Additional Effect Size Measures Helpful in Understanding Lithium and Valproate Trial Results

To the Editor: The article by Oquendo et al. (1) in the October 2011 issue of the Journal deserves commendation for tackling one of the preeminent questions in psychiatry: whether certain psychiatric medications help to reduce suicidal behaviors. Oquendo and colleagues' randomized controlled trial was foresightedly conceived and initiated more than a decade ago, and I applaud the efforts of both the research staff and the patients (some of whom agreed to receive double-blind medication for more than 2 years!) involved in this important study.

However, an unfortunate limitation of the published report from Oquendo et al. is the lack of hazard ratios or other measures of effect size. Such information would be valuable, despite the lack of statistical power, because even underpowered findings are important to incorporate in later meta-analyses. In addition, the Oquendo et al. study is the first trial comparing lithium and valproate with a primary outcome of suicidal behavior, and even an underpowered finding would help inform future research. While it is also true that assumptions concerning proportional hazards may not be met for this study, given the intersection of survival curves in Figure 2, this can sometimes be addressed by breaking the curves into shorter time frames. We would like to request that the authors provide, if possible, hazard ratios or another measure of effect size for their reported outcomes, especially the time to first suicide attempt (since examining suicidal behavior alone, rather than outcomes including prophylactic hospitalizations, has previously been a more sensitive measure for detecting differences in suicidal behavior among medications [2]).

In addition, the authors powered their study for a 1:5 difference in suicidality and argue that even a 1:2 difference might be clinically important. However, I would argue that even a 20% difference in suicide or suicidal behavior risks between psychiatric agents—if reliably demonstrated—may prove clinically significant. A 20% difference was the target for the International Suicide Prevention Trial study (2), and it remains relevant today. While Oquendo et al. are right to address concerns of overdose toxicity, thus far the Baldessarini et al. (3) meta-analysis points to decreases in suicide deaths as well as suicidal behavior in lithium recipients and a decreasing ratio of suicides to suicide attempts. These patterns seem the opposite of what we would expect if overdose toxicity clearly outweighed the possible clinical and behavioral benefits of lithium.

Clearly the important field of psychiatric medication and suicide risk warrants continued investigation; further detail about the results (e.g., hazard ratios) from Oquendo and colleagues' valuable trial would be an important step.