At least 41 studies (at least 1,987 patients) were included in the review, including 18 double-blind RCTs.
Phenobarbital, primidone and phenytoin.
Evidence for these classic anti-epileptic drugs in the treatment of acute mania was weak. No double-blind studies supported the efficacy of barbiturates, and the one single-blind trial that compared phenytoin plus haloperidol with placebo plus haloperidol was limited by the small sample size.
Valproate (10 controlled trials in 838 patients were presented).
One of the RCTs (n=30) found that valproate versus carbamazepine increased the proportion of patients that achieved a 50% or greater reduction on the YMRS (p=0.023). The other RCT (n=30) showed similar reductions in the YMRS when valproate was compared with haloperidol. Non-randomised comparisons of valproate versus placebo (7 comparisons) showed better response rates with valproate. The findings from 3 non-randomised comparisons of valproate versus lithium were inconsistent.
Two double-blind, placebo-controlled parallel-group trials (212 patients) of valproate showed that valproate was significantly better (assessed using YMRS and SADS-MR) than placebo in the treatment of acute mania.
The efficacy of valproate in the maintenance treatment of bipolar disorder was supported by two trials: one double-blind crossover trial involving 12 patients reported an average mean phase interval of 10 months before valproate versus 60 months on valproate; one double-blind, placebo-controlled parallel-group trial involving 372 patients reported a trend for longer duration of prophylaxis with valproate, compared with lithium (p<0.07).
Carbamazepine and oxcarbazepine (7 double-blind RCTs of maintenance treatment in 346 patients). There was limited evidence on the use of carbamazepine in maintenance treatment. The interpretation of the results was hindered by methodological flaws. These included: small and heterogeneous samples; concomitant use of other medications during manic or depressive episodes; differences between the groups in the type of rescue medication; high drop-out rates; insufficient medication; sudden switch from one medication to another; selection bias; and loss of patients' blinding to treatment. The treatment was started during remission and/or during an episode of illness. One double-blind RCT (22 patients) compared carbamazepine with placebo and found no significant difference for carbamazepine versus placebo at 1 year: carbamazepine was effective in 60%, compared with 22% for placebo (p<0.10). Five double-blind RCTs (324 patients) compared carbamazepine and lithium. None of the RCTs appeared to show any consistent significant difference between carbamazepine and lithium.
Lamotrigine (5 double-blind RCTs in 490 patients). Two studies lacked a placebo control group, while another study had low YMRS scores at the beginning of each treatment phase. All three of the placebo-controlled studies reported greater improvement on lamotrigine than on placebo. Three double-blind RCTs compared lamotrigine with placebo. The duration of these RCTs ranged from 6 weeks to 6 months. Two RCTs appeared to find that lamotrigine was significantly better than placebo but p-values for comparisons between the treatments were not reported consistently.
Two double-blind RCTs compared lamotrigine with lithium for 4 weeks. There appeared to be no significant difference between lamotrigine and lithium.
Gabapentin, topiramate and tiagabine.
There was limited information on the clinical usefulness of gabapentin, topiramate and tiagabine. Gabapentin (9 open-label prospective studies in 306 patients): the study results were confounded by concurrent administration of other agents, heterogeneous patient populations, widely varying drug doses, open trials and small sample sizes. Topiramate (5 studies in 145 patients, including 3 open-label add-on studies, 1 open-label monotherapy study and 1 chart review of add-on therapy): no controlled studies were identified. Tiagabine (3 very small open-label add-on studies in 13 patients): one study reported a case of generalised tonic-clonic seizure in a patient with no history of epilepsy, and one case of severe nausea and vomiting. The evidence was limited.