In the December 2009 issue, Gibbons et al1 presented an analysis of medical claims data comparing rates of suicide attempts in patients with bipolar disorder according to whether they had been prescribed antiepileptic drugs (AEDs). This analysis was stimulated by the Food and Drug Administration meta-analysis of placebo-controlled clinical trials of AEDs in seizure disorders and other conditions that showed increased rates of suicidal behavior or ideation in AED-treated subjects. The study's use of nonrandomized data to evaluate its hypotheses is problematic. Because suicide attempts are both a potential adverse effect of AED treatment and a manifestation of bipolar disorder, confounding is a major difficulty in any assessment of this issue that does not use randomized, controlled data. We are concerned about several significant limitations that may undermine the study's conclusions.