BACKGROUND: Developing safe and effective long-term treatments for bipolar disorder remains a major challenge. Given available treatments, patients with bipolar disorder remain unwell in half of long-term follow-up, mostly in depression. As memantine, an N-methyl-D-aspartate (NMDA)-glutamate receptor antagonist used to treat dementia, has been proposed for testing in bipolar disorder, we carried out a 3 + 3-year, mirror-image, chart-review study of the effects of adding memantine to stably continued, but insufficiently effective, ongoing mood-stabilizing treatments. METHOD: Outpatients diagnosed with DSM-IV-TR bipolar disorder (I or II), followed intensively at the Lucio Bini Mood Disorder Center, Rome, Italy, had responded consistently unsatisfactorily to standard treatments (lithium, anticonvulsants, antipsychotics, antidepressants, and electroconvulsive therapy) for ≥ 3 years (2005-2013). Memantine (20-30 mg/d) was added clinically to otherwise stable regimens for another 3 years. On the basis of chart review, we compared morbidity measures and Clinical Global Impressions scale for Bipolar Disorder (CGI-BP) score before versus during memantine treatment. RESULTS: The 30 bipolar I (n = 17) and II (n = 13) subjects showed consistent morbidity for 3 years before memantine, but improved progressively (r = 0.28, P < .01) over 3 years with memantine (23 ± 4.8 mg/d). Markedly decreased (all P values ≤ .01) were (1) percentage of time ill (total, mania, or depression; averaging -75.0%), (2) CGI-BP severity scores (-67.8%), (3) duration of new episodes (-58.6%), and (4) episodes/year (-55.7%). Subjects with previous rapid or continuous cycling were particularly improved (t = 2.61, P = .016). Adverse effects were mild and rare. CONCLUSIONS: Memantine added substantial long-term benefits by preventing or ameliorating depressive as well as mania-like morbidity in previously consistently poorly responsive patients with bipolar disorder. Further testing in randomized, controlled trials is required.