The purpose of this meta-analysis was to examine the efficacy of maintenance treatments for bipolar disorder. Placebo-controlled or active comparator bipolar maintenance clinical trials of ⩾6 months’ duration with at least 15 patients/treatment group were identified using Medline, EMBASE, clinicaltrials.gov, and Cochrane databases (1993 to July 2010). The main outcome measure was relative risk for relapse for patients in remission. Twenty trials (5364 patients) were identified. Overall, lithium and quetiapine were the most studied agents (eight and five trials, respectively). The majority of studies included patients who had previously responded to treatment for an acute episode. All interventions, with the exception of perphenazine+mood stabilizer, showed a relative risk for manic/mixed or depressive relapse below 1.0, although there was variation in the statistical significance of the findings vs. placebo. No monotherapy was associated with a significantly reduced risk for both manic/mixed and depressed relapse. Of the combination treatments, only quetiapine+lithium/divalproex, was associated with a significantly reduced risk vs. comparator (placebo+lithium/valproate) for relapse at both the manic/mixed and depressed poles of bipolar illness. Limitations for the analysis include differences in study durations and definitions of relapse. In conclusion, available maintenance therapies show considerable variation in efficacy. The efficacy of lithium and divalproex has been confirmed, but newer therapies, such as a number of atypical antipsychotics were also shown to be effective in bipolar disorder. Efficacy of all maintenance interventions needs to be balanced against the safety and tolerability profiles of individual agents.