ORIGINAL CONTRIBUTION

Suicide Risk in Bipolar Disorder

During ​Treatment With Lithium and

Divalproex
Frederick K. Goodwin, MD Bruce Fireman, MA Gregory E. Simon, MD Enid M. Hunkeler, MA Janelle Lee, MHA,

DrPH Dennis ​

B​ Revicki, public given mately ​IPOLAR ​year health 1.3% ​DISORDER ​PhD affecting to problem,

1.5% I​ S ​of approxi- A

​ ​the M
​ AJOR ​in any
US population.​1 ​In addition to the per- sonal anguish of affected individuals, bipolar disorder places substantial bur-
dens on the health care, social wel- fare, and criminal justice systems and on families, caregivers, and employ- ers. In
the World Health Organiza- tion’s Global Burden of Disease study, bipolar disorder ranked sixth among all medical
disorders in years of life lost to death or disability.​2 ​Suicide and sui- cide attempts are significant contribu- tors to
that premature mortality and dis- ability. Estimates of the lifetime risk of suicide in patients with bipolar disor- der
range from 8% to 20%, 10 to 20 times that in the US general popula- tion.​3-7​In a review of 31 studies includ- ing
nearly 10000 patients with recur- rent affective disorder (primarily bipolar), the proportion of deaths at- tributable to
suicide ranged from 9% to 60%, with a weighted mean of 18.9%.​8 ​Thus, it is critical to determine if main- tenance
treatment is associated with a reduction in suicide attempts and sui- cide mortality in this high-risk group. Although
a number of mood- stabilizing drugs are commonly used in
Context ​Several studies have suggested that lithium treatment reduces risk of sui- cide in bipolar disorder, but no
research has examined suicide risk during treatment with divalproex, the most commonly prescribed mood-stabilizing
drug in the United States. ​Objective ​To compare risk of suicide attempt and suicide death during treatment with
lithium with that during treatment with divalproex. ​Design and Setting ​Retrospective cohort study conducted at 2
large integrated health plans in California and Washington. ​Patients ​Population-based sample of 20638 health plan
members aged 14 years or older who had at least 1 outpatient diagnosis of bipolar disorder and at least 1 filled pre-
scription for lithium, divalproex, or carbamazepine between January 1, 1994, and De- cember 31, 2001. Follow-up for
each individual began with first qualifying prescription and ended with death, disenrollment from the health plan, or
end of the study period. ​Main Outcome Measures ​Suicide attempt, recorded as a hospital discharge di- agnosis or
an emergency department diagnosis; suicide death, recorded on death certificate. ​Results ​In both health plans,
unadjusted rates were greater during treatment with divalproex than during treatment with lithium for emergency
department suicide at- tempt (31.3 vs 10.8 per 1000 person-years; ​P ​.001), suicide attempt resulting in hos-
pitalization (10.5 vs 4.2 per 1000 person-years; ​P ​.001), and suicide death (1.7 vs 0.7 per 1000 person-years; ​P​=.04).
After adjustment for age, sex, health plan, year of diagnosis, comorbid medical and psychiatric conditions, and
concomitant use of other psychotropic drugs, risk of suicide death was 2.7 times higher (95% confidence in- terval
[CI], 1.1-6.3; ​P​=.03) during treatment with divalproex than during treatment with lithium. Corresponding hazard ratios
for nonfatal attempts were 1.7 (95% CI, 1.2-2.3; ​P= ​ .002) for attempts resulting in hospitalization and 1.8 (95% CI,
1.4-2.2; ​P .​ 001) for attempts diagnosed in the emergency department. ​Conclusion ​Among patients treated for bipolar
disorder, risk of suicide attempt and suicide death is lower during treatment with lithium than during treatment with
divalproex.
JAMA. 2003;290:1467-1473 ​www.jama.com
For editorial comment see p 1517.
©2003 American Medical Association. All rights reserved. ​(Reprinted) JAMA, ​September 17, 2003—Vol 290, No. 11 ​1467
Downloaded From: https://jamanetwork.com/ on 03/17/2019
the long-term management of bipolar disorder, lithium and the anticonvul- sant lamotrigine are the only drugs for
which long-term efficacy has been
established in at least 2 placebo- controlled randomized studies,​9-19 ​and these are the only drugs approved by the US
Food and Drug Administration for
Author Affiliations: ​Department of Psychiatry, George Washington University Medical Center, Washing- ton, DC (Dr Goodwin); Best Practice LLC (Dr
Good- win) and Center for Health Outcomes Research, MED- TAP International (Dr Revicki), Bethesda, Md; Division of Research, Kaiser Permanente
Medical Care Pro- gram, Oakland, Calif (Mr Fireman, Ms Hunkeler, and Dr Lee); Center for Health Studies, Group Health Co- operative, Seattle, Wash
(Dr Simon). ​Financial Disclosures: ​Dr Goodwin has received re- search support from Abbott, Glaxo, Solvay, Janssen,
Pfizer, Lilly, and Forest; has served on the speaker’s bureaus of Glaxo, Solvay, Janssen, Pfizer, Lilly, Ast- raZeneca, and Bristol-Myers Squibb; and has
served as a consultant for Glaxo, Solvay, Pfizer, Lilly, Bristol- Myers Squibb, Elan, and Novartis. Dr Simon has re- ceived a research grant from Abbott.
Corresponding Author and Reprints: ​Frederick K. Goodwin, MD, George Washington University Medi- cal Center, Department of Psychiatry, 2150
Pennsyl- vania Ave NW, Washington, DC 20037 (e-mail: fgood- win@mfa.gwu.edu).
SUICIDE RISK IN BIPOLAR DISORDER
1468 ​JAMA, ​September 17, 2003—Vol 290, No. 11 ​(Reprinted) ​©2003 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 03/17/2019
maintenance treatment of bipolar dis-
using emergency department (ED) visit
95% of prescriptions filled by GHC order. Furthermore, substantial data
and hospital discharge diagnoses. Sui-
members are obtained at GHC phar- have accumulated regarding the anti-
cide deaths were identified using both
macies.​24,25 ​suicide effects of lithium. In a recent
health plan mortality records and rel- meta-analysis of 33 studies (including
evant state death certificate records.
Inclusion/Exclusion Criteria ​primarily patients with bipolar disorder), Baldessarini et al​20 ​found that the an-
METHODS
This retrospective cohort study in- cluded all members aged 14 years or nual rate of attempted or completed sui-
Study Settings
older who had a record of outpatient cide was 0.197% during lithium treat-
Two managed care organizations par-
treatment for bipolar disorder and were ment compared with 2.57% without
ticipated in this study, Kaiser Perma-
enrolled in KP or GHC at any time from lithium, a 13-fold difference. In the
nente (KP) and Group Health Coopera-
January 1, 1994, to December 31, 2001 United States, use of lithium has de-
tive (GHC). Both plans provide
​ BLE ​1​). Institutional review boards clined over the last decade while use of
(​TA
comprehensive medical care, includ-
at GHC and KP approved all study pro- anticonvulsants (especially divalproex
ing mental health care. Kaiser Perma-
cedures. Specific inclusion criteria were and several newer agents) for treat-
nente serves an ethnically diverse popu-
(1) at least 1 outpatient diagnosis of bi- ment of bipolar disorder has steadily in-
lation of more than 3 million persons,
polar disorder type 1 or type 2 on or af- creased. Divalproex generates at least 10
about 30% of the entire population in
ter the 14th birthday (ie,​Diagnostic and t​ imes more sales revenue than does
the San Francisco Bay Area, Sacra-
Statistical Manual of Mental Disorders, ​lithium (Joanne Kearney, Solvay Phar-
mento, and nearby northern California
Fourth Edition[​ ​DSM-IV]d​ iagnosis codes maceuticals, written communication
counties. The group-model portion of
296.4x, 296.5x, 296.6x, 296.7, 296.89, based on IMS Health National Sales Per-
GHC (in which this study was con-
and 296.80) during the study period and spectives audit for fiscal year 2002, Au-
ducted) serves approximately 450000
(2) at least 1 prescription for lithium, di- gust 15, 2003), so industry-supported
members in western Washington State.
valproex, or carbamazepine filled at a KP education regarding divalproex far ex-
Most members of both plans are cov-
or GHC pharmacy during the study pe- ceeds that for lithium.
eredthroughemployer-purchasedplans,
riod. Specific exclusion criteria were (1) Surprisingly few studies address the
but approximately 15% are covered via
diagnosis of schizophrenia (​DSM-IV ​effect of anticonvulsant mood stabiliz-
contracts with Medicare, Medicaid, or
code 295.xx, except 295.7x) on more ers on suicide risk. A single random-
otherlow-incomeprograms.AtKP,there
than 1 occasion at any time during the ized trial by Thies-Flechtner et al​21​com-
are 16 large medical centers and 29 spe-
study period, (2) any diagnosis of schi- pared maintenance treatment with
cialty mental health clinics. At GHC,
zoaffective disorder (​DSM-IV c​ ode lithium or carbamazepine in 175 pa-
there are 24 medical centers and 6 spe-
295.7x) occurring before the first re- tients with bipolar disorder dis-
cialty mental health clinics. The clinics
corded diagnosis of bipolar disorder, or charged from psychiatric inpatient
at both KP and GHC are staffed by psy-
(3) any diagnosis of dementia or cogni- units. There were no suicide attempts
chiatrists, psychologists, licensed clini-
tive disorders occurring before the first or suicide deaths in the lithium group
cal social workers, psychiatric nurse
recorded diagnosis of bipolar disorder. compared with 9 suicide events in the
practitioners, psychiatric nurses, and
Patients with a diagnosis of schizoaf- carbamazepine group, a significant dif-
masters-level psychotherapists.
​ .01. No systematic data are
fective disorder occurring after the first ference at​P=
Kaiser Permanente members are gen-
 bipolar disorder diagnosis were in- available regarding suicide during treat-
erally similar to the non-KP popula-
cluded but censored on the date of the ment with divalproex, currently the
tion of northern California, although
first schizoaffective diagnosis. For each most widely used mood stabilizer in the
poor and elderly persons are some-
patient selected, the period of observa- United States (Sandra Mertz, Abbott,
what underrepresented in KP.​22​Group
tion began with the first filled prescrip- oral communication based on IMS
Health Cooperative members are com-
tion for lithium, divalproex, or carba- Health and Verispan/Scott-Levin data
parable with Seattle-area residents ex-
mazepine during the study period and for July 2003, August 26, 2003).
cept for a higher educational level and
ended with death, disenrollment from This study examined the risk of at-
a lower percentage of high-income resi-
the health plan, or the end of the study tempted and completed suicide among
dents.​23 ​The computerized informa-
period, December 31, 2001. patients treated for bipolar disorder in
tion systems at KP and GHC include 2 large integrated health plans. Admin-
data on all hospital discharges, ED vis-
Outcome Measures ​istrative databases were used to iden-
its, outpatient clinic visits, and outpa-
We identified 3 different classes of sui- tify those treated for bipolar disorder,
tient prescriptions filled in KP or GHC
cide-related events: suicide mortality, assess potential confounding factors,
pharmacies. An estimated 96% of pa-
suicide attempts resulting in hospital- and ascertain periods of exposure to
tients with bipolar disorder at KP had
ization, and suicide attempts or sui- lithium, divalproex, and carbamaze-
pharmacy benefits in 2001. Surveys of
cidal behavior resulting in ED visits but pine. Suicide attempts were identified
GHC members indicate that more than
not hospital admissions.
Re- vision ​codes X60 to X84
SUICIDE RISKand
IN BIPOLAR
Y87.0 for 1999 and DISORDER
2000. State
mortality files were not yet available
Suicide mortality was identified for 2001; conse- quently, the last year
us- ing mortality files from state of the study pe- riod was included in
depart- ments of health for the period analyses of sui- cide attempts but not
January 1, 1994, through December deaths.
31, 2000. Suicide deaths were Suicide attempts were identified
identified based on ​International from computerized records of all ED
Classification of Diseases, Ninth visits or inpatient discharges with an
Revision (​ ​ICD-9)​ codes E950 to E959 ICD-9 ​diagnostic code in the range of
for 1994 to 1998 and on ​Interna- E950 to E959 for the period January 1,
tional Classification of Diseases, 10th 1994, to December 31, 2001. Specific
 check- boxes for “suicide gesture,”
“suicide at- tempt,” and “suicidal (%) unless otherwise indicated. †Percentages ever exposed to each mood stabilizer
ause individual patients could con-
behavior” were in- cluded on KP ED tribute follow-up time to more than 1 exposure group. ‡Data not
encounter forms for the entire study I​ nternational Classification of Diseases, Ninth Revision ​codes for suicide at- tempt
Health Cooperative’s emergency department encounter forms for most of the study
period. For most of the study period,
GHC encounter forms in- cluded no
such codes, so suicide at- tempt could
only be recorded by hand- written combination pe-
diagnosis. Therefore, analyses of ostly combination
suicide attempts identified from ED ncluded exposure
visits included only KP patients. Sui- ition from one
cide attempts identified from ED vis- her.
its and those identified from inpatient
ing Factors
stays were analyzed separately.
were used to
ng specific
Treatment Exposure ​Treatment
ns (present prior to
exposure was measured us- ing
r prescription)
computerized pharmacy records of all
oduce bias by
initial and refill prescriptions. For each
f mood stabilizer
prescription fill or refill, the pe- riod
es in preexisting
of exposure was considered to be- gin
re dis- order,
on the dispensing date and to con-
atic disor- der,
tinue for the expected duration of the
atitis or other
prescription (ie, drug supply days)
anxiety disorders
plus a grace period of either 14 days or
rder, agorapho-
25% of the expected prescription
ess disorder, gen-
duration (whichever was longer).
rder, social pho-
Based on this rule, each day during the
rder not
study period was classified by
ance abuse. Using
exposure to lithium, divalproex,
ods for
carbamazepine, a combi- nation of
tment exposure,
these mood stabilizers, or none of
ere used to
these 3 drugs. Depending on the
use of specific
pattern of medication switches, an
ncluding
individual patient might contribute fol-
cal antipsychotics,
low-up time to any or all of these ex-
​Characteristics of Patients Treated for Bipolar Disorder Between 1994 and 2001 hotics.
pating Health Plans​* Statistical Analysis ​Descriptive
Kaiser Permanente ​ statistics and crude (unad- justed)
Characteristics ​ (n = 16 248)
rates of suicide mortality and suicide
ealth Cooperative ​
(n = 4390) ​Age, mean (SD), y 38.7 (14.6) 37.9 (14.7) Female 10 429 attempt are reported for peri- ods of
5 (67) First mood stabilizer ​Lithium 7121 (44) 2050 (47) Divalproex
​ 7595 (47) 1676 (38) treatment with lithium, dival- proex,
or carbamazepine; combined
zepine 909 (6) 474 (11) Combination 623 (4) 190 (4) Ever exposed to ​Lithium† 8935
treatment (lithium plus divalproex or
9 (59) ​Divalproex† 10 171 (63) 2476 (56) Carbamazepine† 2265 (14) 1020 (23) Other carbamazepine); and no exposure to
opics ​Antidepressants 12 222 (75) 3337 (76) Typical
​ antipsychotics 3420 (21) 1061 (24) any of the 3 drugs. Rates are reported
antipsychotics 5218 (32) 1110 (25) Emergency department−ascertained ​suicide attempts per 1000 person-years of treatment ex-
‡ Suicide attempts resulting in hospitalization 295 (1.8) 43 (1.0) Suicide deaths 47 (0.3) posure. Our focus was on the compari-
 son of lithium with divalproex or car-
bamazepine. It is difficult to interpret
the experience of patients who are un-
treated or who receive combined treat-
ment. We suspect that the former are a
heterogeneous group including those
with relatively mild disease and those
with adherence problems and the lat-
ter are more likely to have severe or
treatment-resistant disease.
Statistical Analysis ​Descriptive
statistics and crude (unad- justed)
rates of suicide mortality and suicide
attempt are reported for peri- ods of
treatment with lithium, dival- proex,
or carbamazepine; combined
treatment (lithium plus divalproex or
carbamazepine); and no exposure to
any of the 3 drugs. Rates are reported
per 1000 person-years of treatment ex-
posure. Our focus was on the compari-
son of lithium with divalproex or car-
bamazepine. It is difficult to interpret
the experience of patients who are un-
treated or who receive combined treat-
ment. We suspect that the former are a
heterogeneous group including those
with relatively mild disease and those
with adherence problems and the lat-
ter are more likely to have severe or
treatment-resistant disease.
Statistical Analysis ​Descriptive
statistics and crude (unad- justed)
rates of suicide mortality and suicide
attempt are reported for peri- ods of
treatment with lithium, dival- proex,
or carbamazepine; combined
treatment (lithium plus divalproex or
carbamazepine); and no exposure to
any of the 3 drugs. Rates are reported
per 1000 person-years of treatment ex-
posure. Our focus was on the compari-
son of lithium with divalproex or car-
bamazepine. It is difficult to interpret
the experience of patients who are un-
treated or who receive combined treat-
©2003 American Medical Association. All rights reserved. ​(Reprinted) JAMA, ​September 17, 2003—Vol 290, No. 11 ​1469
ment. We suspect that the former are a
heterogeneous group including those
with relatively mild disease and those
with adherence problems and the lat-
ter are more likely to have severe or
treatment-resistant disease.
Statistical Analysis ​Descriptive
statistics and crude (unad- justed)
rates of suicide mortality and suicide
attempt are reported for peri- ods of
treatment with lithium, dival- proex,
or carbamazepine; combined
treatment (lithium plus divalproex or
carbamazepine); and no exposure to
any of the 3 drugs. Rates are reported
per 1000 person-years of treatment ex-
posure. Our focus was on the compari-
son of lithium with divalproex or car-
bamazepine. It is difficult to interpret
the experience of patients who are un-
treated or who receive combined treat-
ment. We suspect that the former are a
heterogeneous group including those
with relatively mild disease and those
with adherence problems and the lat-
ter are more likely to have severe or
treatment-resistant disease.
Exact binomial tests​26 ​were used to compare the
unadjusted suicide rate dur- ing
periods of lithium exposure with the
suicide rate during periods of dival-
proex or carbamazepine exposure. For
each type of suicide outcome, Cox
pro-
portionalhazardsregressionmodelswer
e
Exact binomial tests​26 ​were used to compare the
unadjusted suicide rate dur- ing
periods of lithium exposure with the
suicide rate during periods of dival-
proex or carbamazepine exposure. For
each type of suicide outcome, Cox
pro-
portionalhazardsregressionmodelswer
e
Downloaded From: https://jamanetwork.com/ on
03/17/2019
SUICIDE RISK IN BIPOLAR DISORDER
1470 ​JAMA, ​September 17, 2003—Vol 290, No. 11 ​(Reprinted) ​©2003 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 03/17/2019
used to examine risk in relation to ex-
was the reference category for compari- posure to each type of mood stabilizer
sons among mood stabilizers. after adjustment for age, sex, health plan,
Risk sets were blocked by year of di- year of diagnosis, comorbid medical and
agnosis and health plan. Because some psychiatric conditions, and concomi-
patients attempted suicide more than tant use of other psychotropic drugs. Pe-
once, analyses of attempts used the riods of exposure to mood stabilizers and
counting-process specification of the concomitant psychotropic drugs were
Cox model​27 ​with robust variance esti- analyzed as time-dependent variables.
mation.​28 ​Weighted Schoefeld residu- Thus, patients who switched from one
als were examined to assess the propor- mood stabilizer to another contributed
tional hazards assumption that the information to estimates of the suicide
relative risks do not change over time. risks associated with each of the drugs.
For all comparisons of divalproex vs Risk during treatment with lithium alone
lithium—regarding suicide attempts
as well as deaths—the residuals were not significantly correlated with time (​P .​ 10). For relative risk estimates, 95%
confidence intervals (CIs) are re- ported. ​P .​ 05 indicates statistical sig- nificance for all analyses and SAS soft- ware
was used for all analyses.​29
RESULTS ​We identified 20 638 health plan mem- bers treated for bipolar disorder dur- ing the study period (16
248 at KP and 4390 at GHC) who fit the study inclu- sion and exclusion criteria. Demo-
Table 2. ​Numbers and Rates of Suicide Attempts and Suicides During Periods of Exposure to Each Mood-Stabilizing Drug
Lithium Only Divalproex Only Carbamazepine Only Combination None
Kaiser Permanente ​Person-years of observation
Suicide attempts 13 014 8722 1762 1808 23 428 Suicide deaths 11 075 6820 1431 1525 17 782 Suicide attempts ascertained in emergency

department ​No. of events 141 273 39 62 352

Event rate per 1000 person-years 10.8 31.3 22.1 34.3 15.0 ​P ​value​* ​.001 .001 .001 .001 Suicide attempts resulting in hospitalization ​No. of

events 58 104 31 24 122

Event rate per 1000 person-years 4.5 11.9 17.6 13.3 5.2 ​P v​ alue​* ​.001 .001 .001 .37 Suicide deaths ​No. of events 7 12 2 3 23
Event rate per 1000 person-years 0.63 1.8 1.4 2.0 1.3 ​P v​ alue​* ​.05 .55 .22 .13

Group Health Cooperative ​Person-years of observation ​Suicide attempts 3006 1946 754 605 5014 Suicide
​ deaths 2522 1477 605 516 3780

Suicide attempts resulting in hospitalization ​No. of events 9 8 8 6 13

Event rate per 1000 person-years 3.0 4.1 10.6 9.9 2.6 ​P v​ alue​* ​.67 .02 .06 .90 Suicide deaths
No. of events 2 2 0 0 2 Event rate per 1000 person-years 0.8 1.4 0 0 0.5 ​P v​ alue​* ​.94 .99 .99 .99

Both Sites ​Person-years of observation ​Suicide attempts 16 020 10 669 2516 2413 28 442 Suicide
​ deaths 13 597 8297 2036 2040 21 562

Suicide attempts resulting in hospitalization ​No. of events 67 112 39 30 135

Event rate per 1000 person-years 4.2 10.5 15.5 12.4 4.8 ​P v​ alue​* ​.001 .001 .001 .44 Suicide deaths
No. of events 9 14 2 3 25 Event rate per 1000 person-years 0.7 1.7 1.0 1.5 1.2 ​P ​value​* ​.04 .86 .40 .20
*​Risk during treatment with lithium alone is the reference category for comparisons among mood stabilizers.
SUICIDE RISK IN BIPOLAR DISORDER
graphic and clinical characteristics of the study sample at the 2 sites are shown in Table 1. At both sites, the treated
bipolar population was pre- dominantly female and younger than 45 years. At both sites, 70% of pa- tients filled at
least 1 prescription for an antidepressant drug. There were 53 suicides, 338 attempts resulting in hos- pitalization,
and 642 attempts identi- fied in the ED. Of the 53 suicide deaths, 16 (30%) were caused by guns, 14 (26%) by
poisoning, 11 (21%) by hang- ing/suffocation, and 12 (23%) by other methods. ​T​ABLE ​2 ​displays periods of
exposure, number of suicide-related events, and rates of suicide-related events at both study sites. There was a total
of 60060 person-years of obser- vation (ie, a mean follow-up period of approximately 2.9 years per indi- vidual).
There was no exposure to lithium, divalproex, or carbamaze- pine during nearly half (47%) of all per- son-years of
follow-up. Patients were exposed to lithium alone during 27% of follow-up, divalproex alone during 18%,
carbamazepine alone during 4%, and a combination of these mood sta- bilizers during the remaining 4%.
Suicide attempts resulting in hospi- talization occurred 6.2 times more fre- quently than suicide deaths. At KP, sui-
cide attempts ascertained in the ED occurred14.6timesmorefrequentlythan ED suicide deaths. For each outcome ex-
amined at each site, rates of suicide at- tempt and suicide death were substan- tially greater during periods of
exposure to divalproex than that of lithium (Table 2). During periods with no mood stabi- lizer exposure, 8 (32%) of
the 25 sui- cide deaths occurred within 30 days of discontinuation of a mood stabilizer, whereas only 12% of the
total fol- low-up time was within 30 days of dis- continuation (​P​= .01).
After adjustment for age, sex, health plan, year of diagnosis, comorbid medi- cal and psychiatric conditions, and
con- comitant psychotropic drug use, the risk of each of the 3 outcomes was signifi- cantly greater during exposure
to dival- proex than during exposure to lithium; hazard ratios were 2.7, 1.7, and 1.8 for
©2003 American Medical Association. All rights reserved. ​(Reprinted) JAMA, ​September 17, 2003—Vol 290, No. 11 ​1471
Downloaded From: https://jamanetwork.com/ on 03/17/2019
Table 3. ​Risk of Suicide Attempts and Death in Relation to Use of Divalproex or Carbamazepine vs Lithium​*
Divalproex Carbamazepine

Hazard Ratio ​(95% CI) ​P V ​ alue Hazard

​ Ratio ​(95% CI) ​P V​ alue Suicide
​ attempts ascertained in ​emergency department 1.8
​ (1.4-2.2) .001
1.4 (1.0-2.0) .09 Suicide attempts resulting in hospitalization 1.7 (1.2-2.3) .002 2.9 (1.9-4.4) .001 Suicide deaths 2.7 (1.1-6.3) .03 1.5 (0.3-7.0)
.61
Abbreviation: CI, confidence interval. ​*​Hazard ratios for divalproex and carbamazepine use calculated (with lithium use as referent) by Cox regression analy- sis
adjusted for age, health plan, year of diagnosis, medical and psychiatric comorbidities, and concomitant psy- chotropic drugs.
Figure. ​Distribution of Initial Mood Stabilizer Prescriptions According to Year of Initial Bipolar Disorder Diagnosis
80​
90​ 7060​504030​20​100
Year
Lithium Divalproex Carbamazepine
% , noitubirtsi​
D​1994 1995 1996 1997 1998 1999 2000 2001

suicide death, attempt resulting in hos-

proex alone and more than twice as high pitalization, and attempt ascertained in
​ BLE ​3​). In each
as the risk of suicide attempt with the ED, respectively (​TA
lithium alone. Risk of suicide death was of the plans, the risk of each of these 3
not significantly higher during combi- outcomes during exposure to dival-
nation therapy, but the 95% CI for the proex was greater than during exposure
relative risk estimate of 2.1 was very to lithium. Differences between health
wide (0.6-7.7). Compared with peri- plans regarding comparisons of dival-
ods of treatment with only lithium, peri- proexvslithiumwerenotsignificant(for
ods with no mood stabilizer treatment suicide death,​P​=.60;forattemptresult-
wereassociatedwithrelativerisksof2.2, ing in hospitalization, ​P​= .37).
1.6, and 1.4 for suicide death, suicide Results for other comparisons (ie,
attempt ascertained in the ED, and sui- lithium vs carbamazepine, lithium vs
cide attempt resulting in hospitaliza- combination treatment, and lithium vs
tion, respectively. none of the mood stabilizers) were less
Additional secondary analyses exam- consistent or stable. During periods of
inedthepossibilitythatdifferencesinsui- treatment with carbamazepine alone
cide risk associated with exposure to dif- compared with lithium alone, the haz-
ferent mood stabilizers could reflect ardratiosofthe3outcomesrangedfrom
differences in preexisting illness sever- 1.4 to 2.9, but only the hazard ratio of
ity or other factors affecting suicide risk attemptsresultinginhospitalizationwas
(ie, confounding by indication). First, we significantly different from 1 (Table 3).
examined the hypothesis that the selec- During the 4% of follow-up time when
tion of one mood stabilizer rather than patients were exposed to combination
another was determined more by tem- treatment, the risk of suicide attempt—
poral trends in prescribing than by char- ascertained in either the hospital or the
acteristics of individual patients. The ED—was similar to the risk with dival-
F​IGURE ​displays the distribution of first
SUICIDE RISK IN BIPOLAR DISORDER
1472 ​JAMA, ​September 17, 2003—Vol 290, No. 11 ​(Reprinted) ​©2003 American Medical Association. All rights reserved.
Downloaded From: https://jamanetwork.com/ on 03/17/2019
mood-stabilizing drugs prescribed ac-
ference in risk was consistent across all
lithium or an anticonvulsant was more cording to year of initial diagnosis. The
outcome measures (suicide death, at-
strongly influenced by secular trends in ratio of initial filled prescriptions for
tempt resulting in hospitalization, and at-
prescribing than by clinical characteris- lithium to that for divalproex shifted
tempt diagnosed in the ED) and across
tics of individual patients. Despite dra- from approximately 6:1 in 1994 to ap-
the 2 study sites. Results for carbamaz-
matic changes in prescribing patterns proximately1:2in2001.Thisshiftiscon-
epine were qualitatively similar to those
over time, the lower risk during lithium sistent with shifts in prescribing behav-
for divalproex but (reflecting the smaller
treatment was consistent throughout the iorfromlithiumtodivalproexduringthis
sample size) much less precise. The find-
study period. Third, our analyses of those time frame noted in other settings.​30-32
ings reported here are consistent with
switching between mood stabilizers does We also examined whether risk differ-
substantial previous data suggesting that
not suggest that switching from lithium ences between lithium and divalproex
lithium reduces suicide attempts and sui-
to divalproex was preferentially associ- were stable throughout the study pe-
cide mortality.​20 ​The current study
ated with higher suicide risk. Only a ran- riod. For suicide attempts, rates during
complements a previous study​21 ​com-
domized trial can completely exclude the lithiumexposurewereconsistentlylower
paring lithium and carbamazepine. An-
possibility of confounding or bias, but than during anticonvulsant treatment
other small observational study​33 ​found
large observational studies such as this throughout the 8-year period. For sui-
no difference in risk of suicidal behav-
one may be the only realistic option for cide deaths, the risk difference was less
ior during treatment with lithium com-
studying relatively rare outcomes such consistent over time and appeared
pared with treatment with divalproex or
as suicide death. smaller during the first half of the study
carbamazepine, but its sample size was
Although our analyses support the va- period. Given the smaller number of sui-
not adequate to detect 2- or 3-fold dif-
lidity of comparing lithium treatment cidedeaths,thistrendmayhavebeendue
ferences in risk. To our knowledge, this
with divalproex treatment, we urge cau- to chance (​P= ​ .13).Inaddition,weevalu-
is the first study comparing suicide
tion in interpreting comparisons with ated the hypothesis that patients with
deaths and attempts associated with
combination treatment or periods with higher suicide risk were more likely to
lithium and divalproex, the most widely
no mood stabilizer treatment. Deci- be switched from one class of mood sta-
used mood stabilizer in the United States.
sions (by care providers or patients) to bilizer to another by examining suicide
Several limitations are inherent in the
discontinue a mood stabilizer or to ini- risk in the subgroups of patients switch-
use of administrative databases for this
tiate combined treatment are almost cer- ing classes of mood stabilizers (from
type of research. First, we relied on the
tainly related to perceived severity of ill- lithium to divalproex or the reverse) and
treating clinician’s discharge or encoun-
ness and anticipated risk of suicide the subgroups of patients continuing in
ter diagnoses rather than structured
attempt. Consequently, confounding by the same class (lithium or divalproex).
research evaluations to identify patients
indication may seriously bias any com- Among those exposed to divalproex, a
treated for bipolar disorder. Second, we
parison of risk during single-drug treat- previous switch from lithium was asso-
reliedondiagnosisandprescriptiondata
ment (with lithium or divalproex) with ciated with a significantly higher risk of
ratherthanclinicalassessmentstoadjust
that during either combined treatment suicide attempt resulting in hospitaliza-
fordifferencesinillnessseverityorunder-
or no treatment. Furthermore, the total tion (hazard ratio, 2.0; 95% CI, 1.1-3.4).
lying suicide risk. Third, visit or dis-
exposure to combination treatment in But among those switching from dival-
charge diagnosis data may miss a signifi-
this sample was too small to support ac- proex to lithium, there was a similar in-
cant proportion of true suicide attempts.
curate estimates of suicide risk. Given the crease in risk of suicide attempt result-
No observational study can com-
importance of combination treatment for ing in hospitalization (hazard ratio, 1.8;
pletely exclude the possibility of con-
some patients, future research should 95% CI, 0.9-3.6). Therefore, a history of
founding by indication or bias due to un-
 evaluate whether the apparently lower any medication switch was associated
measured differences in illness severity
suicideriskduringlithiumtreatmentalso with a higher risk of suicide attempt, but
or suicide risk in those treated with
occurs during treatment with lithium- this effect did not differ by direction of
lithium and those treated with dival-
anticonvulsant combinations. the switch (from lithium to divalproex
proex or carbamazepine. Still, we made
We included carbamazepine in these vs the reverse).
every attempt to detect and adjust for this
analyses to allow comparison with the
COMMENT
bias in our analyses. First, we con-
only previous randomized compari- trolled for characteristics that might in-
son study of suicide risk during mood In this population-based sample of more
dicate greater severity of illness or sui-
stabilizer treatment.​21​Our findings are than 20000 persons treated for bipolar
cide risk or might influence choice of
consistent with that study, although our disorder, we found that risk of suicide
mood stabilizer (eg, comorbid medical
power to evaluate carbamazepine is low attempt or suicide death was 1.5 to 3
or psychiatric illness, concomitant use
because only 4% of follow-up was in pa- times higher during periods of treat-
of other psychotropic drugs). Second,
tients treated with carbamazepine only. ment with divalproex than during peri-
analyses of time trends in prescribing
The mechanism by which lithium ods of treatment with lithium. This dif-
(Figure) suggest that choice of either
might exert an antisuicide effect is not
serotonin receptor sites inRISK
SUICIDE the rat
IN BIPOLAR
DISORDER
hippocampus.​46 ​This evidence of lower
suicide risk during lithium treatment
clear. Several controlled studies have should be viewed in light of the
demonstrated modest efficacy of declining use of lithium by
lithium in treatment and prevention of psychiatrists in the United States,
bipolar depression.​ 34-38 ​
Randomized particularly among recently trained
trials have also demonstrated that psychiatrists. Many psychiat- ric
lithium reduces aggressive behavior in residents have no or limited expe-
39-41​
prisoners​ and reduces impulsive rience prescribing lithium, largely a
behavior in children and re- flection of the enormous focus on
adolescents.​ 42-44 ​
At the biological level, the newer drugs in educational
suicide death is clearly associated with programs supported by the
reduced functional capacity of cen- pharmaceutical in- dustry. If lithium
45 ​
tral serotonin systems.​ Long-term does have an antisui- cide effect not
lithium administration enhances sero- matched by currently available
tonin turnover, reflected by increased alternatives, then current pre- scribing
re- lease and down-regulation of patterns should be reevalu- ated. At
the least, use of lithium to treat mood
 disorders should be an essential -controlled 18-month
um maintenance
component of training in psychiatry. or hypomanic patients
er pre- sented at:
Author Contributions: ​Dr Goodwin, as principal rican Psychi- atric
in- vestigator, had full access to all of the data in 02; Philadelphia, Pa.
this study and takes full responsibility for the CL, Sachs G, et al. A
integrity of the data and the accuracy of the data nth trial of lamotrigine
analysis. ​Study concept and design: G ​ oodwin, eatment in recently
Fireman, Simon, Hunkeler, Revicki. ​Acquisition of ts with bipolar I
data: F​ ireman, Simon, Hunkeler, Lee. ​Analysis atry​. 2003;60:392-400.
and interpretation of data: ​Goodwin, Fireman, fey EM Jr. Lithium
Simon, Hunkeler, Lee, Revicki. ​Drafting of the ffective illness. ​Am J
manuscript: ​Goodwin, Fireman, Simon, Lee, 203. 1 ​ 4. ​Prien RF, Klett
Revicki. ​Critical revision of the manuscript for carbon- ate and
important in- tellectual content: ​Goodwin, Fireman, affective epi- sodes.
Simon, Hunkeler, Revicki. ​Statistical expertise: ;29:420-425. ​15.
Fireman, Simon, Lee, Revicki. ​Obtained funding: dlewicz J, Fieve RR.
Goodwin. ​Administrative, technical, or material ve disorders. ​Am J
support: G ​ oodwin, Lee. ​Study supervision: -1010. 1 ​ 6. ​Fieve RR,
Goodwin, Fireman, Hunkeler. ​Funding/Support: Lithium pro- phylaxis of
This study was supported by fund- ing from olar II, and uni- polar
Solvay Pharmaceuticals to Best Practice LLC 1976;133:925-929. ​17.
based on a proposal developed by the study n A, et al. Lithium and
investi- gators. ​Role of the Sponsor: ​The xis of unipolar and
sponsor did not participate in decisions regarding spective,
data extraction and data analy- sis and had no rison.
role in interpretation of results, prepa- ration of the 81;17:142-144. 1 ​ 8.
manuscript, or decisions regarding pub- lication. n A, Ramos-Lorenzi
Acknowledgment: ​We acknowledge the Lithium carbonate and
assistance of Mark Goldstein, BA, Department of axis of unipolar and
Psychiatry and Be- n Psychiatry.
ciences, George Washington University owden CL, Calabrese
Center, Washington, DC, and James n- domized,
PhD, Center for Health Studies, Group th trial of dival- proex
operative. outpatients with bi-
Psychiatry.
NCES 1. ​Narrow WE, Rae DS, Robins
essarini RJ, Tondo L,
ed prevalence estimates of mental
cidal patient with
in the United States: us- ing a clinical
uicide risk with lithium.
ce criterion to reconcile 2 surveys’
32:24-38. ​21.
​Arch Gen Psychiatry. 2 ​ 002;59:115-123.
Oerlinghausen B, Seib-
CJL, Lopez AD, eds. ​The Global Burden
ylactic treatment on
e. ​Cambridge, Mass: Harvard University
h major affective
96. ​3. ​Bostwick JM, Pankratz VS.
hiatry.
disorders and suicide risk. ​Am J
ger N. Overcoming the
y. ​2000;157:1925-1932. ​4. ​Sharma R,
c data in medical
R. Mortality in affective dis- order. ​J
h. ​1992;82:703-710.
ord. 1 ​ 994;31:91-96. ​5. ​Guze SB, Robins
23. ​Simon GE, VonKorff M, Barlow W, Pabiniak
and primary affective disorders. ​Br J
C, Wagner E. Predictors of chronic
y. ​1970;117:437-438. ​6. ​Harris EC,
benzodiazepine use in a health maintenance
gh B. Suicide as an outcome for mental
organization sample. ​J Clin Epidemiol.
a meta-analysis. ​Br J Psychiatry. 1 ​ 997;
1996;49:1067-1073. ​24. ​Unutzer J, Simon G,
28. ​7. ​Brodersen A, Licht RW,
Pabiniak C, Bond K, Katon W. The use of
rd P, Olesen AV, Mortensen PB.
administrative data to assess quality of care for
ear mortality in patients with affective
bipolar disorder in a large staff model HMO. ​Gen
ommenced on lithium. ​Br J Psy- chiatry.
Hosp Psychiatry. 2 ​ 000;22:1-10. ​25.
429-433. ​8. ​Goodwin FK, Jamison KR.
SaundersK,StergachisA,VonKorffM.GroupHealth
pressive Ill- ness. N ​ ew York, NY: Oxford
Cooperative of Puget Sound. In: Strom B, ed.
Press; 1990. ​9. ​Prien RF, Kupfer DJ,
Pharma-
A, et al. Drug therapy in the prevention
coepidemiology.​ NewYork,NY:JohnWiley&Sons;19
nces in unipolar and bipolar affective
94. ​26. B
​ lyth CR, Still HA. Binomial confidence
​Arch Gen Psychiatry. ​1984;41:1096-
intervals. ​J Am Stat Assoc. 1 ​ 983;78:108-116. ​27.
Prien RF, Caffey EM Jr, Klett CJ.
Anderson PK, Gill RD. Cox’s regression model
tic ef- ficacy of lithium carbonate in
counting process. ​Ann Stat. 1 ​ 982;10:1100-1120.
pressive ill- ness. ​Arch Gen Psychiatry.
28. ​Lin DY, Wei LJ. The robust inference for the
37-341. ​11. ​Calabrese JR, Bowden CL,
pro- portional hazards model. ​J Am Stat Assoc.
1989;84: 1074-1078. ​29. S ​ AS/STAT Changes and
Enhancements Through Release 8.2.​ Cary, NC:
SAS Institute Inc; 2002. ​30. ​Blanco C, Laje G,
Olfson M, et al. Trends in the treatment of bipolar
disorder by outpatient psychia- trists. ​Am J
Psychiatry. ​2002;159:1005-1010. ​31. ​Fenn HH,
Robinson D, Luby V, et al. Trends in
pharmacotherapy of schizoaffective and bipolar
af- fective disorders: a 5-year naturalistic study.
Am J Psy- chiatry. ​1996;153:711-713. ​32.
LimPZ,TunisSL,EdellWS,JensikSE,TohenM.Medi-
cation prescribing patterns for patients with
bipolar I dis- order in hospital settings: adherence
to published prac- tice guidelines. ​Bipolar Disord.
2001;3:165-173. ​33. ​Yerevanian BI, Koek RJ,
Mintz J. Lithium, anti- convulsants and suicidal
behavior in bipolar disorder. ​J Affect Disord.
2003;73:223-228. ​34. G ​ oodwin FK, Murphy DL,
Dunner DL, Bunney WE Jr. Lithium response in
unipolar versus bipolar depres- sion. ​Am J
Psychiatry. ​1972;129:44-47. ​35. ​Watanabe S,
Ishino H, Otsuki S. Double-blind comparison of
lithium carbonate and imipramine in treatment of
depression. ​Arch Gen Psychiatry. 1 ​ 975;
32:659-668. ​36. ​Worrall EP, Moody JP, Peet M, et
al. Controlled studies of the acute antidepressant
effects of lithium. ​Br J Psychiatry.
1979;135:255-262. ​37. ​Bauer M, Bschor T, Kunz
D, et al. Double-blind, placebo-controlled trial of
the use of lithium to aug- ment antidepressant
medication in continuation treat- ment of unipolar
major depression. ​Am J Psychiatry.
2000;157:1429-1435. ​38. ​Nemeroff CB, Evans
DL, Gyulai L, et al. Double- blind,
placebo-controlled comparison of imipramine and
paroxetine in the treatment of bipolar depres-
sion. ​Am J Psychiatry. 2 ​ 001;158:906-912. ​39.
Sheard MH, Marini JL, Bridges CI, Wagner E. The
effect of lithium on impulsive aggressive behavior
in man. ​Am J Psychiatry. 1 ​ 976;133:1409-1413.
40. ​Morrison SD, Erwin CW, Gianturco DT,
Gerber CJ. Effect of lithium on combative
behavior in hu- mans. ​Dis Nerv Syst.
1973;34:186-189. ​41. T ​ upin JP, Smith DB, Clanon
TL, Kim LI, Nugent A, Groupe A. The long-term
use of lithium in aggres- sive prisoners. ​Compr
Psychiatry. ​1973;14:311-317. ​42. ​Campbell M,
Perry R, Green WH. Use of lithium in children and
adolescents. ​Psychosomatics. 1 ​ 984; 25:95-101,
105-106. ​43. ​Campbell M, Adams PB, Small AM,
et al. Lithium in hospitalized aggressive children
with conduct disor- der: a double-blind and
placebo-controlled study. ​J Am Acad Child
Adolesc Psychiatry. ​1995;34:445-453. ​44. M ​ alone
RP, Delaney MA, Luebbert JF, Cater J, Campbell
M. A double-blind placebo-controlled study of
lithium in hospitalized aggressive children and
ado- lescents with conduct disorder. ​Arch Gen
Psychiatry. ​2000;57:649-654. ​45. ​Mann JJ. A
current perspective of suicide and at- tempted
suicide. ​Ann Intern Med. 2 ​ 002;136:302-311. ​46.
Treiser SL, Cascio CS, O’Donohue TL, Thoa NB,
Jacobowitz DM, Kellar KJ. Lithium increases
seroto- nin release and decreases serotonin
receptors in the hippocampus. ​Science.
1981;213:1529-1531.
23. ​Simon GE, VonKorff M, Barlow W, Pabiniak
C, Wagner E. Predictors of chronic
benzodiazepine use in a health maintenance
organization sample. ​J Clin Epidemiol.
1996;49:1067-1073. ​24. ​Unutzer J, Simon G,
Pabiniak C, Bond K, Katon W. The use of
administrative data to assess quality of care for
bipolar disorder in a large staff model HMO. ​Gen
Hosp Psychiatry. 2 ​ 000;22:1-10. ​25.
SaundersK,StergachisA,VonKorffM.GroupHealth
Cooperative of Puget Sound. In: Strom B, ed.
Pharma-
coepidemiology.​ NewYork,NY:JohnWiley&Sons;19
94. ​26. B
​ lyth CR, Still HA. Binomial confidence
intervals. ​J Am Stat Assoc. 1 ​ 983;78:108-116. ​27.
Anderson PK, Gill RD. Cox’s regression model
counting process. ​Ann Stat. 1 ​ 982;10:1100-1120.
28. ​Lin DY, Wei LJ. The robust inference for the
pro- portional hazards model. ​J Am Stat Assoc.
1989;84: 1074-1078. ​29. ​SAS/STAT Changes and
Enhancements Through Release 8.2.​ Cary, NC:
SAS Institute Inc; 2002. ​30. ​Blanco C, Laje G,
Olfson M, et al. Trends in the treatment of bipolar
disorder by outpatient psychia- trists. ​Am J
Psychiatry. ​2002;159:1005-1010. ​31. ​Fenn HH,
Robinson D, Luby V, et al. Trends in
pharmacotherapy of schizoaffective and bipolar
af- fective disorders: a 5-year naturalistic study.
Am J Psy- chiatry. ​1996;153:711-713. ​32.
LimPZ,TunisSL,EdellWS,JensikSE,TohenM.Medi-
cation prescribing patterns for patients with
bipolar I dis- order in hospital settings: adherence
to published prac- tice guidelines. ​Bipolar Disord.
2001;3:165-173. ​33. Y ​ erevanian BI, Koek RJ,
Mintz J. Lithium, anti- convulsants and suicidal
behavior in bipolar disorder. ​J Affect Disord.
2003;73:223-228. ​34. ​Goodwin FK, Murphy DL,
Dunner DL, Bunney WE Jr. Lithium response in
unipolar versus bipolar depres- sion. ​Am J
Psychiatry. ​1972;129:44-47. 3 ​ 5. ​Watanabe S,
Ishino H, Otsuki S. Double-blind comparison of
lithium carbonate and imipramine in treatment of
depression. ​Arch Gen Psychiatry. 1 ​ 975;
32:659-668. ​36. ​Worrall EP, Moody JP, Peet M, et
al. Controlled studies of the acute antidepressant
effects of lithium. ​Br J Psychiatry.
1979;135:255-262. ​37. ​Bauer M, Bschor T, Kunz
D, et al. Double-blind, placebo-controlled trial of
the use of lithium to aug- ment antidepressant
medication in continuation treat- ment of unipolar
major depression. ​Am J Psychiatry.
2000;157:1429-1435. ​38. ​Nemeroff CB, Evans
DL, Gyulai L, et al. Double- blind,
placebo-controlled comparison of imipramine and
paroxetine in the treatment of bipolar depres-
sion. ​Am J Psychiatry. 2 ​ 001;158:906-912. ​39.
Sheard MH, Marini JL, Bridges CI, Wagner E. The
effect of lithium on impulsive aggressive behavior
in man. ​Am J Psychiatry. 1 ​ 976;133:1409-1413.
40. ​Morrison SD, Erwin CW, Gianturco DT,
Gerber CJ. Effect of lithium on combative
behavior in hu- mans. ​Dis Nerv Syst.
1973;34:186-189. ​41. ​Tupin JP, Smith DB, Clanon
TL, Kim LI, Nugent A, Groupe A. The long-term
use of lithium in aggres- sive prisoners. ​Compr
Psychiatry. ​1973;14:311-317. ​42. ​Campbell M,
Perry R, Green WH. Use of lithium in children and
adolescents. ​Psychosomatics. 1 ​ 984; 25:95-101,
105-106. ​43. ​Campbell M, Adams PB, Small AM,
et al. Lithium in hospitalized aggressive children
with conduct disor- der: a double-blind and
placebo-controlled study. ​J Am Acad Child
Adolesc Psychiatry. ​1995;34:445-453. ​44. M ​ alone
RP, Delaney MA, Luebbert JF, Cater J, Campbell
M. A double-blind placebo-controlled study of
lithium in hospitalized aggressive children and
ado- lescents with conduct disorder. ​Arch Gen
Psychiatry. ​2000;57:649-654. ​45. ​Mann JJ. A
current perspective of suicide and at- tempted
suicide. ​Ann Intern Med. 2 ​ 002;136:302-311. ​46.
Treiser SL, Cascio CS, O’Donohue TL, Thoa NB,
Jacobowitz DM, Kellar KJ. Lithium increases
seroto- nin release and decreases serotonin
receptors in the hippocampus. ​Science.
1981;213:1529-1531.
23. ​Simon GE, VonKorff M, Barlow W, Pabiniak
C, Wagner E. Predictors of chronic
benzodiazepine use in a health maintenance
organization sample. ​J Clin Epidemiol.
1996;49:1067-1073. ​24. ​Unutzer J, Simon G,
Pabiniak C, Bond K, Katon W. The use of
administrative data to assess quality of care for
bipolar disorder in a large staff model HMO. ​Gen
Hosp Psychiatry. ​2000;22:1-10. ​25.
SaundersK,StergachisA,VonKorffM.GroupHealth
Cooperative of Puget Sound. In: Strom B, ed.
Pharma-
coepidemiology.​ NewYork,NY:JohnWiley&Sons;19
94. ​26. ​Blyth CR, Still HA. Binomial confidence
intervals. ​J Am Stat Assoc. 1 ​ 983;78:108-116. ​27.
Anderson PK, Gill RD. Cox’s regression model
counting process. ​Ann Stat. 1 ​ 982;10:1100-1120.
28. ​Lin DY, Wei LJ. The robust inference for the
pro- portional hazards model. ​J Am Stat Assoc.
1989;84: 1074-1078. ​29. S ​ AS/STAT Changes and
Enhancements Through Release 8.2.​ Cary, NC:
SAS Institute Inc; 2002. ​30. ​Blanco C, Laje G,
Olfson M, et al. Trends in the treatment of bipolar
disorder by outpatient psychia- trists. ​Am J
Psychiatry. ​2002;159:1005-1010. ​31. ​Fenn HH,
Robinson D, Luby V, et al. Trends in
pharmacotherapy of schizoaffective and bipolar
af- fective disorders: a 5-year naturalistic study.
Am J Psy- chiatry. ​1996;153:711-713. ​32.
LimPZ,TunisSL,EdellWS,JensikSE,TohenM.Medi-
cation prescribing patterns for patients with
bipolar I dis- order in hospital settings: adherence
to published prac- tice guidelines. ​Bipolar Disord.
2001;3:165-173. ​33. ​Yerevanian BI, Koek RJ,
Mintz J. Lithium, anti- convulsants and suicidal
behavior in bipolar disorder. ​J Affect Disord.
2003;73:223-228. ​34. G ​ oodwin FK, Murphy DL,
Dunner DL, Bunney WE Jr. Lithium response in
unipolar versus bipolar depres- sion. ​Am J
Psychiatry. ​1972;129:44-47. ​35. ​Watanabe S,
Ishino H, Otsuki S. Double-blind comparison of
lithium carbonate and imipramine in treatment of
depression. ​Arch Gen Psychiatry. 1 ​ 975;
32:659-668. ​36. ​Worrall EP, Moody JP, Peet M, et
al. Controlled studies of the acute antidepressant
effects of lithium. ​Br J Psychiatry.
1979;135:255-262. ​37. ​Bauer M, Bschor T, Kunz
D, et al. Double-blind, placebo-controlled trial of
the use of lithium to aug- ment antidepressant
medication in continuation treat- ment of unipolar
major depression. ​Am J Psychiatry.
2000;157:1429-1435. ​38. ​Nemeroff CB, Evans
DL, Gyulai L, et al. Double- blind,
placebo-controlled comparison of imipramine and
paroxetine in the treatment of bipolar depres-
sion. ​Am J Psychiatry. ​2001;158:906-912. ​39.
Sheard MH, Marini JL, Bridges CI, Wagner E. The
effect of lithium on impulsive aggressive behavior
in man. ​Am J Psychiatry. 1 ​ 976;133:1409-1413.
40. ​Morrison SD, Erwin CW, Gianturco DT,
Gerber CJ. Effect of lithium on combative
behavior in hu- mans. ​Dis Nerv Syst.
1973;34:186-189. ​41. ​Tupin JP, Smith DB, Clanon
TL, Kim LI, Nugent A, Groupe A. The long-term
use of lithium in aggres- sive prisoners. ​Compr
Psychiatry. ​1973;14:311-317. 4 ​ 2. ​Campbell M,
Perry R, Green WH. Use of lithium in children and
adolescents. ​Psychosomatics. ​1984; 25:95-101,
105-106. ​43. ​Campbell M, Adams PB, Small AM,
et al. Lithium in hospitalized aggressive children
with conduct disor- der: a double-blind and
placebo-controlled study. ​J Am Acad Child
Adolesc Psychiatry. ​1995;34:445-453. ​44. ​Malone
RP, Delaney MA, Luebbert JF, Cater J, Campbell
M. A double-blind placebo-controlled study of
lithium in hospitalized aggressive children and
ado- lescents with conduct disorder. ​Arch Gen
Psychiatry. ​2000;57:649-654. 4 ​ 5. ​Mann JJ. A
current perspective of suicide and at- tempted
suicide. ​Ann Intern Med. 2 ​ 002;136:302-311. ​46.
Treiser SL, Cascio CS, O’Donohue TL, Thoa NB,
Jacobowitz DM, Kellar KJ. Lithium increases
seroto- nin release and decreases serotonin
receptors in the hippocampus. ​Science.
1981;213:1529-1531.
23. ​Simon GE, VonKorff M, Barlow W, Pabiniak
C, Wagner E. Predictors of chronic
benzodiazepine use in a health maintenance
organization sample. ​J Clin Epidemiol.
1996;49:1067-1073. ​24. ​Unutzer J, Simon G,
Pabiniak C, Bond K, Katon W. The use of
administrative data to assess quality of care for
bipolar disorder in a large staff model HMO. ​Gen
Hosp Psychiatry. ​2000;22:1-10. ​25.
SaundersK,StergachisA,VonKorffM.GroupHealth
Cooperative of Puget Sound. In: Strom B, ed.
Pharma-
coepidemiology.​ NewYork,NY:JohnWiley&Sons;19
94. ​26. B
​ lyth CR, Still HA. Binomial confidence
intervals. ​J Am Stat Assoc. 1 ​ 983;78:108-116. ​27.
Anderson PK, Gill RD. Cox’s regression model
counting process. ​Ann Stat. 1 ​ 982;10:1100-1120.
28. ​Lin DY, Wei LJ. The robust inference for the
pro- portional hazards model. ​J Am Stat Assoc.
1989;84: 1074-1078. ​29. ​SAS/STAT Changes and
Enhancements Through Release 8.2.​ Cary, NC:
SAS Institute Inc; 2002. ​30. ​Blanco C, Laje G,
Olfson M, et al. Trends in the treatment of bipolar
disorder by outpatient psychia- trists. ​Am J
Psychiatry. ​2002;159:1005-1010. ​31. ​Fenn HH,
Robinson D, Luby V, et al. Trends in
pharmacotherapy of schizoaffective and bipolar
 af- fective disorders: a 5-year naturalistic study.
Am J Psy- chiatry. ​1996;153:711-713. ​32.
LimPZ,TunisSL,EdellWS,JensikSE,TohenM.Medi-
cation prescribing patterns for patients with
bipolar I dis- order in hospital settings: adherence
to published prac- tice guidelines. ​Bipolar Disord.
2001;3:165-173. ​33. ​Yerevanian BI, Koek RJ,
Mintz J. Lithium, anti- convulsants and suicidal
behavior in bipolar disorder. ​J Affect Disord.
2003;73:223-228. ​34. G ​ oodwin FK, Murphy DL,
Dunner DL, Bunney WE Jr. Lithium response in
unipolar versus bipolar depres- sion. ​Am J
Psychiatry. ​1972;129:44-47. ​35. ​Watanabe S,
Ishino H, Otsuki S. Double-blind comparison of
lithium carbonate and imipramine in treatment of
depression. ​Arch Gen Psychiatry. 1 ​ 975;
32:659-668. ​36. ​Worrall EP, Moody JP, Peet M, et
al. Controlled studies of the acute antidepressant
effects of lithium. ​Br J Psychiatry.
1979;135:255-262. ​37. ​Bauer M, Bschor T, Kunz
D, et al. Double-blind, placebo-controlled trial of
the use of lithium to aug- ment antidepressant
medication in continuation treat- ment of unipolar
major depression. ​Am J Psychiatry.
2000;157:1429-1435. ​38. ​Nemeroff CB, Evans
DL, Gyulai L, et al. Double- blind,
placebo-controlled comparison of imipramine and
paroxetine in the treatment of bipolar depres-
sion. ​Am J Psychiatry. 2 ​ 001;158:906-912. ​39.
©2003 American Medical Association. All rights reserved. ​(Reprinted) JAMA, ​September 17, 2003—Vol 290, No. 11 ​1473
Downloaded From: https://jamanetwork.com/ on
03/17/2019
Sheard MH, Marini JL, Bridges CI, Wagner E. The
effect of lithium on impulsive aggressive behavior
in man. ​Am J Psychiatry. 1 ​ 976;133:1409-1413.
40. ​Morrison SD, Erwin CW, Gianturco DT,
Gerber CJ. Effect of lithium on combative
behavior in hu- mans. ​Dis Nerv Syst.
1973;34:186-189. ​41. ​Tupin JP, Smith DB, Clanon
TL, Kim LI, Nugent A, Groupe A. The long-term
use of lithium in aggres- sive prisoners. ​Compr
Psychiatry. ​1973;14:311-317. 4 ​ 2. ​Campbell M,
Perry R, Green WH. Use of lithium in children and
adolescents. ​Psychosomatics. ​1984; 25:95-101,
105-106. ​43. ​Campbell M, Adams PB, Small AM,
et al. Lithium in hospitalized aggressive children
with conduct disor- der: a double-blind and
placebo-controlled study. ​J Am Acad Child
Adolesc Psychiatry. ​1995;34:445-453. ​44. ​Malone
RP, Delaney MA, Luebbert JF, Cater J, Campbell
M. A double-blind placebo-controlled study of
lithium in hospitalized aggressive children and
ado- lescents with conduct disorder. ​Arch Gen
Psychiatry. ​2000;57:649-654. 4 ​ 5. ​Mann JJ. A
current perspective of suicide and at- tempted
suicide. ​Ann Intern Med. 2 ​ 002;136:302-311. ​46.
Treiser SL, Cascio CS, O’Donohue TL, Thoa NB,
Jacobowitz DM, Kellar KJ. Lithium increases
seroto- nin release and decreases serotonin
receptors in the hippocampus. ​Science.
1981;213:1529-1531.