Main

There is increasing recognition of the inadequacy of “lithium treatment” in bipolar illness, even with adjunctive antidepressants and neuroleptics (Maj et al. 1989; Aagaard and Vestergaard 1990; O'Connell et al. 1991; Vestergaard 1992; Gitlin et al. 1995). Some bipolar patient subtypes are particularly prone to lithium nonresponsiveness; among these are patients with dysphoric mania and rapid cycling; a negative family history for bipolar illness in first-degree relatives; the episode sequence pattern of depression-mania-well interval (i.e., the D-M-I pattern as opposed to the M-D-I pattern); more than three episodes prior to the initiation of prophylaxis (Sarantidis and Waters 1981; Gelenberg et al. 1989; O'Connell et al. 1991; Denicoff et al. 1997); a history of co-morbid substance abuse; and those patients with a history of head trauma or other such medical co-morbidities as multiple sclerosis, etc.

In addition, it is recognized that patients with initial excellent responses to lithium can begin to develop breakthrough episodes in subsequent extended years of follow-up (Maj et al. 1989; Post et al. 1992). In our series of 66 lithium-refractory patients, 23 (34.9%) seem to have developed their lithium-refractoriness over more prolonged periods of follow-up, in a pattern resembling tolerance.

Another recently described route to refractoriness has been found in patients who were excellent responders to lithium, but then discontinued treatment (on the basis of either noncompliance or physician acquiescence), experienced a relapse, and then failed to re-respond once lithium was reinstituted (Post et al. 1992, 1993; Koukopoulos et al. 1995). This type of lithium discontinuation-induced refractoriness can be quite malignant, with patients failing to re-respond, even when lithium is augmented with a variety of the other alternative treatments described below; or more mild, and just require greater neuroleptic augmentation.

Thus, lithium shows parallels to the role of penicillin in the treatment of infectious diseases. Penicillin was initially heralded as a miracle drug (which it was), until patterns of nonresponse or treatment-resistance emerged, and a variety of alternative treatments were required. In cases of tuberculosis or AIDS-related opportunistic infections, for example, patients can require complex combination therapy to bring these more malignant infections under control. A similar sequence of adjuncts and alternatives to lithium treatment is now required for many patients with lithium nonresponsive affective illness.

Some of these emerging alternatives are described below, two of which (the mood-stabilizing anticonvulsants carbamazepine and valproate) have already gained wide acceptance in clinical practice. Valproate recently has been FDA-approved for first-line treatment of acute mania. Many of the other promising approaches are more investigational, but, in many instances, clinical practice is again ahead of more formal controlled clinical trials.

CARBAMAZEPINE

Nineteen double-blind studies of various methodologies compared with placebo (N = 5), compared with neuroleptics (n = 6), compared with lithium (n = 6), and B-A-B-A (off-on-off-on) designs (n = 2) document the acute antimanic efficacy of carbamazepine or its congener oxcarbazepine (Table 1). Perhaps most revealing is the clear documentation of responsivity in individual patients using double-blind, placebo substitution designs that demonstrate efficacy, reconfirmed with placebo-related exacerbation and re-responsiveness after drug institution on a blind basis (Ballenger and Post 1978; Post et al. 1984a; Post et al. 1996a). Antidepressant effects are much less well documented and require further controlled study (Post et al. 1994). However, in long-term prophylaxis, the ability of carbamazepine to prevent recurrent depressive episodes seems as prominent as the ability to prevent recurrent manias in both controlled and partially controlled double-blind studies as well as in open studies (Table 2), with an overall moderate to marked response rate of approximately 62%. Although many of these double-blind studies have been criticized on methodological grounds, other designs used, such as mirror image designs in treatment-refractory patients and B-A-B-A designs, have been cited by Prien and Gelenberg (1989) as more convincing.

Table 1 Controlled Studies of Carbamazepine (CBZ) and Oxcarbazepine (OXCBZ) in Acute Mania
Full size table
Table 2 Controlled and Partially Controlled Studies of Carbamazepine (CBZ) and Oxcarbazepine Prophylaxis in Affective Illness
Full size table

As with lithium, a subgroup of patients who are initially responsive to carbamazepine alone or with adjunctive treatment with other agents seems to lose efficacy with the development of tolerance (Post et al. 1990; Leverich et al. unpublished data 1998). Loss of responsiveness may have accounted for the large drop-out rate in the follow-up study of Frankenburg et al. (1988)), although the percentage attributable to tolerance is unknown.

In our prospective follow-up of initial responders, 13 of the 29 carbamazepine responders (45%) followed in long-term prophylaxis for an average of 6.9 years developed a loss of efficacy after an average of 2.8 years of carbamazepine treatment (Leverich et al. unpublished data). This high percentage may be unrepresentative of the general population because of the relatively treatment-refractory and rapid-cycling nature of the patients who were started on carbamazepine in the acute inpatient phase of this study. However, it would seem that at least a subgroup of patients may develop tolerance to carbamazepine in the treatment of bipolar illness, similar to the even higher percentage of tolerance development using carbamazepine in the treatment of trigeminal neuralgia (Fromm and Terrence 1987).

Weiss et al. (1995a) have described some of the potential molecular mechanisms involved in the loss of anticonvulsant efficacy to carbamazepine in the treatment of amygdala-kindled seizures in rodents. These data suggest that tolerance to carbamazepine is pharmacodynamic rather than pharmacokinetic and may represent the loss of illness-induced adaptations during carbamazepine treatment. As such, these preclinical data would suggest the potential utility of switching to drugs with different mechanisms of action or returning to carbamazepine after a time-off period. Several cases studies support the possibility of the usefulness of such a maneuver (Pazzaglia and Post 1992; and unpublished data), although larger long-term prospective studies are required in order to develop the optimal intervention paradigms for the treatment of carbamazepine tolerance.

In patients who remain successfully treated with carbamazepine, there does not seem to be a very high rate of discontinuation-related refractoriness as observed with lithium; all 10 patients re-responded during a second double-blind trial. However, one schizoaffective patient who showed a rapid antimanic re-response remained psychotic for a longer period (Post et al. unpublished observations). On this issue, as well, longer-term studies in a larger number of patients are required in order to assess this potential liability more definitively across different psychotropic medications.

The clinical predictors of response to carbamazepine have not been adequately delineated. However, there is preliminary evidence that, as in lithium treatment, rapid-cycling patients do not respond as well to carbamazepine monotherapy as do nonrapid-cycling patients (Okuma 1993; Denicoff et al. 1997); whereas, those with a negative family history for bipolar illness in first-degree relatives may be included among those responsive to carbamazepine (Post et al. 1987), as are those with more severe and discrete episodes of depression and those who have greater decrements in T4 and free T4 during treatment with the drug (Roy-Byrne et al. 1984; Post et al. 1986). In the study of Denicoff et al. (1997), the combination of lithium and carbamazepine proved much more effective for rapid-cycling patients than monotherapy with either agent. Ketter et al. (1996) found that patients with global hypermetabolism, and particularly in the region of the left insula, seem more responsive to carbamazepine as compared with depressed patients with a more classical picture of frontal lobe or left insula hypometabolism. It is these latter patients with the typical depressive pattern of frontal hypometabolism who seem more responsive to the dihydropyridine L-type calcium channel blocker nimodipine.

VALPROATE

Valproate is now FDA-approved for the treatment of acute mania based on several placebo-controlled clinical trials demonstrating efficacy (Pope et al. 1991; Bowden et al. 1994). However, in the study of Bowden et al., only 50% of the patients showed a 50% or greater response at the end of the 3-week study, findings in parallel with those of lithium. These data indicate that although valproate is highly effective as compared with placebo, many patients remain inadequately responsive and require adjunctive medications in order to optimize either the time frame or completeness of clinical response.

In this regard, Keck et al. (1993) have used loading doses of valproate (20 mg/kg/day) from the beginning of treatment and find it generally well tolerated and able to achieve higher blood levels more readily. This is convergent with the data of Bowden et al. (1996), indicating that patients with blood levels over 45 μg/ml were better early responders to valproate. Open studies by a variety of investigators indicate that valproate has prophylactic effects against recurrences of both acute mania and acute depression in a variety of patients subgroups, including those with dysphoric mania, rapid cycling, and a variety of other traditionally inadequately treated subgroups of bipolar patients (Lambert 1984; Emrich et al. 1984; Fawcett 1988; Calabrese and Delucchi 1989, 1990; Post 1989; Schaff et al. 1993; Lambert and Venaud 1995).

In particular, we and others have noted many patients with excellent responses to valproate who previously were unresponsive to both lithium and carbamazepine. Conversely, we have observed some patients who responded to carbamazepine but failed valproate (Post et al. 1984b), further suggesting that response or nonresponse to one anticonvulsant is not necessarily predictive of responsivity to another (Calabrese et al. 1992; Schaff et al. 1993).

In our initial series of 27 patients followed for an average of 5.8 years, we observed that four of 15 (27%) initial responders developed a pattern consistent with a loss of efficacy via tolerance after an average of 2.9 years of valproate treatment (Post et al. unpublished data), and McElroy et al. (personal communication) have also reported a small percentage of patients with loss of efficacy to valproate. Clinical predictors and therapeutic maneuvers remain to be more definitively delineated.

CALCIUM CHANNEL BLOCKERS

A substantial, but partially mixed, literature review supports the use of verapmil in acute mania (Table 3), although several recent studies do not support this perspective (Janicak et al. 1994; Walton et al. 1996). Even the positive double-blind series of Hoschl and Kozeny (1989) in acute mania was unable to demonstrate significant acute antidepressant effects. In light of this ambiguous profile and less than satisfactory response in acute depression, we elected to study a different calcium channel blocker with a different profile of physiochemical and biochemical properties—the dihydropyridine L-type calcium channel blocker nimodipine.

Table 3 Calcium Channel Blockers in Affective Illness
Full size table

Nimodipine has greater lipid solubility and better penetration into the central nervous system (Freedman and Waters 1987) and is reported to be problematic with tolerance in the treatment of migraine as compared with other calcium channel blockers. Moreover, nimodipine has a different profile of anticonvulsant effects in animal models and, in contrast to verapamil, blocks cocaine-induced hyperactivity and its associated dopamine overflow (Pani et al. 1990).

Five of the first 12 patients treated with nimodipine showed clinically relevant improvement using a double-blind B-A-B, and in some instances, B-A-B-A placebo-substitution design (Pazzaglia et al. 1993). Using this design, response could be confirmed in patients with very different cycle frequencies, and appropriate statistical techniques could be applied in single-case analyses, as described by McDermut et al. (1995).

In an extension of the original study of Pazzaglia et al. (1993), we have now studied a total of 36 patients on nimodipine, 30 of whom were evaluable. Ten of these patients showed clinically relevant degrees of response as rated by marked to moderate response on the clinical global impressions scale (CGI) and confirmed by appropriate within-patient statistical analyses (Pazzaglia et al. 1998). Fourteen patients with an incomplete or inadequate response to nimodipine monotherapy were given a trial of blind carbamazepine augmentation. Four of these showed a clinically significant degree of improvement with the addition of carbamazepine as rated by the CGI change score and confirmed with t-tests.

Many of the responsive patients had failed multiple clinical trials of more traditional treatments, including lithium and a variety of other mood-stabilizing and antidepressant modalities. Among these were several patients with ultrarapid and ultra-ultrarapid (ultradian) cycling and several patients with recurrent brief depression. Several of the bipolar patients on either nimodipine monotherapy or combination therapy with carbamazepine were switched from the dihydropyridine nimodipine to the phenylalkylamine verapamil on a blind basis and failed to maintain their same degree of clinical improvement. On a blind basis, they re-responded to nimodipine and continued to maintain this degree of improvement with the transition to another dihydropyridine, isradipine. These data suggest that, at least in these few bipolar patients, there may not be a cross responsivity among all of the L-type calcium channel blockers. The dihydropyridines, with their different site of action inside the calcium channel and different biochemical properties (Triggle 1992), may be preferable to the phenylalkylamines. However, a more formal comparable clinical trial in larger numbers of patients is required to confirm these differential preliminary observations in a few patients with confirmed responses.

LAMOTRIGINE

This newly approved anticonvulsant for adjunctive therapy in refractory epilepsies has received attention in the treatment of bipolar illness. Lamotrigine has interesting clinical properties including inhibition of excitatory amino acid release and blockade of sodium channels (Messenheimer 1995). However, evidence indicates that lamotrigine may possess other actions not shared by carbamazepine (Brodie et al. 1995), because of its additional efficacy in inadequate responders to carbamazepine in both epilepsy studies and affective illness studies, as described below.

Calabrese (1996) reported a 69% response rate when lamotrigine was added to previously ineffective treatments for patients with recurrent affective illness and, in a few cases, used in monotherapy. Particularly good responses to depression were noted and showed a relatively rapid onset, including 27 of 39 (69%) in the first week or two of treatment with the 25-mg dose. This low dose was initiated and fully titrated to average doses of approximately 150 mg/day to avoid the reported high incidence of rash.

Relatively similar degrees of response have been observed by Frye et al. (1998) in our group in the initial patients entering a 6-week randomized comparison of lamotrigine to placebo and gabapentin. Seventeen of 33 (52%) patients responded to lamotrigine as marked or moderate on the CGI-BP. Again, both antidepressant and mood-stabilizing properties were evident in these patients as compared with the placebo phase. Much work remains to delineate the clinical spectrum of efficacy of lamotrigine in refractory affectively ill patients, but the preliminary open observations of Calabrese (1996) and double-blind monotherapy studies of Frye et al. (1998) suggest that this agent could play an important role in the pharmacotherapy of bipolar illness.

GABAPENTIN

Gabapentin has important effects as an adjunctive treatment of refractory epilepsies and has a variety of purported mechanisms of action with indirect effects on γ-aminobutyric acid (GABA), such as affecting the GABA transporter and leading to increased levels of GABA in the central nervous system (Beydoun et al. 1995). Its potential for mood stabilization is just beginning to be explored, with preliminary reports of some success (McElroy et al. personal communication, Suppes et al. personal communication, (Young et al., 1997) in open clinical observations, and responsivity in 9 of the first 33 patients (27%) to be studied in a randomized crossover as compared with placebo and lamotrigine, as noted above (Frye et al. 1998). Because gabapentin is not metabolized in the liver and is excreted in the kidney, and has few pharmacokinetic interactions with other agents, its positive side effects profile and lack of interaction with other drugs makes it a particularly appealing agent for adjunctive therapy, should it prove useful in further clinical trials.

HIGH-POTENCY BENZODIAZEPINES

The high-potency anticonvulsant benzodiazepines, such as clonazepam and lorazepam, are widely used in the treatment of acute mania and in the anxiety and insomnia of acute depression. The overall antimanic and mood-stabilizing effects of these agents has not yet been adequately delineated (Table 4). However, it is recommended that clinical use proceed with these agents in lieu of neuroleptics (with their liabilities for acute Parkinsonian side-effects and longer-term tardive dyskinesia) for patients with mania, agitation, insomnia, and anxiety breaking-through other treatments.

Table 4 Partially Controlled Studies of Clonazepam and Lorazepam in Acute Mania
Full size table

The increasing use of benzodiazepines is to be encouraged in light of the recent data of Sernyak et al. (1994), who reported that once manic patients are placed on neuroleptics, they are often continued on neuroleptics for extended periods of time, putting patients at greater risk for tardive dyskinesia, which is reported to be a very high 20 to 40% incidence rate in most studies of patients with bipolar illness (Hunt and Silverstone 1991). Further clinical trials are needed to delineate the acute antimanic efficacy of benzodiazepines in monotherapy, although their adjunctive use, particularly in place of neuroleptics, would seem to have considerable merit, based even on the modicum of preliminary data available.

ATYPICAL NEUROLEPTICS

The atypical neuroleptics, such as clozapine, are now being used more widely with some success in the treatment of dysphoric mania and rapid cycling (Suppes et al. 1992; Calabrese and Woyshville 1995; Calabrese et al. 1996; Frye et al. 1996). Although seizures are a side effect of clozapine, one study reports that clozapine might have some effects in inhibiting kindled seizure evolution (Graham and Kokkinidis 1993). Denney and Stevens (1995) have also postulated that the microconvulsive properties of clozapine could be related to its profile of clinical efficacy and have noted that patients cotreated with valproate seem to have less robust responses than those who are not so treated. These observations remain to be more systematically documented, but they raise a variety of intriguing possibilities in relation to convulsant and anticonvulsant mechanisms of psychotropic drug action.

Olanzapine and a whole group of atypical neuroleptics (Tamminga and Lahti 1996) are close to FDA approval in addition to risperidone and seroquel, and it is hoped that many of these agents will have positive clozapine-like properties without the liability of blood dyscrasias in the treatment of otherwise refractory bipolar illness.

REPEATED TRANSCRANIAL MAGNETIC STIMULATION (rTMS) OF THE BRAIN

This modality is just beginning to be explored in the treatment of conventional and refractory depression (Table 5). George et al. (1995) in our group first reported that stimulation with rTMS utilizing 20 Hz for 20 min over the left frontal cortex led to improvement in two of the first six patients. This finding led to a randomized design compared with sham, which also showed statistically significant improvement with rTMS (George et al. 1997). These studies were followed by another (Pascual-Leone et al. 1996) reporting 11 of 17 psychotic depressed patients responded to left frontal rTMS but not right frontal cortex or vertex rTMS. Five stimulations in that series led to clinical improvement lasting approximately 2 weeks.

Table 5 Repeated Transcranial Magnetic Stimulation (rTMS) in Affective Illness
Full size table

These very exciting preliminary data require further elucidation in order to define the parameters and localities of stimulation that optimally produce antidepressant effects. Because low-frequency stimulation (1 Hz for 15 min) of the amygdala (“quenching”) has dramatic anticonvulsant effects on kindled seizures (Weiss et al. 1995b), it is possible that low-frequency stimulation may increase inhibition and decrease hyperactivity circuits. However, more recent data indicate that a small direct current of 10-15 μA is the critical variable in quenching (Weiss et al., 1998, unpublished data).

Should preliminary observations with the rTMS be supported in subsequent clinical trials, they would have very interesting implications for the mechanism of action of electroconvulsive therapy (ECT), which had always been presumably linked to the generation of a seizure. In rTMS, a seizure is not generated, and there are no subjective reports of cognitive clouding or impairment similar to that often induced with ECT. It is possible that rTMS is more directly engaging the mechanism generated after a ECT seizure and that these rTMS effects are related to its potential clinical antidepressant profile. The molecular mechanisms require further elucidation, but the clinical and research potential for rTMS in discretely stimulating areas of the brain remains an exciting frontier.

CONCLUSION

Two decades ago lithium, as supplemented by neuroleptics and antidepressants, was virtually the only putative mood-stabilizing agent available for bipolar illness. Increasingly, it has been recognized that lithium alone or in combination with adjunctive strategies is inadequate for a very substantial group of bipolar patients. Now a variety of putative mood-stabilizing agents are available for acute and long-term therapeutics and for devising the optimal treatment algorithms in the most appropriate combination. The sequencing of these agents remains a very important issue for further clinical work and systematic clinical trials in methodological exploration. Over each of the last five 5-year epochs at the National Institute for Mental Health (NIMH), we have found that approximately 80% of our patients were able to be discharged with marked or moderate improvement. Twenty years ago, three-quarters of the patients were discharged from our clinical research unit on monotherapy; in the last 5-year epoch, only 25% of the patients were discharged on monotherapy, and the average was 3 drugs per patient. Thus, rational combination therapy seems to be a requirement for clinical stabilization in many instances (Frye 1996). Defining the principles for optimum combination of these agents (Post et al. 1996b; Post et al. 1997) remains a very important task for future research.