To the Editor: We applaud the extensive effort and recognize the ethically challenging nature of designing and conducting the randomized clinical trial comparing lithium and valproate in the treatment of suicidal behavior (1) reported in the October 2011 issue of the Journal. Such studies are desperately needed to identify pharmacological treatment approaches that reduce suicide. The report by Oquendo et al. (1) should further encourage such experimental approaches. However, our concern is that the article's conclusions could discourage clinicians from prescribing lithium for at-risk bipolar patients.

Lithium therapy has been associated with a reduction in the risk of suicidal behavior, defined as a suicide attempt or completion (2). However, there exists limited evidence to suggest that planning for a suicide, a subtype of suicidal ideation, is attenuated by lithium therapy. In a recent consensus statement, experts argued that combining suicidal thinking and behavior should not be a standard endpoint for randomized controlled trials (3), and ideation per se does not have a well-documented biological basis. Oquendo et al. emphasize that they calculated power to detect a relative risk of 5 or greater, which compares favorably with the relative risks associated previously with not taking lithium versus taking lithium (relative risk=4.91, 95% CI=3.82–6.31) (2). However, sufficient power is only derived from including individuals with a plan for a suicidal act. Can the authors provide a power analysis only for suicidal behavior as an outcome, and discuss conclusions based on the outcome of that analysis?

A current hypothesis about the biological mechanism of lithium action in preventing suicide—consistent with available evidence—is that lithium does not reduce suicidal thoughts, but it reduces acting on such thoughts by decreasing impulsivity, aggression, or decision-making deficits, which are well-defined endophenotypes intermediate to suicidal behaviors (4). It is possible that lithium could have the greatest effect in preventing suicidal acts in patients who have a plan for suicide and that by intervening, as per the experimental design, the therapeutic effects of lithium would not be observed. We recognize that the experimental options were limited by ethical concerns that the authors were unmistakably correct to follow, but that does not reduce the rationale for considering that such a mechanism may be important.