Impulsive aggressive behavior is common in psychiatric disorders and accounts for significant morbidity and mortality. However, little systematic treatment data exist from placebo-controlled trials for this symptom domain. This was a multicenter, randomized, double-blind, placebo-controlled study in which outpatients with a score of > or =15 on the Aggression scale of the Overt Aggression Scale-Modified (OAS-M) and who fulfilled DSM-IV criteria for Cluster B personality disorder (n=96), intermittent explosive disorder (n=116), or post-traumatic stress disorder (n=34) were randomized to divalproex sodium or placebo for 12 weeks duration. Based on average OAS-M Aggression scores over the last 4 weeks of treatment, a treatment effect was not observed in the intent-to-treat data set (combined across the three psychiatric disorders), but was observed in both intent-to-treat and evaluable data sets for patients with Cluster B personality disorders. In the Cluster B evaluable data set, statistically significant treatment differences favoring divalproex were also observed for component items of the OAS-M Aggression score, including verbal assault and assault against objects, as well as OAS-M Irritability score, and Clinical Global Impression (CGI)-Severity at multiple time points throughout the study. No treatment group difference was noted for overall premature discontinuation rate; however, across psychiatric diagnoses, 21 (17%) patients in the divalproex group prematurely discontinued because of an adverse event, as compared to 4 (3%) patients in the placebo group (p <0.001). While a treatment effect was not observed when all diagnostic groups were combined, in a large subgroup of patients with Cluster B disorders, divalproex was superior to placebo in the treatment of impulsive aggression, irritability, and global severity.