 1.54 (95% CI, 1.06-2.23; P=.20); for olanzapine vs placebo, 0.01
patients with Alzheimer disease. The P=.02), and the risk difference was 0.01 (95% CI, 0.00 to 0.03; P = .07); for
quetiapine vs placebo, 0.02 (95% CI,
0.01 to 0.05; P=.22); and for risperi-
Figure 1. Trials Identification and Selection Process done vs placebo, 0.01 (95% CI, 0.01
to 0.02; P=.33). All but 3 trials showed
513 Potentially Relevant Citations From Initial Search risk differences in favor of the placebo
352 From MEDLINE
118 From Cochrane Controlled Trials Register group.
24 Posters/Presentations From Conferences
18 RCT Citations From Miscellaneous Sources
We found 1079 all-cause dropouts
1 RCT From a Recent Unindexed Publication (32.2%) among the drug-treated groups
and 551 (31.4%) among the placebo-
419 Citations Excluded (Not RCTs, Not Placebo-Controlled, treated groups (FIGURE 3). Overall, there
or Redundancy)
325 Excluded From MEDLINE was no significant difference in drop-
94 Excluded From Cochrane outs by meta-analysis, although there
was significant heterogeneity among
94 Citations Retrieved for Detailed Review
dropouts from trial to trial and drug to
24 From Cochrane drug ( 152 =30.89, P =.009; I2 =51.4%).
24 From Posters/Presentations
18 From Miscellaneous Sources The risk differences for dropouts in pa-
1 RCT From a Recent Unindexed Publication tients treated with aripiprazole vs pla-
cebo were 0.07 (95% CI, 0.15 to 0.01;
55 Excluded P=.10); for olanzapine vs placebo, 0.06
40 Excluded From MEDLINE and Cochrane (Not RCTs, Not
Placebo-Controlled, Not Demented, Intra-muscular (95% CI, 0.02 to 0.15; P = .12); for
Administration, Inadequate Data [n = 1], or Redundancy)
quetiapine vs placebo, 0.02 (95% CI,
18 From Cochrane 0.08 to 0.11; P=.73); and for risperi-
11 Posters/Presentations (Redundancy)
4 RCTs From Miscellaneous Sources done vs placebo, 0.03 (95% CI, 0.03
1 Intramuscular Treatment to 0.08; P = .31). The weak statistical
2 Did Not Have Data Available (N = 39, N = 18)
1 Nonspecific Diagnosis and Unavailable Data (N = 815) trends for more patients receiving
olanzapine vs placebo and fewer pa-
39 Prospective Placebo-Controlled RCTs tients receiving aripiprazole vs pla-
6 From Cochrane
cebo to dropout may have contributed
13 From Posters/Presentations to the overall heterogeneity of the ef-
14 Citations From Miscellaneous Sources
1 RCT From a Recent Unindexed Publication fects. There was no association be-
tween risk for death per contrast with
14 Excluded dropouts per contrast (log OR of death
5 RCTs Excluded From MEDLINE (Redundancy With Cochrane)
9 Citations From Miscellaneous Sources (Redundancy With Other
vs log OR of dropouts: r =0.23, df =14,
Sources or Inadequate Information) P =.40).
Subgroup analyses did not reveal het-
15 Placebo-Controlled RCTs Included (Documented by erogeneity between trials of patients of
7 Primary Publications, 13 Posters/Presentations, and
5 Miscellaneous Sources) higher cognitive function (Mini-
Mental State Examination score 10)
RCT indicates randomized controlled trial. compared with lower cognitive function,
2005 American Medical Association. All rights reserved. (Reprinted) JAMA, October 19, 2005Vol 294, No. 15 1937

Downloaded From: on 02/25/2013


Table. Characteristics of Placebo-Controlled Trials and Patients Included in the Analyses

No. of
Patients Key Inclusion Daily Medication Duration, Age, Mean Women, MMSE Score,
Source* Randomized Criteria Dose, mg Location wk (SD) [Range], y % Mean (SD)
CN 138-00438 487 AD with psychosis 2, 5, and 10 dose Nursing home 10 82.5 [56-97] 79 12.4 (4.4)
CN 138-00539-41 256 AD with psychosis 2-15 (mean, 8.6) Nursing home 10 83 [59-96] 76 12.9
CN 138-00640-42 208 AD with psychosis 2-15 (mean, 10) Outpatient 10 81.5 (6.5) [56-99] 72 13.6 (4.3)
HGAO37,43-45 238 AD with psychosis 1-8 (modal dose, Outpatient 8 78.6 [64-94] 66 NA
HGEU6,45 206 AD with agitation, 5, 10, and 15 dose Nursing home 6 82.8 (6.6) [61-97] 61 6.7 (6.4)
delusions, or groups
hallucinations, not
HGGU45-47 494 Dementia with 2.5-10 (mean, 5.2); Outpatient 10 78.4 (7.4) 66 14.5 (5.6)
hallucinations or risperidone:
delusions (78% 0.5-2 (mean, 1.0)
AD, 5% vascular
dementia, 17%
HGIC45,48 268 AD, nonpsychotic, 5 Outpatient 26 78 (8.0) 56 21.5 (3.6)
MMSE score of
HGIV8,45 652 AD with delusions or 1, 2.5, 5, and 7.5 Nursing home 10 76.6 (10.4) 75 13.7 (5.1)
hallucinations dose groups
Ballard33 80 AD with agitation 50-100; also Nursing home 26 83.8 (7.7) 80 NA
5077 US-03911,12 378 Elderly patients with 25-600 (median, Nursing home 10 83.9 (6.5) [66-99] 73 12.8 (5.4)
psychosis, not 97); haloperidol:
bedridden (75% 0.5-12 (median,
AD, 15% vascular 1.9)
dementia, 10%
5077 US-04649 333 Dementia with 100 and 200 dose Nursing home 10 83.2 (7.5) 74 5.4 (4.0)
agitation (73% AD, groups
7% vascular
dementia, 8%
RIS-AUS-057 345 Dementia with 0.50-2 (mean, 0.95) Nursing home 12 82.7 (7.1) 71 5.3 (8.0)
aggression, MMSE
score of 23 (58%
AD, 29% vascular
demetia, 13%
RIS-INT-245,50 344 Dementia, MMSE 0.50-4 (mean, 1.1); Nursing home 12 81 [56-97] 56 8.4 (7.8)
score of 23, haloperidol:
BEHAVE-AD scale 0.50-4 (mean,
8 (67% AD, 26% 1.2)
vascular dementia,
7% mixed)
RIS-USA-634 625 Dementia, MMSE 0.5, 1, and 2 dose Nursing home 12 82.7 (7.7) 68 6.6 (6.3)
score of 23, groups
8 (73% AD, 15%
vascular dementia,
12% mixed)
RIS-USA-23251 473 AD with psychosis, 0.5-1.5 (mean, 1.0) Nursing home 8 83.3 (7.3) 77 13.2 (5.0)
MMSE score of
5-23, ambulatory
with assistance
Abbreviations: AD, Alzheimer disease; BEHAVE-AD, Behavioral Pathology in Alzheimer Disease Rating Scale (a rating scale for behavioral symptoms); MMSE, Mini-Mental State Ex-
amination; NA, not assessed.
*Unique identification code, which identifies the study or the collection of posters, abstracts, unpublished manuscripts, or published trials of the study drug.

1938 JAMA, October 19, 2005Vol 294, No. 15 (Reprinted) 2005 American Medical Association. All rights reserved.

Downloaded From: on 02/25/2013


patients with psychosis of Alzheimer dis- increased risks with individual drugs peridone trial (A. Greenspan, Janssen
ease compared with those trials that did (FIGURE 4). Pharmaceuticals Inc, written commu-
not select patients on this basis, trials that Although not a planned analysis, data nication, December 7, 2004) and from
included inpatients compared with out- were available from the 2 contrasts with the quetiapine trial,11,12 we calculated
patients, or among the 4 drugs in- haloperidol from a 12-week risperi- an RR for haloperidol of 2.07 (95% CI,
cluded. done trial5 and a 10-week quetiapine 0.78-5.51; P=.15).31
trial11,12 and were combined. There were
Ad Hoc Analyses 15 deaths (6.2%) with haloperidol and COMMENT
Using available data on total exposure 9 (3.8%) with placebo among 243 pa- Overall, the use of atypical antipsy-
to drug or placebo, we calculated RRs tients receiving haloperidol and 239 pa- chotic drugs for relatively brief periods
for death by pooling data for each drug, tients receiving placebo. Risk for death of less than 8 to 12 weeks was associ-
resulting in an overall RR of 1.65 (95% was calculated as an OR of 1.68 (95% ated with a small increased risk for death
CI, 1.19-2.29; P = .003) for the atypi- CI, 0.72-3.92; P = .23). Using expo- compared with placebo. The increased
cal drugs combined and weak trends for sure data for haloperidol from the ris- risk only could be identified when the

Figure 2. Deaths by Individual Comparisons by Drugs and Overall Compared With Placebo


Treatment, Placebo, OR
No. of Events/ No. of Events/ (95% CI) Favors Favors
Source Total No. Total No. (Fixed-Effects Model) Treatment Control
CN 138-00438 15/366 3/121 1.68 (0.48-5.91)
CN 138-00539-41 2/131 3/125 0.63 (0.10-3.84)
CN 138-00640-42 4/106 0/102 9.00 (0.48-169.32)

Subtotal 21/603 6/348 1.73 (0.70-4.30)

Test for Heterogeneity 22 = 2.42, I 2 = 17.2% (P = .30)
Test for Overall Effect z = 1.18 (P = .24)

HGAO37,43-45 3/120 2/118 1.49 (0.24-9.06)
HGEU6,45 6/159 0/47 4.02 (0.22-72.73)
HGGU45-47 6/204 1/94 2.82 (0.33-23.75)
HGIC45,48 1/178 1/90 0.50 (0.03-8.13)
HGIV8,45 15/523 2/129 1.88 (0.42-8.30)

Subtotal 31/1184 6/478 1.91 (0.79-4.59)

Test for Heterogeneity 24 = 1.34, I 2 = 0% (P = .85)
Test for Overall Effect z = 1.44 (P = .15)

5077 US-03911,12 4/124 4/125 1.01 (0.25-4.12)
5077 US-04649 16/241 3/92 2.11 (0.60-7.42)
Ballard33 1/26 0/29 3.47 (0.14-88.99)

Subtotal 21/391 7/246 1.67 (0.70-4.03)

Test for Heterogeneity 22 = 0.82, I 2 = 0% (P = .66)
Test for Overall Effect z = 1.15 (P = .25)

HGGU45-47 4/196 1/94 1.94 (0.21-17.58)
RIS-AUS-057 6/167 5/170 1.23 (0.37-4.11)
RIS-INT-245,50 1/115 5/114 0.19 (0.02-1.66)
RIS-USA-23251 9/235 6/238 1.54 (0.54-4.40)
RIS-USA-634 25/462 5/163 1.81 (0.68-4.80)

Subtotal 45/1175 22/779 1.30 (0.76-2.23)

Test for Heterogeneity 24 = 3.69, I 2 = 0% (P = .45)
Test for Overall Effect z = 0.94 (P = .35)

Overall 118/3353 41/1851 1.54 (1.06-2.23)

Test for Heterogeneity 15 2
= 8.45, I 2 = 0% (P = .90)
Test for Overall Effect z = 2.28 (P = .02)

0.01 0.1 1.0 10 100

OR (95% CI)

CI indicates confidence interval; OR, odds ratio.

*Unique identification code which identifies the study or the collection of posters, abstracts, unpublished manuscripts, or published trials of the study drug. The total
number of placebo patients is 1757 and deaths, 40. The trial HGGU placebo group is used for both risperidone and olanzapine comparisons.

2005 American Medical Association. All rights reserved. (Reprinted) JAMA, October 19, 2005Vol 294, No. 15 1939

Downloaded From: on 02/25/2013


atypical drugs were combined in a meta- estimates and also mitigates the likeli- presentations with partial data that had
analysis. The meta-analyses of each drug hood of bias or significant errors in our to be reconstructed. In this respect, it
were not statistically significant, al- acquisition of the available data. The is notable that most of the unpub-
though the point estimates of the ORs fact that 12 of the 15 atypical drug com- lished trials have been completed for
ranged between 1.3 and 1.9. The upper parisons showed more deaths with an- several years and publication is long
bounds of the CIs, however, ranged from tipsychotic drug than placebo lessens overdue.
2.2 to 4.6 and are compatible with the the possibility that there was bias in the The unpublished trials, however,
possibility of moderately increased risks. ascertainment and reporting of the were of similar quality to the pub-
The supplementary analysis of risk rates deaths in any particular trial. Never- lished trials. All were required to be ran-
using time of exposure to drug sup- theless, because any bias no matter how domized and double-blinded. All but
ported this finding. small could be additive to the overall one of the trials were sponsored and
A recent FDA public health advi- effect, it would have been better if clear conducted by drug manufacturers. The
sory27 reporting an increased risk be- unambiguous complete trial reports had likely reasons for the delays in publi-
tween 1.6 and 1.7 is consistent with our been available rather than posters and cation were that most did not show sta-

Figure 3. All-Cause Dropouts by Individual Comparisons by Drugs and Overall Compared With Placebo

All-Cause Dropouts

Treatment, Placebo, OR
No. of Events/ No. of Events/ (95% CI) Favors Favors
Source Total No. Total No. (Random-Effects Model) Treatment Control
CN 138-00438 147/366 56/121 0.78 (0.52-1.18)
CN 138-00539-41 44/131 61/125 0.53 (0.32-0.88)
CN 138-00640-42 18/106 18/102 0.95 (0.47-1.96)

Subtotal 209/603 135/348 0.71 (0.52-0.96)

Test for Heterogeneity 22 = 2.13, I 2 = 5.9% (P = .35)
Test for Overall Effect z = 2.23 (P = .03)

HGAO37,43-45 57/120 57/118 0.97 (0.58-1.61)
HGEU6,45 43/159 11/47 1.21 (0.57-2.60)
HGGU45-47 77/204 19/94 2.39 (1.34-4.26)
HGIC45,48 71/178 24/90 1.82 (1.05-3.18)
HGIV8,45 146/523 38/129 0.93 (0.61-1.42)

Subtotal 394/1184 149/478 1.34 (0.92-1.96)

Test for Heterogeneity 24 = 9.48, I 2 = 57.8% (P = .05)
Test for Overall Effect z = 1.51 (P = .13)

5077 US-03911,12 29/124 36/125 0.75 (0.43-1.33)
5077 US-04649 86/241 32/92 1.04 (0.63-1.72)
Ballard33 3/26 0/29 8.79 (0.43-178.69)

Subtotal 118/391 68/246 0.95 (0.58-1.58)

Test for Heterogeneity 22 = 2.86, I 2 = 30.1% (P = .24)
Test for Overall Effect z = 0.19 (P = .85)

HGGU45-47 61/196 19/94 1.78 (0.99-3.21)
RIS-AUS-057 45/167 56/170 0.75 (0.47-1.20)
RIS-INT-245,50 47/115 40/114 1.28 (0.75-2.18)
RIS-USA-23251 59/235 59/238 1.02 (0.67-1.54)
RIS-USA-634 146/462 44/163 1.25 (0.84-1.86)

Subtotal 358/1175 218/779 1.14 (0.88-1.47)

Test for Heterogeneity 24 = 5.95, I 2 = 32.7% (P = .20)
Test for Overall Effect z = 0.97 (P = .33)

Overall 1079/3353 570/1851 1.07 (0.88-1.30)

Test for Heterogeneity 15 2
= 30.89, I 2 = 51.4% (P = .009)
Test for Overall Effect z = 0.68 (P = .50)

0.01 0.1 1.0 10 100

OR (95% CI)

CI indicates confidence interval; OR, odds ratio.

*Unique identification code, which identifies the study or the collection of posters, abstracts, unpublished manuscripts, or published trials of the study drug. The total
number of placebo patients is 1757 and dropouts, 551. The trial HGGU placebo group is used for both risperidone and olanzapine comparisons.

1940 JAMA, October 19, 2005Vol 294, No. 15 (Reprinted) 2005 American Medical Association. All rights reserved.

Downloaded From: on 02/25/2013


Figure 4. Deaths Based on Total Drug and Placebo Exposures Pooled by Drug


Treatment, Placebo, RR
No. of Events/ No. of Events/ (95% CI) Favors Favors
Drug Total No. Total No. (Random-Effects Model) Treatment Control
Aripiprazole 21/95 6/54 1.99 (0.86-4.62)
Olanzapine 31/217 6/97 2.31 (1.00-5.35)
Quetiapine 21/58 7/36 1.86 (0.88-3.93)
Risperidone 45/201 22/133 1.35 (0.85-2.14)

Overall 118/571 41/320 1.65 (1.19-2.29)

Test for Heterogeneity 23 = 1.63, I 2 = 0% (P = .65)
Test for Overall Effect z = 3.01 (P = .003)

0.1 1.0 10
RR (95% CI)

Total population is patient-years exposure to treatment (drug) or placebo. CI indicates confidence interval; RR, relative risk. Exposure time to treatment for 4 risperi-
done and 3 olanzapine trials was obtained from data presented by the US Food and Drug Administration.23 Exposure time for 1 risperidone trial,46 2 olanzapine trials,8,46
and 1 quetiapine trial33 was estimated from sample sizes, trial lengths, and dropout rates. Exposure time for aripiprazole was calculated from sample sizes and incidence
data provided in a letter from Bristol-Myers Squibb (February 10, 2005),32 and for 2 quetiapine trials from data provided in Schneider et al.31 The total number of
placebo deaths is 40 and placebo exposure is 303 patient-years (see footnote to Figure 2).

tistically significant results on their pri- It was beyond the potential of our risk estimates above by approximately
mary efficacy outcomes, perhaps meta-analysis to review each death, al- 25% and result in a similarly mark-
lessening the enthusiasm of the spon- though careful investigation of each edly lower projected RR for death of
sors to submit the manuscripts. The case is warranted. There is insufficient 1.15 over the course of a years treat-
poster presentations may have repre- information available on individual ment compared with the observed RR
sented a compromise in this regard. cases, causes or circumstances, base- of 1.65 during the length of the trials.
With respect to death as an outcome, line clinical characteristics, medical Unfortunately, there is a lack of any
however, a funnel plot of the log ORs conditions, and concurrent medica- long-term controlled studies that can
against sample size was symmetrical tions. An individual patient meta- resolve this issue.
around the mean overall effect, thus not analysis might be able to identify char- Expressing the absolute risk differ-
providing evidence for selection bias. acteristics associated with mortality ence as an inverse yields a number
Moreover, death and dropout rates did potentially due to drugs. Considering needed to harm of 100 with a very broad
not differ between the 6 published trials that future trials are unlikely, the phar- 95% CI from 53 to 1000, implying that
(death: OR, 1.42; 95% CI, 0.80-2.51; maceutical manufacturers, the own- there may be 1 death due to atypical
and dropouts: OR, 1.05; 95% CI, 0.84- ers of most of these data, might be en- drug use for every 100 patients treated
1.33) and the 9 unpublished trials couraged to allow their data to be over 10 to 12 weeks. Considering that
(death: OR, 1.63; 95% CI, 1.00-2.65; combined and analyzed by an indepen- many of these trials demonstrated that
and dropouts: OR, 1.07; 95% CI, 0.81- dent organization without a material in- these medications are only modestly ef-
1.42). terest in the outcomes. fective with numbers needing to treat
It can be appreciated that excess mor- The absolute risk difference of 1% ex- ranging from 4 to 12 in specific meta-
tality could not have been recognized cess deaths (with upper limits to the CI analyses, the likelihood for helping vs
by examining any individual trial. The of 4% to 5% depending on the drug), harming may be rather modest as well,
events were too sparse and the trials too considering that the trials were about such that for every 9 to 25 persons
small to be able to meaningfully assess 8 to 12 weeks and assuming propor- helped in these trials there possibly will
for a dose response that might make at- tionality in the risk (and that the risk be 1 death.
tribution even more compelling. Sub- is not just an early effect of treat- Considering the consistency of the
group analyses, limited in statistical ment), implies a 4% to 5% risk differ- risks among the trials, it is likely that
power, were unenlightening and did not ence over a year of treatment with an there is increased risk from any of the
provide any evidence that risk might upper bound for some drugs as high as drugs and not from a particular atypi-
differ by drug, psychotic symptoms as 25%. An argument could be made, how- cal drug. This is supported by the ob-
criteria for inclusion, living arrange- ever, that the risk is highest earlier in servation that the risk for haloperidol,
ments, or cognitive impairment. Analy- treatment with risk diminishing over which was randomly and double-
ses by age, sex, ability to ambulate, or time. An assumption of no excess blindedly assigned in 2 of the trials, was
potential interactions could not have deaths over placebo from after the first similar in magnitude to that of the atypi-
been performed unless individual pa- 3 months of treatment and continuing cal drugs, although not statistically sig-
tient data had been available. for a year would effectively decrease the nificant. A fair speculation is that in
2005 American Medical Association. All rights reserved. (Reprinted) JAMA, October 19, 2005Vol 294, No. 15 1941

Downloaded From: on 02/25/2013


frail, often medically ill, elderly pa- during the first 1 to 4 weeks of treat- Obtained funding: Schneider.
Administrative, technical, or material support:
tients with dementia a wide range of ment, lead to the consideration that an- Schneider, Dagerman.
classes of drugs (antidepressants, seda- tipsychotic drugs should be pre- Study supervision: Schneider.
Financial Disclosures: Dr Schneider has served as a
tives, hypnotics, anxiolytics, mood sta- scribed and dosage adjusted with the consultant to AstraZeneca, Eli Lilly, Bristol-Myers
bilizers, anticonvulsants, and cardio- expectation of clinical improvement Squibb, Johnson & Johnson, Novartis, and Pfizer, all
vascular or antihypertensive drugs) within that time. If improvement is not manufacturers of antipsychotic medications. No other
authors reported financial disclosures.
similarly could be associated with this observed, the medication could be dis- Funding/Support: This study was supported by grant
level of risk. This review also demon- continued. Moreover, because a sub- 03-75274 from the Alzheimers Disease Centers of Cali-
fornia and by the University of Southern California Alz-
strates that there is a substantially larger stantial proportion of patients respond- heimers Disease Research Center (NIH AG 05142).
body of placebo-controlled trials of ing may be responding to in-study Role of the Sponsor: The funding organizations did
not participate in the design and conduct of the study,
atypical drugs vs other central ner- effects, increased nursing care, envi- in the collection, analysis, and interpretation of the data,
vous systemacting drugs in very el- ronmental changes, or changes in medi- or in the preparation, review, or approval of the manu-
derly patients (81 years), mainly in cal status, and not actually to medica-
an institutionalized population with de- tion, n of one trials of medication
mentia, and this collection of trials is withdrawal could be undertaken at fre- REFERENCES
larger than has previously been iden- quent intervals to assess continuing 1. 42 CFR 483.
tified in the published literature.9,52 need. Of course, if the risks for serious 2. Alexopoulos GS, Silver JM, Kahn DA, Frances A,
Carpenter D. The Expert Consensus Guideline Se-
An absence of evidence for either ef- adverse events are more related to ini- ries: Treatment of Agitation in Older Persons With
ficacy or safety with nonatypical anti- tiation of medication than to continu- Dementia. Minneapolis, Minn: McGraw-Hill Co; 1998:
psychotic drugs was observed and the ation, starting and stopping might ex- 3. Alexopoulos GS, Jeste DV, Chung H, Carpenter D,
existing trials are not of adequate sta- pose patients to greater risk. Ross R, Docherty J. The Expert Consensus Guideline
Series: Treatment of Dementia and Its Behavioral
tistical power or quality to detect any As a meta-analysis, our results should Disturbances. Minneapolis, Minn: McGraw-Hill Co;
increased risk at the level reported be taken as hypothesis-generating for 2005:1-111.
4. Katz IR, Jeste DV, Mintzer JE, Clyde C, Napoli-
herein with atypical drugs. It is plau- an increased risk for deaths in pa- tano J, Brecher M. Comparison of risperidone and pla-
sible that increased mortality is asso- tients with dementia receiving atypi- cebo for psychosis and behavioral disturbances asso-
ciated with the use of many or all classes cal antipsychotic drugs. Although the ciated with dementia: a randomized, double-blind trial:
Risperidone Study Group. J Clin Psychiatry. 1999;60:
of drugs used to treat these symptoms findings were consistent from trial to 107-115.
and syndromes. In elderly patients, it trial and from drug to drug with no evi- 5. De Deyn PP, Rabheru K, Rasmussen A, et al. A ran-
domized trial of risperidone, placebo, and haloperi-
is likely that any given medication will dence for heterogeneity, they in- dol for behavioral symptoms of dementia. Neurology.
both help and harm, and the safety of cluded relatively small numbers of pa- 1999;53:946-955.
6. Street JS, Clark WS, Gannon KS, et al. Olanzapine
a drug must be considered in the con- tients and sparse events within any 1 treatment of psychotic and behavioral symptoms in
text of known efficacy. Ironically, analy- antipsychotic drug. It is only when all patients with Alzheimer disease in nursing care facili-
ses such as these expose the risks of per- trials are combined that a statistically ties: a double-blind, randomized, placebo-controlled
trial. Arch Gen Psychiatry. 2000;57:968-976.
forming clinical trials in elderly patients significant effect is found. No drug is 7. Brodaty H, Ames D, Snowdon J, et al. A random-
as well and are likely to discourage individually responsible for the effect, ized placebo-controlled trial of risperidone for the treat-
ment of aggression, agitation, and psychosis of
pharmaceutical companies, govern- but rather each contributes to the over- dementia. J Clin Psychiatry. 2003;64:134-143.
ments, and institutions from undertak- all effect. This effect may not be lim- 8. De Deyn PP, Carrasco MM, Deberdt W, et al.
Olanzapine versus placebo in the treatment of psy-
ing future trials in this area. ited to atypical drugs as a class and may chosis with or without associated behavioral distur-
These findings emphasize the need be associated with haloperidol and other bances in patients with Alzheimers disease. Int J Geri-
atr Psychiatry. 2004;19:115-126.
to consider certain changes in some drugs that have not been subjected to 9. Sink KM, Holden KF, Yaffe K. Pharmacological treat-
clinical practices. Antipsychotic drugs efficacy trials in elderly patients with ment of neuropsychiatric symptoms of dementia: a re-
have been dispensed fairly frequently dementia. view of the evidence. JAMA. 2005;293:596-608.
10. Chan W, Lam L, Choy C, Leung V, Li S, Chiu H.
to patients with dementia and used for Author Affiliations: Departments of Psychiatry and the A double-blind randomised comparison of risperi-
long periods. The established risks for Behavioral Sciences (Dr Schneider and Ms Dagerman), done and haloperidol in the treatment of behavioural
Neurology (Dr Schneider), and Department of Preven- and psychological symptoms in Chinese dementia
cerebrovascular adverse events to- tive Medicine (Mr Insel), Keck School of Medicine, and patients. Int J Geriatr Psychiatry. 2001;16:1156-1162.
gether with the present observations Leonard Davis School of Gerontology (Dr Schneider), 11. Tariot PN, Schneider LS, Katz IR, Mintzer JE, Street
University of Southern California, Los Angeles. JS. Quetiapine in nursing home residents with Alzhei-
suggest that antipsychotic drugs should Author Contributions: Dr Schneider had full access to mers dementia and psychosis. Poster presented at:
be used with care in these patients. The all of the data in the study and takes responsibility for 15th Annual Meeting of the American Association for
fact that excess deaths and cerebrovas- the integrity of the data and the accuracy of the data Geriatric Psychiatry; February 24-27, 2002; Orlando,
analysis. Fla.
cular adverse events can be observed Study concept and design: Schneider. 12. Tariot PN, Schneider LS, Katz IR, Mintzer JE, Street
within 10 to 12 weeks of initiating Acquisition of data: Schneider, Dagerman, Insel. JS. Quetiapine in nursing home residents with Alzhei-
Analysis and interpretation of data: Schneider, mers dementia and psychosis [abstract]. Am J Geri-
medication, coupled with observa- Dagerman, Insel. atr Psychiatry. 2002;10(2 suppl 1):93.
tions from individual clinical trials re- Drafting of the manuscript: Schneider, Dagerman, Insel. 13. McShane R, Keene J, Gedling K, Fairburn C, Ja-
Critical revision of the manuscript for important in- coby R, Hope T. Do neuroleptic drugs hasten cogni-
sults that there is substantial improve- tellectual content: Schneider. tive decline in dementia? prospective study with nec-
ment in both drug and placebo groups Statistical analysis: Schneider, Dagerman, Insel. ropsy follow up. BMJ. 1997;314:266-270.

1942 JAMA, October 19, 2005Vol 294, No. 15 (Reprinted) 2005 American Medical Association. All rights reserved.

Downloaded From: on 02/25/2013


14. Chui HC, Lyness SA, Sobel E, Schneider LS. Ex- 29. The Cochrane Library, Issue 1, 2005. Chichester, zole in dementia of the Alzheimers type. Presented
trapyramidal signs and psychiatric symptoms predict England: John Wiley & Sons; 2005. at: 16th Annual Meeting of the American Associa-
faster cognitive decline in Alzheimers disease. Arch 30. Jeste DV, Finkel SI. Psychosis of Alzheimers dis- tion for Geriatric Psychiatry; March 1-4, 2003; Ho-
Neurol. 1994;51:676-681. ease and related dementias: diagnostic criteria for a nolulu, Hawaii.
15. Drevets W, Rubin E. Psychotic symptoms and the distinct syndrome. Am J Geriatr Psychiatry. 2000;8:29- 43. Street JS, Tollefson GD, Tohen M, et al. Olanza-
longitudinal course of senile dementia of the Alzhei- 34. pine for psychotic conditions in the elderly. Psychiatr
mer type. Biol Psychiatry. 1989;25:39-48. 31. Schneider LS, Tariot PN, Mintzer J, Minkwitz M, Ann. 2000;30:191-196.
16. Lopez O, Becker J, Brenner R, Rosen J, Bajaulaiye Zhong K. Cerebrovascular adverse events and queti- 44. Street JS, Kinon B, Stauffer V. Olanzapine in
O, Reynolds C. Alzheimers disease with delusions and apine: a pooled analysis in elderly patients with dementia. In: Tran PV, ed. Olanzapine (Zyprexa): A
hallucinations: neuropsychological and electrocephalo- dementia. Paper presented at: 18th Annual Meeting Novel Antipsychotic. Philadelphia, Pa: Lippincott Wil-
graphic correlates. Neurology. 1991;41:906-912. of the American Association for Geriatric Psychiatry; liams & Wilkins; 2000:416-426.
17. Stern Y, Albert M, Brandt J, et al. Utility of extra- March 3-6, 2005; San Diego, Calif. 45. Cavazzoni P, Young CA, Polzer J, et al. Inci-
pyramidal signs and psychosis as predictors of cogni- 32. Labeling Change for Abilify on Risk of CVA in dence of cerebrovascular adverse events and mortal-
tive and functional decline, nursing home admission, Elderly. Bristol-Myers Squibb Medical Letter. Febru- ity during antipsychotic clinical trials of elderly pa-
and death in Alzheimers disease: prospective analy- ary 10, 2005. tients with dementia. Paper presented at: 44th Annual
ses from the Predictors Study. Neurology. 33. Ballard C, Margallo-Lana M, Juszczak E, et al. New Clinical Drug Evaluation Unit (NCDEU) Meet-
1994;44:2300-2307. Quetiapine and rivastigmine and cognitive decline in ing (modifications added); June 1-4, 2004; Phoenix,
18. Stern Y, Mayeux R, Sano M, Hauser W, Bush T. Alzheimers disease: randomised double blind pla- Ariz.
Predictors of disease course in patients with probable cebo controlled trial. BMJ. 2005;330:874. 46. Deberdt W, Dysken MW, Rappaport SA, et al. Pla-
Alzheimers disease. Neurology. 1987;37:1649-1653. 34. Higgins J, Thompson S, Deeks J, Altman D. Mea- cebo-controlled comparison of olanzapine and ris-
19. Wooltorton E. Risperidone (Risperdal): in- suring inconsistency in meta-analysis. BMJ. 2003;327: peridone in the treatment of psychosis and behav-
creased rate of cerebrovascular events in dementia 557-560. ioral disturbances in patients with dementia. Paper
trials. CMAJ. 2002;167:1269-1270. 35. Deeks J, Altman D, Bradburn M. Statistical meth- presented at: American Geriatrics Society; May 17-
20. Medicines and Healthcare products Regulatory ods for examining heterogeneity and combining re- 21, 2004; Las Vegas, Nev.
Agency. Summary of clinical trial data on cerebrovas- sults from several studies in meta-analysis. In: Egger 47. Kennedy JS, Young CA, Hoffman VP, Feldman
cular adverse events (CVAEs) in randomised clinical M, Davey Smith G, Altman D, eds. Systematic Re- PD, Deberdt W. A placebo-controlled 10-week pro-
trials of risperidone conducted in patients with de- views in Health Care: Meta-Analysis in Context. 2nd spective comparison of the occurrence of falls in de-
mentia: March 9, 2004. Available at: http://medicines ed. London, England: BMJ Publication Group; 2001. mentia: olanzapine versus risperidone. Poster pre- Accessibility verified September 6, 2005. 36. Howanitz E, Wisotzek I. Olanzapine versus pla- sented at: International College of Geriatric
21. Wooltorton E. Olanzapine (Zyprexa): increased cebo in the treatment of behavioral disturbances as- Psychoneuropharmacology; October 10-12, 2002; Bar-
incidence of cerebrovascular events in dementia trials. sociated with vascular dementia. Presented at: 14th celona, Spain.
CMAJ. 2004;170:1395. Annual Meeting of the American Association for Geri- 48. Kennedy JS, Deberdt W, Micca J, et al. The effect
22. US Food and Drug Administration. Safety data on atric Psychiatry; February 23-26, 2001; San Fran- of olanzapine on cognition in patients with Alzhei-
Zyprexa (olanzapine): hyperglycemia and diabetes. cisco, Calif. mers disease without psychosis or agitation. Poster
Available at: 37. Satterlee WG, Reams SG, Burns PR, Hamilton S, presented at: International College of Geriatric Psy-
/2004/zyprexa.htm. Accessibility verified September Tran PV, Tollefson GD. A clinical update on olanza- choneuropharmacology; October 14-16, 2004; Basel,
6, 2005. pine treatment in schizophrenia and in elderly Alzhei- Switzerland.
23. Racoosin JA. Evaluating a safety signal in the post- mers disease patients [abstract]. Psychopharmacol Bull. 49. Zhong K, Tariot P, Minkwitz M, Devine N, Mint-
marking period: cerebrovascular adverse events as- 1995;31:534. zer J. Quetiapine for the treatment of agitation in
sociated with risperidone and olanzapine. Paper pre- 38. Breder C, Swanink R, Marcus R, et al. Dose- elderly institutionalized patients with dementia: a ran-
sented at: 17th Annual Meeting of the American ranging study of aripiprazole in patients with Alzhei- domized, double-blind trial. Poster presented at: 9th
Association for Geriatric Psychiatry; February 21-24, mers dementia. Presented at: 9th International Con- International Conference on Alzheimers Disease
2004; Baltimore, Md. ference on Alzheimers Disease and Related Disorders; and Related Disorders; July 17-22, 2004; Philadel-
24. Health CanadaTherapeutic Products Director- July 17-22, 2004; Philadelphia, Pa. phia, Pa.
ate, Risperdal warning letter, October 11, 2002. Avail- 39. Streim J, Breder C, Swanink R, McQuade R, Iwa- 50. De Deyn PP, DeSmedt G. Risperidone in the treat-
able at: moto T, Carson W. Flexible dose aripiprazole in psy- ment of behavioral disturbances in elderly patients with
/advisories-avis/prof/2002/risperdal_hpc-cps_e chosis of Alzheimers dementia (AD). Poster pre- dementia. Presented at: Eighth Congress of the In-
.html. Accessibility verified September 20, 2005. sented at: 157th American Psychiatric Association ternational Psychogeriatric Association; August 17-
25. 21 CFR 312.32. Annual Meeting; May 1-6, 2004; New York, NY. 22, 1997; Jerusalem, Israel.
26. Mosholder A. FDA oral presentation. Paper pre- 40. Kujawa M, Marcus R, Breder C, et al. Safety pro- 51. Mintzer J, Weiner M, Greenspan A, et al. Effi-
sented at: 42nd Annual New Clinical Drug Evalua- file of aripiprazole in psychosis of Alzheimers demen- cacy and safety of a flexible dose of risperidone ver-
tion Unit (NCDEU) Meeting; June 10-13, 2002; Boca tia: pooled analysis. Poster presented at: 9th Interna- sus placebo in the treatment of psychosis of Alzhei-
Raton, Fla. tional Conference on Alzheimers Disease and Related mers disease. Poster presented at: International College
27. US Food and Drug Administration. FDA Public Disorders; July 17-22, 2004; Philadelphia, Pa. of Geriatric Psychopharmacology; October 14-17,
Health Advisory: deaths with antipsychotics in el- 41. Stock E, Breder C, Goyvaerts H, et al. Safety pro- 2004; Basel, Switzerland.
derly patients with behavioral disturbances. Avail- file of aripiprazole in psychosis of Alzheimers demen- 52. Lee PE, Gill SS, Freedman M, Bronskill SE, Hillmer
able at: tia: pooled analysis. Poster presented at: American Psy- MP, Rochon PA. Atypical antipsychotic drugs in the
/antipsychotics.htm. Accessed April 13, 2005. chiatric Association Annual Meeting; May 1-6, 2004; treatment of behavioural and psychological symp-
28. Atypical antipsychotics in the elderly. Med Lett New York, NY. toms of dementia: systematic review. BMJ. 2004;
Drugs Ther. 2005;47:61-62. 42. De Deyn PP, Jeste DV, Mintzer JE, et al. Aripipra- 329:75.

2005 American Medical Association. All rights reserved. (Reprinted) JAMA, October 19, 2005Vol 294, No. 15 1943

Downloaded From: on 02/25/2013