Revisiting the association between treatment with antidepressants and mania: A nationwide within‐individual study of 3554 patients with bipolar disorder

INTRODUCTION: Antidepressants are commonly used "off-label" for bipolar depression, despite concerns over the risk of potential treatment-emergent mania (or "manic switch"). Treatment-emergent mania is difficult to study with adequate power in clinical trials as it requires a large group of participants and long follow-up. Therefore, naturalistic register-based studies have been applied to assess this phenomenon. Here, we aimed to replicate previous findings and address key methodological limitations that were not previously taken into account. METHODS: We utilized data from nationwide Danish health registries to identify patients with bipolar disorder treated with an antidepressant, either with or without concomitant treatment with a mood stabilizer (drug treatment proxied via redeemed prescriptions). We plotted the incidence of manic and depressive episodes relative to the initiation of antidepressant treatment and compared the incidence of mania in the period prior to and following initiation of antidepressant treatment (within-individual design). RESULTS: In 3554 patients with bipolar disorder initiating treatment with an antidepressant, the number of manic episodes peaked approximately 3 months prior to initiation of antidepressant treatment, and the number of depressive episodes peaked around the initiation of antidepressant prescription. This temporal pattern suggests that antidepressants were used to treat post-manic depression. CONCLUSION: Within-individual designs do not control sufficiently for confounding by indication, when the treatment indication is time-varying. Thus, results from prior within-individual studies of antidepressant treatment in the context of bipolar disorder may be invalid due to time-varying confounding by indication.