ª 2010 The Authors

Bipolar Disorders 2010: 12: 483–493 Journal compilation ª 2010 John Wiley & Sons A/S
BIPOLAR DISORDERS

Original Article

Lamotrigine versus lithium as maintenance

treatment in bipolar I disorder: an open,
randomized effectiveness study mimicking
clinical practice. The 6th trial of the Danish
University Antidepressant Group (DUAG-6)
Licht RW, Nielsen JN, Gram LF, Vestergaard P, Bendz H. Lamotrigine Rasmus W Lichta, Jannie N
versus lithium as maintenance treatment in bipolar I disorder: an open, Nielsena, Lars F Gramb, Per
randomized effectiveness study mimicking clinical practice. The 6th trial Vestergaarda and Hans Bendzc
of the Danish University Antidepressant Group (DUAG-6). a
Mood Disorders Research Unit, Aarhus University
Bipolar Disord 2010: 12: 483–493. ª 2010 The Authors.
Hospital, Risskov, bClinical Pharmacology, IST,
Journal compilation ª 2010 John Wiley & Sons A ⁄ S.
University of South Denmark, Odense, Denmark,
c
The Lund Department of Clinical Sciences,
Objectives: In industry-generated pivotal studies, lamotrigine has been
Psychiatry, University of Lund, Lund, Sweden
found to be superior to placebo and comparable to lithium in the
maintenance treatment of bipolar I disorder. Here, we directly compared
lamotrigine to lithium under conditions similar to clinical routine
conditions.

Methods: Adult bipolar I disorder patients with at least two episodes

within the last five years and an index episode requiring treatment were
randomized to lithium (n = 78; doses adjusted to obtain serum levels of
0.5–1.0 mmol ⁄ L) or to lamotrigine (n = 77; up-titrated to 400 mg ⁄ day)
as maintenance treatments. Randomization took place when clinically
appropriate, and comedication was allowed within the first six months
after randomization. The patients were enrolled from March 2001 to
December 2005, and observations were censored December 2006,
allowing a subgroup of patients to be followed for more than five years.
The primary outcome measure was time to predefined endpoints
indicating insufficient maintenance treatment, and the major secondary
outcome measure was time to any study endpoint. Data were analyzed
primarily by Cox proportional regression models. doi: 10.1111/j.1399-5618.2010.00836.x

Key words: bipolar I disorder – lamotrigine –

Results: For the primary outcome measure, the crude Hazard Rate lithium – maintenance treatment – randomized
Ratio (HRR) (lamotrigine relative to lithium) was 0.92 [95% confidence clinical trial
interval (CI): 0.60–1.40]. When the primary endpoints were broken down
by polarity, the HRRs (lamotrigine relative to lithium) for mania and Received 24 July 2009, revised and accepted for
depression were, respectively, 1.91 (95% CI: 0.73–5.04) and 0.69 (95% publication 29 March 2010
CI: 0.41–1.22). There was no between-group difference in terms of
staying in study [HRR: 0.85 (95% CI: 0.61–1.19)]. Most treatment Corresponding author:
failures occurred within the first 1.5 years of treatment, and, among Rasmus W. Licht, M.D., Ph.D.
patients followed for at least five years, practically no patients were Mood Disorders Research Unit
maintained successfully on monotherapy with either of the drugs. Aarhus University Hospital, Risskov
The lithium-treated patients reported diarrhea, tremor, polyuria, Skovagervej 2
and thirst more frequently. Two cases, probably lamotrigine-related, of 8240 Risskov, Denmark
benign rash occurred. Fax: +45 77893859
E-mail: rasmus.licht@ps.rm.dk
Conclusions: No differences in maintenance effectiveness between
lithium and lamotrigine could be demonstrated. Lamotrigine was better This trial is registered at ClinicalTrials.gov (Identifier
tolerated than lithium, but apparently this did not influence the outcome. #NCT00226135).

483
Licht et al.

In the long-term drug treatment of bipolar disor-
der, lithium has been the mainstay since the 1970s
(1). However, despite the fact that the efficacy of
lithium has been established in various research
settings, including randomized, placebo-controlled
trials (2, 3), many bipolar disorder patients started
on lithium under routine conditions do not achieve
sufficient long-term stabilization (4). A major
reason for this is poor compliance, which to some
extent is related to the considerable burden of side
effects of lithium (5). Thus, alternative prophylac-
tic drugs with higher tolerability are needed.
When the antiepileptic drug lamotrigine was
introduced in psychiatry in the late 1990s, it
appeared to be beneficial for the treatment of
acute bipolar depression, with no risk of inducing
mania (6). Apart from the risk of rash (which can
be minimized by a slow up-titrating dosing strat-
egy) (7), the drug was relatively well tolerated and
easy to manage for the clinician since there were
no requirements for blood monitoring. Subse-
quently, the superiority of lamotrigine (and lith-
ium) over placebo in the maintenance treatment of
bipolar disorder was demonstrated in two similarly
designed industry-generated trials using lithium as
an internal comparator for validating the study
assay (8, 9). However, in these studies, designed to
obtain regulatory approval of lamotrigine for the
indication of maintenance treatment of bipolar
disorder, the study samples were enriched prior to
randomization by selection of patients who toler-
ated lamotrigine and who were able to be main-
tained on lamotrigine monotherapy for at least
one week in the aftermath of an acute episode
without showing signs of major destabilisation.
This type of sample enrichment increases the
chance of detecting a signal in favour of lamotri-
gine over placebo, but at the expense of a
narrowing of the generalizability of the findings.
Additionally, since the lithium arm was incorpo-
rated in a nonenriched way, the comparisons of
lamotrigine with lithium might be difficult to
interpret. For these reasons, we initiated in 2001
(i.e., before the results from these maintenance
studies were available) an effectiveness trial with
no sample enrichment, aiming at comparing the
preventive potential of lamotrigine with that of
lithium under conditions common to clinical
practice. Since there were no clear indications
from the literature that lamotrigine had antimanic
potentials, it was hypothesized that lithium would
be more effective than lamotrigine.
Methods
Design and patients
This was an investigator-sponsored, open, ran-
domized trial, executed within the framework of
the Danish University Antidepressant Group
(DUAG). There was one main centre (Aarhus
University Hospital, Risskov, Denmark) and seven
additional centres (four Danish and three Swedish
centres). Patients with bipolar I disorder diagnosed
according to the DSM-IV, aged at least 18 years
and in need of prophylaxis, were recruited during
or in the aftermath of an index episode, outlined in
detail below. The index episode was treated at the
discretion of the clinician, and, when it was timely,
the patients were randomly allocated to either
lithium or lamotrigine as continuation and pro-
phylactic (i.e., maintenance) treatments. Addi-
tional medications were allowed until month 6
after randomization, and monotherapy failures
(primary endpoint) were not recorded until after
that point in time. The study design is outlined in
Figure 1.
Patients were enrolled (and randomized) from
March 2001 until December 2005 and were fol-
lowed until they reached a study endpoint (see
below) or the final censoring in December 2006.
This allowed a patient follow-up time after ran-
domization of up to 5.8 years, depending on the
date of randomization of the patient.
The index episode was either major depression,
mania, or mixed mania according to the Cincinnati
criteria (10), and patients were required to have
had at least two episodes within the last five years.
The index episode could be with or without
psychotic symptoms, was of a severity requiring
acute drug treatment, and had an onset within the
last year prior to randomization. The diagnostic

This trial was partly supported by unrestricted grants from The Stanley Medical Institute, Bethesda, MD, USA; Pulje til
Psykiatrisk Forskning, Aarhus University Hospital, Risskov, Denmark; and GlaxoSmithKline, Copenhagen, Denmark.
GlaxoSmithKline, Denmark, provided the lamotrigine and lithium tablets. RWL has received a research grant from
GlaxoSmithKline; has received speakers honoraria from Eli Lilly & Co., Janssen Cilag, GlaxoSmithKline, and Bristol-Myers
Squibb; and has served on advisory boards of AstraZeneca and Bristol-Myers Squibb. LFG has received a research grant
from Lundbeck. HB has received a research grant from GlaxoSmithKline. JNN and PV have no conflicts of interest to
disclose.

484

Randomization when
clinically appropriate
Monotherapy failures
recorded hereafter
Various psychotropics
Manic/mixed manic
or depressed episode
6 months
12 months Up to 5.8 years
Recruitment period: from March 2001 Fixed study end (final censoring):
to December 2005 December 2006
Fig. 1. Study design.
evaluation of the current and the past episodes was
performed by using the Present State Examination
(PSE-10) (11). Patients were excluded if written
informed consent could not be obtained, if they
suffered from epilepsy needing antiepileptic ther-
apy, if they previously had experienced sympto-
matic recurrence despite receiving adequate
prophylactic treatment with either lithium or
lamotrigine for at least six months (clinical judg-
ment), if there was current drug or alcohol abuse
that might result in protocol violation, or if there
were any contraindications to study medications.
Pregnant women or women of childbearing age not
using contraceptives were also excluded.
Randomization for maintenance took place
when clinically appropriate, i.e., when the symp-
toms of the (index) episode had begun to decrease
but before full symptomatic stabilization necessar-
ily had been obtained and when the patient had a
capacity for reliably agreeing to maintenance
treatment. This point in time was chosen through
clinical judgment by the investigators with no aid
from operational criteria. After the written
informed consent had been obtained, the patient
received an allocation number, which was written
on the consent form. This form was subsequently
faxed to the DUAG centre (at the University of
Southern Denmark, Odense, Denmark), and there
a person not directly involved in the clinical part of
the study assigned the patient to one of the two
treatment groups according to a computer-gener-
ated randomization plan. The consent form with
the result of the randomization was then faxed
back to the investigator. This procedure rendered
reallocation after randomization impossible. The
randomization was done in blocks of four within
each centre, with the block size kept hidden for the
investigators.
After randomization, the patients were evaluated
at visits with their treating psychiatrist (investiga-
Lamotrigine versus lithium as bipolar maintenance treatment
tor) as clinically needed and in accordance with the
protocol, which required a minimum of fixed visits
at month 1, month 3, and month 6, and subse-
Lamotrigine quently at least one visit per each three months of
follow-up.
The study was approved by the regional Danish
Lithium Biomedical Research Ethics Committee Systems,
the Regional Ethical Review Board in Lund,
Sweden, the Danish Medicines Agency, the
Swedish Medical Products Agency, the Danish
Data Protection Agency, and the Swedish Data
Inspection System.
Treatment
During the index episode, prior to randomization,
the patients were treated at the discretion of the
responsible clinicians.
When randomized to lithium, patients received
lithium citrate (Danish centres) or lithium sulphate
(Swedish centres) (each tablet containing 6 mmol
lithium ion), one tablet twice daily through week 1,
with subsequent adjustment of dose until a serum
lithium level (12-hour value) of 0.5–1.0 mmol ⁄ L
was achieved. The patients randomized to lamo-
trigine were treated with lamotrigine tablets,
25 mg ⁄ day during weeks 1 and 2, 50 mg ⁄ day
during weeks 3 and 4, and 100 mg ⁄ day during
week 5. After week 5 the daily dose was increased
by 100 mg weekly to a maximum of 400 mg daily.
The dose could be reduced if severe side effects
occurred. Lamotrigine was given in the morning.
Current treatment with valproate, carbamazepine,
and oxcarbazepine, all interacting pharmacokinet-
ically with lamotrigine, were discontinued at
randomization.
For patients treated with lithium until random-
ization and assigned to continue this treatment,
and likewise for lamotrigine, the postrandomiza-
tion doses were adjusted according to the require-
ments outlined above. If patients receiving lithium
were assigned to lamotrigine, lithium was tapered
off after randomization over a period of 1–3
months, and if patients receiving lamotrigine were
assigned to lithium, lamotrigine was discontinued
at randomization.
Additional antipsychotic or antidepressant
drugs were allowed within the first six months
after randomization, with no restrictions in terms
of type, number, or dose of medications,
and no attempts were made to encourage the
investigators to discontinue such drugs at ran-
domization. However, the investigators were
encouraged to achieve monotherapy with study
drugs at month 6. Supplemental treatment with
benzodiazepines and benzodiazepine-like drugs,
485
Licht et al.
e.g., zolpidem, was allowed throughout the
study.
Serum levels of lithium and lamotrigine were
determined at least every three months of follow-
up as a control of compliance, and of lithium also
for safety reasons. Blood samples were drawn in
the morning and, for lithium, approximately
12 hours after the evening dose. Pill counts were
also used as a compliance check.
Assessments
In addition to the diagnostic evaluation described
above, clinical information obtained via the initial
patient interview was collected and recorded. At
the same time, i.e., when symptoms of the index
episode had started to decrease and just prior to
randomization, the patients were rated with the
Bech-Rafaelsen Mania Scale (MAS) (12), the Bech-
Rafaelsen Melancholia Scale (MES) (13), and the
Clinical Global Impression Scale (CGI) (14).
At each visit the investigator evaluated whether
an endpoint (see below) had been reached since the
last visit and recorded its date. If the endpoint
indicated insufficient maintenance treatment (see
below), the polarity of symptoms was recorded
based on clinical judgment, i.e., manic symptoms,
depressive symptoms, mixed symptoms, or uncer-
tain polarity.
At inclusion, blood samples were analyzed to
screen for various medical conditions. For
patients randomized to lithium treatment, renal
and thyroid function was monitored every three
and six months throughout the follow-up. Weight
was measured at inclusion and at each visit.
Adverse medical events were queried at each visit
and recorded.
During the execution of the trial, regular study
monitoring of the participating centres according
to the principles of Good Clinical Practice was
performed by the DUAG centre by persons not
directly involved in the clinical part of the trial.
Outcome measures
The primary outcome measure was time to any
of four primary study endpoints (displayed in
Table 1). It is a composite measure which
presumably reflects four different types of mono-
therapy failure in terms of insufficient continuation
and ⁄ or prophylactic treatment response. Since the
patients within the first six months after random-
ization were allowed to be hospitalized and to
receive psychotropics in addition to study drugs
(and benzodiazepines), a primary endpoint (hospi-
talization and ⁄ or additional psychotropics) could
486
Table 1. Study endpoints
Primary endpointsa
1a. Psychotropic treatment in addition to study drugs and
benzodiazepines still required at month 6 (after
randomization)
1b. Hospitalization still required at month 6 (after randomization)
2a. Psychotropic treatment during at least one week (in addition
to study drugs and benzodiazepines) required after month
6 (after randomization)b
2b. Hospitalization during at least one week required after month
6 (after randomization)b
Other endpoints
3. Patient discontinued study drug
4. Consent withdrawn
5. Protocol violation
6. Severe abuse of alcohol or drugs
7. Adverse events (investigator decision)
8. Lost to follow-up
9. Other
Suicide
Suicide attempt
Pregnancy
10. Censoring due to fixed study end
a
Occurring by definition not until month 6 (after randomization).
b
Provided that endpoints 1a or 1b had not been reached.
by definition not occur until month 6 after
randomization.
A secondary outcome measure was time to any
endpoint (Table 1) appearing first and before the
fixed study end (December 2006), i.e., study
survival. Other secondary outcome measures were
time to any of the primary endpoints broken down
by polarity of relapse ⁄ recurrence, i.e., by time to
mania and time to depression.
Tolerability was examined through the recorded
adverse events and through change in weight from
baseline.
Statistical analyses
Data were entered from printed case report forms
into an Access database; after entry, all data were
listed and full proofreading was performed
twice. Before statistical analysis, the data were
transformed into SPSS format and analyses
were performed by SPSS, version 15 (SPSS, Inc.,
Chicago, IL, USA).
For simple between-group comparisons, Mann–
Whitney U-tests, correcting for ties, or chi-square
tests with YatesÕ correction were used with a
significance level of 5%. Time-to-event data, i.e.,
time to endpoints, were analysed through the use
of Cox proportional regression models (primary
analysis), with computing hazard rate ratios
(HRRs) for the treatment effect (lamotrigine rela-
tive to lithium) with 95% confidence intervals (CI).
Observations that were terminated for reasons
other than the endpoints in question were
 Lamotrigine versus lithium as bipolar maintenance treatment

censored. In addition to crude (unadjusted) HRRs, were needed. As a substantial proportion of

HRRs adjusted multivariately for the presence of
confounders and interaction variables were com-
puted. Potential confounders and interaction vari-
ables were identified through performing various
exploratory Kaplan–Meier survival plots, and
these variables (together with treatment) were
entered (all at once) as covariates in a Cox
regression analysis using Wald statistics with a
significance level of 5%. Through these procedures,
potential baseline variables independently predict-
ing outcome were also analyzed. Crude Kaplan–
Meier survival plots were displayed to illustrate the
cumulative probability of not reaching an end-
point. All analyses were based on the total sample
of randomized patients (intention-to-treat popula-
tion).
The study was powered as a superiority trial,
aiming at not overlooking a superiority of lithium
over lamotrigine corresponding to an HRR of 2
(lamotrigine relative to lithium). Based on the
assumption that 31% of lithium-treated patients
would relapse over a period of two years (15),
corresponding to an HR of 0.21 year)1 {HR =
[) ln (survival fraction through the time T) ⁄ T]},
and with an acceptable risk of type I and type II
error of, respectively, 5% and 20%, 114 patients
patients could be expected to drop out without
providing valid information in the primary analy-
sis, a sample size of 150 was the goal.
Results
Study sample
During the enrollment period (March 2001 to
December 2005), 155 patients were randomized.
Out of the 155 patients, 123 (79%) were recruited
from the main centre (Aarhus University Hospital,
Risskov, Denmark). It was planned that the main
centre would recruit 50% of all patients, and the
additional centres would each recruit 10% of the
patients, but the actual distribution reflects an
unforeseen slow recruitment from the additional
centres. Table 2 displays the result of the random-
ization regarding the distribution of patient char-
acteristics at baseline, i.e., at randomization. Apart
from the presence of at least four episodes within
the year prior to randomization, the groups were
similar. As displayed, an index episode of mania
was present in a little less than half of the patients,
and an index episode of depression was present in a
little more than half of the patients. Patients were
relatively ill, with the vast majority of the patients

Table 2. Characteristics of bipolar I disorder study subjects (intention-to-treat population) at randomization

Characteristic Lamotrigine (n = 77) Lithium (n = 78)

Age, years, mean (SD) 38.2 (11.8) 37.3 (11.5)

Females, n (%) 40 (52) 36 (46)
Age of onset, years, mean (SD) 26.2 (12.0) 25.1 (10.8)
Relative with affective disorder, n (%) 50 (65) 43 (55)
Previous episodes
Number of previous episodes, median (IQR) 3 (2–10) 3 (2–8)
At least 4 episodes within the last year, n (%) 15 (20) 7 (9)
History of suicide attempts, n (%) 21 (27) 29 (37)
Prior lithium prophylaxis, n (%) 13 (17) 15 (19)
Prior prophylaxis with any drug, n (%) 24 (31) 27 (35)
Current substance abuse, n (%) 3 (4) 5 (6)
Index episode
Duration of index episode, weeks, mean (SD) 9.8 (6.0) 9.9 (7.2)
Index episode of mania, n (%) 35 (46) 28 (36)
Index episode of mixed mania, n (%) 6 (8) 7 (9)
Index episode of depression, n (%) 36 (47) 43 (55)
Psychotic symptoms, n (%) 21 (27) 25 (32)
Hospitalization, n (%) 61 (79) 68 (87)
Received lithium up to randomization, n (%) 23 (30) 25 (32)
Severity of illness at randomization
CGI score, mean (SD) 3.1 (1.3) 3.0 (1.4)
MAS score, mean (SD) 4.5 (5.6) 3.8 (5.0)
MES score, mean (SD) 9.2 (8.5) 9.0 (8.1)
Remission, clinical judgment, n (%) 14 (18) 18 (23)
CGI score £ 3, n (%) 50 (65) 50 (64)

IQR = interquartile range; CGI = Clinical Global Impression; MAS = Bech-Rafaelsen Mania Scale; MES = Bech Rafaelsen Melancholia
Scale.

487
Licht et al.

being hospitalized during the index episode and
one-third of patients experiencing psychotic symp-
toms during the index episode. Only about one-
third of the patients had received prior prophylaxis
(of more than six monthsÕ duration). Less than
one-fifth of the patients had received prior lithium
prophylaxis, and no patients had received prior
lamotrigine prophylaxis.
The most recent (in relation to endpoint)
serum lithium level (mean ± SD) was 0.69 ±
0.20 mmol ⁄ L, and for lamotrigine the most recent
dose was 379 ± 66 mg. The corresponding serum
level of lamotrigine was 22.5 ± 12.7 lmol ⁄ L. In
the lithium group and in the lamotrigine group,
respectively, 56 (72%) and 51 (66%) patients
received comedication with antidepressants within
the first six months after randomization. As to
comedication with antipsychotics, the numbers
were, respectively, 42 (54%) and 54 (69%). Lith-
ium was tapered off in 23 (30%) patients allocated
to lamotrigine.
Effectiveness
Patient accountability and patient flow are pre-
sented in Table 3 and in Figure 2. With respect to
each of the endpoints, patients were equally
distributed among the treatment groups. Out of
the 27 and 26 patients reaching endpoint 2 (2a or
2b) in the lamotrigine and in the lithium group,
respectively, 6 and 8 patients were hospitalized
(endpoint 2a).
Consistent with the long periods of follow-up
(Table 4), a high proportion of patients reached a
primary or secondary study endpoint before the
final study end. For example, out of the 29 patients
with a follow-up period of at least five years
(Table 4), only 2 patients (both in the lithium
group) did not reach a primary or secondary
endpoint within five years. For interstudy compar-
ative reasons, the major two-year outcome of the
subsample of the 122 patients with a follow-up
period of at least two years after randomization is
shown in Table 5. Again, there were no between-
group differences.
With respect to the primary outcome measure,
i.e., time to endpoint 1 or 2, the crude HRR
(95% CI) (lamotrigine relative to lithium) was
0.92 (0.60–1.40). Corresponding to this, the
Kaplan–Meier plot showed overlapping curves
(Fig. 3).
When the primary endpoints (1 and 2) were
broken down by polarity, i.e., mania versus other
polarity or depression versus other polarity,
lithium performed numerically better than
lamotrigine regarding mania and lamotrigine
better than lithium regarding depression, but
there were no statistically significant differences.
The crude HRR (95% CI) (lamotrigine relative
to lithium) was 1.91 (0.73–5.04) regarding
mania and 0.69 (0.41–1.22) regarding depression.
In the lamotrigine group, 13 (17%) patients
relapsed into mania and 23 (30%) patients
relapsed into depression, whereas in the lithium

Table 3. Disposition of randomized bipolar I disorder study subjects (intention-to-treat population)

Lamotrigine (n = 77) Lithium (n = 78)

Endpoint n (%) n (%)

1. Monotherapy not achieved (or admission) at month 6 16 (21) 18 (23)

Mania 6 (8) 0 (0)
Depression 8 (10) 15 (19)
Mixed episode or uncertain polarity 2 (3) 3 (4)
2. Additional treatment or admission required after month 6 27 (35) 26 (33)
Mania 7 (9) 7 (9)
Depression 15 (20) 16 (21)
Mixed episode or uncertain polarity 5 (7) 3 (4)
3. Patient discontinued study drug 4 (5) 7 (9)
4. Consent withdrawn 6 (8) 11 (14)
5. Protocol violation 3 (4) 1 (1)
6. Severe abuse of alcohol or drugs 0 (0) 1 (1)
7. Adverse events (investigator decision) 3 (4) 5 (6)
8. Lost to follow-up 3 (4) 1 (1)
9. Other 2 (3) 1 (1)
Suicide 0 (0) 1 (1)
Suicide attempt 1 (1) 0 (0)
Pregnancy 1 (1) 0 (0)
10. Censored due to study enda 13 (17) 7 (9)

No statistically significantly between-group differences.

a
Fixed study end (final censoring) at December 2006.

488
 Lamotrigine versus lithium as bipolar maintenance treatment

Lamotrigine Lithium

Randomized patients
n = 77 n = 78

Patients reaching a secondary endpoint before month 6

n = 14 n = 19

Patients reaching month 6

n = 63 n = 59

Patients reaching a primary endpoint at month 6

n = 16 n = 18

Patients achieving monotherapy and outpatient status at month 6

n = 47 n = 41

Patients reaching a primary/secondary endpoint after month 6

n = 27/7 n = 26/8

Patients not reaching a primary/secondary endpoint at study end

n = 13 n=7

Fig. 2. Patient flow.

Table 4. Total sample of randomized bipolar I disorder study subjects subdivided into subsamples (cohorts) according to duration of follow-upa

Cohort Lamotrigine Lithium

1 year follow-up (or more); included up to December 2005 77 78

2 years follow-up (or more); included up to December 2004 62 60
3 years follow-up (or more); included up to December 2003 45 45
4 years follow-up (or more); included up to December 2002 34 33
5 years follow-up (or more); included up to December 2001 15 14

No statistically significant between-group differences.

a
Variable duration of follow-up due to inclusion between March 2001 and December 2005 and fixed study end (final censoring) at
December 2006.

Table 5. Major two-year outcome of the subsample of 122 randomized bipolar I disorder study subjects with duration of follow-up of at least two yearsa

Lamotrigine (n = 62) Lithium (n = 60)

n (%) n (%)

Monotherapy not achieved (or admission) at month 6 (endpoint 1) 12 (19) 12 (20)

Additional treatment or admission required after month 6 (endpoint 2) 21 (34) 19 (32)
Other reasons for study end (endpoints 3–9) 18 (29) 19 (31)
Still in study after 2 years 11 (18) 10 (17)

No statistically significant between-group differences.

a
With the fixed study end (and final censoring) at December 2006, this cohort was characterized by inclusion and randomization before
January 2005.

group, the numbers were, respectively, 7 (9%) Through the multivariate analysis (entering
and 31 (40%). Details on polarity are given in potential confounders and interaction variables),
Table 3. an interaction between the polarity of the index
489
Licht et al.

1.0 similar (Fig. 4). In accordance with this, the crude

HRR (95% CI) (lamotrigine relative to lithium)
was 0.85 (0.61–1.19). When time to discontinuation
0.8 of study medications (endpoints 3 or 7; Table 3)
Lamotrigine (n = 77)
was analysed, no between-group difference was
0.6 Lithium (n = 78) found. Regarding time on study medications, the
patients were no longer followed after reaching an
P(t)

endpoint. Therefore, the study survival curves

0.4
0.2
0.0
0 500 1000 1500 2000
Days of observation (t)
Fig. 3. Estimates of the cumulative probability [P(t)] of ran-
domized bipolar I disorder study subjects (intention-to-treat
population) not reaching a primary endpoint indicating insuf-
ficient maintenance treatment (Kaplan–Meyer survival curves).
episode and treatment was identified. For an index
episode of mania, the adjusted HRR (95% CI) was
2.11 (1.48–5.32), i.e., in favour of lithium, whereas
for an index episode of depression, the adjusted
HRR (95% CI) was 0.57 (0.33–0.96), i.e., in favour
of lamotrigine. There were no interactions with any
other variables, including centre (main centre
versus the others), remission status at randomiza-
tion, having received lithium up to randomization,
or having received previous lithium (or any)
prophylaxis.
When time to any of the endpoints (1 to 9), i.e.,
survival in study, was analysed, censoring at
planned study end only, the Kaplan Meier survival
curves for the two treatment groups again were
1.0
0.8
0.6
Lamotrigine (n = 77)
P(t)
reflect the duration that patients could be kept on
study medications within the context of the study.
The median survival times in days (with quartiles
in parentheses) were, respectively, 232 (184, 384)
and 202 (183, 379) for the lamotrigine group and
lithium group.
Predictors of outcome
A number of previous episodes above 3 was
independently (and statistically significantly)
related to an increased risk of reaching a primary
endpoint (1 or 2) with an adjusted HRR (95% CI)
of 1.66 (1.01–2.74). Likewise, rapid cycling was
associated with an increased risk with an adjusted
HRR (95% CI) of 2.00 (1.11–3.63).
Adverse events
One patient who was allocated to the lithium group
committed suicide by hanging three months after
randomization. The patient had achieved remission
shortly after the randomization and had been
discharged from the hospital. Depression emerged
one month later and the patient was rehospitalized.
Serum lithium levels had been measured regularly
and were adequate. There were no other serious
adverse events.
Seven patients were withdrawn at the discretion
of the treating investigator due to possible drug-
related, treatment-emergent, medical adverse
events, 3 in the lamotrigine group (2 cases of
benign rash and one case of acne) and 4 in the
lithium group (because of acne, severe memory
problems, polyuria, and severe weight gain).

Commonly reported adverse events are pre-

Lithium (n = 78)
0.4 sented in Table 6. Diarrhea, tremor, polyuria,
and thirst were reported statistically significantly
more frequently in the lithium-treated patients
0.2
compared with the lamotrigine-treated patients.
In the lithium group and the lamotrigine group,
0.0 weight gain (mean ± SD) was, respectively,
0 500 1000 1500 2000 1.49 ± 3.98 kg and )0.04 ± 6.84 kg (p = 0.09).
Days of observation (t)

Fig. 4. Estimates of the cumulative probability [P(t)] of ran- Discussion

domized bipolar I disorder study subjects (intention-to-treat
population) remaining in the study (Kaplan–Meyer survival No superiority of lithium over lamotrigine in terms
curves). of overall long-term effectiveness could be demon-
490

Table 6. Adverse events reported by at least 5% of randomized bipolar I
disorder study subjects (intention-to-treat population) within at least one of
the treatment groups
Lamotrigine
(n = 77)
Adverse events n (%)
Headache 11 (14)
Dizziness 12 (16)
Memory difficulties 0 (0)
Rash 6 (8)
Acne 8 (10)
Nausea 6 (8)
Diarrhea 0 (0)
Polyuria 1 (1)
Tremor 4 (5)
Thirst 3 (4)
Weight gain 7 (9)
a
p < 0.05.
strated in this study. Since the minimal superiority
of lithium relative to lamotrigine not to be
overlooked (with a likelihood of 80%) was con-
siderable, i.e., a factor 2, finding no statistically
significant differences indeed would still be com-
patible with differences at lower levels. However,
for both primary and secondary outcome meas-
ures, the HRR estimates were close to unity, and
the Kaplan Meier curves were almost overlapping.
Even though the results of the two previously
mentioned pivotal studies (8, 9) testing the efficacy
of lamotrigine against placebo as maintenance
treatment were reported before the completion of
our study, and with each indicating overall equiv-
alence between lamotrigine and lithium, there were
still reasons to believe that in our design lithium
would be superior to lamotrigine. First, as outlined
in the introduction, our study sample was not
enriched as was the case in the pivotal studies.
Second, also in contrast to the pivotal studies, we
excluded patients with prior nonresponse to lith-
ium maintenance. When interpreting our negative
findings, it should be taken into account that the
patients were characterised by having relatively
many previous episodes and by a high proportion
of them being hospitalized or psychotic during the
index mood episode, which in itself may contribute
to a relatively high risk of recurrence despite any
treatment (16, 17). Within the present sample, the
strong influence of a high number of previous
episodes (and rapid cycling) on outcome could in
fact be demonstrated.
Since the pivotal studies, one recruiting patients
with an index episode of depression (8) and the
other recruiting patients with an index episode of
mania (9), were similarly designed and had com-
parable study samples (besides the index episode),
Lamotrigine versus lithium as bipolar maintenance treatment
an analysis on the combined sample was performed
(18). In the interest of comparison, it should be
noted that this combined sample was comparable
Lithium to the sample of the present study in terms of the
(n = 78) distribution of index episodes. The combined
n (%) analysis confirmed the almost identical outcome
6 (8) of the lamotrigine-treated and the lithium-treated
7 (9) patients in terms of time to any mood episode.
5 (6) However, the risk for depression was numerically
5 (6) lower and the risk for mania numerically higher in
4 (5)
4 (5)
the lamotrigine group compared with the lithium
22 (28)a group, with HRRs of 0.84 and 1.83, respectively,
9 (12)a and with the latter value being statistically signif-
25 (32)a icantly different from unity (18). A similar differ-
22 (28)a ential profile, lamotrigine versus lithium, was also
8 (10)
indicated in the present study, although the
numerical between-group differences found here
were not statistically significant, most likely due to
low power, since our sample size was less than half
that of the combined sample of the pivotal studies.
In line with differential profiles of the study
drugs, we found an interaction between treatment
and index episode, with lamotrigine performing
better than lithium in patients with an index episode
of depression but poorer than lithium among
patients with an index episode of mania. However,
such an interaction could not be demonstrated in
the combined analysis of the pivotal studies (8, 9,
18). Due to this inconsistency and due to the fact
that evaluations of outcome were performed
openly, the interaction may not be a true biological
finding but a methodological artefact. Even though
the primary outcome measure was supposed to
reflect the clinical condition of the patient more
than the beliefs of the patients or the investigators,
the possibility that the results of the pivotal studies,
appearing in the beginning of the patient recruit-
ment in the present study, may have influenced the
investigators cannot be ignored.
As illustrated above, the open design may be a
source of bias. However, it may facilitate the
recruitment of patients, thereby adding to the
representativeness of the study sample. This was
diligently demonstrated in a report by Greil et al.
(19), who, in the context of an open, randomized,
comparative maintenance trial in bipolar disorder,
carefully documented the recruitment and flow of
patients from screening to randomization. In the
present study, we screened many times the number
of subjects randomized, and the majority of the
screened patients were excluded due to prior
insufficiency of prophylactic effect of any of the
study drugs, but unfortunately this was not doc-
umented. Despite the fact that we could not
compare the randomized patients with the poten-
tially eligible patients, our final study sample
491
Licht et al.
appeared fairly similar to that of a recently
described catchment area non-trial treatment
sample of bipolar disorder patients from the
U.K. (20). The only striking discrepancy is that
more than two-thirds of the patients in our study
sample had never received any prior maintenance
treatment. However, this most likely can be
explained by the fact that insufficiency of prior
maintenance treatment with lithium or lamotrigine
was a major exclusion criterion.
In open trials, the dropout rates may also be
minimized. The two-year dropout rate of our
subsample of the 122 patients with a follow-up
period of at least two years (Table 5) was 30%.
This rate was comparable to the rate of 28%
observed in the open study by Greil et al. (21), with
a follow-up period of 2.5 years, and lower than the
rate of 38% in the double-blind combined pivotal
studies, with a duration of follow-up of only
1.5 years (18). The median study survival of the
present sample was roughly 2.5-fold higher than
that of the combined sample of the pivotal studies.
However, this difference may to a large extent be
explained by the difference in the duration of
follow-up.
The study design was employed to attempt to
provide a representative sample with maximal
potential for generalizability. We randomized
patients independently of their remission status,
as long as they were able to agree to the mainte-
nance treatment as suggested by the treating
physicians. Our design differed from the traditional
design of a maintenance trial, where only remitted
patients are randomized, and mimicked initiating
maintenance under common clinical conditions.
The fact that the primary study endpoint (Table 1)
could not occur within the first six months after
randomization also reflects clinical practice as a
maintenance treatment rarely is judged to be
insufficient within the first months of maintenance,
even if the patient is symptomatic or requires
additional medications during this period. If, on
the other hand, such additional medications cannot
be discontinued after a certain period of time, the
maintenance treatment in question can rightly be
considered insufficient as a continuation and ⁄ or
prophylactic (monotherapy) treatment. Obviously,
the vertical drop in the Kaplan–Meier curves in
Figures 3 and 4 about month 6 is due to our choice
of a 6-month duration for this period of time.
Due to the long duration of follow-up in the
present trial, i.e., up to 5.6 years and with 58% of
the patients being followed for at least three years
(Table 4), our data may address the question of the
ideal duration of comparative long-term preventive
drug trials in bipolar I disorder. As most endpoints
492
occurred within the first 1.5 years after randomi-
zation (Figs. 3 and 4), a trial duration of 1.5 years
seems sufficient, in line with epidemiological data
(17). Among the subsample of 29 patients followed
for at least five years, only 2 patients were
maintained successfully on monotherapy after five
years, confirming what is known from pharma-
coepidemiological surveys, i.e., that monotherapy
over time is a rare exception.
In terms of tolerability, diarrhea, tremor, poly-
uria, and thirst were reported statistically signifi-
cantly more frequently in lithium-treated patients
compared with lamotrigine-treated patients,
although there were no more overall study with-
drawals among the lithium-treated patients than
among the lamotrigine-treated patients. The pat-
tern of side effects among the lithium-treated
patients indicates that the serum level of lithium
was sufficiently high. Among the lamotrigine-
treated patients, 2 probably drug-related cases of
benign rash occurred. In 2 of the 8 patients on
lamotrigine who developed acne, a causal relation
to the drug was likely (22). It is worth noting that
lamotrigine was well tolerated despite the relatively
high target dose of 400 mg. This dose was twice the
dose now approved for bipolar disorder mainte-
nance in most countries, and was chosen to give
lamotrigine a fair chance in terms of efficacy.
In conclusion, no differences among bipolar I
disorder patients in terms of overall effectiveness
between maintenance treatment with lithium and
lamotrigine could be demonstrated. However, the
findings are compatible with the notion that
lithium may be more effective than lamotrigine in
the prevention of mania and lamotrigine more
effective than lithium in the prevention of depres-
sion. Most treatment failures occurred within the
first 1.5 years of treatment, and among patients
with a duration of follow-up of at least five years,
practically no patients were maintained success-
fully on monotherapy with either drug. Lamotri-
gine was better tolerated than lithium, but
apparently this did not influence the outcome.
Due to the different profiles of lamotrigine and
lithium, a future maintenance trial evaluating a
combination of the two drugs against each of the
drugs given in monotherapy would indeed be
warranted. The design and execution of such a
trial could glean inspiration from the recently
published BALANCE trial (23, 24).
Acknowledgements
The DUAG-6 Study was conducted within the framework of
the Danish University Antidepressant Group (DUAG). The
authors thank the following investigators for data collection
 Lamotrigine versus lithium as bipolar maintenance treatment
and medical care of patients at each centre: Maria C. Stuhr,
Else H. Bøgh, Elisabeth Tehrani, Ulla Klaening, Erik R.
Larsen, Inger G. Brødsgaard, Gitte Hausmann, Claus Z.
Jensen, and Bente Stadil (University Hospital of Aarhus,
Risskov, Denmark); Ib S. Thomsen, Jørgen Iversen, and
Bjarne Christensen (Aalborg Psychiatric Hospital, Aalborg,
Denmark); Jens K. Larsen and Lise Tornbjerg (Department of
Psychiatry, Gentofte Hospital, Copenhagen, Denmark);
Birgitte B. Bendsen (Department of Psychiatry, Frederiksberg
Hospital, Copenhagen, Denmark); Lars R. Petersen and Klaus
Martiny (Department of Psychiatry, Hilleroed General Hospi-
tal, Hillerød, Denmark); Hans-Peter Mofors (Sollentuna
Hospital, Stockholm, Sweden); Ingrid Ternfrid (University
Hospital, Lund, Sweden); and Maia Alvariza (Karolinska
Institute, Stockholm, Sweden). The authors also thank Loni
Ledderer (secretary of the DUAG) and Joan Nisted for their
assistance with administration, coordination, randomization,
and data management; Henrik Horneberg for his assistance
with randomization; and Birthe Riddersholm for her help with
providing study medications and her assistance with the data
monitoring. Finally, the authors thank Loni Ledderer, Klaus
Martiny, Jens K. Larsen, Gary Evoniuk, and Walter Paska for
their helpful comments to the manuscript and Morten
Frydenberg for statistical advice.
References
1. Müller-Oerlinghausen B, Berghöfer A, Bauer M. Bipolar
disorder. Lancet 2002; 359: 241–247.
2. Baldessarini RJ, Tondo L, Hennen J, Viguera AC. Is
lithium still worth using? An update of selected recent
research. Harv Rev Psychiatry 2002; 10: 59–75.
3. Geddes JR, Burgess S, Hawton K, Jamison K, Goodwin
GM. Long-term lithium therapy for bipolar disorder:
systematic review and meta-analysis of randomized con-
trolled trials. Am J Psychiatry 2004; 161: 217–222.
4. Licht RW, Vestergaard P, Brodersen A. Long-term
outcome of patients with bipolar disorder commenced
on lithium prophylaxis during hospitalization: a complete
15-year register-based follow-up. Bipolar Disord 2008; 10:
79–86.
5. Licht RW, Vestergaard P, Rasmussen NA, Jepsen K,
Brodersen A, Hansen PE. A lithium clinic for bipolar
patients: 2-year outcome of the first 148 patients. Acta
Psychiatr Scand 2001; 104: 387–390.
6. Calabrese JR, Bowden CL, Sachs GS, Ascher JA, Mona-
ghan E, Rudd GD. A double-blind placebo-controlled
study of lamotrigine monotherapy in outpatients with
bipolar I depression. Lamictal 602 Study Group. J Clin
Psychiatry 1999; 60: 79–88.
7. Calabrese JR, Sullivan JR, Bowden CL et al. Rash in
multicenter trials of lamotrigine in mood disorders: clinical
relevance and management. J Clin Psychiatry 2002; 63:
1012–1019.
8. Calabrese JR, Bowden CL, Sachs G et al. A placebo-
controlled 18-month trial of lamotrigine and lithium
maintenance treatment in recently depressed patients with
bipolar I disorder. J Clin Psychiatry 2003; 64: 1013–1024.
9. Bowden CL, Calabrese JR, Sachs G et al. A placebo-
controlled 18-month trial of lamotrigine and lithium
maintenance treatment in recently manic or hypomanic
patients with bipolar I disorder. Arch Gen Psychiatry 2003;
60: 392–400.
10. McElroy SL, Keck PE Jr, Pope HG Jr, Hudson JI, Faedda
GL, Swann AC. Clinical and research implications of the
diagnosis of dysphoric or mixed mania or hypomania. Am
J Psychiatry 1992; 149: 1633–1644.
11. World Health Organization. Schedules for Clinical Assess-
ment in Neuropsychiatry: Version 2.0. Geneva: World
Health Organization, 1994.
12. Bech P, Bolwig TG, Kramp P, Rafaelsen OJ. The Bech-
Rafaelsen Mania Scale and the Hamilton Depression
Scale. Acta Psychiatr Scand 1979; 59: 420–430.
13. Bech P. The Bech-Rafaelsen Melancholia Scale (MES) in
clinical trials of therapies in depressive disorders: a 20-year
review of its use as outcome measure. Acta Psychiatr Scand
2002; 106: 252–264.
14. Guy W. ECDEU Assessment Manual for Psychopharma-
cology. Kensington: George Washington University, Bio-
metric Laboratory, 1976.
15. Goodwin FK, Jamison KR. Manic-Depressive Illness.
New York: Oxford University Press, 1990.
16. Perlis RH, Ostacher MJ, Patel JK et al. Predictors of
recurrence in bipolar disorder: primary outcomes from the
Systematic Treatment Enhancement Program for Bipolar
Disorder (STEP-BD). Am J Psychiatry 2006; 163: 217–224.
17. Tohen M, Waternaux CM, Tsuang MT. Outcome in
mania: a 4-year prospective follow-up of 75 patients
utilizing survival analysis. Arch Gen Psychiatry 1990; 47:
1106–1111.
18. Goodwin GM, Bowden CL, Calabrese JR et al. A pooled
analysis of 2 placebo-controlled 18-month trials of
lamotrigine and lithium maintenance in bipolar I disorder.
J Clin Psychiatry 2004; 65: 432–441.
19. Greil W, Ludwig-Mayerhofer W, Steller B et al. The
recruitment process for a multicenter study on the long-
term prophylactic treatment of affective disorders. J Affect
Dis 1993; 28: 257–265.
20. Raymont V, Bettany D, Frangou S. The Maudsley Bipolar
Disorder Project. Clinical characteristics of bipolar disor-
der I in a catchment area treatment sample. Eur Psychiatry
2003; 18: 13–17.
21. Greil W, Ludwig-Mayerhofer W, Erazo N et al. Lithium
versus carbamazepine in the maintenance treatment of
bipolar disorders – a randomised study. J Affect Dis 1997;
43: 151–161.
22. Nielsen JN, Licht RW, Fogh K. Two cases of acneiform
eruption associated with lamotrigine. J Clin Psychiatry
2004; 65: 1720–1722.
23. Geddes JR, Goodwin GM, Rendell J et al. Lithium plus
valproate combination therapy versus monotherapy
for relapse prevention in bipolar I disorder (BALANCE):
a randomised open-label trial. Lancet 2010; 375: 385–
395.
24. Licht RW. A new BALANCE in bipolar I disorder. Lancet
2010; 375: 350–352.
493