Long-Term Treatment of Mood Disorders

Robert F. Prien and James H. Kocsis

This chapter deals with research findings, issues, and strategies relating to the long-term treatment of unipolar and bipolar mood disorders (see Neuroendocrinology of Mood Disorders). There is increasing recognition of the importance of looking beyond the acute episode in treatment planning. This is evidenced by recent clinical practice guidelines developed by the U.S. Agency for Health Care Policy and Research (AHCPR) and the American Psychiatric Association (APA) that provide information on continuation and maintenance treatment of depression and encourage strategic planning to meet the long-term treatment needs of the patient (6, 17). A set of APA guidelines for bipolar disorder also emphasize the need for long-term therapy (5).

For the purpose of this chapter, the term unipolar depression refers to a depressive disorder with no history of a manic or unequivocal hypomanic episode. Bipolar disorder refers to a mood disorder with a history of at least one manic episode, sometimes labeled as bipolar I to distinguish it from bipolar II that requires at least one episode of major depression and a history of hypomania but not mania.

A significant proportion of individuals with major depressive disorder have a course of illness characterized by recurrence and/or chronicity. Longitudinal naturalistic studies and placebo-controlled therapeutic trials indicate that over 50% of individuals who have an initial episode of major depression will have one or more recurrences (17). Bipolar disorder presents an even higher risk of recurrence. Patients with an initial episode of mania have an 80% to 90% probability of a recurrence (29). Recurrences for both unipolar and bipolar disorder tend to occur more frequently with each successive episode (8). These recurring episodes not only cause significant disruption in social, familial, and occupational functioning but may be fatal. Approximately 15% of individuals with a mood disorder commit suicide.

Although the prevention of recurrence is a priority objective for treatment research and practice, it should not divert attention from the disruption that can occur during the intervals between major episodes. The classical portrayal of recurrent mood disorders as a series of discrete episodes with minimal or no symptoms between episodes has been modified by studies showing that many patients with major mood disorder have a course characterized by insidious onset of episodes with incomplete symptom recovery and impaired functioning between episodes (38, 78). An estimated 20% to 35% fail to recover fully from any given episode (6, 38). The incomplete recovery may be from inadequate treatment of the episode, a preexisting chronic disorder, or the progressive deteriorative effects of recurring episodes.

The treatment of mood disorders can be categorized into three interlocking phases: acute, continuation, and maintenance. Fig. 1 illustrates these phases in a schematized episode of unipolar depression. This figure was developed to conceptually frame the sequence of significant change points in the course of a depressive disorder rather than to set standard practice guidelines (43). The relationship of the phases and possible outcomes represented in the figure are discussed below.

The acute phase of treatment aims at achieving remission of the acute symptoms of the episode. Remission may be defined as the return to the individual's premorbid or well self (72). In research, a remission often is declared when the individual no longer meets diagnostic criteria for the syndrome and has no more than minimal symptoms (25). When a treated patient shows significant improvement (with or without a remission), a treatment response is declared. In the research literature, a treatment response often is defined in terms of reduction in the severity of the syndrome. Usually, a clinically significant response requires at least a 40% to 50% reduction in the baseline level of severity. Failure to achieve a treatment response within 6 to 8 weeks with an adequate dose of an antidepressant or within 12 to 16 weeks with a psychotherapy targeted for symptom relief should cause one to consider an alternative treatment.

Continuation treatment will be discussed more fully in the next section. Briefly, the purpose of this phase of treatment is to stabilize the remission and prevent a return of the symptoms of the acute phase, that is, a relapse. The pharmacotherapy that led to remission is usually maintained during the continuation phase. It is recommended that continuation therapy be maintained for at least 4 months if there is no reappearance of acute symptoms (17, 61). Typically, however, there are symptomatic breakthroughs that prolong the continuation phase, sometimes for as long as 9 to 12 months.

The declaration of a recovery designates that the episode is over and presents the option of either discontinuing treatment or continuing it as a maintenance therapy with the aim of preventing a new episode. Following recovery, a subsequent episode traditionally is labeled a "recurrence" rather than a "relapse" to signify that it represents a new episode rather than a continuation or reemergence of the original episode.

Continuation treatment is based upon the assumption that episodes of depression or mania do not necessarily end with the treatment-induced remission of acute symptoms. One theory is that antidepressant and antimanic drugs may suppress symptoms without immediately correcting the postulated biological process or pathophysiology underlying the episode (33, 61). As a result, even with successful treatment there may be a gap of months between the disappearance of overt symptoms and the end of the episode.

There is evidence from over a dozen placebo-controlled studies to support the need for continuation of pharmacotherapy following remission of symptoms for both unipolar and bipolar disorders (58, 61, 72). All of the studies used a double-blind discontinuation design in which an antidepressant or lithium was used to achieve remission, after which approximately half of the sample received a placebo and the other half continued to receive the medication. All provide relapse rates for the 6 months following randomization to continuation treatment. Overall, approximately 50% of the patients treated with placebo relapsed, compared to only 20% receiving medication. Most of the relapses occurred within 3 months.

Discontinuation studies for electroconvulsive therapy (ECT) show a rate of relapse similar to that occurring after withdrawal of antidepressants or lithium (4, 58, 68). A standard practice is to initiate continuation treatment with an antidepressant following the ECT-induced remission of acute symptoms. However, patients do not appear to benefit from continuation treatments that were ineffective in treating prior acute episodes. Sackeim et al. (68) suggested that one alternative is to continue ECT on an intermittent basis following symptomatic response. Recent retrospective reports tend to find continuation ECT helpful. Continuation ECT is endorsed by an APA report (4).

The most critical decision with continuation treatment is how long treatment should be maintained to ensure that the episode is over. The decision is not a problem if the patient is to be maintained on the same treatment for prevention of new episodes. However, if the intention is to discontinue the medication after recovery from the episode, uncertainty as to when the episode is over may result in premature withdrawal of treatment, resulting in relapse or unnecessary prolongation of treatment.

When a medication is successful in suppressing manifest symptoms, it is difficult to determine when the episode is over and continuation treatment is no longer required. A further complication is raised by Hollen et al. (33) who suggested that there may be two types of continuation treatment: (1) suppressive and (2) curative. Suppressive treatment blocks manifest symptoms without alleviating the underlying episode, whereas curative treatment alters the episode by shortening the time period that the episode would have run if left untreated. Although these categories have heuristic significance, they require further research to differentiate one from the other in making treatment decisions.

With major depression, a frequent practice is to continue medication for 4 to 12 months following remission of symptoms and then gradually reduce dosage while carefully monitoring the patient for signs of an emerging episode (6, 17). The recommendation for continuing medication for at least 4 months is based upon findings indicating that risk of relapse is significantly reduced after the patient has remained in remission for at least 16 to 20 weeks (23, 49, 61, 64). A recent review of both controlled and naturalistic data indicates that a continuation treatment period of 4 to 5 months reduces relapse rates to approximately 20% (45). This finding is used sometimes to define the boundary between remission of symptoms and recovery from the episode. For some patients, however, 4 to 5 months is not enough. The presence of residual (subsyndromal) symptoms is a sign that the patient may still require continuation treatment. Reports from three multicenter collaborative studies and several other trials suggest that even mild or moderate symptoms signal that the patient is at high risk for relapse if treatment is withdrawn (40, 48, 61, 73).

In making decisions about withdrawal of continuation therapy, the clinician should be wary about relying solely upon clinical symptoms. A study on maintenance drug therapy coordinated by the National Institute of Mental Health (NIMH) found that the Global Assessment Scale (19) which evaluates both clinical symptoms and impairment of functioning was a more sensitive predictor of relapse than assessment instruments that focus exclusively on clinical symptoms. This finding and data from other studies (78) tend to support the need for instruments that can be used to evaluate both clinical state and functional status in predicting outcome for the continuation and maintenance phases of treatment. There should also be increased attention to the numerous other risk factors for relapse, including chronicity that does not respond to appropriate therapies, patient noncompliance, comorbid medical illnesses, and psychiatric disorders, especially personality disorders, and ongoing or renewed psychosocial stressors (10, 45).

During the past decade, there has been increased effort to identify a biological marker of recovery from a depressive episode (a state-dependent marker). A valid state marker would not only aid in determining when continuation treatment can be safely withdrawn but also might be useful in detecting an emerging episode before the appearance of overt symptoms. Some progress has been made with rapid eye movement (REM) dysregulation, slow-wave sleep (SWS) alterations, and dysregulations in the hypothalamic–pituitary–adrenal axis and hypothalamic–pituitary–thyroid axis (10, 44, 74). Kupfer (44) has proposed an innovative biological model of recurrent depression that attempts to differentiate biological measures associated with persistent features of depressive disorder (type 1) from episodic features (type 2). Type 2 episodic changes appear to be strongly correlated with REM sleep dysregulation and level of acute stress whereas type 1 alterations seem more associated with SWS alterations and growth hormone regulation. Data suggest that the model may allow predictions for acute and long-term treatment and provide a better understanding of the pathophysiology of depressive disorder. However, currently this and related research is not at a stage where it can be applied in clinical practice. This concurs with the conclusion from a European Consensus Conference (8) that no single biological marker could be recommended for determination of the endpoint of a depressive episode or the evaluation of treatment outcome.

There are three basic dosage strategies for continuation treatment: (a) maintaining the dose at the level that effectively treated acute symptoms; (b) reducing the dose to a fixed level, for example, to half of the acute therapeutic level or to the equivalent of 150 mg/day of imipramine; and (c) reducing the dose in graduated steps, with a rapid buildup at the signs of an emerging relapse. This has been a controversial issue. However, recent research with recurrent depression suggests that patients should continue to receive the full therapeutic dose used to attain remission of acute symptoms, providing that there are no significant adverse reactions or dose-related negative effects on patient compliance (6, 17, 24). A similar finding has been reported for treatment of dysthymia (41).

Care should be taken in withdrawing continuation medication following completion of the episode. Rapid discontinuation of heterocyclic antidepressants, particularly tertiary amine tricyclics, can produce withdrawal effects such as sleep disturbance, gastrointestinal symptoms, behavioral activation, agitation, and irritability (66). Many of these withdrawal symptoms are indistinguishable from an emerging relapse and may lead the clinician to misinterpret the need for continued medication. Also, there is evidence suggesting that rapid withdrawal of an antidepressant may produce an increased risk of early mania in susceptible patients (29). Current convention is that the dose of heterocyclic agents be tapered over a period of a month or more. It is not clear if a similar tapering schedule is needed for newer antidepressants. Fluoxetine with its long life does not appear to require a slow taper. In any case, the patient should be monitored during and after withdrawal to ensure that recovery from the episode is complete. Similar care should be applied to the management of bipolar disorder with lithium.

The need for continuation treatment following standardized short-term psychotherapies (cognitive therapy, interpersonal therapy, and behavioral treatment approaches) has not been as well studied as continuation therapy for medication. Published data from naturalistic follow-up studies suggest that the relapse rate following successful acute-phase cognitive therapy is on the order of 40% to 60%, similar to that for medication (17). Two studies that extended cognitive therapy for 8 months following successful acute treatment of major depression report low relapse rates of 19% (35) and 23% (12).

A recent study of cognitive therapy reports that patients who show only partial remission during acute treatment are at higher risk for relapse than patients who achieve a full remission (73). The relapse rate during the 6 months following a 20-week program of cognitive therapy was 52% for patients with partial remission and only 9% for patients with a full remission. The investigators strongly recommend that models of longer term psychotherapy be developed for depressed patients who do not show full remission during time-limited cognitive therapy. The need for longer term models for short-term psychotherapies is further emphasized by a follow-up study of patients from the NIMH Treatment of Depression Collaborative Research Program who had been successfully treated with cognitive behavior therapy, interpersonal therapy, or imipramine over a 16-week period (70). The major finding is that 16 weeks of these specific treatments is insufficient for most patients to sustain remission and achieve full recovery from the episode.

There is only one study of continuation interpersonal therapy. It found that patients who received continuation interpersonal therapy had better social, vocational, and marital adjustment after a 1-year follow-up than patients who had no continuation therapy (77). This study, conducted in the 1970s, has not been replicated.

Continuation treatment for bipolar illness may pose a different set of problems than exist for unipolar disorder. Because of the high risk of recurrence, patients with bipolar disorder are likely to be maintained on medication following recovery from an episode. Thus, there may be no need for the clinician to judge when the episode is over and when continuation medication can be safely withdrawn. However, there may be a need for adjustment of treatment during or after the continuation phase. Lithium, the treatment of choice for pure mania, often is supplemented by neuroleptics for more rapid onset of action or by anticonvulsants or benzodiazepines if the patient fails to respond adequately to lithium alone (63). With the waning of acute symptoms, consideration should be given to discontinuing the neuroleptic before it becomes a routine adjunct to lithium for long-term maintenance and exposes the patient to risk of tardive dyskinesia or other adverse reactions. Other combinations also may warrant dismantling to determine whether the patient needs more than one medication to sustain improvement. Other problems arise with patients who have dysphoric mania, bipolar major depression, or acute rapid cycling. There is only limited information on the appropriate long-term treatment for these subtypes of bipolar disorder (see the section on maintenance treatment).

There is ample evidence supporting the efficacy of antidepressants and mood stabilizers for the maintenance treatment of mood disorders. The major issues are (a) who should receive treatment; (b) what treatments should be used; and (c) how long treatment should continue.

A critical factor in deciding whether or not to initiate maintenance treatment is the likelihood of a recurrence in the near future and the impact it might have on the patient's career, family relations, and overall functioning. Most investigators who have published an opinion on the use of maintenance therapy recommend that patients with major depression should have two or three well-defined episodes before receiving maintenance treatment. The AHCRP Clinical Practice Guideline (17) strongly recommends maintenance treatment for individuals who have three or more episodes. The Guideline also recommends maintenance medication for individuals with two episodes who have one or more of the following characteristics: (1) a family history of bipolar disorder or recurrent depression; (2) early onset of the first episode (before age 20); and (3) both episodes occurred during the past 3 years and were sudden and severe or life-threatening. The World Health Organization (WHO) consensus statement on the pharmacotherapy of depressive disorders also recommends that maintenance therapy be initiated after three episodes, particularly if there have been two episodes in the last 5 years (79). There is general agreement that patients who have only a single episode of major depression, mild episodes, or a lengthy interval between episodes (e.g., 5 years) probably should not be started on maintenance treatment. The exception is the patient for whom a second episode would have severely disruptive or life-threatening consequences.

Naturalistic studies tend to support the aforementioned recommendations, confirming that patients with a history of three or more episodes are at high risk for an early recurrence or unremitting chronic course (40). A study by Maj et al. (47) assessing the pattern of recurrence after recovery from an episode of major depression found that a history of three or more episodes was the most powerful predictor of an early recurrence. In a study of over 400 patients followed for 20 years, Angst (7) reports that patients who have two major episodes within 5 years have a 70% probability of developing two or more episodes during the following 5 years.

In general, bipolar disorder generates more urgency for early intervention than unipolar disorder. The WHO consensus statement recommends that patients with bipolar illness receive maintenance treatment after two episodes (79). Goodwin and Jamison (29) suggest more stringent guidelines. They recommend that maintenance treatment be considered after the initial manic episode if the episode has a sudden onset, is highly disruptive or life-threatening, or occurs in an adolescent, especially if there is substantial genetic loading.

There are factors other than the frequency and severity of prior episodes that should be considered in decisions about maintenance treatment. These include the rapidity of onset of prior episodes, contraindications to treatment, extenuating circumstances that may have contributed to prior episodes, and, perhaps most important, the patient's willingness to commit himself or herself to the treatment program.

There have been numerous placebo-controlled trials evaluating long-term pharmacotherapy in unipolar disorder. Table 1 lists studies that used a placebo control and had a duration of at least 1 year. The 15 listed studies evaluated 11 medications, 10 of which were found to be superior to placebo in at least one study. Ideally, a design for determining the efficacy of an antidepressant in preventing recurrence should have four components: (1) a sample at high risk for recurrence; (2) a placebo control; (3) a minimum 4-month period of remission or symptom-free interval preceding randomization to maintenance treatment; and (4) at least a 2-year maintenance phase. Three of the studies listed in Table 1 satisfied all four components (23, 37, 65). Eight met all but one requirement (11, 16, 27, 28, 49, 53, 60). The remaining four failed to satisfy two or more requirements (15, 18, 20, 67). Six studies did not include a defined symptom-free interval (16, 18, 20, 28, 60, 67). All six were multisite trials in which patients who showed a satisfactory response during acute treatment were randomly assigned to maintenance treatment without an intervening stabilization of remission or a symptom-free period. Two of these studies made no distinction between a relapse and a recurrence (28, 67).

Despite the variation in design of the maintenance trials, the significant difference between drug and placebo in the large majority of studies provides evidence that antidepressants and mood stabilizers such as lithium can reduce the occurrence of new episodes in unipolar disorder. In most studies, the recurrence rate for placebo was more than twice that for active medication. The findings also show that none of the medications is a panacea. Even the most favorable results indicate failure in at least one-fourth of patients. Overall, there is no solid evidence that any one treatment is superior to another in preventing recurrences. Success with a given medication may vary from one study to another, often due to differences in study design and application of treatment. The classic example is imipramine which was evaluated in four of the studies. The recurrence rates were 22% (23), 46% (62), 48% (60), and 83% (37). Corresponding rates of recurrence for placebo were 78%, 71%, 92%, and 100%. The best results with imipramine were obtained by Frank et al. (23) who, paradoxically, used the patient sample at highest risk for recurrence among the four studies. Patients had to have at least three prior episodes (including the current episode) to enter the study. The low incidence of recurrence (22%) in this study may have been due, in part, to the use of the full therapeutic dose of imipramine throughout the maintenance phase. The mean dose at the start of the maintenance treatment was 200 mg/day. By contrast, the mean dose for the other three studies ranged from 135 to 138 mg/day, with doses seldom exceeding 150 mg/day.

In contemporary practice, the medication that induces remission of the acute episode generally is used for maintenance therapy. Unless the original medication presents special risks when used on a long-term basis, the advantages of continuing the medication for maintenance therapy clearly outweigh the disadvantages. Replacing a medication that is producing a good response can create logistical, compliance, and ethical problems. Also, there may be a loss of control over symptoms during and after replacement, particularly if the episode has not run its course.

It is important that the clinician carefully consider the side-effect profile of medication to be used for maintenance treatment and the effect it may have on patient compliance. Reactions that may be viewed by the clinician as being relatively mild may be troublesome enough for the patient to stop treatment. Examples include weight gain, dry mouth, and sexual dysfunction with antidepressants, and weight gain, hand tremor, and polyuria with lithium. An average of 4% of the patients receiving long-term treatment with an antidepressant discontinue medication because of unwanted effects (66). These unwanted effects are not significantly different from those occurring during treatment of the acute episode. The newer antidepressants, particularly the serotonin-selective reuptake inhibitors, appear to have an advantage over the standard tricyclic antidepressants (TCAs) in terms of attrition from side effects.

Side effects are not the only factor contributing to noncompliance. Other factors are dislike of having one's mood controlled by a drug, feeling well and seeing no need for medication, and the inconvenience in taking daily medication. The clinician and patient should discuss these and other problems that might interfere with the treatment program and come to an agreement as to how they may best be handled.

Lithium has long been the cornerstone for the maintenance treatment of bipolar disorder but has generated conflicting findings as a long-term treatment for unipolar disorder (58). A NIMH consensus development conference on maintenance therapy for mood disorders (51) concluded that both lithium and TCAs are effective maintenance treatments for unipolar recurrent depression, each with advantages for certain patients. Lithium has potent antimanic properties, whereas TCAs and other antidepressants are ineffective in preventing manic episodes and may even precipitate hypomania, mania, or rapid cycling in vulnerable patients (76). Because of the devastating effects of a manic episode, lithium may be the preferred treatment when there is suspicion of a previously unrecognized bipolar disorder or higher than usual risk of a subsequent manic episode. A major risk factor is a prior hypomanic episode that may have been missed in evaluating the patient. Other factors increasing the risk of a manic episode are a family history of mania, a high frequency of depressive recurrences, and early age of onset of the mood disorder (55). Overall, 10% to 15% of patients originally diagnosed as unipolar eventually develop a manic episode (51).

The advantages of TCAs and other antidepressants is that they are used far more frequently than lithium in treating acute unipolar depression and, thus, are a more logical choice for maintenance treatment. Also, there is evidence that lithium may not be as effective as an antidepressant in preventing severe depression (28, 62). Adding lithium to an antidepressant regimen for the long-term treatment of unipolar disorder has demonstrated no significant advantages over an antidepressant alone (62).

Psychotherapies designed for the treatment of depressed patients (cognitive and interpersonal) and certain behavioral, marital, brief dynamic, and group therapies have demonstrated efficacy in alleviating acute depressive symptoms. However, there has been only one randomized controlled trial evaluating the efficacy of psychotherapy as a maintenance treatment. The 3-year study by Frank et al. (23) found that monthly maintenance interpersonal therapy (IPT) delayed, but did not ultimately prevent, a new episode. Interpersonal therapy was found to be significantly less effective than maintenance imipramine. Furthermore, the combination of IPT and imipramine was no more effective than imipramine alone. The investigators emphasize that the delay in onset of a new episode with IPT may be of value to patients who cannot tolerate medication or wish to interrupt medication for selected clinical conditions such as pregnancy. They also question whether more frequent IPT sessions (e.g., twice monthly) would produce better results. Another 3-year study using a similar design to evaluate the combination of IPT and nortriptyline with elderly depressed patients is underway and should provide further information on the use of maintenance psychotherapy alone and in combination with medication (64).

From a clinical standpoint, maintenance psychotherapy is not recommended as the sole treatment for preventing recurrences unless the patient needs to avoid pharmacotherapy. The value of combining psychotherapy and pharmacotherapy for long-term maintenance has not been established in controlled trials. Nonetheless, many practitioners believe that psychotherapy may aid in preventing recurrence by improving social and occupational functioning and the capability to cope with life events that may precipitate a new episode. Psychosocial programs may also improve compliance with medication.

Lithium is the medication of choice for the maintenance treatment of bipolar disorder and is the only therapy to be compared to placebo for this indication. Several studies conducted in the 1970s found that lithium was significantly more effective than placebo in preventing both manic and depressive episodes (9, 16, 22, 60). The pooled mean failure rate for lithium was 33% compared to 81% for placebo. More recent studies of lithium therapy report a less positive outcome. A multicenter study sponsored by the NIMH found that only one-third of patients receiving maintenance lithium treatment remained episode-free over a 18- to 24-month period (62). Similar results were reported for a multisite study conducted by the Medical Research Council in England (28). One explanation for the poorer outcome with lithium compared to earlier years is the change in patient selection for clinical studies. Many of the earlier studies were initiated when lithium was still an investigational drug and study centers were among the few places where lithium was available. These early trials included a high proportion of patients with classic manic-depressive disorder characterized by clear-cut onset and recovery and good interepisode functioning. In the current era, many of the classic cases are treated satisfactorily in the community and are not referred to the university medical centers or receiving hospitals that conduct most of the maintenance treatment studies. Thus, patient samples available for maintenance trials tend to include a larger proportion of difficult-to-treat patients who have failed to respond to traditional therapies in the community and may be unresponsive to lithium.

Recent follow-up studies and reviews of the literature on the long-term use of lithium across the entire spectrum of bipolar disorder suggest that lithium may be an inadequate treatment for as many as 50% or more of bipolar patients (26, 52, 75). However, these newer estimates of lithium's efficacy should be interpreted with caution. Partial or unsatisfactory response to lithium can have many causes that need to be addressed before labeling the medication a failure. These include misdiagnosis, poor compliance unrelated to adverse reactions, inadequate dose, hypothyroidism, and failure to allow sufficient time for lithium to exert full prophylactic effect.

Numerous studies and reviews have identified clinical features associated with lithium response (29, 46, 59, 63). Maintenance lithium appears to be most effective with patients who have an uncomplicated manic episode, good functioning between episodes, and a family history of bipolar disorder. Negative predictors of response to lithium include a history of dysphoric mania or rapid/continuous cycling, severe or chronic depression, moodincongruent psychotic features, alcohol or drug abuse not associated with mood change, and co-occurring personality disorder. A pattern of course of illness characterized by a cycle starting with depression followed by mania and euthymia tends to respond poorly to lithium whereas a sequence of mania-depression-well interval tends to respond positively (21, 42). Reports also suggest that patients who have three or four prior episodes are less likely to respond to lithium therapy than patients with fewer episodes (26, 56).

The anticonvulsants carbamazepine and valproate have emerged as the leading alternatives to lithium. The results from nearly a dozen maintenance studies of carbamazepine have reported positive results, suggesting that the compound may be beneficial with patients who fail to respond adequately to lithium (54, 63). However, studies comparing carbamazepine to either lithium or placebo have significant methodological flaws that confound interpretation of data (59). The strongest evidence for the maintenance efficacy of carbamazepine comes from longitudinal studies in which carbamazepine is periodically replaced by a placebo and from mirror image studies in which the course of illness during treatment with carbamazepine is compared to the course of illness during an equivalent period preceding carbamazepine (59). Most of these studies suffer from small sample size. Valproate provides the same difficulties, although a recent 3-week study comparing valproate to lithium and placebo in patients with classic mania indicate that valproate is as effective as lithium and superior to placebo (13). Several open studies and naturalistic trials have evaluated valproate as a maintenance treatment (14, 63) but, to date, there is no published study that systematically compares valproate with lithium, carbamazepine, or other treatments for bipolar disorder.

In the absence of clinical predictors for identifying patients who will respond to one anticonvulsant and not the other, the choice depends upon the clinician's preference and the side-effect profile. Usually, the anticonvulsant is used as an adjunct to lithium, thus avoiding the problem of sequential discontinuation of lithium and initiation of the anticonvulsant. An unresolved issue is whether an anticonvulsant alone or in combination with lithium is an appropriate first line maintenance treatment for certain subgroups of patients (e.g., rapid cyclers) or should be limited to patients who are resistant to lithium.

The most frequently used medications other than lithium for long-term treatment of bipolar disorder are the neuroleptics. They often are added to lithium for rapid control of manic episodes that occur during lithium maintenance treatment. Although neuroleptics are not routinely recommended for long-term use in bipolar disorder, they may be helpful for patients who suffer repeated breakthrough manias while receiving lithium. In such cases, the risk of tardive dyskinesia and other adverse effects of neuroleptic treatment must be balanced against the highly disruptive and life-threatening consequences of repeated manic attacks. Evidence suggesting that patients with major mood disorders are at higher risk for tardive dyskinesia than patients with schizophrenia makes this decision all the more difficult (36). Several case studies suggest that clozapine may be useful in the treatment of refractory bipolar disorder, including rapid cycling. The availability of newer atypical neuroleptics such as risperidone may provide protection against manic recurrence without the risks associated with clozapine and the traditional compounds, but this is yet to be determined.

Breakthrough depression occurring during maintenance treatment with lithium poses special problems. Initial steps include a reevaluation of the patient's lithium plasma level, thyroid function, and physical condition. The possibility of noncompliance in taking medication should be carefully explored. There should also be a reevaluation of life situations or psychosocial stressors that may be contributing to the depression state. Although some breakthrough depressions can be treated successfully by enhancing thyroid function or increasing the dose of lithium to levels of 1.2 mEq/L or higher (63), the standard practice is to add an antidepressant to the lithium regimen. This practice has been criticized by some investigators who claim that antidepressants may precipitate rapid or continuous cycling or mania in susceptible patients (42, 76). This risk, however, must be balanced against the suffering, occupational and social impairment, and risk of suicide of a persisting severe depression. The efficacy of newer antidepressants for breakthrough depression has not been established; bupropion appears to be the most promising (31). Although anticonvulsants may be useful in preventing the development of rapid cycling, their efficacy in combating primary bipolar depression has not been demonstrated. Electroconvulsive therapy may be required for severe depression but should be used with caution if the patient remains on lithium. Administered with lithium, ECT has been reported to cause memory loss and neurological abnormalities (71).

Breakthrough rapid cycling or dysphoric (mixed) mania during lithium treatment is also a problem. As with breakthrough depression, there should be medical evaluation with particular attention to thyroid status. Carbamazepine or valproate may be useful in attenuating rapid cycling and mixed states (5). Other strategies reported to aid in stabilization of mixed or rapid cycling states are withdrawal of antidepressants, neuroleptics, or other potential cycle-inducing medications, the use of adjunct clorgyline, thyroid treatment, and repetitive sleep deprivation (5, 29, 63). None of these strategies have been well studied and should be carefully monitored. In particular, the withdrawal of an antidepressant may raise ethical concerns about allowing a patient to endure severe depression. Substance abuse may pose an additional complication in patients with rapid cycling or mixed states and needs to be dealt with aggressively.

It is clear that pharmacotherapy alone does not meet the needs of many bipolar patients. Even with adequate medication treatment, many patients fail to show full recovery from acute episodes and display symptomatic, interpersonal, and social deficits during interepisode periods. There are several reports citing the value of combined pharmacological–psychosocial treatment following remission or recovery from a bipolar episode (5, 34, 63). Most involve family or marital-based therapeutic intervention and psychoeducational programs. Research has not yet provided a standardized therapy tailored to maintenance treatment that has demonstrated definitive efficacy in a controlled trial. The slow progress in developing and evaluating adjunct psychosocial maintenance therapies may be due, in part, to the earlier misconception that most bipolar patients have a "clear" recovery between episodes and do not benefit from nonbiological treatments. As a result, the development of long-term psychotherapeutic treatments for bipolar disorder lags far behind similar efforts for schizophrenia as well as unipolar depression. It is noteworthy that despite the popularity of cognitive therapy for the treatment of depression, there have been few efforts to apply this treatment following an episode of mania. This may be a fertile ground for cognitive-behavioral therapeutic approaches.

One innovative maintenance treatment under study is Interpersonal and Social Rhythm Therapy developed by Frank et al. (personal communication). Its basic goal is to attenuate or prevent recurrences by stabilizing the social rhythms of bipolar patients. The primary focus of this intervention is on sleep rhythms, exercise rhythms, patterns of social interaction, and patterns of interpersonal, intellectual and emotional stimulation. It is being evaluated as an adjunct to lithium in a 2-year controlled maintenance study.

An important decision facing the clinician is what dose to use for maintenance treatment. With the exception of lithium, there are few dose–response studies for long-term therapy. A standard practice with antidepressants is to maintain the patient on a dose equivalent to 75 to 150 mg/day of imipramine. This level often represents a 40% to 60% reduction in the dose that was used to treat the acute episode. The study on maintenance treatment of recurrent depression by Frank et al. (23) suggests that the common practice of reducing dosage following recovery from the episode warrants reconsideration. Full-dose treatment tended to be more successful that half-dose treatment in a 3-year subset of the trial (24). Other maintenance studies suggest a difference in outcome relating to dose. Rouillon et al. (67) found that a 75-mg dose of maprotiline was superior to a 37.5 mg dose. However, 75 mg is only a half-standard therapeutic dose. It is possible that even better results may have been obtained with the standard dose of 150 mg. A study of long-term use of phenelzine (65) favored the use of a 60-mg dose over a 45-mg dose, although there was not a statistically significant difference. Montgomery et al. (49) achieved highly successful results with a 40-mg maintenance dose of fluoxetine, twice the dose level eventually approved for marketing. The aforementioned studies emphasize the need for dose–response trials that extend beyond the treatment of acute symptoms to the prevention of relapse and recurrence.

The maintenance dose range for lithium has been studied more carefully than that for the antidepressants. The dose generally is targeted at 0.6 to 0.8 mEq/L with a level of 0.5 not uncommon for elderly patients (51, 63). Some patients may benefit from higher levels. A controlled trial by Gelenberg et al. (26) found that the risk of recurrence was 2.6 times greater for patients assigned a low dose (0.4–0.6 mEq/L) than a standard dose (0.6–0.8 mEq/L). However, there were two other critical findings. First, patients with three or more prior episodes did poorly at both levels, suggesting that even higher doses may be required for this subgroup. The finding may also mean that patients with a history of frequent recurrences are not responsive to lithium at any dose level. Second, the standard dose produced more side effects than the low dose, a factor that can adversely affect compliance. There are no studies evaluating the maintenance dose for other treatments used for bipolar disorder. Because of the difficulty in conducting these studies, it is questionable whether such trials will be undertaken. The most likely candidates are valproate and carbamazepine, both of which are being evaluated by pharmaceutical companies in the United States as a treatment for acute mania.

The decision of how long to continue maintenance treatment is not easy to determine from existing data. There is evidence that some patients with a long duration of illness who have not suffered a recurrence for several years may not require maintenance treatment. Angst et al. (7) suggest that approximately one of three patients with bipolar disorder and one of eight with unipolar disorder who are over age 65 with a long history of illness appear to stop having recurrences. However, there are no valid predictors for identifying patients who no longer require maintenance treatment. Overall, existing data indicates that recurrent major depression and bipolar disorder can be a lifelong illness requiring treatment. The only way to determine whether the patient still needs medication is to gradually withdraw medication and carefully monitor the patient for signs of a recurrence. Schou (69) suggests that after the patient has remained well for 3 or 4 years, the clinician and patient should discuss whether the medication should be continued, taking into account the course of illness before and during treatment, the risk and consequences of a subsequent episode, and the patient's tolerance to the medication. Special precautions should be taken with patients having a history of episodes characterized by suicide attempts, need for hospitalization, or severe disruption of familial or career functioning, particularly if the episodes had a rapid onset.

Even with the most careful review of course of illness, discontinuation of medication can provide unanticipated difficulties. For example, Post (55) reports that some patients who have maintained long-term stability on lithium and discontinue treatment have a recurrence and then fail to respond to reinstated lithium, even when administered at higher doses. This phenomenon is labeled discontinuation-induced refractoriness. Post conceptualizes that the occurrence of new episodes despite renewal of previously effective lithium treatment is a manifestation of the kindling model. Adding one or more episodes to the cumulative load may impact the neurobiological mechanisms mediating mood dysregulation such that the disorder is no longer responsive to lithium. Although others have also observed this disturbing development, it is by no means clear that it is common. Its application to decisions regarding clinical treatment awaits further research.

The diagnostic category of dysthymic disorder was developed by the American Psychiatry Association DSM-III (3). By definition, the disorder is long standing (at least 2 years duration in adults) and is characterized by a persistent or intermittent depressive syndrome of insufficient severity to satisfy the criteria for major depression. It soon became apparent that many patients with dysthymic disorder subsequently develop a major depressive episode, a condition commonly referred to as "double depression" (39). A naturalistic follow-up study by Keller et al. (39) found that a high proportion of patients with double depression had a remission of the superimposed episode of major depression during a 2-year period. However, overall, the outcome was relatively poor. A high percentage failed to remit from the underlying dysthymia and there was a high recurrence rate for episodes of major depression. It is noted, however, that treatment, particularly pharmacotherapy, was less than optimal for most of the patients. Subsequent placebo-controlled studies suggest that antidepressants are efficacious in treating both double depression and pure dysthymia (dysthymia without a super-imposed major depression) (32, 41, 50). All have been short-term studies, mostly of 6-week duration.

Overall, there has been relatively little research on the long-term treatment of dysthymic disorder. In a continuation treatment study, patients with double depression who responded to short-term phenelzine treatment were randomly assigned to continue phenelzine or receive a placebo for 6 months (30). Four of five patients assigned to placebo relapsed compared to none of seven who continued to receive active medication. An ongoing large-scale study by Kocsis and coworkers is evaluating the use of desipramine as a maintenance treatment for dysthymic disorder. The sample consists of patients who require acute treatment for either double depression or pure dysthymic disorder. Patients who respond to desipramine and remain in remission for 16 weeks are randomly assigned to continue desipramine or receive a placebo for 2 years. Preliminary results show that only 3 of 23 desipramine-treated patients (13%) have had a recurrence compared to 11 of 21 patients receiving placebo (52%). There is no significant difference in rate of recurrence between the double depression and pure dysthymic subgroups, suggesting the need for long-term treatment for both subgroups of chronic depression.

Despite the increased attention to the long-term treatment of mood disorders during the past few years, the development and evaluation of maintenance treatments has progressed at only a moderate pace. One problem is that the pharmaceutical industry, which fuels much of the therapeutic drug development for mood disorders, is required for regulatory purposes only to demonstrate the efficacy of an investigational drug during the acute episode. Efficacy trials for antidepressants and mood stabilizers average 6 to 8 weeks in duration, despite the fact that most episodes of major depression and bipolar disorder extend for much longer periods. Also, partial remission of symptoms often is accepted as confirmatory evidence of the therapeutic efficacy of a new compound if the compound shows a statistically significant superiority over a placebo. Unfortunately, it cannot be assumed that partial remission will convert into a full remission with continued treatment. Nor can it be assumed that the efficacy of a drug in treating an acute episode necessarily translates into long-term efficacy in preventing relapse or recurrence.

Efficacy studies of psychotherapy also tend to focus on the acute episode. Many of the studies use follow-up evaluations after discontinuation of treatment to determine whether the therapy has an enduring effect on the course of the disorder. These follow-up trials are difficult to interpret. In most cases, the researcher has no control over the patient's treatment and evaluations are conducted at infrequent intervals under nonblind conditions.

In general, longitudinal studies of therapeutic strategies are limited by the considerable expense, sample size requirements, and staffing resources needed to conduct well-controlled long-term trials. Sample attrition is a particularly pressing problem. Most studies of maintenance treatment enter patients during an acute episode and randomize them to long-term treatment groups following recovery from the episode. The attrition from the time of admission into the acute phase to entrance into the maintenance phase averages about 50% in reported studies. This heavy loss of patients often creates samples that are not representative of the population defined by study intake criteria and confounds generalization of results.

The chapter on long-term treatment of mood disorders in Psychopharmacology: the third generation of progress (57) identified several research needs, many of which are still relevant. The needs identified in the earlier chapter included (a) evaluation of the maintenance efficacy of drugs other than lithium for bipolar disorder and the TCAs and lithium for unipolar disorder; (b) long-term studies to identify treatment-responsive subgroups and areas of functioning that benefit from psychotherapy alone and in combination with medication; (c) determination of the efficacy of long-term treatments in attenuating the psychopathology and impairment in functioning that often is present during interepisode periods; (d) treatment strategies for alleviating or managing dysthymic disorder; (e) identification of biological markers for determining when continuation treatment can be safely withdrawn and for detecting an emerging recurrence before the appearance of clinical symptoms; (f) the effective dose range of lithium and antidepressants for maintenance treatment; (g) the use of maintenance ECT for patients who are refractory to medication and responsive to ECT; (h) the appropriate treatment for rapid-cycling and breakthrough depression during lithium maintenance therapy; and (i) development and evaluation of programs to reduce the high rate of attrition from noncompliance.

There have been significant advances with some of the aforementioned needs. There has been progress in evaluating alternatives to lithium and the TCAs, particularly with the anticonvulsants and serotonin reuptake inhibitors. There are also increased efforts to extend the clinical evaluation of investigational and newly marketed drugs beyond the acute episode. Traditional antidepressants are also being reevaluated in long-term trials. Other areas that have received attention are dosing strategies for long-term treatment, biological markers, treatment of rapid cycling, and treatment for dysthymia. Research on the extended use of combined pharmacotherapeutic–psychosocial treatments also has progressed, but is still more than a decade behind multimodality research for schizophrenia. A promising sign is the development of extended models of short-term psychotherapies for use as continuation or maintenance treatments.

A continuing under-served area of research is the study of the effect of maintenance treatments on interepisode functioning and subsyndromal psychopathology. Too often, the interepisode period is viewed as a well interval not requiring extensive study. Clinical instruments used in controlled clinical trials continue to be selected primarily for their capacity to identify and describe major episodes. As a result, the treatment of interepisode functioning and subthreshold symptoms often is neglected in the study design or in the report of results, omitting what could be useful information for judging the efficacy of individual treatments. This omission assumes particular importance in light of findings that subthreshold depressive symptoms and poor functional status can have a significant impact on a subsequent course of illness. The degree to which maintenance treatments effect the overall course of unipolar and bipolar disorder in domains extending beyond relapse and recurrence will be a challenge for the next generation of research.

published 2000