The n e w e ng l a n d j o u r na l of
original article
From Risk MR Pharmacovigilance Services,
Zaragoza (A.A., F.M.A.); the Department
of Psychiatry, Hospital Universitario de la
Princesa, Universidad Autónoma de Ma-
drid, and Centro de Investigación Bio-
médica en Red de Salud Mental, Instituto
de Salud Carlos III, Madrid (J.L.A.-M.) —
all in Spain; and RiskMR, Bridgewater, NJ
(C.E.W., F.M.A.). Address reprint requests
to Dr. Arana at Risk MR Pharmacovigi-
lance Services, Jerónimo Zurita 14, 1-D,
50001 Zaragoza, Spain, or at arana.riskmr@
gmail.com.
N Engl J Med 2010;363:542-51.
Copyright © 2010 Massachusetts Medical Society.
542
Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
m e dic i n e
Suicide-Related Events in Patients Treated
with Antiepileptic Drugs
Alejandro Arana, M.D., Charles E. Wentworth, M.S., José L. Ayuso-Mateos, M.D.,
and Felix M. Arellano, M.D.
A bs t r ac t
Background
A previous meta-analysis of data from clinical trials showed an association between
antiepileptic drugs and suicidality (suicidal ideation, behavior, or both). We used
observational data to examine the association between the use or nonuse of anti-
epileptic drugs and suicide-related events (attempted suicides and completed sui-
cides) in patients with epilepsy, depression, or bipolar disorder.
Methods
We used data collected as part of the clinical care of patients who were representa-
tive of the general population in the United Kingdom to identify patients with epi-
lepsy, depression, or bipolar disorder and to determine whether they received anti-
epileptic drugs. We estimated the incidence rate of suicide-related events and used
logistic regression to compute odds ratios, controlling for confounding factors.
Results
In a cohort of 5,130,795 patients, the incidence of suicide-related events per 100,000
person-years was 15.0 (95% confidence interval [CI], 14.6 to 15.5) among patients
without epilepsy, depression, bipolar disorder, or antiepileptic-drug treatment, 38.2
(95% CI, 26.3 to 53.7) among patients with epilepsy who did not receive antiepileptic
drugs, and 48.2 (95% CI, 39.4 to 58.5) among patients with epilepsy who received
antiepileptic drugs. In adjusted analyses, the use of antiepileptic drugs was not as-
sociated with an increased risk of suicide-related events among patients with epi-
lepsy (odds ratio, 0.59; 95% CI, 0.35 to 0.98) or bipolar disorder (1.13; 95% CI, 0.35
to 3.61) but was significantly associated with an increased risk among patients with
depression (1.65; 95% CI, 1.24 to 2.19) and those who did not have epilepsy, depres-
sion, or bipolar disorder (2.57; 95% CI, 1.78 to 3.71).
Conclusions
The current use of antiepileptic drugs was not associated with an increased risk of
suicide-related events among patients with epilepsy, but it was associated with an
increased risk of such events among patients with depression and among those who
did not have epilepsy, depression, or bipolar disorder.
n engl j med 363;6  nejm.org  august 5, 2010
The New England Journal of Medicine
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
 Suicide-Related Events and Antiepileptic Drugs
S
uicide is the 13th leading cause of
death worldwide,1 and attempted suicide is
a major cause of injury.2 Psychiatric disor-
ders (especially affective conditions) increase the
risk of suicide.3-5 Epilepsy increases both the risk
of suicide among patients with psychiatric disor-
ders6 and the risk of the development of psychi-
atric illness.7 The early onset of epilepsy, female
sex, psychiatric illness, temporal-lobe epilepsy,
and inadequate neurologic follow-up are risk fac-
tors for suicide among patients with epilepsy.6,8
Depression is common in patients with severe
epilepsy,8 but the severity of epilepsy does not
necessarily correlate with the risk of suicide.5
The risk of suicide increases soon after the onset
of epilepsy5 and then decreases gradually. Patients
with long-standing epilepsy or dysphoric condi-
tions appear to have an increased risk of suicide
soon after gaining control of their seizures.9
In January 2008, the Food and Drug Admin-
istration (FDA) issued a safety warning on the
risk of suicidality associated with antiepileptic
drugs.10 The warning summarized the results of a
meta-analysis of placebo-controlled clinical trials
of 11 antiepileptic drugs; this meta-analysis showed
a risk of the development of suicidality (primar-
ily suicidal behavior or ideation) that was twice
as high among patients who received antiepilep-
tic drugs as among patients who received pla-
cebo. The risk increased soon after the initiation
of treatment, persisted through week 24, and was
elevated regardless of the type of antiepileptic
drugs received and the indication for their use.
The assessment of suicidality in the clinical
trials that were included in the meta-analysis
was subject to several limitations such as the
lack of systematic or standardized language to
define suicidal ideation and behavior across
clinical trials.11 We examined the association be-
tween antiepileptic drugs and suicide-related
events (defined as suicide attempts and completed
suicides), using data collected as part of clinical
practice in the United Kingdom.
Me thods
Study Design and Oversight
The study was designed by the investigators in
Spain. The data were collected by the second au-
thor, and the analyses were performed by the
first and second authors. The last author wrote
n engl j med 363;6  nejm.org  august 5, 2010
The New England Journal of Medicine
Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
the manuscript. All authors reviewed the manu-
script and decided before the study began to sub-
mit the manuscript for publication.
Source Population and Data
We performed the study using The Health Im-
provement Network (THIN) database, which is rep-
resentative of the general population in the Unit-
ed Kingdom12 and includes more than 6.7 million
patients. THIN data are entered at general prac-
titioners’ offices and are based on the daily rec­
ord keeping of these practices. The data provide
anonymous demographic, medical, and prescrip-
tion information on individual patients, and they
provide a longitudinal medical record for each
patient.13
Study Population
Patients were eligible for inclusion in the study
population if they were enrolled in a clinical prac-
tice for at least 6 months during the study period
(from July 1, 1988, through March 31, 2008). A
history of a suicide attempt is the major risk fac-
tor for a subsequent suicide attempt.14 Thus, we
excluded patients with a family history of suicide
and a personal history of one or more suicide at-
tempts.
The study had two components. In the descrip-
tive analysis, we estimated the crude incidence
rate of suicide-related events among patients with
epilepsy, depression, or bipolar disorder, with or
without the use of antiepileptic drugs. We then
performed a case−control analysis to examine
the association between the use or nonuse of
antiepileptic drugs and suicide-related events.
Definition of the Cohorts
In the descriptive analysis, we defined the refer-
ence group as patients who did not have epilepsy,
depression, or bipolar disorder and did not re-
ceive antiepileptic drugs. The other cohorts are
listed in Table 1.
Cohort Follow-up
Patients were assigned to cohorts according to
the presence of an event that defined a given co-
hort. The index date was the date when the last
Read code15 that would include a patient in a co-
hort was recorded. Patients were followed until
death, loss to follow-up, or the end of the study
period, whichever came first.
543
 The n e w e ng l a n d j o u r na l of m e dic i n e

tempts) for validation to determine the positive

Table 1. Study Cohorts.
Patients who did not have epilepsy, depression, or bipolar disorder but who
received antiepileptic drugs
Patients with epilepsy who did not receive antiepileptic drugs
Patients with epilepsy who received antiepileptic drugs
Patients with depression who did not receive antiepileptic drugs
Patients with depression who received antiepileptic drugs
Patients with bipolar disorder who did not receive antiepileptic drugs
Patients with bipolar disorder who received antiepileptic drugs
Patients with epilepsy and depression who did not receive antiepileptic drugs
Patients with epilepsy and depression who received antiepileptic drugs
Patients with epilepsy and bipolar disorder who did not receive antiepileptic drugs
Patients with epilepsy and bipolar disorder who received antiepileptic drugs
predictive value of the diagnoses used. The vali-
dation process included a questionnaire sent to
the general practitioner, as well as a review of the
patient’s medical records and death certificate
when available (see the Supplementary Appendix,
available with the full text of this article at NEJM
.org). The questionnaires and records were re-
viewed by one of the investigators, who was un-
aware of the patients’ exposure to antiepileptic
drugs. The positive predictive value was 97% for
suicide-related events and 87% for completed
suicide. This method of validating cases based
on THIN data has been successfully used in other
studies.16

Selection of Controls
Exposure to Medications For the case−control component of the study, we
The antiepileptic drugs of interest were those randomly selected five controls from the cohort
drugs identified in the FDA’s meta-analysis which for each case patient, matched according to age
were available in the United Kingdom: carbamaz­ at the index date (±5 years), sex, and clinical prac-
epine, gabapentin, lamotrigine, levetiracetam, tice. Using risk-set control sampling, we defined
oxcarbazepine, pregabalin, tiagabine, topiramate, the date of the end of the follow-up period for the
valproate, and zonisamide (felbamate was not in- controls as the date when the case patient received
cluded because it was not marketed in the United a diagnosis. There were no unmatched case pa-
Kingdom). Current use was defined as the inter- tients. For the analysis of the use or nonuse of
val between the date of the prescription of an antiepileptic drugs within each category of ill-
antiepileptic drug and 75 days after that date. ness, another set of controls, matched according
After that interval, as long as no other antiepilep- to age, sex, clinical practice, and psychiatric or
tic drug was prescribed, the patient was consid- neurologic condition, was selected. A total of 103
ered to be a previous user. Patients without codes case patients did not match any controls.
for antiepileptic drugs were classified as nonusers.
Statistical Analysis
Case Identification and Characterization We described the characteristics of the various
Cases of suicide-related events were based on cohorts using frequency tables. We calculated
codes for suicide, attempted suicide, and inten-the incidence rates and 95% confidence intervals
tional self-inflicted injuries plus suicide. A com-
for suicide-related events in each of the cohorts
pleted suicide was defined as a code for suicidal-
with the use of a Poisson regression model.
ity followed by a code for death in the following The relationship between the use or nonuse
month and a final date of any administrative of antiepileptic drugs and suicide-related events
activity in the database or disenrollment within(expressed as odds ratios and 95% confidence
6 months after the suicidality code. If the disen-
intervals) was assessed with the use of a fixed-
rollment date occurred more than 6 months after effects conditional logistic-regression model. In
a suicidality code, we reviewed the patient’s pro-
addition to matching for age, sex, clinical prac-
file. Patients with a last medical or other health-
tice, and condition when applicable, we controlled
related code that was recorded within 1 month for potential confounders (Table 2) with the use
after the suicide date were also considered to of a backward-elimination process by stepwise
have completed suicide. deletion of the confounder that made the small-
est change in the exposure−effect estimate on
Case Validation deletion, provided that the effect of the con-
We selected a random sample of 218 cases of founder was 10% or less of the estimated effect.
suicide-related events (86 suicides and 132 at- Any confounder with an effect of more than 10%

544 n engl j med 363;6  nejm.org  august 5, 2010

The New England Journal of Medicine

Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
 Suicide-Related Events and Antiepileptic Drugs

Table 2. Demographic Characteristics and Risk Factors for Suicide-Related Events, According to Use or Nonuse
of Antiepileptic Drugs.*
No Epilepsy, Depression,
Bipolar Disorder, or Use Antiepileptic Epilepsy and Use
of Antiepileptic Drugs Drugs Only Epilepsy Only of Antiepileptic Drugs
Variable (N = 4,514,366)† (N = 77,319) (N = 16,120) (N = 39,325)
Male sex — no. (%) 2,273,135 (50.4) 34,978 (45.2) 8,565 (53.1) 20,868 (53.1)
Age — no. (%)
<20 yr 1,413,006 (31.3) 3,975 (5.1) 3,715 (23.0) 7,935 (20.2)
20–34 yr 1,204,371 (26.7) 7,506 (9.7) 4,786 (29.7) 8,815 (22.4)
>34–64 yr 1,326,817 (29.4) 34,689 (44.9) 5,564 (34.5) 14,031 (35.7)
>64 yr 500,482 (11.1) 30,848 (39.9) 1,762 (10.9) 8,008 (20.4)
Data missing 69,690 (1.5) 301 (0.4) 293 (1.8) 536 (1.4)
Mean age — yr 32.0±23.1 56.5±20.6 35.3±20.9 41.3±23.5
Duration of time in study — yr 6.2±5.3 3.9±3.9 5.4±4.9 5.4±4.7
Drug use in previous yr — no. (%)
Antipsychotic agent 53,894 (1.2) 7,910 (10.2) 488 (3.0) 2,588 (6.6)
Antidepressant agent 103,343 (2.3) 23,034 (29.8) 810 (5.0) 3,002 (7.6)
Lithium 1,146 (<0.1) 402 (0.5) 6 (<0.1) 38 (0.1)
History — no. (%)
Drug abuse 1,232 (<0.1) 80 (0.1) 15 (0.1) 25 (0.1)
Diagnosis of mental disorder 171,846 (3.8) 13,845 (17.9) 1,794 (11.1) 4,174 (10.6)
Alcohol abuse 26,189 (0.6) 2,106 (2.7) 620 (3.8) 1,199 (3.0)
Chronic disease score‡
Mean score 0.45±0.97 3.26±1.96 0.68±1.20 2.09±1.51
Score — no. (%)
0 3,443,036 (76.3) 1,776 (2.3) 10,801 (67.0) 973 (2.5)
1 509,052 (11.3) 16,023 (20.7) 2,236 (13.9) 18,973 (48.2)
2 329,417 (7.3) 12,478 (16.1) 1,650 (10.2) 7,111 (18.1)
3 137,521 (3.0) 14,830 (19.2) 788 (4.9) 5,644 (14.4)
4 59,001 (1.3) 12,887 (16.7) 375 (2.3) 3,468 (8.8)
5 36,339 (0.8) 19,325 (25.0) 270 (1.7) 3,156 (8.0)

* Plus−minus values are means ±SD.

† This was the reference category.
‡ Chronic disease scores can range from 0 to 5, with higher scores indicating a higher frequency of chronic disease.

on the estimate was kept in the model. To con- follow-up. Table 2, and Table 1 in the Supplemen-
trol for the severity of chronic disease, we included
tary Appendix, show the demographic character-
a modified version of the chronic disease score istics of the patients. Overall, 48.6% of the pa-
in the model.17,18 All analyses were conducted tients were male, and the mean (±SD) age was
with the use of SAS software, version 9.1 (SAS 33.7±23.1 years. The mean follow-up time for all
Institute) and Stata software, version 7.0 (Stata). the cohorts was 6.2±5.2 years. A total of 66,925
patients had epilepsy; 16,120 of these patients
R e sult s did not have a diagnosis of depression or bipolar
disorder and did not receive antiepileptic drugs.
Characteristics of the Patients A total of 435,790 patients with depression only
The cohorts included a total of 5,130,795 pa- and 3814 with bipolar disorder only did not re-
tients, with a total of 31,527,585 patient-years of ceive antiepileptic drugs. A total of 77,319 pa-

n engl j med 363;6  nejm.org  august 5, 2010 545

The New England Journal of Medicine
Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 3. Incidence of Suicide-Related Events, According to Cohorts and Use or Nonuse of Antiepileptic Drugs.

Patients with First

Suicide-Related Crude Incidence Rate
Cohort All Patients Event Person-Yr (95% Confidence Interval)
no. of events/
no. of patients 100,000 person-yr
No epilepsy, depression, or bipolar disorder
No use of antiepileptic drugs (reference group) 4,514,366 4239 28,170,361 15.05 (14.60−15.51)
Any use of antiepileptic drugs 77,319 120 304,807 39.37 (32.64−47.08)
Epilepsy only (no depression or bipolar disorder)
No use of antiepileptic drugs 16,120 33 86,333 38.22 (26.31−53.68)
Any use of antiepileptic drugs 39,325 103 213,507 48.24 (39.38−58.51)
Depression only (no epilepsy or bipolar disorder)
No use of antiepileptic drugs 435,790 3271 2,534,502 129.06 (124.70−133.60)
Any use of antiepileptic drugs 30,772 232 130,854 177.30 (155.21−201.64)
Bipolar disorder (no epilepsy)
No use of antiepileptic drugs 3,814 48 22,324 215.02 (158.54−285.08)
Any use of antiepileptic drugs 1,809 40 9,065 441.26 (315.24−600.87)
Epilepsy and depression (no bipolar disorder)
No use of antiepileptic drugs 3,392 28 16,353 171.22 (113.78−247.47)
Any use of antiepileptic drugs 7,870 96 38,452 249.66 (202.23−304.88)
Epilepsy and bipolar disorder
No use of antiepileptic drugs 47 0 210 0
Any use of antiepileptic drugs 171 2 819 244.24 (29.58−882.28)
All patients 5,130,795 8212 31,527,585 26.05 (25.49−26.62)

tients received antiepileptic drugs but did not
have epilepsy, depression, or bipolar disorder. The
indications for the use of antiepileptic drugs in
these patients were not known, although pain
and pain-related diagnoses (e.g., herpes zoster)
were documented in the records of 18.7% of these
patients in the 30 days before prescription of the
antiepileptic drugs.
Incidence of Attempted and Completed
Suicides
A total of 8212 patients attempted suicide, includ-
ing 464 patients who completed suicide. The in-
cidence rate of suicide-related events among pa-
tients without epilepsy, depression, or bipolar
disorder who did not receive antiepileptic drugs
(the reference group) was 15.0 (95% confidence
interval [CI], 14.6 to 15.5) per 100,000 person-
years (Table 3). Details of the incidence of suicide
according to the patient’s psychiatric or neuro-
logic condition are available in Table 2 in the
Supplementary Appendix. The incidence rate of
suicide-related events among patients who did
not have epilepsy, depression, or bipolar disorder
was 57.9 (95% CI, 42.8 to 76.5) per 100,000 per-
son-years for current users of antiepileptic drugs
and 32.3 (95% CI, 25.2 to 40.7) per 100,000 per-
son-years for previous users of antiepileptic drugs
(Table 4 in the Supplementary Appendix). Among
patients with depression alone, the incidence rate
of suicide-related events was 129.1 (95% CI, 124.7
to 133.6) per 100,000 person-years for those who
were not currently receiving antiepileptic drugs
and 177.3 (95% CI, 155.2 to 201.6) per 100,000
person-years for those who were receiving anti-
epileptic drugs. Among patients with bipolar dis-
order alone, the incidence rate of suicide-related
events was 215.0 (95% CI, 158.5 to 285.1) per
100,000 person-years for those who were not re-
ceiving antiepileptic drugs and 441.3 (95% CI,
315.2 to 600.9) per 100,000 person-years for those
who were receiving antiepileptic drugs. Among

546 n engl j med 363;6  nejm.org  august 5, 2010

The New England Journal of Medicine

Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
 Suicide-Related Events and Antiepileptic Drugs
patients with epilepsy alone, the incidence of
suicide-related events was 38.2 (95% CI, 26.3 to
53.7) per 100,000 person-years for those who
were not receiving antiepileptic drugs and 48.2
(95% CI, 39.4 to 58.5) per 100,000 person-years
for those who were receiving such treatment. Fig-
ure 1 shows the probability of no suicide-related
events over time in the various cohorts. The un-
derlying illness was more strongly associated
with suicide-related events than was the use or
nonuse of antiepileptic drugs.
Case–Control Analysis
The results of the case−control analysis were ad-
justed for age; duration of illness; status with
respect to previous use of antiepileptic drugs,
lithium, antipsychotic drugs, or antidepressants;
presence or absence of a history of alcohol abuse
or a mental disorder; and chronic disease score.
With adjustment for these factors, epilepsy and
depression were associated with a 50% increase
in the risk of suicide-related events; the risk as-
sociated with bipolar disorder was higher (Table
5 in the Supplementary Appendix). The risk of
suicide-related events was increased with a history
of alcohol abuse (odds ratio, 6.45; 95% CI, 5.68 to
7.32), a mental disorder (odds ratio, 2.75; 95% CI,
2.54 to 2.98), lithium use (odds ratio, 2.48; 95%
CI, 1.76 to 3.49), antipsychotic-drug use (odds ra-
tio, 1.64; 95% CI, 1.51 to 1.77), and antidepres-
sant-drug use (odds ratio, 3.69; 95% CI, 3.42 to
3.99). Chronic diseases were clearly associated with
suicide-related events, with a 20% increase in risk
per unit of increase in the chronic disease score.
The current use of antiepileptic drugs in pa-
tients without epilepsy, depression, or bipolar
disorder was associated with an increased risk of
suicide-related events (odds ratio, 2.57; 95% CI,
1.78 to 3.71). The risk of suicide-related events
was elevated both among patients with epilepsy
who received antiepileptic drugs and among
those who did not; among patients with epilepsy
and no coexisting depression or bipolar disorder
who did not receive antiepileptic drugs, the odds
ratio was 3.34 (95% CI, 2.34 to 4.78), and among
patients with epilepsy who received antiepileptic
drugs, the odds ratio was 2.31 (95% CI, 1.77 to
3.02). Similarly, the risk of suicide-related events
was increased among patients with depression
who were not receiving antiepileptic drugs (odds
ratio, 1.58; 95% CI, 1.43 to 1.74) and among
n engl j med 363;6  nejm.org  august 5, 2010
The New England Journal of Medicine
Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
1.00
Probability of No Suicide-Related Events
0.99
0.98
Reference group
Epilepsy alone
0.97 Epilepsy and antiepileptic drugs
Antiepileptic drugs
Depression alone
0.96
Depression and antiepileptic drugs
Bipolar disorder and antiepileptic drugs
0.95 Bipolar disorder alone
0.00
0 10 20
Years since Treatment Initiation
Figure 1. Kaplan−Meier Estimates of the Probability of No Suicide-Related
Events over Time, According to Cohort.
those who were receiving such treatment (odds
ratio, 2.06; 95% CI, 1.36 to 3.11) and among
patients with bipolar disorder who were not re-
ceiving antiepileptic drugs (odds ratio, 2.44; 95%
CI, 1.48 to 4.03) and among those who were (odds
ratio, 3.77; 1.24 to 11.43) (Table 4 and Fig. 2). As
compared with patients with bipolar disorder
who received neither lithium nor antiepileptic
drugs, the odds ratio for suicide-related events
was 0.56 (95% CI, 0.11 to 1.71) among patients
with bipolar disorder who received antiepileptic
drugs and lithium and 0.74 (95% CI, 0.45 to 1.19)
among those who received lithium alone.
In a series of separate case−control analyses,
we observed no significant increase in the risk
of suicide-related events among current users of
antiepileptic drugs as compared with nonusers
in the subgroup of patients with epilepsy alone
(odds ratio, 0.59; 95% CI, 0.35 to 0.98), the sub-
group of patients with bipolar disorder alone
(odds ratio, 1.13; 95% CI, 0.35 to 3.61), or the
subgroup of patients with epilepsy and depres-
sion (odds ratio, 1.24; 95% CI, 0.56 to 2.72).
Among patients with depression alone, the use
of antiepileptic drugs was significantly associ-
ated with an increased risk of suicide-related
events (odds ratio, 1.65; 95% CI, 1.24 to 2.19)
(Table 5). The effects of the use of various anti-
epileptic drugs are shown in Table 6 in the
Supplementary Appendix.
547
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 4. Combined Effect of Use of Antiepileptic Drugs and Various Conditions on the Risk of Suicide-Related Events,
Adjusted for Covariates.*

Case
Cohort Patients Controls Odds Ratio (95% CI)
number
No epilepsy, depression, or bipolar disorder
No use of antiepileptic drugs (reference group) 5,948 45,620 1.00
Current use of antiepileptic drugs 76 84 2.57 (1.78−3.71)
Epilepsy only (no depression or bipolar disorder)
No use of antiepileptic drugs 64 168 3.34 (2.34−4.78)
Current use of antiepileptic drugs 134 380 2.31 (1.77−3.02)
Depression only (no epilepsy or bipolar disorder)
No use of antiepileptic drugs 3,475 4,448 1.58 (1.43−1.74)
Current use of antiepileptic drugs 239 159 2.06 (1.36−3.11)
Bipolar disorder only (no epilepsy)
No use of antiepileptic drugs 60 42 2.44 (1.48−4.03)
Current use of antiepileptic drugs 40 7 3.77 (1.24−11.43)
Epilepsy and depression (no bipolar disorder)
No use of antiepileptic drugs 33 37 1.60 (0.91−2.81)
Current use of antiepileptic drugs 93 59 2.82 (1.71−4.67)
Epilepsy and bipolar disorder
No use of antiepileptic drugs 0 1 0
Current use of antiepileptic drugs 2 0 Infinite

* Odds ratios were derived from one conditional logistic-regression model with matching according to age (±5 years),
sex, and clinical practice. Odds ratios were adjusted for age, duration of disease, previous use of antiepileptic drugs,
lithium, antipsychotic drugs, or antidepressants; presence or absence of a history of alcohol abuse or a mental disor-
der; and chronic disease score.

Discussion antiepileptic drugs (Table 2); we controlled for

Our analyses of observational data collected as
part of clinical practice in the United Kingdom
confirmed the previously reported increased risk
of suicide-related events associated with epilepsy,
depression, and bipolar disorder.19 Our findings
suggest that treatment with antiepileptic drugs
does not confer an additional risk of suicide-
related events among patients with epilepsy.
The crude incidence of suicide-related events
among patients with epilepsy who did not receive
antiepileptic drugs was 38.2 per 100,000 person-
years, and the incidence was slightly higher
(48.2 per 100,000 person-years) among patients
with epilepsy who received antiepileptic drugs.
The most likely explanation for the difference
between the unadjusted and adjusted findings is
that patients who received antiepileptic drugs
were older and had more coexisting conditions
and risk factors than those who did not receive
these factors in the case−control analysis.
The risk of suicide-related events was in-
creased among patients who received antiepilep-
tic drugs for indications other than epilepsy,
bipolar disorder, or depression (odds ratio, 2.57;
95% CI, 1.78 to 3.71). It is not possible to be
certain about the indications for the use of anti-
epileptic drugs in these patients, but it is likely
that for some patients the indications were pain-
related (e.g., herpes zoster). Pain, especially chron-
ic pain, has been associated with an increased
risk of suicide.20 In patients with depression, the
risk was also higher among current users of anti-
epileptic drugs than among nonusers. Although
a causal role of antiepileptic drugs is possible, it
is also possible that the use of antiepileptic
drugs in these patients is a marker of severe
depression or the presence of another condition
that may be associated with an increased risk of
suicide-related events.21

548 n engl j med 363;6  nejm.org  august 5, 2010

The New England Journal of Medicine

Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
 Suicide-Related Events and Antiepileptic Drugs
A study has suggested that the risk of suicide
associated with bipolar disorder is lower during
treatment with lithium than with other treat-
ments.22 This finding was also observed in our
study; however, the number of patients who re-
ceived lithium was too low (29 patients overall)
to draw firm conclusions. Among patients with
bipolar disorder in our study, the odds ratio for
suicide-related events was 0.56 (95% CI, 0.11 to
1.71) among those who received antiepileptic
drugs and lithium and 0.74 (95% CI, 0.45 to 1.19)
among those who received lithium alone, as
compared with patients who received neither
lithium nor antiepileptic drugs. However, the
number of case patients who received lithium
was low (29 overall).
In general, our results do not confirm the
findings previously reported by the FDA.10 The
FDA study was a meta-analysis of data from
placebo-controlled clinical trials of the use of
antiepileptic drugs across a number of indica-
tions for up to 24 weeks. Among the 142 cases
of suicide-related events included in the meta-
analysis, 4 (2.8%) were completed suicides and
38 (26.8%) were suicide attempts. Our study fo-
cused on these “harder” end points that are of
greatest clinical concern and involved a longer
follow-up (mean, 6.2 years). Unlike the FDA meta-
analysis, our results suggest that in patients
with epilepsy, the use of antiepileptic drugs is
not associated with an increased risk of suicide
attempts or completed suicide. Our results were
similar for patients with bipolar disorder: we did
not detect a significant effect of antiepileptic
drugs on suicide-related events among patients
with this condition, which is associated with a
high risk of suicide,23 although our results had
wide confidence intervals. We could not rule out
a large association, since the upper bound of the
95% confidence interval for the odds ratio ex-
ceeded 3, but we also could not rule out a “pro-
tective” effect, since the lower confidence limit
was 0.35. Our findings in bipolar disorder were
consistent with those of other studies.24 Differ-
ences between the results of a meta-analysis
conducted by the FDA and epidemiologic studies
of suicidality have also been reported with regard
to antidepressants.25 Reasons for these differ-
ences include ascertainment bias11 and the in-
creased frequency of suicide in the first month
after initiation of antidepressant treatment.26
Since most clinical trials have a short duration
n engl j med 363;6  nejm.org  august 5, 2010
The New England Journal of Medicine
Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
100.0
No use of antiepileptic drugs Current use of antiepileptic drugs
10.0
Odds Ratio
1.0
0.1
No Disorder Epilepsy Depression Bipolar Disorder
Figure 2. Effect of Current Antiepileptic-Drug Use on the Risk of Suicide-­
Related Events.
The risk is shown according to disease status (epilepsy, depression, bipolar
disorder, or none of these conditions). The bars indicate 95% confidence
intervals.
of treatment and epidemiologic studies have a
long follow-up, the incidence of these early-occur-
ring suicides becomes diluted in epidemiologic
studies as compared with clinical trials.
The incidence rate of completed suicide in our
population was 1.5 per 100,000 patient-years. In
2007, the incidence rate was 16.8 per 100,000
among men and 5.0 per 100,000 among women
in the United Kingdom.27 Potential explanations
for the lower incidence of suicide in our cohort
include the fact that we calculated the incidence
rate of suicide after excluding from the sample
patients with a family history of suicide or a per-
sonal history of suicide attempts and the fact
that suicide has been shown to be underreported
in THIN.28 Underreporting would not have a
major effect on the associations observed in our
case−control analyses unless it affected some
cohorts more than others. The high positive pre-
dictive value for the diagnosis of suicide-related
events in the case validation supports the valid-
ity of our study.
An increased risk of suicide-related events
among patients receiving specific antiepileptic
drugs as compared with patients receiving topi-
ramate for any indication has recently been re-
ported.29 This study addressed a different ques-
tion and used substantially different methods.
For a patient who is being treated with antiepi-
leptic drugs, the risk of suicide results from the
combination of the risk associated with the ill-
ness prompting the use of these drugs and the
549
 The n e w e ng l a n d j o u r na l of m e dic i n e

Table 5. Association between Current Use of Antiepileptic Drugs and Suicide-Related Events in Each of the Cohorts,
Adjusted for Covariates.*

Case
Cohort Patients Controls Odds Ratio (95% CI) P Value
number
Epilepsy only (no depression or bipolar disorder)
No use of antiepileptic drugs 64 282 1.00
Current use of antiepileptic drugs 104 399 0.59 (0.35−0.98) 0.04
Depression only (no epilepsy or bipolar disorder)
No use of antiepileptic drugs 3475 17,747 1.00
Current use of antiepileptic drugs 99 157 1.65 (1.24−2.19) 0.001
Bipolar disorder only (no epilepsy)
No use of antiepileptic drugs 60 65 1.00
Current use of antiepileptic drugs 17 14 1.13 (0.35−3.61) 0.84
Epilepsy and depression (no bipolar disorder)
No use of antiepileptic drugs 33 64 1.00
Current use of antiepileptic drugs 61 85 1.24 (0.56−2.72) 0.60
Epilepsy and bipolar disorder
No use of antiepileptic drugs 0 0 1.00
Current use of antiepileptic drugs 1 0 Infinite (0.00) Infinite

* Odds ratios were derived from conditional logistic-regression models matched according to diagnostic category, age
(±5 years), sex, and clinical practice. Odds ratios were adjusted for age; duration of disease; previous use of antiepilep-
tic drugs, lithium, antipsychotic drugs, or antidepressants; presence or absence of a history of alcohol abuse or a men-
tal disorder; and chronic disease score. A total of 103 case patients did not match any controls.

risk associated with the drugs themselves. Our
results, stratified according to the indication for
drug use, suggest that illness carries more im-
portance than the use of antiepileptic drugs. The
results of our analysis of individual antiepileptic
drugs were imprecise, with wide confidence in-
tervals, but they point to differences in risk as-
sociated with antiepileptic drugs used for vari-
ous indications.
To minimize confounding, we excluded from
the analysis those patients with a history of sui-
cide-related events, so it is theoretically possible
that our results cannot be extrapolated to this
high-risk population. Although we tried to limit
confounding, some of the results may be partially
attributable to confounding by indication.30 For
example, the increased risk observed among
patients with pain may be due to an effect of
antiepileptic drugs or may simply reflect the fact
that patients with pain, which is associated with
an increased risk of suicide, received antiepilep-
tic drugs. The way we built our cohorts, with no
overlap of subjects between cohorts, may have
led to immortal time bias. This bias was avoided
in the case−control analysis.
In conclusion, our findings do not provide
support for an association between antiepileptic
drugs and suicide-related events among patients
receiving antiepileptic drugs for epilepsy. How-
ever, we did observe an association between the
current use of antiepileptic drugs and suicide-
related events among patients with depression
and among patients who did not have epilepsy,
depression, or bipolar disorder.
Disclosure forms provided by the authors are available with
the full text of this article at NEJM.org.
We thank Delories Dunn, Dr. Janet Price, and Dr. Randal Mar-
shall for their suggestions for revisions of an earlier version of
the manuscript.

References
1. DeLeo D, Bertolote J, Lester D. Self-
directed violence. In: Krug EG, Dahlberg
LL, Mercy JA, Zwi AB, Lozano R, eds.
World report on violence and health. Ge-
neva: World Health Organization, 2002: 2. Moscicki EK. Epidemiology of suicid-
183-212. (Accessed July 9, 2010, at http:// al behavior. In: Silverman MM, Maris RW,
www.who.int/violence_injury_prevention/ eds. Suicide prevention: toward the year
violence/world_report/en/index.html.) 2000. New York: Guilford, 1985:22-35.

550 n engl j med 363;6  nejm.org  august 5, 2010

The New England Journal of Medicine

Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
 Suicide-Related Events and Antiepileptic Drugs
3. Idem. Epidemiology of completed and
attempted suicide: toward a framework
for prevention. Clin Neurosci Res 2001;1:
310-23.
4. Mortensen PB, Agerbo E, Erikson T,
Qin P, Westergaard-Nielsen N. Psychiatric
illness and risk factors for suicide in Den-
mark. Lancet 2000;355:9-12.
5. Nilsson L, Ahlbom A, Fahraman BY,
Asberg M, Tomson T. Risk factors for sui-
cide in epilepsy: a case control study. Epi-
lepsia 2002;43:644-51.
6. Christensen J, Vestergaard M, Mor­ten­
sen PB, Sidenius P, Agerbo E. Epilepsy and
risk of suicide: a population-based case-
control study. Lancet Neurol 2007;6:693-8.
7. Qin P, Xu H, Laursen TM, Vestergaard
M, Mortensen PB. Risk for schizophrenia
and schizophrenia-like psychosis among
patients with epilepsy: population based
cohort study. BMJ 2005;331:23.
8. Hesdorffer DC, Hauser WA, Olafsson
E, Ludvigsson P, Kjartansson O. Depres-
sion and suicide attempt as risk factors
for incident unprovoked seizures. Ann
Neurol 2006;59:35-41.
9. Blumer D, Montouris G, Davies K,
Wyler A, Phillips B, Hermann B. Suicide in
epilepsy: psychopathology, pathogenesis
and prevention. Epilepsy Behav 2002;3:
232-41.
10. Katz R. Briefing document for the July
10, 2008 advisory committee meeting to
discuss antiepileptic drugs (AEDs) and
suicidality. Memorandum. (Accessed July 9,
2010, at http://www.fda.gov/ohrms/dockets/
ac/08/briefing/2008-4372b1-01-FDA-Katz
.pdf.)
11. Posner K, Oquendo MA, Gould M,
Stanley B, Davies M. Columbia Classifi-
cation Algorithm of Suicide Assessment
(C-CASA): classification of suicidal events
in the FDA’s pediatric suicidal risk analy-
sis of antidepressants. Am J Psychiatry
2007;164:1035-43.
n engl j med 363;6  nejm.org  august 5, 2010
Downloaded from nejm.org on April 20, 2016. For personal use only. No other uses without permission.
Copyright © 2010 Massachusetts Medical Society. All rights reserved.
12. Hippisley-Cox J, Coupland CA, Vino­
gradova Y, Robson J, Brindle P. The per-
formance of the QRISK cardiovascular
risk prediction algorithm in an external
UK sample of patients from General Prac-
tice: a validation study. Heart 2008;94:
34-9.
13. The Health Improvement Network
Web site. (Accessed July 9, 2010, at http://
www.thin-uk.com.)
14. American Foundation for Suicide Pre-
vention. Risk factors for suicide. (Accessed
July 9, 2010, at http://www.afsp.org/index
.cfm?page_id=05147440-E24E-E376-
BDF4BF8BA6444E76.)
15. NHS Centre for Coding and Classifi-
cation. The Read codes, version 3. London:
Stationery Office, 1996.
16. Arellano FM, Arana A, Wentworth
CE, Fernandez-Vidaurre C, Schlienger RG,
Conde E. Lymphoma among patients with
atopic dermatitis and/or treated with top-
ical immunosupressants in the United
Kingdom. J Allergy Clin Immunol 2009;
123:1111-6.
17. Von Korff M, Wagner EH, Saunders K.
A chronic disease score from automated
pharmacy data. J Clin Epidemiol 1992;45:
197-203.
18. Arellano FM, Yood MU, Wentworth
CE, et al. Use of cyclo-oxygenase 2 inhibi-
tors (COX-2) and prescription non-steroi-
dal anti-inflammatory drugs (NSAIDs) in
UK and USA populations. Pharmacoepi-
demiol Drug Safe 2006;15:861-72.
19. Harris EC, Barraclough B. Suicide as
an outcome for mental disorders: a meta-
analysis. Br J Psychiatry 1997;170:205-28.
20. Braden JB, Sullivan MD. Suicidal
thoughts and behavior among adults with
self-reported pain conditions in the Na-
tional Comorbidity Survey Replication.
J Pain 2008;9:1106-15.
21. Vigo DV, Baldessarini RJ. Anticonvul-
sants in the treatment of major depressive
The New England Journal of Medicine
disorder: an overview. Harv Rev Psychia-
try 2009;17:231-41.
22. Goodwin FK, Fireman B, Simon GE,
Hunkeler EM, Lee J, Revicki D. Suicide
risk in bipolar disorder during treatment
with lithium and divalproex. JAMA 2003;
290:1467-73.
23. Kessler RC, Chiu WT, Demler O, Wal-
ters EE. Prevalence, severity, and comor-
bidity of 12-month DSM-IV disorders in
the National Comorbidity Survey Replica-
tion (NCS-R). Arch Gen Psychiatry 2005;
62:617-27.
24. Gibbons RD, Hur K, Brown CH, Mann
JJ. Relationship between antiepileptic
drugs and suicide attempts in patients
with bipolar disorder. Arch Gen Psychia-
try 2009;66:1354-60.
25. Gibbons RD, Brown CH, Hur K, Mar-
cus SM, Bhaumik DK, Mann JJ. Relation-
ship between antidepressants and suicide
attempts: an analysis of the Veterans
Health Administration data sets. Am J
Psychiatry 2007;164:1044-9.
26. Jick H, Kaye JA, Jick SS. Antidepres-
sants and the risk of suicidal behaviors.
JAMA 2004;292:338-43.
27. Suicides: UK suicide rates continue to
fall. Newport, United Kingdom: Office
for National Statistics, January 2009. (Ac-
cessed July 9, 2010, at http://www.statistics
.gov.uk/pdfdir/sui0109.pdf.)
28. Hall G. Validation of death and sui-
cide recording on the THIN UK primary
care database. Pharmacoepidemiol Drug
Safe 2009;18:120-31.
29. Patorno E, Bohn RL, Wahl PM, et al.
Anticonvulsant medications and the risk
of suicide, attempted suicide, or violent
death. JAMA 2010;303:1401-9.
30. Bell GS, Mula M, Sander JW. Suicidal-
ity in people taking antiepileptic drugs:
what is the evidence? CNS Drugs 2009;23:
281-92.
Copyright © 2010 Massachusetts Medical Society.
551