Evidence synthesis methods enabling direct and indirect comparisons over the entire set of relevant clinical data produce quantitative point estimates for the treatments contrasts between competing interventions, and provide a hierarchical rank ordering between them. We aimed to provide evidence-based guidance on the efficacy and all-cause discontinuation of antimanic treatments.

We conducted a network meta-analysis within a Bayesian framework. We searched all standard literature databases without language restrictions up to 15 January 2014 to identify reports of short-term, randomized, blinded trials of putative antimanic drugs as monotherapy for adults with bipolar-I mania.

Altogether, 14256 manic patients randomized to one of 18 active treatments or placebo provided 95 direct comparisons on 128 data points. For the primary outcome, standardized mean difference as Hedges’ g (standardized mean difference; SMD), the hierarchies indicated by surface under the cumulative ranking (SUCRA) probabilities were in agreement with the point estimates for all antimanic drugs identified as effective. For the 12 effective antimanic drugs on clinical use, SMDs against placebo ranged from 0.32 to 0.66 without superiority of one over another, except for risperidone v. aripiprazole and valproate. Aripiprazole, olanzapine, quetiapine, risperidone, and valproate had less all-cause discontinuation rates than placebo. Sensitivity analysis by drug class indicated similar efficacy profiles for haloperidol, second-generation antipsychotics, and mood stabilizers.

Hierarchical rank ordering by comparative efficacy and risk of all-cause discontinuations should help to guide antimanic treatment choices by clinicians, healthcare policy makers, and guideline developers.