Joseph R. Calabrese and Charles Bowden

I. INTRODUCTION

Although many believe lithium (LI) remains the treatment of choice for bipolar disorder, only 60-80% of classic bipolar patients have a satisfactory clinical response. Fewer than half of all bipolar patients have classical, elated syndromes. When the response rate of LI is considered across the wider spectrum of bipolar disorders permitted by the most recent edition of the Diagnostic and Statistical Manual of Mental Disorders (Diagnostic and Statistical Manual of Mental Disorders 1995), it is less than 50% of patients who respond to it (Gitlin et al 1995, Swann et al 1997, Vestergaard et al 1998). Among mixed manics who account for 16 to 67% of all bipolar patients, only 30-40% respond to LI (Calabrese et al 1993 a). Rapid cycling patients who constitute 13-20% of all bipolar disorders also have low (20-30%) response rates to LI (Bauer et al 1994, Dunner and Fieve 1974, Kukopulos et al 1980). Although there are numerous other predictors of LI non-response, these two variants appear to account for the largest proportion of those who exhibit poor response to LI. Serendipitously, these two predictors of LI non-response may be putative predictors of positive response to both carbamazepine (CBZ) (Post et al 1989) and valproate (VPA) (Calabrese et al 1993 b, Freeman et al 1992, Swann et al 1997).

Although controlled trials have demonstrated the efficacy of CBZ in the acute treatment of mania, initial responsiveness is often lost over time (Post et al 1990, Frankenburg et al 1988). Recent data also suggest that only one-third of acutely manic patients have moderate to marked improvement during the first two months of treatment with either LI or CBZ as monotherapy (Small et al 1991). In addition, more severe mania was noted to predict nonresponse to CBZ. Such findings suggest that large numbers of bipolar patients are non-responsive to LI and that substantial numbers of both classic and rapid cycling bipolar patients eventually become resistant to CBZ. Recent multicenter controlled data demonstrating that divalproex sodium has efficacy equal to lithium in the acute management of mania has led to this drug's approval by the US Food and Drug Administration in 1995 for use in the management of manic episodes associated with bipolar disorder (Bowden et al 1994, Pope et al 1991). The data regarding the efficacy of divalproex sodium in the prevention of relapse and recurrence in bipolar disorder is unclear (Bowden et al 1997).

At present, lithium, divalproex sodium, and the antipsychotic medications have all been approved for use in the management of the manic phase of bipolar disorder. However, there have been no medications approved for use in the management of the depressed phase of the illness and less is known about the antidepressant effects of lithium, and to an even greater extent, divalproex sodium; there have been no published studies on the acute efficacy of lithium in bipolar I depression in the last 20 years. Although controversial, lithium (Fieve et al 1968) and divalproex sodium (Calabrese et al 1990) appear to be less effective in the treatment of the depressed phase of the illness, as well as depressive symptoms associated with mania (Swann et al 1997). The adjunctive use of antidepressant medication is common, but this practice can put patients with bipolar disorder at increased risk for the development of hypomania, mania, or cycle acceleration (Post et al 1997). This has led to the development of another anticonvulsant, lamotrigine, for use in the treatment of bipolar I depression (Calabrese et al in press a, Calabrese et al in press b).

This chapter will critically review recent information regarding the efficacy of lithium and its use in the treatment of bipolar disorder and the rapidly developing data regarding the evolving role of CBZ, divalproex sodium, and more recently, lamotrigine. In addition, there will be some discussion of the various methodologies employed in studies of bipolar disorder. Extrapolation of results generated from these controlled, blinded clinical trials to clinical practice is difficult. It is plausible that results from the better designed trials emphasized here may actually underestimate patient outcome. Factors such as the enrollment of refractory patients in trials with monotherapy arms that allow for only the limited use of adjunctive medications depart substantially from the practice of psychiatry in the community.

II. LITHIUM

Acute Mania: Recent studies have indicated that approximately half of acutely manic patients respond well to LI (Kukopulos et al 1980, Small et al 1988, Small et al 1991). The benefits of response appear to be more compromised by adverse effects than previously thought (Grof et al 1987, Vestergaard and Aagaard 1991). Such reports have led to reassessment of the data on which approval of LI was based and new studies to clarify these issues, utilizing improved research methodologies and patient samples that reflect the full diagnostic spectrum of bipolar disorder in the current nomenclature. The evidence for the efficacy of LI in acute mania largely came from three double-blind, placebo controlled crossover studies of a total of 78 patients (Goodwin and Jamison 1990). Outcome criteria were variable across the studies and diagnostic criteria were not consistently specified. In one, patients with any degree of improvement on a nurse rated 7 point global scale were reported as responders. In two of the studies initial assignment to study medication was not randomized. No placebo controlled study of LI in acute mania was published from 1971 through 1993.

The one parallel group, randomized, double-blind study of LI in acutely manic patients reported that 49% had moderate or better improvement in manic symptoms during the three week trial, a rate of response significantly greater than the 28 % seen in placebo treated patients (Bowden et al 1994). The efficacy of LI did not differ from that of VPA. Thirty-one percent of LI-treated patients had some degree of worsening of manic symptoms at the end of the trial. The percentage of patients discontinuing due to medication intolerance was non-significantly higher among LI-treated than VPA-treated patients (11% vs. 6%). No neuroleptics were used during the study and the only other psychotropic medication allowed was lorazepam during the first week of study medication. These results generally conform to recent impressionistic data, and indicate that although LI is clearly effective, substantial numbers of patients remain unresponsive during monotherapy of an acute episode.

Clinical improvement with LI is relatively slow, with an initial response generally occurring 7 to 14 days after initiation of LI. Initial improvement may not occur sooner than the fourth week for some patients. Because of aggressive, disruptive behavior of the acutely manic patient, adjunctive medications are often required during this lag period. Lithium plasma concentrations required for acutely manic patients are often higher than needed for maintenance therapy. Although a loading dose strategy is ineffective, dosage should be increased as rapidly as tolerated until response occurs or a plasma level of around 1.2 mEq/l is reached. The side effects of LI are generally less apparent during acute treatment than maintenance therapy.

Acute Bipolar Depression: Six of seven placebo controlled studies have reported significant antidepressant effects of LI (Goodwin and Jamison 1990). However, methodological problems limit generalizability. Most included unipolar as well as bipolar depressed patients. Most studies comparing tricyclic antidepressants with lithium lacked placebo (PBO) controls. The single study which compared LI, imipramine and PBO in bipolar depressed patients found imipramine more effective than LI (Fieve et al 1968). The protective effect of LI against subsequent episodes appears to be lower for depressive episodes than for manic episodes (Davis 1976, Dunner and Fieve 1976). Additionally, most depressive episodes in bipolar patients will occur in patients already taking LI, or alternative mood stabilizing agents. Therefore, early use of other antidepressant agents is important if the patient does not respond to an increase in LI dosage.

Prophylaxis: Maintenance phase studies of LI are more extensive, better designed, and more conclusive than acute treatment studies regarding LI's efficacy in preventing relapses (Baastrup et al 1970, Davis 1976). However, no PBO-controlled maintenance treatment study of LI's efficacy in bipolar disorder has been published in over 20 years. Several recent naturalistic studies suggest that at present a substantial number of bipolar patients have inadequate responses to LI. Harrow and colleagues reported that 40 percent of LI-treated patients had a manic relapse during a mean 1.7 year period of treatment, a result similar to that of patients on no medication (Harrow et al 1990). Coryell reported that whereas lithium treatment yielded fewer relapses than no medication for the first 32 month of prophylaxis, subsequent relapse rates did not differ with or without lithium (Coryell et al 1997). Gitlin and colleagues reported that 17% of bipolar patients had good outcomes over 2 to 5 years. Average mood symptomatology was much more strongly associated with occupational, social and family disruption than were relapses (Gitlin et al 1995). Tohen and colleagues also reported that 46 percent of patients remained stable for four years after their first episode, with outcome being unrelated to treatment (Tohen et al 1990). Other recent prospective, open, naturalistic trials also report good outcomes in approximately on-third of patients (Maj et al 1998, Vestergaard et al 1998).

Issues of dosage and approach to discontinuation of LI have been clarified by recent studies. Discontinuation of LI therapy in bipolar patients is followed by a high rate of relapse, with most episodes being manic rather than depressed (Suppes et al 1991). This study by Suppes et al, as well as other withdrawal study data, strongly suggest that LI's prophylactic effect is principally in reducing the frequency of manic rather than depressive episodes (Nilsson and Axelsson 1990). Furthermore, abrupt discontinuation is followed early on by a much higher rate of relapse than is discontinuation over two or more weeks (Faedda et al 1992). Relapse rates following discontinuation were 50 percent higher among bipolar I than bipolar II patients (Baldessarini et al 1996). Among patients with two or fewer prior episodes, maintenance LI levels of 0.8 to 1.0 mEq/l were associated with lower rates of relapse than were levels of 0.4 to 0.6 mEq/l. Relapse rates into depression did not differ between the two groups. The severity of adverse effects was greater in the higher plasma level group (Gelenberg et al 1989).

Predictors of Response to Lithium: As evidence has developed that a substantial percentage fail to do well with acute or maintenance treatment with LI, and as alternative treatments for bipolar disorder have developed, it has become important to know which illness characteristics are associated with a favorable, or unfavorable response to LI, as well as alternative treatments. Acutely, patients with elated, less severe episodes without psychotic features, also referred to as classical mania, appear to have the best response to LI during an acute manic episode (Goodwin and Jamison 1990). Patients who move from a depressive episode into a period of euthymia, and then mania do better than those who move directly from a depressive episode into a manic episode (Grof et al 1987). It is not well established that these differences hold equally during maintenance treatment. Patient with more lifetime illness episodes consistently have less good longterm outcome with lithium (Gitlin et al 1995).

Several syndromal variants of bipolar disorder are associated with relatively low response rates to LI. The best studied is mixed mania, with more than half a dozen studies consistently indicating lower acute and chronic response rates to LI treatment (Goodwin and Jamison 1990). Rapid cycling patients also respond less well to LI (Dunner and Fieve 1976, Kukopulos et al 1980). This illness course of bipolar disorder will be defined in the DSM-IV as 4 or more depressions, hypomanias, or manias in the prior 12 month period, with episodes being demarcated by a switch to an episode of opposite polarity or by a period of remission. This phenomenon appears to be late in onset, occurs most commonly in bipolar type II females, and is not usually associated with antidepressant use (Bauer et al 1994, Dunner and Fieve 1976, Kukopulos et al 1980). Rapid cycling, and possibly mixed states, appear to be non-familial modifiers of state and course that transiently come and go during the natural history of bipolar disorder and its treatment (Coryell et al 1992). When these variants are present during treatment with LI therapy, prognosis worsens and the morbidity and mortality associated with the illness increases (Calabrese et al 1993 b, Fawcett et al 1987). As evidence has emerged that antidepressant medications increase the likelihood of rapid cycling, it is prudent to be cautious in interpreting the data about rapid cycling, as a component of the poor response may have been secondary to recent antidepressant therapy. Patients with bipolar disorder which appears to be secondary to other medical disorders respond poorly to lithium. Patients with comorbid substance respond relatively poorly to lithium therapy, although this may be simply because of the conservative management of their substance abuse.

During maintenance therapy with LI, the development of hypomania is strongly predictive of a subsequent full manic episode, with 65 percent having a subsequent major manic relapse. By contrast, minor depressive episodes were not usefully predictive of either subsequent manic or depressive episodes, as 61 percent of the minor depressive episodes were not followed by any major relapse (Keller et al 1992).

Pharmacokinetics and Adverse Effects of Lithium. The clearance of LI is delayed by most non-steroidal anti-inflammatory drugs. Several of the non-steroidal anti-inflammatory agents have been marketed in recent years, with most not assessed regarding effects on LI clearance. A conservative approach is to monitor LI plasma levels closely for several months following introduction of non-steroidal anti-inflammatory drugs.

Lithium consistently impairs urinary concentrating ability. The ensuing frequent day and night time urination are often functionally impairing to the patient. Use of sustained release lithium is associated with less renal concentrating impairment (Miller et al 1985). If such symptoms remain prominent, one should consider discontinuing LI and utilizing VPA or CBZ, or possibly, adding a diuretic that reduces the concentration impairment. Both loop diuretics and thiazides are effective. Although thiazides usually increase the plasma concentration of lithium, this can be managed by dosage adjustment. It is plausible, though not studied, that the reduction in distal tubule exposure to LI associated with thiazide use could reduce long-term risk of renal impairment.

Although several recent reviews have drawn sanguine conclusions regarding the effect of long-term LI treatment on renal function, several other studies have reported a small number of cases of renal insufficiency, probably at incident rates no higher than 1 percent (Schou 1988, Stancer and Forbath 1989). Duration of lithium use, dosage and plasma level, and prior experiencing of polydipsia are possible, but not clearly established, risk factors in cases of renal insufficiency. Regular monitoring of creatinine levels in plasma is warranted indefinitely, and consultation indicated if creatinine levels rise to and remain above 1.6 mg/100 ml.

Long-term LI use is also associated with a significant increase in subclinical hypothyroidism, a risk which may be greater in women. The duration of LI use was positively correlated to antithyroid antibody titer (Calabrese et al 1985). It thus seems advisable to check thyroid function regularly in LI-treated patients, perhaps semiannually for the first two years of treatment. Some authorities recommend addition of T4 to the regimen of any LI treated bipolar patient if indices of thyroid function are in the low normal range, such as a thyroid stimulating hormone level greater than 5 microinternational units. This may be particularly important if the patient exhibits signs of poor control of illness, such as subsyndromal depression.

The risks of use of LI in pregnancy remain an unresolved issue. Earlier studies suggested that LI increased the risk of the cardiovascular anomaly Ebstein's atresia. The low base rate of Ebstein's atresia of approximately 1 in 20,000 live births makes conclusive study of this risk difficult. A recent study suggested no increased risk, but was of such a small number of subjects that its ability to recognize two to three fold increases in risk was negligible (Zalstein et al 1990). Another recent study of 148 women using lithium during the first trimester of pregnancy indicated rates of major congenital malformations did not differ between the lithium (2.8%) and control (2.4%) group and concluded that lithium is not an important human teratogen (Jacobson et al 1992). Since neuroleptics may control some aspects of manic symptoms, it may be advisable to try discontinuing LI as soon as pregnancy is recognized and, for any manic symptoms emergent during the first trimester, try neuroleptics first with electroconvulsive therapy as an additional consideration.

Although the most frequently reported subjective side effects of lithium have been well recognized (excessive thirst, polyuria, memory problems, tremor, and weight), the implications of lithium's adverse effects, especially their role in contributing to non-compliance, have been reconsidered (Goodwin and Jamison 1990). It seems likely that the absence of alternatives to LI inclined psychiatrists to try to manage adverse effects which seriously interfered with function with some form of continued LI use. As evidence for the efficacy of alternative mood stabilizers has emerged, earlier consideration of alternative regimens has become prudent. Recent studies underscore the frequency and degree to which LI impairs cognitive functions, and the contribution of such adverse effects to poor compliance (Goodwin and Jamison 1990).

III. CARBAMAZEPINE

The clinical efficacy of CBZ (carbamazepine or oxcarbamazepine) in bipolar disorder has been documented under controlled conditions in at least 288 acutely manic patients, 31 acutely depressed patients, and 116 patients for prophylaxis. Additionally, another 581 bipolar patients were treated with CBZ in open studies. Carbamazepine has been established as having marked acute and prophylactic antimanic and antimixed states properties with poor to moderate antidepressant responses. Predictors of response data suggest this drug's profile may complement that of lithium and have particular application in the management of mixed states and rapid cycling.

Acute Mania: Sixteen controlled studies generating acute antimanic efficacy data in a total of 571 predominately bipolar manic patients indicate that CBZ is superior to PBO and at least equal in efficacy to LI and neuroleptics; Six comparisons each between CBZ and placebo, neuroleptics, and LI were made. Although reports of CBZ's efficacy appear in the literature as early as 1971 (Takezaki and Hanaoka, 1971), the first PBO-controlled study of this drug's acute antimanic efficacy was published by Ballenger and Post in 1978 (Ballenger and Post, 1978). Reporting their results to date in 1980 (Ballenger and Post 1978), they found that better than 50% of manic patients experienced a moderate-to-marked improvement relative to their placebo baseline; and again in 1987 (Post et al 1987), aggregate experience with 19 acutely manic patients indicated that 12 had a moderate-to-robust response. While these findings are in support of CBZ's acute antimanic efficacy, Post and colleagues observe that as a group, the responders remained symptomatic, and that response in this study represents clinical improvement and not necessarily remission (Post et al 1987). Other placebo-controlled studies (Desai et al 1987, Emrich et al 1985, Klein et al 1984, Moller et al 1989, Muller and Stoll 1984) demonstrate the superiority of CBZ over PBO, but frequently adjunctive therapy with neuroleptics (Klein et al 1984, Moller et al 1989, Muller and Stoll 1984) or LI (Desai et al 1987) was used.

Due in part to the morbidity associated with the acutely manic patient, the majority of controlled studies in this population compare CBZ with an active treatment, such as LI or a neuroleptic. Post and colleagues report a comparison between CBZ, LI, and neuroleptics in which equal efficacy was demonstrated (Post et al 1989). Following a placebo lead-in, nineteen patients received CBZ, nineteen received LI, and seventeen received neuroleptics. By three weeks, all three groups showed substantial improvement with respect to the PBO phase, and convergence to the same degree of symptomatic severity despite higher degrees of baseline mania in the CBZ and neuroleptic groups. Other neuroleptic-controlled studies find that CBZ is at least as effective (Grossi et al 1984, Moller et al 1989, Muller and Stoll 1984, Okuma et al 1981), or more effective (Brown et al 1989, Stoll et al 1985) than neuroleptics. Okuma and colleagues carried out the largest double blind comparison to a neuroleptic in their study of 55 patients randomized to either CBZ or chlorpromazine (Okuma et al 1979). Rescue medications, such as benzodiazepines, were permitted in most of these studies (Grossi et al 1984, Moller et al 1989, Muller and Stoll 1984, Okuma et al 1981).

Controlled studies of CBZ versus LI in the management of acute mania give results paralleling the findings of the CBZ versus neuroleptic literature. Specifically, CBZ was found to be at least as efficacious as LI in five studies (Lenzi et al 1986, Lerer et al 1987, Okuma 1990, Post et al 1989, Small et al 1991). Small and colleagues carried out the largest (n = 48) and best designed double blind parallel group comparison of CBZ to LI. Although they found that CBZ was equal in efficacy to LI, only one-third of patients in each cohort were described as responders. There is additional evidence that CBZ is at least as effective acutely as LI from a maintenance study (Pope et al 1991), which had an acute phase in which 20 of 29 bipolar patients treated with CBZ continued the trial past two months, as opposed to 16 of 27 LI-treated patients. Three studies permitted neuroleptics as rescue medication (Lenzi et al 1986, Okuma 1990, Pope et al 1991) and one permitted antidepressants (Post et al 1987). While these studies support the conclusion that CBZ and LI have comparable antimanic efficacy, it should be noted that a substantial number of LI non-responders are often among the participants in such studies. Thus, interpretation of the above results might be skewed by the presence of patients with inherently more severe illness. For example, the study of Small and colleagues (Small et al 1991), which contains a high proportion of lithium and other treatment non-responders, only demonstrated response in one-third of patients treated with either CBZ or LI as monotherapy. There is some evidence, however, that LI and CBZ may possess synergistic action that converts some failures of each drug as monotherapy into responders when used in combination. Kramlinger and Post added LI in a blinded fashion to seven patients responding poorly to CBZ, of whom six were previous LI failures, the remaining patient responding to LI and neuroleptics in combination (Kramlinger et al 1989). Complete remission was seen in five (one of whom also required haloperidol), marked improvement in one, and no response in one (who was one of four rapid cyclers).

Acute Bipolar Depression: There is much less controlled data addressing the acute efficacy of CBZ in the treatment of bipolar depression than there is for mania. Placebo-controlled studies conducted by Post and colleagues report acute antidepressant efficacy in bipolar depression, but the response is less robust that seen in acute mania. Specifically, in their study of 1980 (Ballenger and Post 1980), three of seven bipolar depressed patients had a good response to CBZ; and in their 1986 update of their experience treating bipolar depression (Post et al 1986), they report on 24 patients with bipolar depression, of which 16 were BP I and eight were BP II. Fifteen of the 24 patients exhibited moderate improvement.

Prophylaxis: Seven controlled studies produced prophylactic data on 217 bipolar patients, of whom 116 were treated with CBZ. Carbamazepine was compared to PBO in two studies, and to LI in five. While CBZ is superior to PBO, the results relative to LI suggest equal efficacy overall, however, some evidence suggested that CBZ-treated patients relapsed earlier in the course of prophylaxis than their LI-treated counterparts.

In 1978, Post and colleagues published the first PBO-controlled prophylactic study of CBZ using a cross over design in 10 patients (Ballenger and Post 1978). Several years later, Okuma and colleagues replicated these findings, noting moderate-to-marked responses in six of 10 CBZ patients as opposed to two of nine PBO patients followed for a year (Okuma 1983). Both studies permitted rescue medication to manage breakthrough episodes.

Although still arguing for the efficacy of CBZ, prophylactic studies comparing this agent to LI have been less impressive than the acute studies. While two studies found that the two treatments were equally efficacious (Lusznat et al 1988, Placidi et al 1986), other studies (Coxhead et al 1992, Small et al 1991, Watkins et al 1987) demonstrate superior prophylactic efficacy of LI; Patients given CBZ appeared to relapse earlier. Placidi and colleagues (Placidi et al 1986) found that although both drugs were effective in two-thirds of patients, a higher dropout rate for patients with mood incongruent psychotic features was observed in the lithium cohort. Watkins and colleagues observed that "LI significantly lengthens the time in remission" (Watkins et al 1987). Small and colleagues (Small et al 1991) reported that all CBZ patients dropped out during the first 24 weeks of prophylaxis, whereas LI group drop-outs occurred over the first year of follow-up. Coxhead and colleagues (Coxhead et al 1992) stated that nearly all CBZ drop-outs occurred in the first month of follow-up, but may have been due to precipitous withdrawal of LI (as all patients entering prophylaxis were initially stabilized on LI). As in the placebo studies, rescue medications were permitted and included neuroleptics (Lusznat et al 1988, Placidi et al 1986) and antidepressants (Lusznat et al 1988, Placidi et al 1986, Watkins et al 1987).

In addition to the above controlled studies, there are many open studies which in the main support the conclusions drawn from the controlled studies. By and large, over the broad spectrum of bipolar disorder, the literature suggests that CBZ is a safe and effective alternative and adjunct to LI in the acute and prophylactic management of both classic bipolar disorder, as well as atypical lithium non-responsive variants. Carbamazepine appears to have its greatest clinical impact on the manic phase of the illness as opposed to the depressed phase, but may lose some prophylactic efficacy over time (Frankenburg et al 1988, Post et al 1990, Tohen et al 1990).

Predictors of Response to Carbamazepine: Findings in the controlled literature bearing on predictors of response have been uneven. Initially, it seemed that predictors of LI nonresponse, particularly dysphoria complicating mania and rapid cycling, predicted a differential response to CBZ relative to LI (Ballenger and Post 1980, Post et al 1984, Post et al 1989). Summarizing predictors of response in mania, Ballenger (Ballenger 1988) notes that rapid and circular continuous cycling patients respond much better to CBZ than to LI; in fact, their response rate rivals that of patients without rapid cycling who are treated with LI. He indicates that the predictors of response of bipolar illness to CBZ include mania complicated by dysphoria/anxiety (a subset of mixed states), rapid cycling, and greater severity of illness. Indeed, for mixed states, the work of Post and colleagues (Post et al 1989, Post et al 1990) supports the use of CBZ in this LI-resistant population. For rapid cycling, a focused review of the literature reveals 19 open studies and 3 controlled studies yielding 119 and 6 rapid cyclers respectively, in whom acute response rates were 32% for depression and 51% for mania. Prophylactically, responses were 60% for the depressed and 64% for mania. It should be noted that the great majority of these patients required acute and prophylactic adjunctive medications including neuroleptics, antidepressants, and LI. Accordingly, these combination therapy predictors of response data may not extend to the drug's use in monotherapy.

Findings at variance with the prevailing view of predictors of response include acute work by Klein and colleagues which suggests CBZ does not have a unique response profile (Klein et al 1984). Stromgren and Boller's comprehensive review of CBZ treatment of bipolar illness (Stromgren and Boller 1985) reported that "scrutiny" of the case histories of the literature reviewed gave no impression of pretreatment clinical differences between responders and nonresponders to CBZ. Lovett and colleagues (Lovett et al 1986) evaluated age of illness onset, predominant affective typology, or sequential pattern and found no correlation with response. In contrast to Post's prior data, Small and colleagues (Small et al 1991) found that decreasing severity of mania predicted positive response to not only LI, but CBZ as well. Coxhead and colleagues (Coxhead et al 1992) found that previous morbidity predicted future morbidity, regardless of assignment to CBZ or LI groups. This is consistent with recent data suggesting that (Okuma 1993) patients without rapid cycling respond better to CBZ than those with rapid cycling. These findings indicate that the initial optimism regarding the efficacy of CBZ in the management of the traditionally LI refractory patient may need to be tempered somewhat, particularly in the case of rapid cycling. These observations were further embellished by Greil et al (1998) who recently reported on prospective randomized data suggesting that LI was superior to CBZ in treating classical bipolar patients, whereas non-classical patients (bipolar II, bipolar NOS, or those with mood incongruent delusions or comorbidity) were more responsive to carbamazepine which appeared to have a broader spectrum of efficacy.

IV. VALPROATE

There now are over 1,707 psychiatric patients that have undergone various trials of valproate. Of these, 438 acutely manic patients underwent controlled trials, 493 controlled maintenance trials, 528 open longitudinal trials for bipolar or schizoaffective disorder - bipolar subtype patients, and 258 open longitudinal trials for miscellaneous diagnoses including schizophrenia, unipolar depression, and personality disorders.

Acute Mania: [Footnote: The text in the chapter will now refer to all similarly active preparations of the compound as "valproate (VPA) including divalproex sodium, sodium VPA, valproic acid, valpromide, and dipropylacetamide.] Well-designed studies have demonstrated the efficacy of VPA in comparatively large samples of acutely manic patients (Bowden et al 1994, Pope et al 1991). These studies are consistent with the findings of an extensive open longitudinal literature dating back to the mid-1960s. Prospective studies have documented poor to moderate acute and prophylactic efficacy in the management of the depressed phase of the bipolar disorder, but these were not controlled and await replication (Calabrese et al 1990).

Eight controlled studies have evaluated the antimanic efficacy of VPA in 438 acutely manic patients in randomized double blind trials (Brennan et al 1984, Post et al 1984, Emrich et al 1985, Pope et al 1991, Freeman et al 1992, Bowden et al 1994, McElroy et al 1996, Muller-Oerlinghausen et al 1998). All of the PBO-controlled studies found VPA superior (Bowden et al 1994, Brennan et al 1984, Emrich et al 1985, Pope et al 1991). Most (Bowden et al, Brennan et al 1984, Emrich et al 1985), but not all (Freeman et al 1992), lithium controlled studies found VPA to be equal in efficacy. Of the eight, four had PBO arms (Brennan et al 1984, Emrich et al 1985, Pope et al 1991, Bowden et al 1994), two had LI arms (Freeman et al 1992, Bowden et al 1994), one a haloperidol active comparator (McElroy et al 1996), one restricted recruitment to patients shown to be inadequately responsive to prior lithium therapy (Pope et al 1991), and one evaluated valproate as augmentation (Muller-Oerlinghausen et al 1998).

Emrich first documented the efficacy of VPA in a controlled study. He administered VPA and PBO to five acutely manic patients in a double blind ABA crossover fashion and noted that four of five had marked responses (Emrich et al 1985). Brennan et al replicated these initial findings in 1984 by documenting marked efficacy in six of eight acutely manic patients that received VPA and PBO in a double blind ABA crossover fashion (Brennan et al 1984). In a case study, Post et al then showed VPA and phenytoin non-response in one patient who responded to CBZ in a double blind crossover fashion. They concluded that non-response to one anticonvulsant did not predict non-response to others in bipolar disorder (Post et al 1987). The first well-designed double blind PBO controlled study of VPA was published in 1991 by Pope and colleagues. Thirty-six patients were randomly assigned to treatment for 7-21 days with no other medications permitted after day 10. Valproate was significantly superior to PBO. The 17 patients randomized to VPA demonstrated a 54% decrease in scores on the Young Mania Rating Scale as compared with a median 5.0% decrease among the 19 patients receiving PBO. Significant differences also emerged on the Global Assessment Scale and the Brief Psychiatric Rating Scale. Substantial antimanic effects appeared within 1-4 days of achieving therapeutic VPA concentration. No adverse effects occurred more frequently with VPA than with PBO. Although responders had an older age of onset, shorter duration of illness, and higher VPA levels, rapid cycling and dysphoric mania did not predict response during this short trial (Post et al 1984). Up until Pope et al published their data, there were no studies available in the literature comparing the efficacy of VPA to PBO. Although this study only randomized 36 patients, its sample size compared favorably to the three double-blind placebo controlled crossover studies of lithium which totalled 78 and was superior in design in that it employed random assignment to parallel groups. Freeman studied 27 acute manics who were randomized into a double-blind parallel group trial comparing LI to VPA (Freeman et al 1992). Favorable response to VPA was associated with high pre-treatment depression scores. The coexistence of high pre-treatment depression scores was found to be a predictor of response to valproate. They determined VPA and LI were both effective in improving manic symptoms, though LI was slightly more efficacious overall, and found VPA to be superior to LI in the management of acute episodes of mania accompanied by co-existing depression, ie. mixed states or dysphoric mania. This study's unusually high LI response rate (90%) suggested the need for a PBO.

The completion of a multicenter acute double-blind comparison of VPA to LI and PBO study (n = 179) increased the number of acutely manic patients studied in a double-blind fashion from a total of 77 to 256 (Bowden et al 1994). Patients were randomly assigned to parallel groups (VPA 40%, PBO 40%, LI 20%) and studied over a 21 day experimental period. Rescue medications (up to 4gm/d chloral hydrate and/or 2 mg/d lorazepam) were permitted up to day 10. The Mania Rating Scale, composed of the Manic Syndrome Scale and the Behavior and Ideation Scale from the Schedule for Affective Disorders and Schizophrenia, constituted the primary efficacy variable. Results were similar for the intent-to-treat sample, composed of all patients who received at least one dose of study drug or PBO, and the evaluable sample, composed of patients who received the study medication for at least eight days and had adequate serum levels. Valproate-treated patients had significantly greater improvement than PBO-treated patients on the Manic Syndrome Scale from day 5 through the end of the 21 day trial. From day 10 on, the Mania Rating Scale and the Behavior and Ideation Scale indicated significantly greater improvement for VPA-treated patients compared with PBO. Lithium-treated patients had profiles of response similar to those of the VPA-treated patients on every efficacy variable throughout the entire study. Categorical outcome among the three groups was also analyzed. Mild, moderate, and marked improvements were defined as 30%, 40%, and 50% improvement from baseline to the final evaluation on efficacy measures. At all levels, the response to the two active medications was approximately twice that of response to PBO and was highly clinically significant.

Analysis of individual behavioral items indicated that manic mood, reduced sleep, increased activity, generalized motor hyperactivity, elation/grandiosity, and psychosis improved significantly more in VPA- than PBO-treated patients and that manic mood improved significantly more in LI- than PBO-treated patients. During the 21 day trial, both drugs were relatively well-tolerated. However, there was a trend towards more LI-treated patients dropping out prematurely for intolerance to treatment than in the PBO group. Forty-eight percent of the VPA group, 61% of the LI group, and 64% of the PBO group failed to complete all 21 days of treatment.

Recently, McElroy and colleagues have carried out a controlled study which compared the efficacy of divalproex oral loading to haloperidol (McElroy et al 1996). Thirty-six consecutive hospitalized patients with bipolar disorder, manic or mixed phase and with psychotic features, were randomly assigned to either divalproex 20 mg/kg/day or haloperidol 0.2 mg/kg/day for six full days, without other psychotropic agents except lorazepam up to 4 mg/day for management of agitation. Response was measured daily by a blind rater using the Young Mania Rating Sale and the Scale for Assessment of Positive Symptoms. Divalproex oral loading and haloperidol were equally effective in acutely reducing manic and psychotic symptoms. The authors concluded that divalproex oral loading produced a rapid onset of both antimanic and antipsychotic effect comparable to that of haloperidol, but with minimal side effects.

Very recently, Muller-Oerlinghausen and colleagues (1998) completed a multicenter comparison of VPA as an adjunct to neuroleptic medication (haloperidol and/or perazine) in the treatment of acute hospitalized episodes of mania in 136 patients with either bipolar disorder or schizoaffective disorder - bipolar subtype (Muller-Oerlinghausen et al 1998). Daily adjustments of neuroleptics was permitted, but the dose was decreased by one-third on study days 5, 10, and 15. This prospective, randomized, double-blind, placebo-controlled augmentation study increased the number of acutely manic patients studied in a double-blind fashion to 438. This study reported a valproate response rate of 70% (as defined by a 50% reduction in the Young Mania Rating Scale) which exceeds those response rates reported by all other studies. Of note, however, was that the placebo-response rate in this study (45%) was also the highest ever reported in an acute study, and presumably attributable due to the study's augmentation design.

Prophylaxis: Recently, Bowden and colleagues have submitted data on the prophylactic effects of divalproex in bipolar I disorder from a large multicenter trial (Bowden et al 1997, Bowden et al submitted). The objective of this trial was to compare the efficacy of divalproex with lithium and placebo in the prevention of relapse. This trial employed a randomized, double-blind, parallel-group design during a one-year prospective follow-up period. A total of 372 outpatients were studied in 37 academic medical centers. Patients came from an initial group of 571 subjects enrolled in an open trial of discretionary treatment of a manic episode that began within three months of randomization, and had improved by the time of randomization. Patients were randomized to divalproex, lithium, or placebo in a 2:1:1 ratio, with the taper of the antecedent open trial medication over the first two weeks of the 52-week randomized period. Dosages of divalproex and lithium were adjusted to maintain a concentration of 71-125 ug/ml for divalproex, and 0.8 - 1.2 mEq/L for lithium during the maintenance period. Changes in the Mania Rating scale, the Depressive Syndrome scale, and the Global Assessment scale derived from the Schedule for Affective Disorders and Schizophrenia.

Both survival analyses and mean change from baseline analyses during the maintenance period were conducted. Intent-to-treat analyses for efficacy were based on data from 187, 90, and 92 patients in the divalproex, lithium, and placebo groups, respectively. In 24%, 31%, and 38% of the above groups, treatments were prematurely terminated due to the development of a manic or depressive episode, with fewer premature terminations from divalproex than placebo (P = 0.01). The overall rates of premature termination were 62%, 76%, and 75% of the above groups, with lower rates for divalproex than either lithium (P = 0.03) or placebo (P = 0.045). The mean durations in maintenance treatment for the three groups were 198, 152, and 165 days, respectively, with the difference significant between divalproex and lithium treated groups (P = 0.015). Survival analyses of time to manic relapse or a depressive episode for the three treatment groups favored divalproex compared to lithium (P = 0.05) and placebo (P = 0.06). There were no significant differences between the three groups in time to relapse into a manic episode.

There are 11 open longitudinal studies in the literature that have investigated the prophylactic use of VPA in a total of 786 psychiatric patients (Brennan et al 1984, Calabrese et al 1993 b, Emrich et al 1985, Hayes 1989, Lambert 1984, Lambert and Venaud 1992, Puzynski and Klosiewicz 1984, Semadeni 1967, Sovner 1989, Vencovsky et al 1987, Zapletalek et al 1988). Of these 786, there were 528 with bipolar disorder or the bipolar subtype of schizoaffective disorder with enough detail to allow for conclusions. Those of particular importance are described. The French investigator, Lambert, first began using VPA in 1966. He administered the drug to 393 psychiatric patients: 244 affective disorders (141 bipolar), 27 schizophrenics, and 122 character disorders (Lambert 1984). In open trials, 141 bipolar patients received an average of 900 - 1800 mg of VPA monotherapy for 6-163 months. Although his studies were methodologically compromised, he noted marked acute and prophylactic antimanic effects, with poor to moderate antidepressant effects. Unfortunately, he did not comment on the degree to which rapid cycling and mixed states predicted response to VPA.

Recently, Lambert and Venaud conducted a novel open prospective longitudinal study by randomly assigning patients for two years to either LI or VPA (Lambert and Venaud 1992). Of 150 patients studied, 121 were bipolar and 29 unipolar. Rapid cyclers, circular continuous cyclers, and patients with mood-incongruent psychotic symptoms were excluded. Episodes frequency was compared between groups and then within groups (the two years prior to study was compared to the two year experimental period). When outcome or tolerance of the treatment was judged poor by the investigator, a switch to the alternative experimental agent was permitted. The main outcome measure was the number of episodes of mania or depression occurring during the follow-up. The number of episodes was 39 (0.51 per patients) in the VPA group and 42 (0.61) in the LI group. Episodes for patients receiving VPA decreased from 4.12 during the two years prior to study to 0.51 during the experimental period. Episodes for LI decreased from 3.92 to 0.61. Therefore, efficacy was similar according to the number of episodes between groups. For both drugs, efficacy was slightly higher in preventing mania than depression. Although VPA was equal in efficacy to LI, at the end of the study there were fewer dropouts in the VPA group (10%) as compared to the LI group (25%). The rates of dropouts were comparable during the first year and then the disparity showed up during the second year of study, particularly towards the end of the study. Patients leaving the study due to inefficacy were the same for the two treatments. Treatment of four subjects initially treated with VPA was changed due to lack of efficacy. Treatment of 10 subjects with LI was changed, and of the 10, four were due to lack of efficacy and six due to poor tolerance. Poor tolerance was evaluated by combining the important and very important side effects, 4% for VPA and 10% for LI. This lower incidence of important or very important side effects in the VPA group was evident at three months and led to VPA never being replaced by LI, while LI was replaced by VPA six times. Further information regarding side effect profiles was not reported. The above literature suggests VPA is superior to PBO and equal in efficacy to LI. Preliminary data suggests VPA is better tolerated than LI.

The safety and efficacy VPA literature was compromised by its open, and sometimes retrospective designs, but distinguished by the large sample sizes evaluated by Lambert and his colleagues. Lambert's more recent use of open naturalistic designs with random assignment to parallel groups was a major advance and will probably usher in their more widespread use. There remains a needs for studies that compare one of the anticonvulsants to LI in the acute and prophylactic treatment of homogeneous cohorts of rapid cyclers and those patients with mixed states (dysphoric mania/hypomania). Until such studies are conducted, the treatment of choice for such variants of illness will remain controversial.

Predictors of Response to Valproate: In addition to studies evaluating outcome in classic bipolar patients, there are now data from 6 published open trials that have assessed the efficacy of VPA in the management of a total 147 rapid cycling bipolar patients (Calabrese et al 1993 b, Emrich et al 1985, Herridge and Pope 1985, Klosiewicz 1985, McElroy et al 1988, Puzynski and Klosiewicz 1985). Seventy-one rapid cyclers were treated with combination therapy and 76 with VPA monotherapy. Marked acute and prophylactic antimanic effects, marked antimixed state effects, and poor to moderate acute and prophylactic antidepressants effects were observed. Of the 147 patients, only 101 (Calabrese et al 1993 b) were a homogeneously patient population evaluated prospectively. The results from prospective studies of the acute and prophylactic efficacy of valproate in the management of bipolar rapid cycling (Calabrese et al 1990) have recently been updated (Calabrese et al 1993 b) with samples adequate in size to allow for predictors of response. One-hundred one bipolar patients were given a trial of VPA in a prospective, longitudinal, naturalistic design. Predictors of good antimanic response included decreasing or stable episode frequencies and non-psychotic mania. Predictors of good antidepressant response included non-psychotic mania worsening over the years of the illness and absence of borderline personality disorder comorbidity. Episode frequency did not predict response. Although there appeared to be a minimum therapeutic concentration of 50-80 ug/ml, there was no correlation between dose, level and response at the upper limits of the therapeutic range. The data suggested that although comorbidity with borderline personality negatively impacted on the management of depression, it did not do so with hypomania/mania and mixed states (Calabrese et al 1993 b). These predictors are consistent with recent data indicating that both decreasing/stable episode frequencies (Post et al 1990) and decreasing mania severity predict good outcome to CBZ (Small et al 1991). Calabrese's initial report of valproate's acute and prophylactic antimixed state efficacy was replicated by Freeman and colleagues with a double blind parallel group comparison of LI to VPA (Freeman et al 1992).

V. Pharmacokinetics and Adverse Effects of Carbamazepine and Valproate

Carbamazepine and VPA possess pharmacokinetics different from LI, and as a result, unique drug-drug interactions occur that affect clinical course and outcome. Three features of the pharmacology of CBZ and VPA incur particular clinical relevance: absorption, protein binding, and metabolism. Since LI is not protein bound and excreted by the kidney unchanged, the last two of these concerns are not relevant to its use. In contrast, the pharmacology of anticonvulsants is somewhat more complicated, and these three pharmacologic concerns frequently have clinical consequence.

The general pharmacokinetic features of orally administered CBZ and VPA are clinically relevant (Table 1). Normally, the side effects associated with anticonvulsants are either related to peak serum levels, i.e. tremors, or the speed and location of absorption, i.e. nausea and stomach cramps. Anticonvulsants are absorbed quickly, completely, and primarily in the stomach and are more likely to cause nausea and stomach cramps. Anticonvulsants like CBZ and either sodium valproate or valproic acid, but not divalproex sodium, have with steep absorption curves are more likely to cause such side effects such as tremors at the upper limits of the therapeutic range. In contrast to VPA, there is only one preparation of CBZ readily available, so one is unable to change preparations in response to toxicity.

The absorption profiles of the different preparations of VPA vary substantially, provide flexibility, and influence the likelihood of such side effects as nausea, stomach cramps, or tremors. For example, the enteric coated preparation of VPA (divalproex sodium) causes nausea or stomach cramping much less frequently than the older preparations, valproic acid or sodium VPA. Divalproex sodium, which is actually the dimer of sodium VPA and valproic acid, reaches peak serum levels in 4-6 hrs and is better tolerated than valproic acid, which is absorbed rapidly (1-3 hrs). The gastric absorption of valproic acid and sodium valproate is rapid and complete. Divalproex sodium is enteric coated and dissolves at the more basic pHs associated with the small intestine, thereby delaying its release and absorption. Occasionally, patients are unable to take VPA because of difficulty swallowing the capsules (valproic acid) or tablets (divalproex sodium); others cannot tolerate the aftertaste associated with the elixir (sodium VPA). The more recently released preparation of divalproex sprinkle is a hard gelatin, pull-apart capsule containing divalproex sodium-coated particles designed to be administered intact or opened to sprinkle on soft foods. Where as VPA absorption with enteric-coated divalproex is delayed, the sprinkle preparation is a smooth release formulation since particles are released over an extended period of time. This minimizes fluctuation in levels between doses and makes it possible to be administered once daily in some patients. This preparation also offers additional flexibility for children and older adults since it can be sprinkled on food for consumption. In addition to these currently available preparations, an intravenous preparation that reaches peak serum level in 10-20 minutes is currently under study.

Most anticonvulsants are protein bound. Protein binding occurs predominantly on serum albumin and is proportional to the albumin concentration. The proportion of unbound, free, or bioactive drug is increased in patients with low levels of circulating albumin (caused by severe malnutrition, liver or renal failure, etc.) Unbound drug is important because it affects drug clearance. When using drugs that have concentration-dependent kinetics, such as VPA, it is important to note that an increase in the amount of unbound drug increases the rate of metabolism and ultimately lowers the blood concentration through what appears to be accelerated metabolism. For this reason, VPA does not have a truly linear relationship between total blood level and dose. The relationship between increasing dose and blood levels breaks down at the upper limits of the therapeutic range. Although blood levels are of critical importance in the dosing of VPA at the lower limits of the therapeutic range, at higher doses total blood levels provide only a rough guide. In the absence of toxicity, higher doses of VPA should be titrated against side effects rather than levels. This is in notable contrast to LI, in which the upper limit of the therapeutic range is well-defined. In other cases, competition for binding sites becomes important when two anticonvulsants with significant protein-binding properties are co-administered. The more highly protein-bound drug, i.e. VPA, increases the free or unbound fraction of less highly protein-bound drug, i.e. CBZ, leading to occult CBZ toxicity.

Carbamazepine is metabolized by a liver microsomal P450 cytochrome oxidase to CBZ-10,11-epoxide. This metabolite is active, stable, and can produce toxic symptoms when significant levels accumulate. Carbamazepine is also metabolized through conjugation with glucuronides and sulfates. The ability of CBZ to autoinduce and heteroinduce drug metabolism complicates its routine clinical use. During the continuation phase of treatment, CBZ autoinduces its single dose T 1/2 of 20-30 hours down to a maintenance dose T 1/2 of 10-15 hours at some poorly defined interval between 3 weeks and 6 months. Not infrequently, this results in a return to subtherapeutic blood levels and cycling. Patients appear to have relapsed from a previously remitted state, and at this time some clinicians incorrectly consider the phenomenon of tolerance. Carbamazepine may also exhibit enzyme induction with many other psychiatric medications, including haloperidol, VPA, and clonazepam, as well as oral contraceptives and most medications that employ P450 cytochrome oxidase as their main route of metabolism. For these reasons, before CBZ is prescribed to a female of child-bearing potential currently taking oral contraceptives, routinely administered low-dose pills should be changed to higher doses. A variety of other drugs, including fluoxetine, erythromycin, calcium channel blockers, cimetidine, and propoxyphene, cause clinically significant increases in CBZ levels and occasionally produce a picture similar to tricyclic antidepressant toxicity.

The metabolism of VPA is unusually complex and many active metabolites are produced. There are two major routes of metabolism of VPA: the P450 microsomal enzyme system and the mitochondrial beta oxidation system. Metabolites are formed through both pathways and it is currently believed that the 2-en-VPA metabolite has antiepileptic properties and 4-en-VPA is a toxin. The more toxic metabolites are produced through the P450 pathway, and it is this pathway that tends to be enhanced by the co-administration of CBZ. Therefore, it can take significantly higher dosages of VPA to produce effective drug levels when the drug is prescribed in combination therapy. Conversely, VPA is known to inhibit the metabolism of 10,11-epoxide CBZ. This metabolite is active, not usually included in blood monitoring, and may cause occult toxicity. With the exception of the co-administration of CBZ, VPA's drug-drug interactions rarely result in clinically significant problems in the management of bipolar patients.

The co-administration of CBZ and VPA in the treatment of patients with bipolar disorder occasionally leads to a series of drug-drug interactions that include the heteroinduction of VPA metabolism by CBZ, the displacement of protein-bound CBZ by VPA, and the inhibition of the metabolism of the primary metabolite of CBZ (10,11-epoxide CBZ) by VPA. When VPA is added to CBZ, anticipate increased CBZ effect, as well as the eventual reduction in T 1/2 of VPA. When CBZ is added to VPA, anticipate decreasing VPA levels.

Although the side effect profile of CBZ and most other anticonvulsants can be loosely categorized as dose related, non-dose related, and idiosyncratic, one occasionally sees dose related side effects that do not subside until the dose is substantially decreased or the drug discontinued, ie. CBZ-related fatigue and VPA-related hair thinning or stomach cramping. The dose related side effects of CBZ are benign and include leukopenia (15%), lethargy, double vision, cognitive changes, ­ liver functions tests, hyponatremia. The non-dose related side effects include nausea/vomiting and fluid retention. The idiosyncratic side effects include aplastic anemia (seen in less than 1 in 125,000), hepatic failure, Stevens-Johnson syndrome, rash (15%), and fetal drug effects. Nausea, benign leukopenia, fatigue, tremors, weight gain, cognitive changes can be ameliorated by a decrease in dose, but occasionally beta blockers are required for treatment of tremors. Fetal drug effects can be minimized by use of folic acid supplements during pregnancy.

The dose related side effects of VPA are benign and include nausea, vomiting, stomach cramps, increased liver functions tests, lethargy, hair thinning, benign thrombocytopenia. Non-dose related side effects include weight gain. Idiosyncratic side effects include hepatic failure (seen in less than 1 in 118,000), pancreatitis, agranulocytosis, fetal drug effects. Nausea, vomiting, or stomach cramps can be managed by decreased dose, using divalproex sodium rather than valproic acid or sodium VPA, histamine type 2 blockers, or if necessary switching to the sprinkle preparation. Tremors can be managed by decreasing dose or use of beta blockers. Fatigue can be managed by decreasing dose. Hair thinning can be managed by decreasing the dose or zinc and selenium supplements. The benign thrombocytopenia and the benign increase in liver function tests can be managed by decreasing dose. Fetal drug effects of VPA can also be minimized by use of folic acid supplements during pregnancy.

VI. LAMOTRIGINE

Lamotrigine is an antiepileptic drug of the phenyltriazine class which has been shown to be effective as add-on treatment and monotherapy treatment of partial seizures (Schachter et al 1995). In 14 clinical reports involving 207 patients with bipolar disorder (66 with rapid cycling), lamotrigine was observed to possess moderate to marked efficacy in depression, hypomania, and mixed states; efficacy in hospitalized mania was unclear (see Calabrese et al 1998 for a review).

In the largest of these clinical reports, the spectrum of activity of lamotrigine was examined in a 48-week, open-label, prospective trial of lamotrigine in 75 patients with either bipolar I or II disorder (Calabrese et al In Press a). Lamotrigine was used as add-on therapy (n = 60) or monotherapy (n = 15) in patients presenting in depressed, hypomanic, manic, or mixed states. Of the 40 depressed patients included in the analysis, 48% exhibited a marked response and 20% a moderate response as measured by reductions in the 17-item Hamilton Depression Scale (HAM-D) scores. Of the 31 presenting hypomanic, manic, or mixed, 81% displayed a marked response and 3% a moderate response on the Mania Rating Scale (MRS). The magnitude of overall observed improvement was substantial with the depressed patients exhibiting a 42% decrease in HAM-D scores from baseline to endpoint, and the patients presenting hypomanic/manic/mixed exhibiting a 74% decrease in MRS scores from baseline to endpoint. The most common drug-related adverse events included dizziness, tremor, somnolence, headache, nausea, and rash. Rash was the most common adverse event resulting in drug discontinuation (9% of patients). One patient developed a serious rash which required hospitalization. Lamotrigine appeared to exhibit a broad spectrum of efficacy in the management of the depressed, hypomanic, manic, and mixed phases of bipolar disorder.

Acute Bipolar Depression: The findings of these clinical reports have been replicated by two recent controlled studies. A study was conducted to evaluate the efficacy and safety of two doses of lamotrigine compared with placebo in the treatment of a major depressive episode in patients with bipolar I disorder (Calabrese et al, In Press b). Outpatients with bipolar I disorder experiencing a major depressive episode ((n = 195) received lamotrigine (50 or 200 mg/day) or placebo as monotherapy for 7 weeks. Psychiatric evaluations, including the Hamilton Rating Scale for Depression (HAMD), the Montgomery-Asberg Depression Rating Scale (MADRS), the Mania Rating Scale (MRS), and the Clinical Global Impressions Scale for Severity (CGI-S) and Improvement (CGI-I), were completed weekly. Lamotrigine 200 mg/day demonstrated significant antidepressant efficacy on HAMD-17, HAMD Item 1, MADRS, CGI-S, and CGI-I compared with placebo. Improvements were seen as early as week 3. Lamotrigine 50 mg/day also demonstrated efficacy compared with placebo on several measures. The proportion of patients exhibiting a marked response on CGI-I was 51%, 41%, and 26% for lamotrigine 200 mg/day, lamotrigine 50 mg/day, and placebo groups, respectively. Adverse events and other safety results were similar across treatment groups (including switch rates), except for the higher rate of headaches in the lamotrigine groups. The conclusion was that lamotrigine monotherapy was an effective and well-tolerated treatment for bipolar depression.

In a preliminary NIMH communication which compared the efficacy of gabapentin, lamotrigine, and placebo in a double-blind cross-over design, Frye and colleagues (Frye et al 1998) reported that lamotrigine, but not gabapentin, was superior to placebo in mixed cohort of patients with bipolar I and II disorder, as well as recurrent major depression.

Pharmacokinetics and Adverse Effects of Lamotrigine: Lamotrigine has no significant effect on the hepatic cyclo-oxygenase system or on the metabolism of other antiepileptic drugs. It is moderately bound to plasma proteins (55%), has no active metabolites, and has no effect on the pharmacokinetics of oral contraceptives (see Table 1). Routine monitoring of chemistries, complete blood counts, blood pressure, and heart rate is not necessary. Body weight remains relatively stable with chronic lamotrigine use. In the management of epilepsy, there does not appear to be a correlation between blood level and efficacy. Valproate inhibits the metabolism of lamotrigine by immediately and significantly competing for metabolism through glucuronidation. As a result of this, the mean half-life of lamotrigine is immediately increased to about 70 hours and steady state lamotrigine plasma concentrations are increased. Therefore, when lamotrigine is added to ongoing treatment with valproate, the initial dose of lamotrigine should be much lower than the monotherapy dose. When valproate is added to ongoing treatment with lamotrigine, the daily dose of lamotrigine should be immediately decreased by approximately one-half. Carbamazepine induces the metabolism of lamotrigine by gradually decreasing the mean half-life of lamotrigine's half-life to about 12 hours and reducing steady state lamotrigine plasma concentrations. Therefore, when lamotrigine is added to ongoing treatment with carbamazepine, the initial dose of lamotrigine is higher than the monotherapy dose of lamotrigine. When carbamazepine is added to ongoing treatment with lamotrigine, the dose of lamotrigine must be gradually increased by 100%.

The most common side effects associated with the use of lamotrigine in bipolar disorder include dizziness, tremor, somnolence, headache, nausea, and rash (Calabrese et al In Press a). Rash is the most common adverse event resulting in drug discontinuation. The nature of the reported adverse events in bipolar disorder are consistent with those observed in patients with epilepsy. In open-label and placebo-controlled epilepsy clinical trials involving 3,501 patients, rash has been observed to occur in 10% of lamotrigine and 5% of placebo patients. Rash led to drug discontinuation in 3.8% of lamotrigine patients and hospitalization in 0.3% (Physician's Desk Reference). The spectrum of lamotrigine-associated cutaneous reactions include simple morbilliform (measle-like) rash, hives, angioedema, Stevens-Johnson syndrome, toxic epidermal necrolysis, and hypersensitivity reaction syndrome. The rash typically occurs within 2-8 weeks of initiation of treatment. The incidence of rash is higher in patients taking concomitant valproate, and higher if the recommended initial lamotrigine dose and titration schedules are exceeded. The prevalence of serious rash resulting in hospitalization is significantly higher (approximately 1 in 100) in children (< 16 years of age) than in adults. The consensus of clinical opinion is that, unless an alternative etiology can be clearly identified, the drug should be discontinued when patients present with rash of any kind regardless of severity, since the clinician is unable to predict which cutaneous reactions will become serious. Strict adherence to the recommended dose escalation schedule may diminish the likelihood of rash (Physicians Desk Reference, 1997). Lamotrigine is a Category C teratogen which indicates that studies in women and animals are not available.

VII. OTHER ANTICONVULSANTS

Recently, a number of clinical reports have been published describing a total of 216 patients which appear to suggest that gabapentin may have mood stabilizing properties in treatment-refractory bipolar patients, and for this reason. Gabapentin does not bind to plasma proteins, is not appreciably metabolized, does not induce hepatic enzyme activity, and does not appear to alter the pharmacokinetics of commonly used anticonvulsant drugs or oral contraceptives (See Table 1). In addition, the pharmacokinetics of gabapentin are not altered substantially by concomitant administration of others anticonvulsants drugs. As is evident, gabapentin has unique pharmacokinetic properties. Like lithium, it is not appreciably metabolized by the liver. If the double-blind studies document clinical efficacy in bipolar disorder, its clinical use will be substantially less complicated that currently available mood stabilizers (see Ferrier and Calabrese in press for a review).

Topiramate (TPM), a sulfamate-substituted monosaccharide, is a new anticonvulsant with a broad spectrum of activity in epilepsy (Langtry et al 1997) (see Table 1 for pharmacokinetic profile). Topiramate, carbamazepine, and valproate appear to share some similar putative mechanisms of action. Like carbamazepine and valproate, TPM inhibits voltage-activated Na+ channels; like valproate, TPM enhances GABA-mediated neuroinhibition. Topiramate also blocks glutamate activity at non-NMDA receptors and inhibits some of the isoenzymes of carbonic anhydrase (reference), although the latter is not believed to be an important contributor to the drug's anticonvulsant activity. Since TPM possesses these similarities to carbamazepine and valproate, two anticonvulsants with efficacy in bipolar disorder, preliminary open label studies have been undertaken in treatment refractory cohorts of patients with bipolar disorder. In a total of 3 preliminary communications involving a total of 87 patients with bipolar disorder (Calabrese JR et al personal communication, McElroy SE et al personal communications, Marcotte D et al personal communications). This drug may have antimanic and some mood stabilizing effects in some patients. Side effects include paresthesias, nausea, constipation, decreased appetite/weight loss, dizziness, tinnitus, and cognitive changes. The ability of TPM to produce weight loss appears to be directly correlated with pre-treatment body weight.

VIII. Combination Mood Stabilizer Therapy

Although there exists numerous anecdotal references highlighting the necessity of combination mood stabilizer therapy in the clinical management of bipolar disorder, there are few studies that have compared the efficacy of combination therapy to monotherapy. One such study compared the prophylactic efficacy of LI, CBZ, and the combination (Denicoff et al 1997a). These investigators randomly assigned 52 patients with bipolar disorder in a double blind design for an intended one year of treatment with LI or CBZ, a crossover to the opposite drug in the second year, and then a third year on the combination. Their results suggested that LI was more effective than CBZ in the prophylaxis of mania but not depression, and the combination of LI and CBZ was better than either monotherapy.

A subgroup of patients from this study who were non-responsive or intolerant to the combination of and CBZ were given the option of entering an additional year of treatment (Denicoff et al 1997b). Of the 18 evaluable patients, six (33%) had moderate to marked responses to VPA plus LI; four of these six had not responded to any previous treatment conditions. When CBZ was added to the combination of LI and VPA for seven patients, three responded to the triple therapy. The authors concluded that some patients with bipolar disorder refractory to LI and CBZ receive prophylactic benefit from VPA when used with LI.

Despite the absence of controlled studies employing random assignment to parallel groups, recently published expert consensus surveys indicate that combination mood stabilizer therapy is commonly employed in the pharmacotherapy of treatment refractory patients with bipolar disorder (Kusumakar et al 1997, Frances et al 1996).

IX. METHODOLOGIC ISSUES

Although one can work towards treatment with bimodal monotherapy, it is clear that many patients require combination therapy during acute episodes of mania or depression. In fact, during the course of the illness most patients will eventually require adjunctive or augmentation pharmacotherapy with either other bimodal mood stabilizers or benzodiazepines, antidepressants, and/or antipsychotic agents. Informed combination therapy (polypharmacy) is routine and the "state of the art" in longitudinal management of this illness. Despite the above and the pre-existing literature, the Food and Drug Administration requires that studies be designed to evaluate monotherapies against PBO so as to gauge study reliability. Accordingly, the design and completion of studies comparing different acute combination therapy paradigms is normally preceded by acute monotherapy designs with PBO arms. This practice has slowed the development of pharmacotherapy for bipolar disorder.

In addition to the above, it has become increasingly clear that major mood disorders should be viewed longitudinally rather than solely in terms of the acute episode. Unfortunately, the rigor of the scientific literature evaluating the prophylactic pharmacotherapy of bipolar disorder is seriously encumbered by the 3 mood states that accompany this illness and multiple patterns of presentation. Due to the illness's inherent complexity, the methods are not available that allow controlled acute and prophylaxis studies to simultaneously focus on both antidepressant, antimanic, and antimixed prophylactic efficacy with similar degrees of rigor. Naturalistic longitudinal studies more adequately address this need. Optimally, these prospective naturalistic studies, such as the one recently published by Lambert and Venaud, might openly randomize patients to different treatment groups and then allow for either crossovers or augmentation for nonresponders. If these designs are not employed, the remaining patient population at the end of maintenance studies will be so small and atypical, as to not have general relevance to the pool of patients normally treated in the general population. Studies involving rapid cycling and mixed states (dysphoric mania/hypomania, mixed mania) are particularly challenging because these phenomena appears to migrate unpredictably throughout the natural course of the illness.

Study designs employing PBO arms are feasible for research subjects who are hospitalized for acute studies, but not always practical for evaluations of maintenance therapies. Placebo response rates in recent acute hospitalized mania studies have now been noted to range from 10.5% to 45%. The lowest placebo-response rate (10.5%) was observed in a study that skewed enrollment towards lithium non-responders (Pope et al 1991), whereas intermediate response rates of 21% (Bowden et al 1994) and 24% (Tohen et al 1998) have been observed in studies enrolling unselected populations of patients, and more recently, 45% for a study which employed an augmentation design (Muller-Oerlinghausen et al 1998). A survey of response rates in early double-blind PBO controlled LI maintenance studies in classic bipolar patients has been conducted. Whereas 55-100% of patients given PBO relapse over 5 months to 2 years, 0 to 49% of those given LI relapse (Davis 1976). The data of Schou suggest the "half-life" of the untreated patient in remission is approximately 4 1/2 months. More recently, and consistent with this finding, a metanalysis of 14 maintenance studies involving 257 patients with bipolar type I illness reported a 50% mania relapse rate after five months when LI was rapidly discontinued (Suppes et al 1991). Given the remarkably high relapse rate for classic bipolar patients given PBO during these early maintenance studies and the human cost in suicide attempts documented in these studies, it was not clear that the additional methodologic rigor obtained from a PBO arm merited the additional risk. However, a recent multicenter maintenance study (Bowden et al 1997) in bipolar I disorder reported a high PBO response rate which highlighted the importance of PBO arms in studies which use recent nomenclatures and employ random assignment to double blind parallel treatment groups.

The designs employed in bipolar maintenance studies have evolved greatly over the last 28 years. Consequently, there has been minimal or no consensus set for methods used to demonstrate the ability of a new putative mood stabilizer to prevent relapse and recurrence in bipolar disorder. The methods that have evolved the most include enrollment procedures, randomization schemes, use of outcome measures, statistical analyses, and country-specific commercial/regulatory issues. Until recently, controlled trials of patients with bipolar disorder has been viewed as labor-intensive, expensive, not practical, and generally not funded by the National Institute of Mental Health or the pharmaceutical industry. With the recent completion of the first placebo-controlled bipolar maintenance trial in over a quarter of a century (Bowden et al 1997), this impression has begun to change.

published 2000