For the roughly 1-2% of adults living with bipolar I disorder, the gap between staying well and experiencing a devastating mood episode often hinges on medication adherence — a notoriously fragile variable. Quantifying precisely how much a long-acting injectable antipsychotic improves those odds adds clinically actionable evidence to a treatment landscape where benefit-risk communication remains underdeveloped.
This post hoc analysis drew from a 52-week double-blind randomized withdrawal trial (NCT01567527) comparing aripiprazole once-monthly 400 mg (AOM 400) against placebo in 265 patients with bipolar I disorder. At one year, 73.5% of patients on AOM 400 remained free from any mood episode recurrence, versus 48.9% on placebo — yielding a number needed to treat (NNT) of 5. Episode-type breakdowns showed NNT values of 5 for manic episodes and 4 for mixed episodes, while protection against depressive recurrence was not statistically significant. Crucially, discontinuation due to treatment-emergent adverse events was actually lower in the AOM 400 group (17.4%) than placebo (25.6%), producing a negative number needed to harm — meaning tolerability favored the active drug. The resulting likelihood-to-be-helped-or-harmed ratio of 200 reflects a strongly favorable benefit-risk profile.
An NNT of 5 over 52 weeks is clinically meaningful by psychiatric standards — comparable to well-established maintenance treatments in schizophrenia and substantially better than many oral mood stabilizers. The negative NNH is particularly notable and somewhat counterintuitive: it suggests patients on placebo were dropping out due to adverse events at a higher rate, likely reflecting the burden of relapse-associated symptoms being miscoded as adverse events. Several important caveats apply. This is a post hoc analysis of a randomized withdrawal design — patients who tolerated and responded to AOM 400 in an open-label run-in phase were then randomized, introducing selection bias that inflates benefit estimates. The sample size of 265 is modest, and the depressive episode finding was non-significant, which matters given that depressive phases dominate bipolar I morbidity. Still, for maintenance psychiatry, this analysis provides a rare, quantified benefit-risk framework that clinicians can use meaningfully in shared decision-making conversations.