Relapse in patients with bipolar disorder is a common problem. More than 90% of patients will have a recurrent episode, and it is also the cause of long-term morbidity and mortality [1, 2]. Due to the chronic nature of this illness, there is a need for long-term maintenance treatment to prevent or attenuate the risk of relapse [3]. Lithium is the classic mood stabilizer used in treating acute and chronic bipolar disorder. However, studies have suggested that an estimated 20–40% of patients may have an inadequate prophylactic response to this drug [4, 5]. Undesirable side effects and a narrow therapeutic index also make it less than ideal for many patients. This has led to efforts to investigate the potential utility of other mood stabilizers in maintenance therapy.

Carbamazepine (CBZ), the first anticonvulsant used in bipolar disorder, was recognized as a useful medication in the 1970s [6–8]. Moreover, it was found to be compatible with lithium in the treatment of acute mania in small-scale trials [9, 10]. In Taiwan, CBZ has been increasingly used to treat bipolar disorder [11]. Its efficacy for treating acute mania was recently reconfirmed following the completion of large-scale, randomized, placebo-controlled trials in North America using extended-release CBZ [12, 13].

Despite the widespread clinical use of mood-stabilizer combinations for the long-term treatment of patients with bipolar disorder, research on the risks and benefits of this practice is limited. In recent decades, only a few studies have focused on the use of CBZ for the maintenance of bipolar disorder [14–18]. The results of a 2-year follow-up study showed approximately equal effects when comparing CBZ and lithium in the prevention of mood episodes [18]. In a small-scale clinical trial comparing CBZ and placebo [19], the prophylactic effects of CBZ were superior in bipolar patients. However, several meta-analyses were unable to reach a conclusion about the efficacy of CBZ in prophylaxis [20–22]. In addition, there are a number of limitations when interpreting previous double-blind, controlled studies [23]. Firstly, most of these studies were based on a small sample size (up to 60 patients). Second, follow-up periods were relatively short (up to 2.5 years). Studies concerned with long-term efficacy tend to be more rigorous when their follow-up periods are longer. Unfortunately, studies that address the above-mentioned limitations are scarce. Therefore, we aimed to evaluate the clinical usage and long-term efficacy of CBZ in patients with bipolar I disorder treated with CBZ over an extended period. We reviewed the charts of a sample of 129 patients maintained on long-term CBZ therapy for up to 10 years. This paper describes the demographics, clinical characteristics and outcome data, and analyses the correlates of the outcome.

To our knowledge, this is the longest outcome study that addresses the long-term use of CBZ in bipolar patients. We collected data from 129 patients exhibiting good compliance and attendance, which implies a reasonable quality of data. We used chart review to minimize the effect of recall bias, and all patients were evaluated by experienced clinicians. It is also worth noting that the long follow-up period (mean duration of 10 years) in this study could provide information that other studies lack. Mean plasma level of patients was 7.8 ± 5.9 μg/mL, which was within suggested therapeutic range for maintenance therapy.

We found that the frequency of mood episodes and hospitalization decreased significantly after CBZ treatment. We were also surprised to find that almost half of the patients (48.8%) had no more mood episodes after CBZ treatment, showing the efficacy of CBZ, either as monotherapy or combined with other mood stabilizers or antipsychotics, in the prophylactic treatment of bipolar disorder. It is difficult to compare our findings with other studies since there has been no other 10-year naturalistic follow-up of bipolar disorder patients treated with CBZ. The response rates for CBZ in previous studies varied. Kleindienst and Greil [24] found that classical bipolar patients treated with CBZ had a hospitalization rate of about 62% versus 26% in those treated with lithium in a randomized clinical trial with an observation period of 2.5 years. The better remission rate in our study might be related to the naturalistic design of the study, which allowed for a combination of other medications during the study period.

A second aim of our study was to determine the correlates of the CBZ prophylactic response. In previous studies, CBZ response was related to “non-classical disease”, and patients with suicidal behavior, lithium-refractory disease, and mixed episodes. However, in our study, no specific factors were correlated with the response to CBZ treatment with the exception that males had a better response rate. Since it was impossible to attain some of the clinical variables in this retrospective review, the response may be correlated with other factors that were not considered in this study.

The adverse events reported by the patients were also reviewed. The most common side effects coded were dizziness, fatigue and somnolence (24%). Most patients in our study were able to tolerate the side effects of long-term use of CBZ. Patients who did not receive concomitant lithium or valproic acid had significantly less body weight increment. This result confirms previous reports [14, 21] suggesting that one benefit of CBZ include the low propensity toward weight gain and evidence of good tolerability with long-term treatment.

It is worth noting that about 40% of patients were treated concomitantly with lithium or valproic acid. Most patients needed benzodiazepines (74.4%) and antipsychotics (68.2%) in maintenance therapy. Given the naturalistic study design, we were unable to control the adjuvant medication. However, adjuvant medication usage did not differ between our CR and non-CR groups. In a review article by Keck and McElroy [25], the authors suggested that combination therapy may provide better long-term prevention of illness relapse and recurrence in many patients with bipolar disorder.

The main limitation of this study was the potential for a confounding bias due to its observational design. Since this study was designed as a retrospective chart review, it did not include patients who were initially treated with CBZ but discontinued use later for any reason. It was difficult to estimate the percentage of patients who dropped out of initial CBZ treatment, so the results should be interpreted with caution, and may only reflect the prophylactic effect in those who initially responded to acute treatment with CBZ. It is likely that patients who did not respond well to CBZ or who did not tolerate CBZ were more likely to drop out. In a 6-month, multicenter, open-label evaluation of beaded, extended-release CBZ capsule monotherapy in bipolar disorder patients with manic or mixed episodes [26], 68.8% of patients discontinued treatment early due to the lack of efficacy or side effects. This result reflects the fact that only a portion of patients adhere to CBZ according to perceived efficacy or tolerability of side effects. Our study merely showed that if a patient had a fair response to CBZ, they were likely to respond well in the continuing follow-up period. As mentioned earlier, due to the naturalistic setting, medications were not controlled during relapse. In addition, some variables were not reviewed (e.g., rapid cycling, or type of mood episode), which limited further analysis of clinical outcomes. Patients who were prescribed with CBZ during their first episode were coded as ‘zero episode prior to CBZ treatment’, which included 42 of the subjects. It is possible that patients with bipolar disorder who had just one episode over a long period of time may have a milder bipolar illness. However, after excluding these 42 subjects from analysis, the results remained the same (data not shown).

As demonstrated, our longitudinal study proved that CBZ is efficacious and tolerable in the maintenance therapy of bipolar disorder in naturalistic clinical practice, either as monotherapy or in combination with other medications. Patients who have a good response to CBZ tend to continue to do well over the long term and their subsequent course of illness appears to be improved compared to their course of illness prior to CBZ treatment. This should be validated by more controlled long-term follow-up studies.