Directory UMM :Data Elmu:jurnal:B:Biological Psichatry:Vol48.Issue6.2000:

Bipolar Disorder Therapeutics: Maintenance Treatment
Gary S. Sachs and Michael E. Thase
Although most of the care received by bipolar patients
occurs during the maintenance phase, relatively little
empirical data is available to guide long-term treatment
decisions. We review literature pertaining to key questions
related to use of pharmacotherapy in the maintenance
phase of bipolar disorder. The few double-blind trials with
a reasonable sample size are restricted to bipolar I
patients and address a modest range of questions mostly
related to use of lithium. One rigorous multicenter trial
found valproate to have prophylactic benefit. Other studies with valproate alone and in combination suggest
efficacy equivalent to lithium and perhaps greater than
carbamazepine. Data available for combination treatment
are sparse but moderately encouraging.
Maintenance treatment with standard antidepressant medications appears destabilizing for some bipolar patients,
particularly following a mixed episode. Although some
bipolar patients may benefit from combined treatment with
a mood stabilizer and a standard antidepressant medication, current knowledge does not allow confident selection
of the bipolar patients who might benefit. Clozapine and
perhaps other atypical antipsychotics are promising options for maintenance treatment but have not been evaluated in double-blind trials. The numerous other agents

used in maintenance treatment are primarily adjuncts to
lithium, valproate, or carbamazepine, and information
about them is largely anecdotal and uncontrolled.
Study design for maintenance trials remains an imperfect
art. Conclusions must be drawn cautiously, given the
limited generalizability of study designs that accession
samples enriched with presumed treatment responders,
randomize patients after brief periods of partial remission,
abruptly taper prior treatment, make no attempt to distinguish relapse from recurrence, use no formal outcome
assessments, or report hospitalization as the only outcome
criterion. Biol Psychiatry 2000;48:573–581 © 2000 Society of Biological Psychiatry
Key Words: Bipolar disorder, maintenance, relapse prevention, prophylaxis, continuation, combination treatment

From the Department of Psychiatry, Harvard Medical School, and Partners Bipolar
Treatment Center, Massachusetts General Hospital, Boston (GSS) and the
Department of Psychiatry, University of Pittsburgh School of Medicine,
Western Psychiatric Institute and Clinic, Pittsburgh, Pennsylvania (ME).
Address reprint requests to Gary S. Sachs, M.D., Massachusetts General Hospital,
Clinical Psychopharmacology Unit, WACC 815, 15 Parkman St., Boston MA
02114.

Received February 25, 2000; revised July 7, 2000; accepted July 12, 2000.

© 2000 Society of Biological Psychiatry

Introduction

M

aintenance phase treatment comprises the majority
of care received by nearly all bipolar patients.
Unfortunately, practitioners seeking evidence on which to
base treatment decisions will find research on maintenance
phase treatment is both sparse and difficult to interpret.
Expert opinion remains the only basis for answering many
fundamental questions. We review the available literature
pertaining to key questions related to use of medications
for maintenance phase treatment of patients with bipolar
illness.

What Is Maintenance Phase Treatment?

Given the variable course of bipolar illness, the discussion of maintenance treatment starts by necessity with
definition. Maintenance phase treatment is often defined as spanning the time from when the patient has
recovered from an acute episode to the time of onset of
a new acute episode. Once a patient has recovered, the
primary focus of treatment shifts from resolution of
acute symptoms to prophylaxis. Conceptually this shift
in therapeutic focus corresponds to reaching the physiologic end point in the course of the acute episode,
when the pathophysiology underlying the acute symptomatology has been corrected. Whether a true recovery
is reached because the episode has run its natural course
or because of treatment, the high rates of relapse
observed following early discontinuation of treatment
strongly support viewing the recovered state as distinct
from the physiologic state at the point of initial clinical
remission. In the absence of objective markers informing us that the pathology has been corrected, the
maintenance phase is defined operationally. Following
a tradition borrowed from unipolar depression research
(Frank et al 1991), the maintenance phase follows a
period of sustained remission during which acute phase
treatment is continued for a period of time. Such
continuation phase treatment increases the likelihood

that treatment will be sustained throughout the course of
the acute episode and lessens the likelihood that subsequent discontinuation will be closely followed by
relapse.
0006-3223/00/$20.00
PII S0006-3223(00)00991-4

574

BIOL PSYCHIATRY
2000;48:573–581

This crucial phase of treatment remains under studied.
What little empirical data exists comes mostly from
studies of unipolar depression. That literature suggests that
the continuation phase, the period of time between the
onset of substantial clinical remission of acute depression
or mania and the beginning of the maintenance phase,
should not be less than 8 weeks (Prien and Kupfer 1986)
and those with residual symptoms are at high risk for
relapse (Mindham et al 1973). Prien and Kupfer (1986)

found that patients who were symptom free (defined as
Global Assessment Scale [GAS] score $ 71) for at least
16 weeks had a significantly lower relapse rate (18%) than
those who were free of symptoms for less than 16 weeks
(59%). In fact, patients who were well for 16 weeks or
more did no better with active treatment (imipramine and
lithium) than with placebo. Hlastala et al (1997) reported
data indicating the conditional probability of remaining
well (defined as average Hamilton Depression Scale
score # 7 and Bech–Rafaelsen scores # 7) over 8
consecutive weeks for bipolar patients in their first week
of remission following an acute episode was zero. The
median times to reach recovery criteria (8 consecutive
weeks in remission) following manic, depressive, and
mixed episodes were 20 weeks, 40 weeks, and 50 weeks,
respectively.
Accepting that the median duration of the acute episode
is 12– 42 weeks, we must recognize the limitations of
studies that randomize patients after periods of remission
as brief as 1 week. Withdrawal of medication during the

early months of recovery appears analogous to early
removal of sutures following physical trauma. This point
should be kept in mind when reviewing the literature.
Some otherwise rigorous published studies claim to report
results of maintenance treatment, but can only be observing relapse because subjects are randomized to discontinue
acute phase treatment after very brief periods of relative
euthymia.

Who Should Have Maintenance Treatment?
This question generates considerable debate. Many
experts recommend offering maintenance treatment to
all patients who have suffered even a single acute manic
episode, whereas others feel it is appropriate to reserve
long-term maintenance treatment for patients who have
experienced two or more manic episodes. Experts with
the former opinion point out the 95% lifetime risk of
recurrence and studies showing loss of acute and
prophylactic benefit (at least of lithium) among patients
with as few as three episodes (Gelenberg et al 1989;
Swann et al 1999). Experts who hold the latter opinion

point out that the median duration of remission following the first acute episode is more than 4 years (Angst

G.S. Sachs and M.E. Thase

1981; Kreapelin 1921; Roy-Byrne et al 1985; Zis et al
1980) and the paucity of evidence justifying chronic
treatment for bipolar patients beyond 32 weeks (Coryell
et al 1987).
Another aspect of selecting patients for maintenance
treatment relates to the likelihood that patients will discontinue treatment regardless of the psychiatrist’s recommendations. Patients who discontinue lithium treatment,
particularly when discontinuation is abrupt, appear to be at
increased risk for recurrence. Guy Goodwin and colleagues have attempted to quantify the risks and benefits
of lithium maintenance and recommend against offering
maintenance treatment to patients judged unlikely or
unwilling to adhere to treatment for at least 2 years
(Goodwin 1994). It is unclear when in the course of
treatment the liability to such discontinuation phenomena
develops or if similar phenomena apply to other maintenance treatments. In a large prospective study, Bowden et
al (2000) found no evidence of discontinuation phenomena following discontinuation of acute treatment with
lithium or divalproex.

Although current literature does not resolve the issue of
who should have treatment, from a practical point of view
the areas of agreement extend so broadly as to largely
render the question itself a matter of little clinical significance. First, by the time most patients receive a bipolar
diagnosis they have already suffered multiple episodes.
There is general agreement that patients with three or more
episodes warrant long-term maintenance therapy. Second,
since many clinical experts recommend continuation of
effective acute phase treatments for a period of 1 year
beyond the onset of clinical remission, a substantial
portion of bipolar patients will relapse within the period of
prudent continuation of treatment. Third, a surprisingly
large proportion of patients simply never achieve the
period of sustained euthymia considered a sufficient prerequisite to consideration of treatment discontinuation
(Hlastala et al 1997; Maj et al 1998; Peselow et al 1994;
Prien et al 1984). Thus two distinct subgroups exist among
the population of patients receiving long-term care: those
who have achieved a stable remission and those who
have not. Fourth, serious consideration of when to
recommend long-term treatment becomes pointless in

the absence of measures to improve treatment adherence. Johnson and McFarland (1996) have demonstrated that among patients prescribed lithium half
discontinued treatment in less than 3 months. Until the
improved adherence techniques used in manualized
psychosocial interventions (Cochran 1984) become integrated into routine practice, discussion of long-term
treatment may be best postponed in favor of attention to
sustaining the therapeutic effort through the first 3
months of remission.

Bipolar Disorder: Maintenance Therapeutics

Which Treatments Are Effective in the
Maintenance Phase?
The answer differs depending on which evidence is
considered. Double-blind controlled trials appear to provide consistent strong evidence of prophylactic efficacy
for lithium. Coryell’s analysis of this data suggests the
benefit of prophylaxis might be confined to the first 32
weeks following recovery from an acute episode (Coryell
et al 1997). The 58 respondents participating in an expert
consensus guideline development survey (Sachs et al
2000); however, unanimously recommend continuing

maintenance treatment with lithium, valproate, and/or
carbamazepine when such treatment appears effective
during the acute phase. Less data are available for valproate and carbamazepine. Data from naturalistic follow-up raises concerns about treatment effectiveness
(Coryell et al 1997; Harrow et al 1990; Maj et al 1998;
O’Connell et al 1991; Peselow et al 1994; Sachs et al
1994; Tohen et al 1990).

Randomized Controlled Trials
Early lithium prophylaxis studies generally show a strong
benefit for lithium maintenance. Overall, the 10 available
reports of placebo-controlled trials (for a review, see Goodwin and Jamison 1990) involved 514 patients and reveal
recurrence rates of 81% for placebo-treated patients and 34%
for lithium-treated patients. Unfortunately, serious shortcomings limit the generalizability of this literature. First, most of
the studies were published before 1975 (Baastrup et al 1970;
Coppen et al 1991; Cundall et al 1972; Melia 1970; Stallone
et al 1973), and few used formal assessment instruments
(Bowden et al 2000; Prien et al 1984). Second, with one
exception (Bowden et al 2000), all use an “enriched design,”
which limits participation to subjects with known positive
response. Third, all studies in which subjects were randomized after apparently successful long-term response to lithium

and the control condition begins with abrupt lithium discontinuation are likely to overestimate the protective effect of
lithium due to the effect of lithium discontinuation (Suppes et
al 1993). Fourth, response rates for the prophylactic portion
of the trial are reported as a percentage of patients who
achieve remission criteria rather than as a percentage of
patients entering the initial acute treatment phase.
Although eight maintenance study reports are available
showing an overall response rate of 72% in bipolar and
schizoaffective patients treated with carbamazepine (Ballenger and Post 1978; Coxhead et al 1992; Greil and
Kleindienst 1999; Greil et al 1997, 1998; Lusznar et al
1988; Okuma et al 1981; Placidi et al 1986; Small et al
1995; Watkins et al 1987), results for carbamazepine are
not impressive. Three studies (Coxhead et al 1992; Okuma
et al 1981; Watkins et al 1987) found the outcome of

BIOL PSYCHIATRY
2000;48:573–581

575

maintenance treatment with carbamazepine no different
than the outcome of lithium maintenance. Greil found
better results with lithium maintenance than with carbamazepine maintenance (Greil and Kleindienst 1999; Greil et al
1997, 1998). In the only placebo-controlled study, however, Okuma et al (1981) reported that the benefit of
carbamazepine over a placebo did not reach statistical
significance. Several reports suggest the apparent benefit
of carbamazepine may fade after years of apparent success. These suggestions are hardly unique to carbamazepine but do not suggest robust long-term benefit either.
The most valuable data come from the few studies with
samples larger than 80 subjects. Interpretation of these
studies requires detailed consideration of each study design (Table 1). The National Institute of Mental Health
(NIMH) Collaborative Study (Prien et al 1984, 1988)
compared long-term outcome of treatment with lithium,
imipramine, or combined lithium and imipramine for
patients with bipolar disorder (n 5 117) and unipolar
disorder (n 5 150). Bipolar patients (n 5 216) included
in this trial entered an open preliminary treatment phase in
which depression, manic, and mixed episodes were treated
with lithium and imipramine along with any other treatments deemed appropriate. Subjects were randomized to
double-blind maintenance treatment with lithium, imipramine, or continued combination treatment, if they were
able to meet three conditions: 1) patients were able to
discontinue other acute treatments and remain on stable
maintenance doses of both lithium (with serum levels of
$0.6 mmol/L) and imipramine ($75 mg/day) for 2
consecutive months, 2) their GAS rating was above 60,
and 3) neither the depression nor the mania score on the
Raskin Severity of Depression and Mania Scale exceeded
7. Those randomized were not necessarily well. In fact,
15% of the randomized sample (n 5 18 of 117) never met
recovery criteria (remain well 8 consecutive weeks).
Treatment success (completed at least 1 year without a
recurrence) was significantly more common in the lithium-alone (33%) and lithium plus imipramine (33%)
groups than the imipramine-alone group (8%). The rate of
depression was nearly the same across the three groups;
however, in the imipramine-alone group 53% experienced
a manic or mixed recurrence, compared with 26% and
28% for the patients receiving lithium and lithium plus
imipramine, respectively. Patients who entered the study
in a mixed or manic episode enjoyed significantly higher
success rates in the lithium (53%) and the combination
(47%) groups, compared with the imipramine-alone group
(8%). A similar but less robust trend is observed for
patients entering the study depressed, with success rates
favoring lithium (22%) and the combination (18%) over
imipramine (9%) but not reaching statistical significance.
Rapid cycling patients and those with mixed episodes

G.S. Sachs and M.E. Thase

Divalproex superior to the
placebo
GAS, Global Assessment Scale; SADS-C, Schedule for Affective Disorders and Schizophrenia (Change Version).

Any
Bowden et al 2000

In remission on lithium
monotherapy
(N 5 157)
Index episode within 12
weeks (N 5 571)
Gelenberg et al 1989

Stabilize on prior dose
of lithium 2 months

SADS-C subscales,
depression and
mania
SADS-C subscales,
depression and
mania

4 months (2 months prior
to entry 1 2 months
prospectively)
None, meets remission
criteria $ 7 days

Lithium (90)
Divalproex (187)
Placebo (94)

2.6 times higher for patients
in 0.4 – 0.6 mmol/L group

Significantly higher in
imipramine group

Outcome recurrence rates
Treatment (n)

Lithium 1 imipramine (36)
Lithium (42)
Imipramine (36)
Lithium 0.8 –1.0 mmol/L (47)
Lithium 0.4 – 0.6 mmol/L (47)
None (15% never met
recovery criteria)
GAS $ 60
RSDM # 7
Any

Open treatment
Sample accessioned

Acute episode at entry
(N 5 216)
Prien et al 1984

Study

Prerandomization

Criteria for
remission

Minimum time from
recovery to
randomization

Postrandomization

BIOL PSYCHIATRY
2000;48:573–581

Table 1. Double-Blind Prophylaxis Trials with More than 90 Bipolar Subjects

576

had extremely poor outcomes. Among rapid cycling patients (n 5 9) only three were randomized and none met
criteria for treatment success. All 12 patients treated with
imipramine alone following a mixed episode (Prien et al
1988) suffered recurrences, and almost all were mania
(75%) or mixed episodes (17%). The recurrence rate for
patients with mixed episodes who received lithium or the
combination was significantly lower (38% and 41%,
respectively). Combination treatment offered no advantage over lithium monotherapy.
Results for prophylaxis in the unipolar sample include a
placebo control group and are relevant for the observation
of mania or mixed episodes in 6% of placebo-treated
patients as well as 8% in the imipramine and 5% in the
combination group but none of the lithium-treated patients. Severity of the index episode was not a significant
influence on outcome in the bipolar sample. Among
unipolars, however, all three active treatments appear to
have prophylactic benefit for patients following a moderate index episode, but among unipolar subjects with severe
index episodes a clear disadvantage for lithium alone is
evident.
These studies found no disadvantage due to the addition
of a standard antidepressant to the lithium maintenance
regimen. Quitkin et al (1978, 1981), however, found that
bipolar patients maintained on lithium alone experience
superior overall outcome, compared with those receiving
lithium and imipramine. The discrepancy may be an
artifact of the NIMH Collaborative Study design and the
frequency with which patients received low doses of
imipramine. The Collaborative Study simply did not
randomize those patients with poor response to combination treatment during the stabilization phase.
It is worth noting that the statistical methodology used
in the above studies involved comparing the proportion of
patients suffering a recurrence. A reanalysis (Shapiro et al
1989) of the NIMH Collaborative Study data using the
Kaplan–Meier product-limit method of survival analysis
found similar overall results and results for subjects with
index manic episodes (Shapiro et al 1989). For subjects
with index depressive episodes, however, median remission duration was significantly longer for combination
treatment (7.8 months), compared with lithium (3.4
months) and imipramine (4.8 months). Maintenance studies now routinely employ survival analysis. To date, study
analyses carried out using survival analysis have not
detected significant differences in any subsequent parallelgroup maintenance study; however, some of these same
studies (Bowden et al 2000; Greil et al 1997) have
detected significant differences using the simpler comparison proportions methodology.
Gelenberg and colleagues (1989) reported results from a
multicenter study that randomized bipolar patients who

Bipolar Disorder: Maintenance Therapeutics

were well on lithium for 6 months to maintenance treatment targeting serum lithium to either a standard range
(0.8 –1.0 mmol/L) or a low range (0.4 – 0.6 mmol/L).
The two main findings from this study were a 2.6-fold
higher risk of relapse among patients randomized to the
low serum lithium group and the observation that the
protective benefit of lithium was only evident among
subjects with two or fewer episodes before study entry.
Furthermore, higher serum lithium levels were associated
with higher rates of adverse effects and lower rates of
compliance. Among the sample randomized to the 0.8 –1.0
mmol/L range, 50% were found to have lithium levels
below the 0.8 on half their visits.
Reanalysis of the Gelenberg et al results (Rosenbaum et
al 1992) taking into account the baseline lithium level calls
the original conclusion into question. There was no difference in outcome between those patients who started
with and maintained lithium treatment in the low range
(n 5 11) and those who started and maintained lithium in
the standard range (n 5 41). Patients who entered the
study with lithium in the standard range and were randomized to the low range had a risk of recurrence threefold
higher than that experienced by either of the other two
groups. This group for whom lithium dosage was reduced
by 50% at randomization accounts for nearly all relapses
in the original study. This post hoc reanalysis suggests a
strong influence of abrupt reduction as well as abrupt
discontinuation and makes it clear that this effect needs to
be taken into account in the design and interpretation of all
maintenance phase research.
The only placebo-controlled parallel-group study published in the past 25 years is a large multicenter trial with
rigorous methodology. Bowden et al (2000) enrolled 571
bipolar patients into the open phase of a double-blind
study that required at entry an index episode of mania or
hypomania within the preceding 12 weeks. Patients (both
well and ill) entering the open prerandomization treatment
phase received any treatment deemed appropriate by the
study psychiatrist. To enter the randomized phase, subjects
had to sustain remission criteria (Schedule for Affective
Disorders and Schizophrenia [Change Version] subscale
scores of ,13 and ,11 for depression and mania, respectively) for at least one visit after discontinuing all treatments except lithium or valproate within 3 months of
study entry. Those subjects (n 5 372) meeting remission
criteria on consecutive evaluations at least 6 days apart
were randomized 2:1:1 to receive valproate, lithium, or a
placebo in the double-blind treatment phase. Over the
course of the double-blind treatment median serum levels
were 1.0 6 0.48 mEq/L for lithium and 84.8 6 29.9
mg/mL for valproate (mean 6 SD). Overall, significantly
more subjects relapsed in the placebo group (38%) than in
the valproate group (24%; p 5 .017). Relapses on

BIOL PSYCHIATRY
2000;48:573–581

577

lithium (31%) were not significantly different from either
group. Other results fail to meet the traditional (p , .05)
levels for statistical significance; however, several trends
are worth noting. Median times to 50% survival without
any mood episode were 40 weeks, 24 weeks, and 28 weeks
for valproate, lithium, and the placebo, respectively (p 5
.06). These results are striking because they contradict
expectations in some important ways. There was no
evidence of mania following discontinuation of lithium or
divalproex at the conclusion of the prerandomization
phase. In contrast to expert opinion (Sachs et al 2000),
which generally favors lithium over valproate for acute or
prophylactic treatment of depressive symptoms, in this
study the trends favor valproate. The mean lithium levels
during this maintenance study were substantially higher
(1.0 6 0.48 mmol/L) than the level reported effective in
the acute mania study (0.8 mmol/L) but offered no
advantage over the placebo for prevention of mania or
depression.

Naturalistic Data
Maj et al (1998) reported the outcome of lithium treatment
for a large cohort of bipolar patients (n 5 402) over a
5-year period. The results show 39% of patients discontinued treatment or were lost to follow-up. Among those
who remained on lithium treatment over the entire 5 years
(n 5 247), 46% showed at least partial benefit (at least
a 50% reduction in the rate of hospitalization) and 38.1%
had no recurrences. In contrast to this relatively encouraging data, only 14.2% were well enough to be considered
euthymic. Maj and colleagues found better response to
lithium among patients with fewer prior episodes and
fewer prior hospitalizations but did not find a relationship
between maintenance serum lithium level and outcome. In
this study rapid cycling patients and those with psychosis
had particularly poor outcomes.
Peselow et al (1994) reported results of a naturalistic
study for 305 bipolar patients who received lithium prophylaxis for up to 5 years at a clinic associated with
Bellevue Hospital. Perhaps the most interesting result in
the article is reported in Methods. The inclusion criteria
required a period of remission lasting at least 6 months
while the patient received lithium monotherapy. Although
about 1200 bipolar patients were treated with lithium at
the clinic over a 15-year period, only about 25% met this
entry criteria. Observed rates of recurrence (depression
and mania) were relatively low, with 83% remaining
episode free at 1 year and 37% at 5 years. Dropout rates
were high and may reflect the policy of withholding
adjunctive antimanic and antidepressant treatment from
clearly symptomatic patients and not meeting rigorous
operational criteria for recurrence. By 22 months the

578

G.S. Sachs and M.E. Thase

BIOL PSYCHIATRY
2000;48:573–581

sample was 50% depleted, and only 56 subjects (18%)
entered the last year of the follow-up.

Antipsychotics
Despite widespread clinician recognition of potential risks
associated with chronic neuroleptic use, bipolar patients
often receive neuroleptic maintenance treatment (Sachs
1990; Sernyak et al 1994). Given the frequency of this
practice, it is surprising that there are no controlled studies
reporting any advantage for neuroleptic maintenance in
bipolar patients. Recurrence of mania or psychotic symptoms is frequently encountered in the course of tapering
neuroleptics and provides a clinical rationale for this
treatment.
The only prospective studies of neuroleptic maintenance in bipolar patients suggest that this approach is
seldom beneficial. A 2-year double-blind crossover study
compared the neuroleptic flupentixol with a placebo as
adjuncts to lithium maintenance. Patients receiving flupentixol experienced more episodes of depression than did
placebo-treated patients (Sernyak et al 1994). Sachs
(1990) observed no significant difference in outcome for
bipolar patients continuing lithium and a neuroleptic
versus those openly switched to lithium and clonazepam,
but a higher incidence of depressive relapse was observed
in the neuroleptic-treated patients. Among patients suffering breakthrough mania during lithium maintenance, Peselow et al (1994) found a higher frequency of depression
associated with open adjunctive neuroleptic treatment
(37.5%), as compared with open adjunctive treatment with
benzodiazepine (14.3%) or carbamazepine (7.7%).
In a review of lifecharts, Littlejohn et al (1994) found
dramatic reduction in rates of affective episodes and
readmission during periods when bipolar patients (n 5
18) were treated with depot neuroleptic, as compared with
periods of oral neuroleptic use.
The atypical antipsychotic agents, particularly clozapine, appear to offer better results. In case reports and
open series, bipolar patients refractory to neuroleptics,
anticonvulsants, and lithium are reported to derive substantial improvement that persists in the maintenance
phase when atypical agents are added to their treatment
regimen. In an open randomized follow-up study Suppes
et al (1999) found long-term benefit when clozapine was
added to the regimen of refractory bipolar patients. Although in need of replication in rigorous double-blind
studies, these very encouraging results justify the use of
clozapine for treatment-refractory bipolar patients.
Other atypical antipsychotics are not well studied for
maintenance treatment. Tohen et al (1999) reported encouraging results of open treatment with olanzapine for
patients leaving an acute mania trial. Through a 1-year

follow-up, average mania ratings continued to improve
over several months and remained low. None of the
patients developed tardive dyskinesia, one third required
the addition of lithium, and another third received treatment with fluoxetine. Experience with risperidone in
bipolar disorder is limited at this point, but like olazapine,
risperidone has promising open reports (Ghaemi and
Sachs 1997). Guille et al (in press) found impressive
sustained improvement in prospective ratings on the Clinical Global Impressions scale (mean D CGI $ 1.0) after
open additon of clozapine, rispieridone, or olanzapine to
the treatment regimen of refractory bipolar patients. These
findings require replication in controlled trials. Expert
consensus opinion supports using atypical antipsychotic
agents, particularly risperidone and olanzapine, as alternatives to typical neuroleptics and to clozapine in this
population.

Antidepressants
Use of antidepressants alone as maintenance treatment for
bipolar disorder appears to be hazardous (Altshuler et al
1995, 1999; Kukopulos et al 1980; Prien et al 1984; Wehr
and Goodwin 1979). Open reports indicate 31–70% of
bipolar patients treated openly with antidepressants will
experience affective switch (Goodwin and Jamison 1990).
The use of antidepressants as adjuncts to lithium, valproate, and carbamazepine is common but controversial.
There are no prospective randomized comparisons other
than the results noted above from the NIMH Collaborative
Study. This data set demonstrates no harm from maintenance treatment with imipramine in combination with
lithium but used doses of imipramine so modest that
antidepressant activity was not evident either.
Altshuler et al (1999) used blind raters to review
lifecharts of treatment refractory bipolar patients and
found negative outcomes were common (61%) following
the addition of a standard antidepressant medication to the
treatment regimen. In addition to mania (35%), another
26% experienced cycle acceleration. In a separate study,
Altshuler et al (1995) report more frequent recurrence and
shorter time to recurrence in bipolar patients who tapered
antidepressant medications, as compared with those who
continued use of standard antidepressant agents. Peselow
et al (1994) also reported that adding standard antidepressants increased the mean period of remission (D 5 17.3
months, p 5 .01) in lithium-treated bipolar patients who
had suffered a breakthrough depressive episode. During
the 35 months of follow-up, 41.9% did not meet criteria
for recurrence; manic episodes were reported in 14.0% and
depression in 30.2%. The absence of prospective doubleblind data makes it impossible to know which patients
might benefit from long-term treatment with standard

Bipolar Disorder: Maintenance Therapeutics

antidepressant and which will experience a destabilization
due to antidepressant medication.

BIOL PSYCHIATRY
2000;48:573–581

579

it is not known whether raising the dose or initiating
combination therapy is the wisest plan for the next
maintenance phase.

Thyroid
There are no double-blind studies reporting the use of
thyroid hormone in bipolar patients. In an open trial,
high-dose levothyroxine was added to the mood-stabilizing treatment regimen of 11 rapid cycling bipolar patients
(Bauer and Whybrow 1990). Ten of the 11 experienced
clear-cut improvement in the depressive phase, and among
the seven with mania at baseline, five responded. Interestingly, response in these patients was independent of initial
thyroid status. At the time of clinical response, however,
thyroxine and free thyroxine were found to be above the
upper limit of normal in nearly all subjects, and loss of
benefit followed when patients tapered thyroid below
supranormal levels (with some patients having tapered
thyroid under blind conditions). Like thyroid augmentation of antidepressant therapy, hypermetabolic thyroid
treatment is carried out as an adjunct to a standard
mood-stabilizing regimen.

Other Combination Treatments
When bipolar patients experience recurrence during
monotherapy with lithium, carbamazepine, or valproate,
expert consensus supports combining these agents. Many
agents have been used in combination with these primary
mood-stabilizing treatments, but little high quality experimental data are available as yet to support use of
combination maintenance treatment (for reviews, see Freeman and Stoll 1998; Solomon et al 1998).
Solomon (1998) reported pilot data from a randomized
trial suggesting benefit of combined treatment with lithium and valproate over lithium and a placebo. Denicoff
and colleagues (1997) report outcomes for 24 patients who
had failed to respond to 1-year trials of lithium, carbamazepine, and the combination of lithium and carbamazepine. The sample was reduced 25% by dropout due to
side effects (n 5 3) or noncompliance (n 5 3). Of the
18 evaluable subjects, one had a marked response (5%)
and five (28%) were considered moderate responders. A
triple therapy trail consisting of lithium, valproate, and
carbamazepine was carried out in seven subjects, and of
these, three (43%) were considered responders.
These trials suggest that combination treatment offers
some refractory patients incremental improvement. A few
have marked response and about 40% seem to derive little
or no benefit. The data do not address the merits of
combination treatment for less refractory bipolar illness.
For instance, when a patient experiences a breakthrough
episode after 1–2 years of treatment with a therapeutic but
submaximal dose of lithium, valproate, or carbamazepine,

Other Putative Mood Stabilizing Agents
Numerous agents have been described as beneficial based
on case reports, small controlled trials, and/or studies
reporting only acute outcome. As yet the data do not
warrant routine recommendation of unconventional treatments such as calcium channel blocking agents, Omega 3
fatty acids, or donepezil for maintenance treatment. The
Expert Consensus Guidelines (Sachs et al 2000) suggest
limiting use of such agents to cases in which well
established agents fail to produce acceptable benefit.

Summary
In summary, the available empirical data, while admittedly
scant, provide the most reliable basis for clinical practice.
Withdrawal of treatment during the early stage of remission is clearly inappropriate. The ideal dose and duration
of maintenance treatment are not known. Higher dosages
of prophylactic medication have not been proven to be
more effective for patients who relapse while maintained
within the standard therapeutic range. Higher doses are
associated with increased adverse effects and decreased
compliance. It is conceivable but unproven that some
patients will do as well managed with episodic acute and
continuation therapy as they would with sustained prophylactic treatment. Since continued wellness may reflect
treatment benefit or quiescent illness, it is not possible to
know which patients might be managed effectively without prophylactic medication. It is relatively easy, however,
to know which patients are not candidates for episodic
treatment. These include patients with clinically significant residual symptoms, a history of clinical remissions
lasting less than 2 years, or a history of breakthrough
episodes during prophylaxis. There appears to be no doubt
that patients with frequent or severe relapses should be
offered prophylactic treatment.

The authors acknowledge grant support for their work from the Theodore
and Vada Stanley Foundation, the National Institute of Mental Health
(Grant No. DS-1998-0001), and the Mood and Anxiety Disorder Institute
at the Massachusetts General Hospital.
Aspects of this work were presented at the conference “Bipolar
Disorder: From Preclinical to Clinical, Facing the New Millennium,”
January 19 –21, 2000, Scottsdale, Arizona. The conference was sponsored by the Society of Biological Psychiatry through an unrestricted
educational grant provided by Eli Lilly and Company.

580

BIOL PSYCHIATRY
2000;48:573–581

References
Altshuler LL, Gitlin M, Frye M, Calcagno J, Goodman R, Mintz
J (1999, June): Risk of depressive relapse in bipolar patients
when antidepressants are discontinued. Poster presented at the
Third International Conference on Bipolar Disorder, Pittsburgh.
Altshuler LL, Post RM, Leverich GS, Mikalavskas K, Rosoff A,
Ackerman L (1995): Antidepressant-induced mania and cycle
acceleration: A controversy revisited. Am J Psychiatry 152:
1130 –1138.
Angst J (1981): Course of affective disorders. In: Van Praag HM,
Lader MH, Rafaelson OJ, Sachar EJ, editors. Handbook of
Biological Psychiatry. New York: Marcel Dekker, 225–242.
Baastrup PC, Poulsen JC, Schou M, Amdisen A (1970): Prophylactic lithium: Double blind discontinuation in manic-depressive and recurrent-depressive disorders. Lancet 2(7668):326 –
330.
Ballenger J, Post RM (1978): Therapeutic effects of carbamazepine in affective illness: A preliminary report. Commun
Psychopharmacol 2:159 –175.
Bauer MS, Whybrow PC (1990): Rapid cycling bipolar affective
disorder. II. Treatment of refractory rapid cycling with
high-dose levothyroxine: A preliminary study. Arch Gen
Psychiatry 47:435– 440
Bowden CL, Calabrese JR, McElroy SL, Gyulai L, Wassef A,
Petty F, et al (2000): Randomized placebo-controlled trial of
divalproex versus placebo and lithium in maintenance therapy
of bipolar disorder. Arch Gen Psychiatry 57:481– 489.
Cochran SD (1984): Preventing medical noncompliance in the
outpatient treatment of bipolar affective disorder. J Consult
Clin Psychol 52:873– 878.
Coppen A, Standish-Barry H, Bailey J, Houson G, Silcocks P,
Hermon C (1991): Does lithium reduce the mortality of
recurrent mood disorders? J Affect Disord 23:1–7.
Coryell W, Winokur G, Solomon D, Shea T, Leon A, Keller M
(1997): Lithium and recurrence in a long-term follow-up of
bipolar affective disorder. Psychol Med 27:281–289.
Coxhead N, Silverstone T, Cookson J (1992): Carbamazepine
versus lithium in the prophylaxis of bipolar affective disorder.
Acta Psychiatr Scand 85:114 –118.
Cundall RL, Brooks PW, Murray LG (1972): A controlled
evaluation of lithium prophylaxis in affective disorders.
Psychol Med 2:308 –311.
Denicoff KD, Smith-Jackson EE, Bryan AL, Ali SO, Post RM
(1997): Valproate prophylaxis in a prospective clinical trial of
refractory bipolar disorder. Am J Psychiatry 154:1556 –1558.
Esparon J, Kolloori J, Naylor GJ, McHarg AM, Smith AHW,
Hopwood SE (1986): Comparison of the prophylactic action
of flupenthixol with placebo in lithium treated manic-depressive patients. Br J Psychiatry 148:723–725.
Frank E, Prien RF, Jarrett RB, Keller MB, Kupfer DJ, Lavori
PW, et al (1991): Conceptualization and rationale for consensus definitions of terms in major depressive disorder. Remission, recovery, relapse, and recurrence. Arch Gen Psychiatry
48:851– 855.
Freeman M, Stoll AL (1998): Mood stabilizer combination: A
review of safety and efficacy. Am J Pyschiatry 155:12–21.
Gelenberg AJ, Kane JM, Keller MB, Lavori P, Rosenbaum JF,

G.S. Sachs and M.E. Thase

Cole K, Lavelle J (1989): Comparison of standard and low
serum levels of lithium for maintenance treatment of bipolar
disorder. N Engl J Med 321:1489 –1493.
Ghaemi SN, Sachs GS (1997): Long-term risperidone treatment
in bipolar disorder: 6-month follow up. Int Clin Psychopharmacol 12:333–338.
Goldberg JF, Harrow M, Grossman LS (1995): Course and
outcome in bipolar affective disorder. Am J Psychiatry
152:379 –384.
Goodwin FK, Jamison KR (1990): Maintenance medical
treatment. In: Goodwin FK, Jamison KR, editors. Manic
Depressive Illness. New York: Oxford University Press,
665–774.
Goodwin G (1994): Recurrence of mania after lithium withdrawal. Implications for the use of lithium in the treatment of
bipolar affective disorder. Br J Psychiatry 164:149 –152.
Greil W, Kleindienst N (1999): The comparative prophylactic
efficacy of lithium and carbamazepine in patients with bipolar
I disorder. Int Clin Psychopharmacol 14:277–281.
Greil W, Kleindienst N, Erazo N, Muller-Oerlinghausen B
(1998): Differential response to lithium and carbamazepine in
the prophylaxis of bipolar disorder. J Clin Psychopharmacol
18:455– 460.
Greil W, Ludwig-Mayerhofer W, Erazo N (1997): Lithium
versus carbamazepine in the maintenance treatment of
bipolar disorders—a randomised study. J Affect Disord
43:151–161.
Guille C, Sachs GS, Ghaemi SN (in press): A naturalistic
comparison of clozapine, risperidone, and olanzapine in the
treatment of bipolar disorder. J Clin Psychiatry.
Harrow M, Goldberg JF, Grossman LS, Meltzer HY (1990):
Outcome in manic disorders. Arch Gen Psychiatry 47:665–
671.
Hlastala SA, Franke E, Mallinger AG, Thase ME, Ritenour AM,
Kupfer DJ (1997): Bipolar depression: An underestimated
treatment challenge. Depress Anxiety 5:73– 83.
Johnson RE, McFarland BH (1996): Lithium use and discontinuation in a health maintenance organization. Am J Psychiatry
153:993–1000.
Kreapelin E (1921): Manic Depressive Insanity and Paranoia
[Barclay RM, translator]. Edinburgh: E & S Livingstone.
Kukopulos A, Reginaldi D, Laddomada P, Floris G, Serra G,
Tondo L (1980): Course of the manic-depressive cycle and
changes caused by treatment. Pharmakopsychiatr Neuropsychopharmakol 13:156 –167.
Littlejohn R, Leslie F, Cookson J (1994): Depot antipsychotics in
the prophylaxis of bipolar affective disorder. Br J Psychiatry
165:827– 829.
Lusznar RM, Murphy DP, Nunn CMH (1988): Carbamazepine
vs lithium in the treatment and prophylaxis of mania. Br J
Psychiatry 153:198 –204.
Maj M, Pirozzi R, Magliano L, Bartoli L (1998): Long-term
outcome of lithium prophylaxis in bipolar disorder: A 5-year
prospective study of 402 patients at a lithium clinic. Am J
Psychiatry 155:30 –35.
Melia PI (1970): Lithium prophylaxis. Lancet 2(7680):983.
Mindham RH, Howland C, Shepherd M (1973): An evaluation of
continuation therapy with tricyclic antidepressants in depressive illness. Psychol Med 3:5–17.

Bipolar Disorder: Maintenance Therapeutics

O’Connell RA, Mayo JA, Flatow L, Cuthbertson B, Obrien BE
(1991): Outcome of bipolar disorder on long-term treatment
with lithium. Br J Psychiatry 159:123–129.
Okuma T, Inanga K, Otsuki S, Sarai K, Takahashi R, Hazama H,
et al (1981): A preliminary double-blind study on the efficacy
of carbamazepine in prophylaxis of manic-depressive illness.
Psychopharmacol (Berl) 73:95–96.
Peselow ED, Fieve RR, Difiglia C, Sanfilipo MP (1994): Lithium
prophylaxis of bipolar illness: The value of combination
treatment. Br J Psychiatry 164:208 –214.
Placidi GF, Lenzi A, Lazzerini F, Cassano GB, Akiskal HS
(1986): The comparative efficacy and safety of carbamazepine versus lithium: A randomized, double-blind 3-year trial
in 83 patients. J Clin Psychiatry 47:490 – 494.
Prien RF, Himmelhoch JM, Kupfer DJ (1988): Treatment of
mixed mania. J Affect Disord 15:9 –15.
Prien RF, Kupfer DJ (1986): Continuation drug therapy for major
depressive episodes: How long should it be maintained? Am J
Psychiatry 143:18 –23.
Prien RF, Kupfer DJ, Mansky PA, Small JG, Tuason VB,
Voss CB, Johnson WE (1984): Drug therapy in the
prevention of recurrences in unipolar and bipolar affective
disorders. Report of the NIMH Collaborative Study Group
comparing lithium carbonate, imipramine, and a lithium
carbonate-imipramine combination. Arch Gen Psychiatry
41:1096 –1104.
Quitkin F, Rifkin A, Kane J, Ramos-Lorenzi JR, Klien DF
(1978): Prophylactic effect of lithium and imipramine in
unipolar and bipolar II patients: A preliminary report. Am J
Psychiatry 135:570 –572.
Quitkin FM, Kane J, Rifkin A, Ramos-Lorenzi JR, Nayak DV
(1981): Prophylactic lithium carbonate with and without
imipramine for bipolar 1 patients. A double-blind study. Arch
Gen Psychiatry 38:902–907.
Rosenbaum JF, Sachs GS, Lafer B, Kane JM, Keller MB,
Gelenberg AJ (1992, May): High rates of relapse in bipolar
patients abruptly changed from standard to low plasma
lithium levels in a double-blind, controlled trial. Presented at
the 32nd annual meeting of the New Clinical Drug Evaluation
Unit, Key Biscayne, FL.
Roy-Byrne P, Post RM, Uhde TW, Porcu T, Davis D (1985): The
longitudinal course of recurrent affective illness: Life chart
data from research patients at the NIMH. Acta Psychiatr
Scand 71(suppl 317):1–34.
Sachs GS (1990): Use of clonazepam for bipolar affective
disorder. J Clin Psychiatry 51(suppl):31– 4, 50 –53.
Sachs GS, Lafer B, Truman CJ, Noeth M, Thibault AB (1994):
Lithium monotherapy: Miracle, myth and misunderstanding.
Psychiatr Ann 24:299 –306.
Sachs GS, Printz DJ, Kahn DA, Carpenter D (2000): The Expert
Consensus Guideline Series: Medication Treatment of Bipolar Disorder. Postgrad Med April:1–104.

BIOL PSYCHIATRY
2000;48:573–581

581

Sernyak MJ, Griffin RA, Johnson RM, Pearsall HR, Wexler BE,
Woods SW (1994): Neuroleptic exposure following inpatient
treatment of acute mania with lithium and neuroleptic. Am J
Psychiatry 151:133–135.
Shapiro D, Quitkin FM, Fleiss JL (1989): Response to maintenance therapy in bipolar illness: Effect of index episode. Arch
Gen Psychiatry 46:401– 405.
Small JG, Klapper MH, Marhenke JD, Milstien V, Woodham
GC, Kellarns JJ (1995): Lithium combined with carbmazepine or haloperidol in the treatment of mania. Psychopharmacol Bull 31:265–272.
Solomon DA, Ryan CE, Keitner GI, Miller IW, Shea MT, Kazim
A, Keller MB (1997): A pilot study of lithium carbonate plus
divalproex sodium for the continuation and maintenance
treatment of patients with bipolar I disorder. J Clin Psychiatry
58:95–99.
Solomon DA, Ryan CE, Keitner GI, Ryan CE, Miller IW (1998):
Lithium plus valproate as maintenance polypharmacy for
patients with bipolar I disorder: A review. J Clin Psychopharmacol 18:38 – 49.
Stallone F, Shelley E, Mendlewicz J, Fieve RR (1973): The use
of lithium in affective disorders III: A double blind study of
prophylaxis in bipolar illness. Am J Psychiatry 130:1006 –
1010.
Suppes T, Baldessarini RJ, Faedda GL, Tondo L, Tohen M (1993):
Discontinuation of maintenance treatment in bipolar disorder:
Risks and implications. Harv Rev Psychiatry 1:131–144.
Suppes T, Webb A, Paul B, Carmody T, Kraemer H, Rush AJ
(1999): Clinical outcome in a randomized 1-year trial of
clozapine versus treatment as usual for patients with treatment-resistant illness and a history of mania. Am J Psychiatry
156:1164 –1169.
Swann AC, Bowden CL, Calabrese JR, Dilsaver SC, Morris DD
(1999): Differential effect of number of previous episodes of
affective disorder on response to lithium or divalproex in
acute mania. Am J Psychiatry 156:1264 –1266.
Tohen M, Sanger TM, McElroy SL, Tollefson GD, Chengappa
KN, Daniel DG, et al (1999): Olanzapine versus placebo in
the treatment of acute mania. Olanzapine HGEH Study
Group. Am J Psychiatry 156:702–709.
Tohen M, Waternaux CM, Tsuang MT (1990): Outcome in
mania: A 4-year prospective follow-up of 75 patients utilizing
survival analysis. Arch Gen Psychiatry 47:1106 –1111.
Watkins SE, Callender K, Thomas DR, Tidmarsh SF, Shaw DM
(1987): The effect of carbamazepine and lithium on remission
from affective illness. Br J Psychiatry 150:180 –182.
Wehr T, Goodwin FK (1979): Tricyclics modulate frequency of
mood cycles. Chronobiologia 6:377–385.
Zis AP, Grof P, Webster M, Goodwin FK (1980): Prediction of
relapse in recurrent affective disorder. Psychopharmacol Bull
16:47– 49.