OBJECTIVE To evaluate the efficacy and safety of olanzapine, divalproex, and placebo in a randomized, double-blind trial in mild to moderate mania (DSM-IV-TR criteria). METHODS The study was conducted from October 2004 to December 2006. A total of 521 patients from private practices, hospitals, and university clinics were randomly assigned to olanzapine (5-20 mg/day), divalproex (500-2500 mg/day), or placebo for 3 weeks; those completing continued with a 9-week double-blind extension. Efficacy (mean change in Young Mania Rating Scale [YMRS] total score was the primary outcome) and safety were assessed. RESULTS After 3 weeks of treatment, olanzapine-treated (N = 215) and placebo-treated (N = 105) patients significantly differed in YMRS baseline-to-endpoint total score change (p = .034; least squares [LS] mean: -9.4 and -7.4, respectively). Such changes were not significantly different between olanzapine vs. divalproex (N = 201) or divalproex vs. placebo. After 12 weeks of treatment, olanzapine- and divalproex-treated patients significantly differed in YMRS baseline-to-endpoint changes (p = .004; LS mean: -13.3 and -10.7, respectively). Of observed cases, 35.4% (35/99; 3 weeks) to 57.1% (28/49; 12 weeks) had valproate plasma concentrations lower than the recommended valproate therapeutic range, but these patients' YMRS scores were lower than those of patients with valproate concentrations above/within range. Compared with divalproex, after 12 weeks, olanzapine-treated patients had significant increases in weight (p < .001) and in glucose (p < .001), triglyceride (p = .003), cholesterol (p = .024), uric acid (p = .027), and prolactin (p < .001) levels. Divalproex-treated patients had significant decreases in leukocytes (p = .044) and platelets (p < .001) compared with olanzapine after 12 weeks of treatment. The incidence of potentially clinically significant weight gain (>/= 7% from baseline) was higher with olanzapine than with divalproex (3-week: p = .064, 6.4% vs. 2.7%; 12-week: p = .002, 18.8% vs. 8.5%; respectively). CONCLUSIONS Olanzapine was significantly more efficacious than placebo but not divalproex at 3 weeks and significantly more efficacious than divalproex at 12 weeks. Olanzapine-treated patients had significantly greater increases in weight and in glucose, cholesterol, triglyceride, uric acid, and prolactin levels than divalproex-treated patients. BACKGROUND clinicaltrials.gov Identifier: NCT00094549.