Antiepileptic Drugs in Mood-Disordered Patients


  • The first and second authors contributed equally to the article.

Address correspondence and reprint requests to Dr. H. Grunze at Department of Psychiatry, LMU Munich, Nussbaumstr. 7, 80336 Munich, Germany. E-mail:


Summary:  Bipolar disorder is a common, recurrent, often severe mental disorder that, without adequate treatment, is associated with high rates of morbidity and mortality. We review the evidence on the efficacy of a spectrum of antiepileptic drugs (AED) in bipolar disorder. Most studies have been carried out with carbamazepine (CBZ), valproate (VPA), and lamotrigine (LTG). All three of these AEDs have been shown to be of value in the management of patients with bipolar illnesses. VPA and CBZ seem to exert stronger antimanic effects and, to a lesser degree, acute antidepressant efficacy. LTG seems to be effective against depression and mania, with a more robust activity against depression. No firm evidence supports a role for vigabatrin, tiagabine, topiramate, or levetiracetam in these disorders.

Mood changes are a common symptom in epilepsy patients, and effects on mood have been observed for a long time during treatment with antiepileptic drugs (AEDs). Whereas negative effects on mood have been described quite frequently with the first generation of AEDs, especially bromides and then later barbiturates, it was in the 1960s when the first observations were published describing positive and mood-stabilizing effects of AEDs. In 1967, Turner (1) published an observational study entitled, “The usefulness of diphenylhydantoin in treatment of non-epileptic emotional disorder,” separating for the first time the mood effects of AEDs from their antiepileptic efficacy. At the same time or soon afterward, the first reports on the mood-stabilizing efficacy of carbamazepine (CBZ) (2) and valpromide (3) were published. Since the 1990s, an increasing tendency to test every new AED not only for antiepileptic potencies, but also for mood-stabilizing efficacy, has been observed. Now the portfolio of AEDs with proven or potential usefulness in treating bipolar patients is quite respectable. We provide a condensed overview on the efficacy of older and new AEDs most frequently used in bipolar patients.


CBZ has been licensed since the 1980s in several countries as an alternative treatment in those bipolar patients either refractory to lithium treatment or having the side effects of lithium. Surprisingly, the antimanic efficacy of CBZ has only just been demonstrated in a placebo (PBO)-controlled and methodologically unambiguous study (4), and until this time, the antimanic effect of CBZ had been deduced only from comparator trials against neuroleptics (5–8), lithium (9,10) or valproate (VPA) (11). In prophylactic treatment, the claim of CBZ efficacy also is largely based on comparator trials against lithium. In the so-called MAP study, Greil et al. (12) compared the prophylactic efficacy of CBZ against lithium in a randomized but not blinded study lasting for 2.5 years. Lithium was numerically better than CBZ with regard to the rate of rehospitalization and recurrence of mood episodes but reached statistical significance only in the combined outcome of recurrence and use of concomitant medications (13). In a more detailed analysis, however, Greil et al. (14) found that the favorable effect of lithium was only true for bipolar disorder with a classic course. As soon as atypical features such as rapid cycling, dysphoric or psychotic mania, or comorbidities complicated the illness, CBZ was numerically better than lithium.

The relatively better efficacy of lithium in the long-term course of the disorder also was the major finding of a recent study comparing CBZ and lithium in the prophylactic treatment over a 2-year period (15).Whereas relapses under lithium occurred almost exclusively during the first 3 months after randomization, patients taking CBZ carried a constant risk of a new episode of ∼40% per year, suggesting different mechanisms of action of these two substances. This also may support a biologic rationale for why combination treatment with lithium and CBZ together seems to outperform either treatment alone, especially in rapid-cycling patients, as demonstrated in a 1-year, double-blind study by Denicoff et al. (16).


Acute antimanic efficacy of VPA, used mostly as a loading treatment, has been demonstrated in several controlled studies (for an overview, see refs 17–19). In a three-arm study comparing VPA, lithium and PBO by Bowden et al. (20), both VPA and lithium were significantly better than PBO treatment, whereas no difference was found between these two agents. Subanalysis of patients, however, revealed that VPA appeared to have a broader spectrum of efficacy, whereas lithium was especially effective in classic manic patients but not in mixed patients (21). VPA also showed efficacy in patients in whom depressive symptoms or a rapid-cycling course complicated the illness.

A more recent study (22) compared VPA with PBO as add-on medication to standard neuroleptics (haloperidol or perazine) for 3 weeks. The total amount of neuroleptic comedication was significantly less from week 2 onward in patients treated with VPA. With regard to the antimanic efficacy, the combination of a neuroleptic and VPA also was significantly better than the neuroleptic treatment alone. Thus, VPA appears to be not only effective in monotherapy but also in combination and, when prescribed, may allow reducing concomitant neuroleptic medication. Further comparator studies against haloperidol (23) or against CBZ (11) also underlined the antimanic efficacy of VPA, which appears to be equal to lithium and classic neuroleptics and to be slightly better than CBZ.

Two recent studies comparing VPA with olanzapine produced inconsistent results. The study of Tohen et al. (24) found olanzapine to be superior to VPA, while equal efficacy was seen in the study of Zajecka et al. (25).

Three controlled studies examined the prophylactic efficacy of VPA. In the pivotal study of Bowden, VPA was compared with PBO and lithium in a 1-year randomized double-blind study (26). VPA was only numerically better than lithium and PBO, whereas no difference was found between the latter two. The negative outcome of the study was mainly due to an unusually high PBO response, with almost 40% of PBO patients staying healthy for the study period (compared, for example, with <20% in the 1-year olanzapine maintenance study). Some secondary outcome parameters, however, were significantly in favor of VPA. In comparison to patients taking PBO, VPA patients were more likely to stay relapse free for 1 year, maintaining complete response without subsyndromal symptoms, and were less likely to have a new depressed episode. In another subanalysis, it was found that those patients who had a previous response to VPA in acute mania also were significantly more likely to stay episode free when randomized to VPA compared with being randomized to lithium or PBO. Obviously, acute responsiveness to VPA is a potential predictor for long-term prophylactic efficacy. Another study comparing VPA and lithium for 1 year in a head-to-head-comparison (27) focused mainly on pharmacoeconomic parameters, but the secondary outcome parameters for efficacy also were supportive of prophylactic efficacy of VPA. Finally, VPA was compared with olanzapine in an extension of the already cited 3-week study in acute mania (28). During the additional 44-week follow-up, maintenance of the antimanic response was not statistically significant different for both substances. However, when it comes to prophylactic treatment of rapid-cycling patients, the efficacy of VPA appears unsatisfactory. From open studies, a good response of rapid-cycling patients had been assumed for acute manic and mixed episodes and also for the prophylaxis of both manic and depressive episodes (29). In a randomized double-blind study, conducted by Calabrese et al. (30), recently manic or hypomanic patients were started on a combination treatment with lithium and VPA, while other psychotropics were tapered off. When patients were stable with this combination treatment for 6 months, they were randomized to a 20-month double-blind continuation with either VPA or lithium. However, fewer than a fourth of patients reached randomization criteria. Poor compliance, adverse events, and nonresponsiveness, especially refractory depression, were the main reasons for dropping out. With only 60 patients remaining in the study, it was not possible to show a significant advantage of any substance. VPA was numerically better, with 50% of patients still remaining in the study after 45 weeks, compared with 18 weeks with lithium.

In conclusion, the older AEDs, CBZ and VPA, have an established role in mood stabilization in bipolar patients. More recently, phenytoin (PHT) also has shown both acute antimanic (31) and prophylactic (32) efficacy in two controlled studies. However, especially CBZ and, to a lesser degree, VPA, have side effects that may limit the compliance of patients. Thus it is important to develop more efficacious and better-tolerated alternatives for mood stabilization.


Treatment with carbamazepine (CBZ) may be limited because of the side effects induced by the 10,11-epoxide, by cytochrome induction, and by subsequent interaction with other psychotropic drugs. Oxcarbazepine (OXC), in contrast, is better tolerated and shows less interaction with other medications. Two double-blind, but otherwise not very well controlled, studies demonstrated comparable efficacy of OXC and haloperidol and lithium, respectively, in acute mania (33). However, more rigorously controlled trials according to modern standards are still missing. In a recent prospective open trial of OXC monotherapy in acute mania (34), it appeared that the drug was useful only in mild to moderate, but not in severe mania. No clear evidence exists for antidepressant properties or prophylactic properties of OXC in bipolar patients.


During the development of lamotrigine (LTG), the drug was observed to improve mood, alertness, and social interaction in some epilepsy patients (35). These early observations prompted interest in the evaluation of LTG in the treatment of mood disorders.

In 16 open clinical reports, LTG showed moderate to marked efficacy in depression, hypomania, and mixed states; efficacy in more severe and/or hospitalized mania was unclear. A series of multicenter, PBO-controlled studies have been completed, designed to evaluate the efficacy and dose–response relation of LTG in both acute and maintenance treatment of bipolar illness (Table 1).

Table 1. Multicenter trials of lamotrigine in mood disorders
TrialMood stateaStudy designbTreatment armsDurationOutcome
  1. BP, bipolar disorder; CGI, Clinical Global Impressions scale; HAM-D, Hamilton Rating Scale for Depression; LI, lithium; LTG, lamotrigine; MADRS, Montgomery-Asberg Depression Rating Scale; NS, no significant difference observed; PBO, placebo

  2. aMood states were DSM-IV diagnosed.

  3. bStudy 601 was a stand-alone open-label trial. All other studies were controlled trials (N = 1,198).

  4. c,d,e,f,g,h,i GlaxoSmithKine, unpublished studies.

601BPI and II hypomania, mixed episode, or depressionOpenLTG add-on (N = 75)4868–84% responded
60236BPI depressionAcute monotherapyLTG, 200 mg/d (N = 63,
LTG, 50 mg/d (N = 66),
PBO (N +65)
603cBPI and II depressionAcute monotherapyLTG, 100–300 mg/d10
(N = 103), PBO (N = 101)
60538BPI depressionControlled monotherapy continuation after open stabilizationLTG, 50–400 mg/d (N = 219)76LTG>PBO, LI>PBO
LI (N = 120),
PBO (N = 121)
60637BPI mania index episodeControlled monotherapy continuation after open stabilizationLTG, 100–400 mg/d76LTG>PBO, LI>PBO
(N = 59), LI (N = 46),
PBO (N = 70)
609dBPI maniaAcute, monotherapyLTG, 50 mg/d (N = 84),
LI (N = 36),
PBO (N = 95)
610eBPI maniaAcute, add-onLTG, 200 mg/d (N = 74)
Li (N = 78)
611fBPI and II rapid cyclingProphylaxis add-onLTG, 100–500 mg/d
(N +68), PBO (N = 69)
613gUnipolar acute, depressionMonotherapyLTG, 200 mg/d (N = 145),
desipramine (N = 147),
PBO, (N = 145)
 8HAM-D: NS,
614BPII rapid cyclingControlled monotherapy continuation after open stabilizationLTG, 100–300 mg/d
(N = 92), PBO (N = 88)
20022hUnipolar depressionAcute, monotherapyLTG, 200 mg/d (N = 74),
PBO (N = 75)
20025iUnipolar depressionAcute, monotherapyLTG, 200 mg/d (N = 151)
PBO (N = 150)
Total LTG (N = 1,273),
LI (N = 280),
Desipramine (N = 147),
PBO (N = 1056)

The first study (36) in this series was conducted to evaluate efficacy and safety of two dosages of LTG monotherapy (50 and 200 mg) compared with PBO in treatment of depressive episodes in patients with Bipolar I disorder. Patients were randomly assigned to treatment with target dose of either LTG, 50 mg/day (n = 66); LTG, 200 mg/day (n = 63); or PBO (n = 66). The LTG dose was escalated according to predetermined schedule. Chloral hydrate, lorazepam, temazepam, or oxazepam were used as needed to control agitation, insomnia, or hostile behavior during the first 3 weeks of the treatment.

Psychiatric evaluations included Hamilton Rating Scale for Depression (HAM-D), Montgomery-Åsberg Depression Rating Scale (MADRS), Mania Rating Scale (MRS), and the Severity of Illness and Improvement subscales of Clinical Global Impression Scale (CGI-S and CGI-I, respectively) and were completed at 4 days, and then weekly.

LTG at dosage of 200 mg/day demonstrated significant efficacy on HAM-D item1, MADRS, CGI-S, and CGI-I compared with PBO. Improvements were seen as early as week 3, when subjects were taking only 50 mg daily. LTG at a dosage of 50 mg/day also demonstrated efficacy compared with PBO on several measures. The proportion of patients exhibiting a marked response on the CGI-I was 51%, 41%, and 26% for LTG, 200 mg, 50 mg, and PBO, respectively. The rate of switching into hypomania, mania, or mixed states was not significantly different between LTG (4.6–5.4%) and PBO (5%) without concurrent psychotropic medication. The only side effect that was significantly more common in patients randomized to LTG compared with PBO was headaches; no significant differences were noted in the prevalence of benign rash, and no serious rashes were observed. Of 65 patients randomized to PBO, 3% cycled into hypomania, none into mania, and <1% into mixed states. Of 129 patients randomized to LTG, <1% cycled into hypomania, 3% into mania, and <1% into mixed states. These data suggest that LTG monotherapy is an efficacious and well-tolerated treatment for bipolar depression, and that the switch rate associated with the use of LTG does not exceed that typically seen with the natural course of illness.

Two further studies (37,38)—both of which were paired, randomized, double-blind, parallel-group, PBO-controlled multicenter studies—comprised a screening phase of ≤2 weeks; an 8- to 16-week open-label phase, during which LTG was initiated as monotherapy and other psychoactive drugs were tapered and discontinued; and an 18-month double-blind phase. During the double-blind phase, patients received LTG, lithium, or PBO as maintenance therapy. The primary efficacy variable in both studies was time to intervention for any mood episode. One of the studies (37) used the conventional maintenance design and limited enrollment to patients who were currently or recently manic, hypomanic, or had mixed states. The other study (38) used an innovative design by limiting enrollment to patients who were currently or recently depressed. The findings in both studies are concordant and showed that both lithium and LTG were significantly superior to PBO on time to intervention for any mood episode (p = 0.018 for previously manic; p = 0.030 for previously depressed). LTG-treated patients showed longer survival-in-study times than controls (p = 0.030 for the previously manic and p = 0.003 for the previously depressed). Lithium-treated formerly depressed patients also showed longer survival-in-study times (p = 0.022), and a nonsignificant trend was seen for lithium-treated formerly manic patients to show this as well (p = 0.070). LTG, but not lithium, was superior to PBO at prolonging the time to a depressive episode (p = 0.015 for previously manic; p = 0.047 for previously depressed). Lithium, but not LTG, was superior to PBO at prolonging the time to manic, hypomanic, or mixed episodes (p = 0.006 for previously manic; p = 0.026 for previously depressed). No evidence in these studies was seen of worsening mania or depressive symptoms or of accelerated cycling frequency during LTG treatment.

As a result of the data generated from these two pivotal maintenance trials, LTG was approved in the United States by the Food and Drug Administration (FDA) in 2003 for use as a maintenance therapy in patients with bipolar I disorder who have received standard treatment for acute mood episodes. More than 30 countries worldwide have approved LTG for use in the prevention of depressive episodes in bipolar disorder.

Calabrese et al. (39) conducted a large study consisting of two phases: an open-label, 8- to 12-week preliminary phase and a 26-week randomized phase. Enrolled patients had a diagnosis of bipolar disorder I or II with rapid cycling, as defined by Diagnostic and Statistical Manual of Mental Disorders, fourth edition (40). Subjects who met criteria for mood stabilization at the end of the preliminary phase were randomly assigned in a 1:1 ratio to treatment with LTG or matching PBO in a double-blind fashion. This design allowed the investigators to compare the efficacy of LTG monotherapy with that of PBO in preventing mood episodes in patients with rapid-cycling bipolar disorder and to evaluate the safety of LTG in this population. Three hundred twenty-six patients were enrolled in the preliminary phase, and 182 were randomized to receive PBO (n = 89) or LTG (n = 93). The LTG dosage was as high as 500 mg/day, with a mean dose of 216 mg/day. Both arms in the treatment group did not show any statistical difference in time to additional pharmacotherapy for a developing mood episode. A significant difference was noted in the overall survival (time to drop out for any reason) (p = 0.036). Analysis also indicated a 6.7-week difference in median survival significantly favoring LTG. During the 6-month randomized phase, 41% of patients taking LTG, compared with 26% taking PBO (p = 0.028) were stable and did not experience any mood episode. The results suggested that LTG is useful for prevention of mood episodes in patients with rapid-cycling bipolar disorder.

Frye et al. (41) conducted a controlled comparative study to evaluate the efficacy of LTG and gabapentin (GBP) monotherapy versus PBO for refractory mood disorders. The results showed a marked antidepressant response in 45% of LTG-treated patients compared with 19% in the PBO group. GBP did not differ from PBO.

The demographic profile of the study participants evaluated in all three treatment phases (N = 31) included 18 women and 13 men; 11 bipolar I, and 14 bipolar II; 23 rapid cycling and two non–rapid cycling; and six unipolar patients. This study first randomized patients to parallel double-blind treatment with GBP monotherapy, LTG monotherapy, or PBO.

After the initial study period, the study used a crossover design in which the remaining patients were then crossed over to alternative treatments in two subsequent crossover periods. A 1-week crossover period was implemented between phases. Patients were treated only with the study medications except for the following: four patients who continued receiving stable levothyroxine supplementation for corrected primary hypothyroidism; one patient who continued receiving diuretic therapy for essential hypertension; and two patients, each of whom continued receiving stable prior medications (i.e., triiodothyronine and clonazepam). LTG was started at a dose of 25 mg daily for the first week before titration up to 300–500 mg daily in the fifth through sixth weeks. The initial GBP dose was 900 mg daily, which was titrated to 4,800 mg by the fifth through sixth weeks. After completion of all three phases of the study, patients who had responded to a particular phase (I, II, or III) were offered the option of returning to that phase, still on a blinded basis, for response confirmation.

Overall improvement was assessed with the CGI modified for bipolar illness (CGI-BP). Other ratings used in the evaluation included prospective self- and observer-rated life charting, Hamilton Depression Inventory, Young Mania Rating Scale, Spielberger State Anxiety Scale, and Brief Psychiatric Rating Scale. CGI-BP change determinations were made by a consensus of blinded research physicians and clinicians, both in comparison with the previous phase of illness and the worst phase of documented illness. Prior treatment exposure and documented treatment failures, including therapeutic level with inadequate response, clinical intolerance, or affective relapse, were the following: for lithium, 28 (90%) of 31 patients experienced prior exposure and 28 (100%) of 28 patients experienced prior treatment failure; for VPA, 26 (79%) of 31 and 21 (81%) of 26; and for CBZ, 20 (58%) of 31 and 14 (70%) of 20, respectively.

The mean daily doses at week 6 were 274 ± 128 mg for LTG and 3,987 ± 856 mg for GBP. No difference was found in LTG and GBP doses between responders and nonresponders. The response rates based on the overall CGI rating of much or very much improved were the following: LTG, 52% (16 of 31); GBP, 26% (eight of 31), and PBO, 23% (seven of 31). Only a trend level of significance was reached for the CGI response rates for separate manic and depressive components of the illness.

The response rates for mania were the following: LTG, 44% (11 of 25); GBP, 20% (five of 25); and PBO, 32% (eight of 25). For depression, the response rates were the following: LTG, 45% (14 of 31); GBP, 26% (eight of 31); and PBO, 19% (six of 31).

The response rate observed during just phase I was highly similar to that for the whole study using all three phases of the crossover trial: 50% (five of 10) for LTG, 33% (three of nine) for GBP, and 18% (two of 11) for PBO. In addition, when the response data were analyzed as a function of a positive response in the preceding phase, only 23% of LTG responders, 50% of GBP responders, and none of PBO responders were also partial responders in the previous phase, and would thus have entered the next phase somewhat improved. This finding further indicates that a greater percentage of LTG-responsive patients were responders in the preceding phase. Moreover, on supplementary ratings of illness severity, no significant differences were seen at baseline.

No differences were noted between the response rates based on gender. Both agents were generally well tolerated, with the exception of one patient, in whom a rash developed with LTG after the 6-week study phase was over. This rash occurred in week 15 (during the continuation treatment), progressed to toxic epidermal necrolysis, and required the patient to be hospitalized in an intensive care burn unit. Although this patient recovered fully, more recent data suggest that this study used a titration schedule that unnecessarily targeted unusually high doses of LTG.

A pair-wise comparison (LTG vs. GBP: F= 5.884; p = 0.021) showed that patients lost weight when they received LTG relative to the weight gained when they received GBP.

This study suggests that LTG monotherapy is superior to both GBP and PBO treatments in patients with refractory mood disorders. Although crossover designs have the potential for producing carryover and other confounding effects, this did not appear to occur in this analysis and did not affect the interpretation of the outcome data based on a variety of considerations. These data suggest that the response rates were not confounded by the treatment phase, nor did carryover effects on the severity of illness influence them.


GBP, which has shown efficacy in anxiety disorders (42,43), was tested in a double-blind randomized study against PBO for antimanic efficacy (44). However, the outcome of the study was negative. As the previously cited study of Frye et al. (41) was also not supportive of GBP treatment, no further rigorously controlled studies were conducted.

Levetiracetam (LEV) is now an established treatment in epilepsy. The first open studies of LEV suggest moderate antimanic efficacy (45), but controlled studies are still to be conducted.

In an open study in acutely manic, acutely depressed, and rapid-cycling patients of the Stanley Foundation Bipolar Network, add-on treatment with zonisamide (ZNS) was evaluated. ZNS appeared to have moderate antimanic and mood-stabilizing efficacy. In addition, patients lost weight while being treated with ZNS to a similar degree as they did in previous studies with topiramate (TPM) (46).

TPM was considered a promising candidate as a mood stabilizer after open studies (47,48). However, four double-blind randomized and PBO-controlled acute mania studies could not prove efficacy of this drug (49).

Finally, tiagabine (TGB) has been examined in two open studies, where it did not show antimanic efficacy (50) and, if at all, very limited mood-stabilizing properties (51). A high incidence of side effects was noted, including provocation of epileptic seizures in patients without epilepsy. Thus, similar to VGB, TGB is not of interest for treating bipolar patients because of the potential risk of severe side effects.


Several AEDs have become cornerstones for the treatment of bipolar patients. Some of them (e.g., VPA and CBZ) may exert stronger antimanic effects and, to a lesser degree, acute antidepressant efficacy than others.

LTG seems to possess long-term efficacy in all phases of bipolar disorder; the magnitude of its clinical effect in the depressed phase is considered to be larger than its effects in mania, hypomania, and mixed states. Conventional mood stabilizers such as lithium and VPA appear to have less efficacy in preventing depressive relapses. In clinical practice, LTG is frequently used in combination with other mood stabilizers that specifically target symptoms of mania (52).

It is still unclear whether the latest generation of AEDs (for example, LEV) will become possible treatments in bipolar patients. However, it is predictable that future drugs developed for epilepsy will be tested sooner or later in bipolar patients, and some of them may become established treatments, such as CBZ, VPA, and LTG.