Interruption of ongoing treatment with benzodiazepines, antidepressants, antipsychotics and mood stabilisers including lithium can be followed by clinically significant withdrawal reactions within hours or days, as well as later increases in relapses or recurrences of the illness being treated. Such observations support the view that stopping treatment is not equivalent to being untreated. With lithium, antipsychotics and antidepressants, there is consistent evidence that abrupt or rapid discontinuation is followed by earlier clinical worsening than with more gradual removal of treatment. Moreover, treatment discontinuation can complicate interpretation of responses to changes in treatment, including in clinical practice and in experimental treatment trials. Notably, terminating preceding treatments can lead to both discontinuation and carry-over effects that can have an impact on the interpretation of observed outcomes.

Can timely diagnosis and treatment of depression reduce the risk of suicide? Studies of treatment effects on mortality in major mood disorders remain rare and are widely considered difficult to carry out ethically. Despite close associations of suicide with major affective disorders and related comorbidity, the available evidence provides little support for sustained reductions of suicide risk by most mood-altering treatments, including antidepressants. Studies designed to evaluate clinical benefits of mood-stabilizing treatments in bipolar disorders, however, provide comparisons of suicidal rates with and without treatment or under different treatment conditions. This emerging body of research provides consistent evidence of reduced rates of suicides and attempts during long-term treatment with lithium. This effect may not generalize to proposed alternatives, particularly carbamazepine. Our recent international collaborative studies found compelling evidence for prolonged reduction of suicidal risks during treatment with lithium, as well as sharp increases soon after its discontinuation, all in close association with depressive recurrences. Depression was markedly reduced, and suicide attempts were less frequent, when lithium was discontinued gradually. These findings indicate that studies of the effects of long-term treatment on suicide risk are feasible and that more timely diagnosis and treatment for all forms of major depression, but particularly for bipolar depression, should further reduce suicide risk.

Bipolar disorders are prevalent, often severe, and disabling illnesses with elevated lethality largely due to suicide. Suicide rates average ∼1% annually, or perhaps 60 times higher than the international population rate of 0.015% annually. Suicidal acts typically occur early in bipolar disorders and in association with severe depressive or mixed states. The high lethality of suicidal acts in bipolar disorders is suggested by a much lower ratio of attempts:suicide (∼3:1) than in the general population (∼30:1). Risk factors can help to identify patients at increased suicidal risk, but ongoing clinical assessment is essential to limit risk. Empirical short-term interventions to manage acute suicidal risk include close clinical supervision, rapid hospitalization, and electroconvulsive therapy. Remarkably, however, evidence of the long-term effectiveness of most treatments against suicidal behavior is rare. A notable exception is lithium prophylaxis, which is associated with consistent evidence of major (∼80%), sustained relative reductions of risk of suicides and attempts, and lower lethality (increased attempts:suicide ratio). Such benefits are unproved for other treatments commonly used to treat bipolar disorder patients, including anticonvulsants, antipsychotics, antidepressants, and psychosocial interventions. Applying available knowledge systematically, with close and sustained clinical supervision, can enhance management of suicidal risk in bipolar disorders patients.

Bipolar depression remains a major challenge for psychiatric therapeutics. It is associated with disability and excess mortality, and accounts for three-quarters of the time spent in morbid states by treated patients with bipolar disorder. Major limitations of research on the treatment of depression in bipolar disorder include a paucity of short-term and lack of long-term trials, probably reflecting concern about inducing mania. In addition, polytherapy with multiple drugs appears to be widespread, but it is virtually untested for efficacy and safety. Here, we summarise the evidence concerning efficacy of treatment of bipolar depression with antidepressants, mood-stabilising anticonvulsants, lithium and second-generation antipsychotics.

Longitudinal assessment of the course of major psychiatric disorders has been advanced by studies from onset, but only rarely have large numbers of patients with a range of psychotic and major affective disorders been studied simultaneously and systematically from illness-onset. The decade-long McLean-Harvard First Episode Project & International Consortium for Bipolar Disorder Research has systematically followed-up large numbers of patients with DSM-IV bipolar or psychotic disorders from first hospitalization. Major findings among patients with bipolar I disorder include: [a] full functional recovery from initial episodes was uncommon, and full symptomatic recovery, much slower than early syndromal recovery; [b] risks of relapse, recurrence, and switching were very high in the first two years; [c] most early morbidity was depressive-dysphoric, as reported in mid-course; [d] initial depression or mixed-states predicted more later depressive and overall morbidity, whereas initial mania or psychosis predicted later mania and a better prognosis; [e] based on within-subject modeling, most patients did not show progressive cycling over time, and illness-course was rather chaotic within and among patients; [f] treatment-latency or episode-counts were unassociated with responsiveness to long-term mood-stabilizing treatment; [g] very high rates of suicidal behavior and accidents occurred early; [h] early substance-use comorbidity associated with anxiety; [i] factor-analysis of prodromal symptoms predicted bipolar disorder much better than non-affective psychotic disorders. Project findings indicate that the course of bipolar I disorder is much less favorable than had been believed formerly, despite clinical treatment with modern mood-stabilizing and other treatments.

We reviewed available research findings, including meta-analyses on effects of lithium-treatment associated with rates of suicidal behavior in bipolar disorder or unipolar major depressive disorder patients, and for comparisons of lithium to mood-stabilizing anticonvulsants. Data from meta-analyses consistently indicate marked reductions of suicidal behavior and mortality during long-term treatment with lithium salts in bipolar disorder patients, and possibly also in unipolar, recurrent major depressive, perhaps even more effectively than with anticonvulsants proposed as mood-stabilizers. Suicidal risk is frequently associated with dysphoric-agitated symptoms, anger, aggression, and impulsivity-all of which may respond better to treatment with lithium or other mood-stabilizing medicines than to antidepressants. In these conditions, antidepressant treatment may not provide a beneficial effect on risk of suicidal thoughts and perhaps attempts, particularly in juveniles, whereas, lithium, perhaps even more than anticonvulsants, seems to be remarkably effective in the preventing suicidal behavior. The mechanism of action is not well defined and may be associated with either a prevention of mood recurrences or a more specific “antisuicidal” activity.

Declaration of Interest: Dr. Tondo has received research support from Janssen and Eli Lilly Corporations and has served as a consultant to Glaxo-SmithKline and Merck Corporations. Dr. Baldessarini has recently been a consultant or investigator-initiated research collaborator with: AstraZeneca, Auritec, Biotrofix, Janssen, JDS-Noven, Lilly, Luitpold, NeuroHealing, Novartis, Pfizer, and SK-BioPharmaceutical Corporations. Neither author is a member of pharmaceutical speakers’ bureaus, nor do they or any family member hold equity positions in biomedical or pharmaceutical corporations.