Alan C. Swann MD
Pat R. Rutherford, Jr. Professor and Vice Chair for Research, Department of Psychiatry, University of Texas Health Science Center, Houston, Texas
Valproic acid preparations have been used in manic episodes, depressive episodes, and maintenance treatment of bipolar disorder in children, adolescents, and the elderly. Other psychiatric uses of valproate have included treatment of personality disorders, anxiety disorders, psychotic disorders, posttraumatic stress disorder, and impulsive aggression. The quantity and rigor of evidence vary widely. We discuss reported uses of valproic acid and their supporting evidence.
Efficacy of Monotherapy
Based on early case reports of its effectiveness in mania, Emrich et al. gave valproate, using a placebo-controlled design, to five patients experiencing manic episodes. Four of the five improved substantially (1). Subsequently, the same investigators observed positive responses to valproate in seven lithium-resistant patients (2).
These and similar reports led other investigators to use double-blind, parallel-group study designs to assess the effectiveness of valproate. Freeman et al. compared valproate with lithium in 27 hospitalized patients, finding that overall response was similar but that valproate was more effective in patients who also had depressive symptoms (3). Valproate was reported more effective than carbamazepine in a study of 30 patients.
Pope et al. assessed the efficacy of valproate in patients who were either resistant to lithium or unable to tolerate it. Valproate was substantially more effective than placebo; the reductions in Young Mania Rating Scale scores were 54% for valproate and 5% for placebo (4). This strikingly low placebo response rate probably was a result of selecting previously treatment-resistant patients.
Bowden et al. subsequently carried out a larger study comparing valproate with lithium or placebo. Effectiveness of valproate was similar to that of lithium, and both were more effective than placebo (5). Interestingly, this was the first parallel-group, placebo-controlled study of lithium in manic episodes; earlier lithium studies had used crossover designs. The results of this study demonstrated also that optimal effectiveness of valproate required a blood level of at least 45 µg/mL, with a point of diminishing return at approximately 100 to 120 µg/mL (6).
It may be a fortunate patient whose manic episode responds optimally to a single agent. In the aforementioned studies, as usually is the case with structured, controlled efficacy studies, “responders” whose mania scores had improved substantially were still symptomatic, on average, at the end of the study period.
Subtypes of Mania
Manic episodes associated with mixed states or with rapid cycling appear relatively resistant to lithium and to have poorer overall outcomes than episodes without these features. Neither mixed states nor rapid cycling appear to have adverse effects on response to valproate. Two controlled studies found valproate to be at least as effective in manic episodes with depressive features as it was in pure mania (3,7). A prospective, open-label study found divalproex to be effective in treating manic and mixed states associated with rapid cycling (8). There also are case reports of effectiveness in 48-hour cycling (9,10). One study showed that, in addition to preventing affective cycling, valproate prevented the associated changes in norepinephrine excretion, plasma cortisol, and growth hormone in a patient with 48-hour mood cycles (9).
A review of treatment combinations in bipolar disorder showed that the most frequently reported, effective, and practical combination appeared to be lithium and valproate (11). Ketter et al. found an apparent synergy between valproate and carbamazepine in a blinded study of a patient
responding to neither agent alone (12). This was later confirmed in 12 patients with bipolar disorder, although the combination was not as effective in patients with schizoaffective disorder (13). Similarly, lithium and valproate in combination were reported to be more effective than either agent alone, although there also were more side effects (14, 15, 16).
The role of antipsychotic treatments relative to other treatments in manic episodes requires critical examination. Muller-Oerlinghausen et al. carried out an important study in which patients undergoing manic episodes were treated with an antipsychotic drug (17). The patients were randomized to receive, in addition, valproate or placebo. The treating physician periodically was given the option of reducing the dose of antipsychotic medicine if the patient's clinical status allowed. Patients randomized to receive valproate required lower doses of antipsychotic and yet had superior symptomatic improvement compared with those given antipsychotic medicine alone (17).
Economics of Treatment
Practical benefits of lithium, valproate, and carbamazepine have been compared using meta-analysis and pharmacoeconomic studies. Interpretation of these studies is strongly affected by the quality of data and accuracy of underlying assumptions. Meta-analysis showed similar antimanic effects for valproate, lithium, and carbamazepine, but the authors found valproate or carbamazepine to be better tolerated than lithium (18). Pharmacoeconomic studies found valproate had the benefit of shorter hospital stays because of the feasibility of rapid dose escalation and better response in mixed states. Length of hospital stay was similar in divalproex and in the combination of lithium plus carbamazepine, whereas either lithium or carbamazepine alone was associated with longer hospitalizations (19). Cost of treatment was approximately 9% per year less overall with divalproex than with lithium (19).
Biology of Antimanic Response
Acute administration of valproate to rats was shown to decrease hyperactivity associated with combined administration of chlordiazepoxide and amphetamine, proposed as a model for mania; this effect was reduced by γ-aminobutyric acid (GABA) receptor antagonists (20). Valproate also reduces motor responses to acute methylphenidate in doses that do not have independent effects on motor behavior (21). These effects are consistent with reported enhancement of GABA function by valproate (22, 23, 24). In clinical studies, low plasma GABA was shown to predict positive antimanic outcome in women (25), and pretreatment plasma GABA was shown to correlate negatively with symptom change (26). Based partly on effects of valproate, it has been proposed that bipolar disorder was associated with deficient GABAergic function (27).
Although evidence from animal models suggests that valproate may be effective in treating depressive episodes, the published experience with valproate in depression is relatively small, and there are no published conventional placebo-controlled studies. A review of early experience with valproate revealed a response rate of approximately 30% in depressive episodes, but this consisted largely of patients whose depressions had been refractory to other treatments or combinations, so cannot be compared realistically with other response rates (28). More recent reports found valproate to be effective in depression with atypical features (largely hypersomnia) (29) and in treating residual depression in partial responders to other treatments (30). A prospective, open-label study in major depressive disorder, based on the idea that GABA-mimetic agents would have general antidepressant activity, reported most patients responding to valproate (based on >50% improvement in Hamilton Depression scores) at 4 weeks and two-thirds responding at 8 weeks (31).
There is a surprising lack of data in acute bipolar depression, but two randomized studies suggest effectiveness under maintenance conditions. Young et al. found that, for breakthrough depression in patients receiving either lithium or divalproex, addition of the other mood-stabilizing agent was as effective as adding paroxetine (32). Although patients appeared to tolerate the combination of mood stabilizer plus paroxetine better than combined mood stabilizers, combined mood stabilizers may be the better long-term choice because of the potential for better protection against relapse. During a 12-month controlled maintenance study, Bowden et al. found that subjects randomized to divalproex had a lower rate of relapse to depression than did those given placebo, and had a significantly lower rate of interepisode depressive symptoms than did subjects given lithium (33).
Biology of Antidepressant Response
Valproate appears effective in reducing immobilization in the forced swim test, an animal model of depression (34,35). This appears to be a combination of direct GABAA- (34) and indirect 5-hydroxytryptamine (5-HT1A)-mediated effects (35). The latter is consistent with other observations that valproate administration increased brain 5-hydroxyindoleacetic acid, a possible indicator of serotonin utilization, in stressed rats (36).
MAINTENANCE IN BIPOLAR DISORDER
Early open-label studies suggested that valproate might have prophylactic effectiveness in bipolar disorder (37, 38, 39). These studies included series of patients with frequent recurrence and schizoaffective characteristics (40). Calabrese
and Delucchi carried out a prospective, open-label study of divalproex in 78 patients with rapid cycling and found that, over an average of approximately 16 months, this treatment was effective in preventing recurrences of manic, mixed, or depressive states (8). A randomized, open-label comparison of lithium with valpromide maintenance found that patients given valpromide had slightly fewer episodes and significantly better toleration of treatment, and that lithium nonresponders improved when switched to valpromide, but valpromide nonresponders did not improve when switched to lithium (41).
In a continuation of their study of lithium and carbamazepine as maintenance treatments, Denicoff et al. reported that those not stable on lithium or on lithium plus carbamazepine improved when valproate was added (42).
These results encouraged Bowden et al. to conduct a controlled maintenance study of divalproex in bipolar disorder, the first placebo-controlled maintenance study of this illness in over 20 years (33). Subjects were treated for acute manic episodes and were converted, if possible, to monotherapy with lithium or divalproex. Those able to meet recovery criteria on monotherapy (approximately two-thirds of those originally entering the study) were randomized to receive divalproex, lithium, or placebo. Perhaps because of cautious study design and exclusion of severely ill subjects, placebo performed better than it had in the early lithium-placebo studies. Despite a low overall rate of relapse into mania, subjects randomized to divalproex fared better than those receiving placebo in outcome measures, including retention in the study and relapse into depression, and better than those given lithium in retention in the study and in interepisode depressive symptoms (33).
Biology of Maintenance Effects in Bipolar Disorder
Behavioral sensitization to stimulants may be related to the recurrent course of bipolar disorder and to its association with substance abuse. Repeated treatment with valproate, at a dose having no direct effects on motor behavior, prevented the induction of behavioral sensitization to methylphenidate (21). Repeated valproate given after sensitization to methylphenidate had developed prevented the expression of the sensitized motor response (43). These effects may be consistent with the effectiveness of divalproex in subjects who had a large number of previous episodes of illness (44).
COMORBIDITIES, BIPOLAR SPECTRUM, OR RELATED CONDITIONS
Bipolar Disorder and Substance Abuse
Substance abuse is a troubling aspect of bipolar disorder, occurring in as many as 60% of patients and complicating treatment (45). Perhaps because of its association with mixed states (46), which frequently are lithium resistant (7,47, 48, 49), treatment outcome in these patients is believed to be relatively poor. A group of 20 patients with bipolar disorder and substance abuse experienced a decrease in “craving” during open-label divalproex treatment (50). A retrospective chart review study suggested that valproate treatment was associated with a reduction in severity of substance abuse (51). A small, prospective, open-label study suggested that patients with substance abuse and dysphoric mania experienced marked improvement in both mania and substance abuse with divalproex treatment (52). None of these studies compared valproate with other treatments. More recently, Goldberg et al. investigated the effect of substance abuse on outcome in a large group of patients hospitalized for treatment of mania (53). Patients with substance abuse were twice as likely to achieve remission with divalproex or carbamazepine than with lithium treatment. The differential effect appeared to result from substance abuse rather than the associated mixed states, and combinations of lithium with anticonvulsant were no more effective than anticonvulsant alone (53).
Valproate has anxiolytic effects in some, but not all, animal models of anxiety (54, 55, 56). These effects may involve GABA because they usually are prevented by GABAA receptor antagonists (55,56). In a group of 13 patients with mood instability and panic attacks that had been refractory to other treatments, 10 achieved substantial improvement in mood instability, depressive symptoms, anxiety, and frequency of panics with 8 weeks' divalproex treatment (57).
Personality and Bipolar Spectrum Disorders
Valproate may be effective in patients whose presentation lies on the fringes of bipolar disorder. In a group of “bipolar spectrum” patients with cyclothymia or “hyperthymia,” valproate treatment was associated with reduction in mood symptoms and in “noxious personality traits.” Another open-label study found divalproex to be effective in improving mood symptoms, impulsivity, and anxiety in four of eight patients with borderline personality disorder (58). More recently, Hollander et al. (59) reported that valproate was more effective than placebo in treating affective and behavioral symptoms of borderline personality disorder over a 10-week period. Low doses (125 to 500 mg) were reported to be effective in milder forms of illness, including cyclothymic and “mild rapid cycling”; patients not responding to these doses improved with conventional doses (60).
Posttraumatic Stress Disorder
Valproate reduces catecholamine and behavioral changes in animals exposed to stress, suggesting that it may be effective
in stress-related disorders. Two open-label studies and one case report suggest that valproate preparations may be effective in reducing hyperarousal, intrusive symptoms, affective symptoms including depression, and impulsivity in patients with posttraumatic stress disorder (61, 62, 63).
Valproate has been used extensively as a treatment for impulsive aggression. One review found 17 studies involving 164 subjects, approximately three-fourths of whom had at least 50% reductions in the target symptom (64). Antiaggressive effects of valproate may be complementary to those of serotonin-enhancing treatments because a group of patients with assorted personality disorders whose aggressive behavior did not improve during treatment with fluoxetine at up to 60 mg/day had a marked reduction during 4 weeks of divalproex treatment (65).
Until recently, there were no placebo-controlled studies of valproate in aggression. Donovan et al. found valproate to reduce aggression symptom scores and mood lability in a series of 10 adolescent subjects who had explosive aggression and labile mood severe enough to disturb major life areas but who did not meet criteria for bipolar disorder (66). They then carried out a placebo-controlled crossover study in similar subjects. Eight of 10 subjects initially given divalproex improved, compared with 0 of 10 randomized to placebo; when subjects were crossed to the other treatment, valproate was effective in 6 of 7 who initially had not responded to placebo, whereas placebo was effective in 2 of 8 who previously had responded to valproate (67).
The diagnosis of schizoaffective disorder is very unstable, and many patients with this diagnosis ultimately receive a diagnosis of schizophrenia or bipolar disorder (68). Use of valproate in treating schizoaffective disorder, usually in combination with antipsychotic treatments, has increased markedly since 1994 (69). Retrospective reviews suggest effectiveness in up to 75% of these patients (70,71), especially if they have seizures or other neurologic problems (72). We know of no prospective, controlled studies in these patients.
The use of mood stabilizers, especially valproate, has increased substantially in patients with schizophrenia, essentially doubling between 1994 and 1998 in the New York State psychiatric hospital system (73). Evidence supporting effectiveness of valproate has been mixed but includes some tantalizing studies. Unlike carbamazepine, valproate has little or no effect on haloperidol levels (74). A review of previous studies, essentially all open-label, suggested that GABAergic treatments like valproate and benzodiazepines were useful adjunctive treatments in schizophrenia (75). An early study comparing addition of valproate or placebo with haloperidol treatment found improvement in hostile belligerence but not in psychotic symptoms (76). More recently, Wassef et al. carried out an open-label study in which divalproex added to haloperidol, either early or late in treatment, resulted in substantial improvement in psychosis symptoms and a briefer hospitalization (77). On this basis, the same authors carried out a small (n = 12), randomized clinical trial comparing haloperidol combined with placebo or with divalproex; improvement was significantly greater in the group receiving divalproex despite the small sample (78).
Largely because it increases GABA function, there have been multiple trials of valproate in tardive dyskinesia. As with other treatments, results have varied, likely in part because of heterogeneity of subjects and of other medicines that they were receiving. Randomized placebo-controlled trials have produced both positive (79) and negative (80) results. Valproate may be more effective when antipsychotic treatments are continued (81). A meta-analysis of randomized clinical trials in tardive dyskinesia found valproate to be more effective than placebo, although the effect size was not large (82).
Interestingly, valproate is effective in animal models of tardive dyskinesia (83). Valproate reduces prolactin levels in schizophrenic patients with tardive dyskinesia, but not in those without it (84). It was shown to decrease cerebrospinal fluid (CSF) homovanillic acid (a dopamine metabolite) and to increase CSF cyclic guanosine monophosphate in patients with tardive dyskinesia, interpreted as consistent with redressing a dopamine-acetylcholine imbalance (85). These data suggest that, at least in certain patients or under proper conditions, valproate may be a useful adjunct in tardive dyskinesia. No data have been published specifically addressing tardive dyskinesia in bipolar disorder.
CHILDREN AND ADOLESCENTS
The diagnosis and treatment of bipolar disorder or of bipolar-like conditions in children is not yet well defined. Results of treatment studies are very preliminary. A group of 59 children with “manic-like” presentations previously not responding to stimulant, antidepressive, or antipsychotic treatments responded to divalproex (86). A comparison of treatment studies in children or adolescents (mean age, 11.2 years) revealed that valproate, lithium, and carbamazepine
were all potentially effective, with effect sizes of 1.63, 1.06, and 1, respectively (87). In a group of 15 adolescents with mania, divalproex treatment was associated with greater than 75% reduction in mania rating scores in 10 (88).
VALPROATE IN GERIATRIC PSYCHIATRY
There are many reports that valproate was effective in open trials involving geriatric patients, but few data from controlled studies. A case series reported impressive effectiveness and good tolerability in seven patients with mean age of 66 years, most of whom had multiple other medical illnesses and treatments (89).
Valproate has been used extensively to treat behavioral problems associated with dementia in the elderly; many of the symptoms treated resemble those of mania or include impulsive aggression (90). One comparison with lithium treatment in nursing home residents found that lithium treatment was substantially more expensive, largely because of a higher cost of medical adverse events and monitoring (91). A comparison of efficacy found that valproate and lithium were similarly effective if drug levels were adequate (92). Doses and drug levels in case series demonstrating successful treatment have ranged from 375 to 750 mg/day in one prospective series of patients 71 to 94 years of age (93), to 743 mg/day (52.9 mg/L) in a series of patients averaging 71 years of age who were treated for 3 years (94), to 1,650 mg/day (64 mg/L) in a series of 25 patients whose age averaged 77 years (95). When response to a lower dose is inadequate, a higher dose should be considered, if tolerable (90). The incidence of cognitive, hepatic, or other side effects with valproate treatment has been low (94,96,97).
Although overall rates of toxicity are low, a small number of older patients treated with valproate have developed reversible drug-induced parkinsonian symptoms (98,99). These symptoms were reported to resolve with L-DOPA treatment (99) or discontinuation of valproate (98,99).
PREDICTORS OF RESPONSE
Characteristics of classic bipolar disorder predict good response to lithium, and departure from these characteristics predicts unfavorable outcome. The same is not true for valproate: Presence of substance abuse (53), manic-depressive symptoms (7), neurologic abnormalities (72), and poor course of illness with many previous episodes or rapid cycling all predict better outcome with valproate than with lithium treatment. Substance abuse is associated with depressive features in manic episodes, but effects of substance abuse and depressive features on response to lithium versus valproate may be independent (7,53). There is no clear evidence that presence or absence of these characteristics alters response to valproate; outcome with this drug may be independent of them. Apparent improved outcome in patients with many episodes (44) is consistent with effects of valproate on behavioral sensitization to stimulants (21,43).
Other than data on response in acute mania, there is little evidence about predictors of treatment outcome in bipolar disorder. Information on relative predictors of response among anticonvulsants, or among predictors in maintenance treatment or treatment of depressive episodes, would be valuable.
USE AND PREPARATIONS OF VALPROATE
Treatments in bipolar disorder are more effective if therapeutic levels of medicine are achieved quickly (100). Keck et al. showed that subjects responding to divalproex started at 20 mg/kg/day had a 50% improvement in mania rating scores within 48 hours after achieving a threshold valproate level of 50 µg/mL (101). The same group found 20 mg/kg/day divalproex to be equivalent, in improvement of mania and psychosis scores, to haloperidol at 0.2 mg/kg/day (102). More rapid oral loading with an initial dose of 30 mg/kg/day for 2 days appears feasible in some patients, producing a wide range of valproate levels (56 to 124 µg/mL) at 3 days, but additional clinical benefit of this method remains to be demonstrated (103,104).
Intravenous valproate may be an even more efficient way to achieve effective tissue concentrations quickly. Three “neuropsychiatric” patients with mania-like states tolerated and responded to intravenous valproate (105). In a series of patients with severe bipolar disorder, intravenous valproate was tolerated well and was effective for mania but not for depression. One patient improved who had not responded to oral divalproex loading (106).
Oral treatment may be given as valproic acid, as sodium valproate, or as its divalproex complex. In a series of 150 patients with bipolar disorder, those treated with valproate were twice as likely as those given divalproex to have gastrointestinal side effects, and three times as likely to discontinue treatment because of side effects (107). Among 12 patients who discontinued valproate treatment because of gastrointestinal side effects and were placed on divalproex, all but 2 were able to remain in treatment. The oral loading and placebo-controlled studies described previously were all carried out using divalproex. In this review, we have used the term valproate to refer to any of its preparations, including divalproex and valpromide, although in the United States divalproex is the form most commonly used.
Based on the preceding considerations, the best method of instituting treatment depends on the clinical context. When time is of the essence, because of severity of manic symptoms or other behavioral disturbances, it may be most efficient to start divalproex at 20 to 30 mg/kg/day followed
by adjustments as indicated by symptomatic response and side effects (101,102). Under other circumstances, more gradual dose escalation starting at 750 mg/day (or less in elderly patients) may produce better toleration. Lower doses may be indicated in patients with liver disorders or who are taking medicines that can inhibit the oxidative metabolism of valproate; higher doses may be needed if the patient is taking a drug, like carbamazepine or other aromatic anticonvulsants, that can induce oxidation of valproate. In any case, the final dose depends on the balance between symptomatic improvement and side effects. Although a threshold valproate level of at least 45 µg/mL appears to be required for effectiveness (6), the actual optimal level will be different for every patient, usually within a range of 50 to 125 µg/mL. The valproate blood level therefore is a secondary consideration relative to response and side effects, but is useful as a pharmacokinetic benchmark.
Valproate preparations appear effective for acute mania and a range of possibly related problems. Studies in bipolar disorder are summarized in Table 88.1. The strongest supporting evidence is for acute mania and in maintenance for depression, with somewhat less supporting evidence in acute depression or as maintenance for mania. There also is evidence for effects in behavioral problems associated with affective lability, aggression, and impulsivity across a range of clinical contexts, as shown in Table 88.2. Valproate often may be optimally effective as part of judicious combination treatments. Although no single treatment is effective for all patients or circumstances, and controlled studies in bipolar depression and maintenance are sparse, valproate has a substantial breadth of use in bipolar and related disorders.
TABLE 88.1. STUDIES OF VALPROATE PREPARATIONS IN BIPOLAR DISORDERa
TABLE 88.2. PSYCHIATRIC STUDIES OF VALPROATE PREPARATIONS OUTSIDE BIPOLAR DISORDERa