PURPOSE/BACKGROUND: We previously conducted a randomized, double-blind, controlled, 12-week study evaluating the effect of add-on dextromethorphan (DM), a noncompetitive N-methyl-D-aspartate receptor antagonist, on patients with bipolar disorder (BD) treated using valproate (VPA), which showed negative clinical differences. The genetic variation between each individual may be responsible for interindividual differences. The catechol-O-methyltransferase (COMT) gene has been a candidate gene for BD. In the current study, we investigated whether the COMT Val158Met polymorphism predicts treatment response to VPA + add-on DM and to VPA + placebo. METHODS/PROCEDURES: Patients with BD (n = 309) undergoing regular VPA treatments were randomly assigned to groups given either add-on DM (30 mg/d) (n = 102), DM (60 mg/d) (n = 101), or placebo (n = 106) for 12 weeks. The Hamilton Depression Rating Scale and Young Mania Rating Scale were used to evaluate clinical response during weeks 0, 1, 2, 4, 8, and 12. The genotypes of the COMT Val158Met polymorphism were determined using polymerase chain reaction plus restriction fragment length polymorphism analysis. To adjust for within-subject dependence over repeated assessments, multiple linear regression with generalized estimating equation methods was used. FINDINGS/RESULTS: When stratified by the COMT Val158Met genotypes, significantly greater decreases in Hamilton Depression Rating Scale scores were found in the VPA + DM (30 mg/d) group in patients with the Val/Met genotype (P = 0.008). CONCLUSIONS: We conclude that the COMT Val158Met polymorphism may influence responses to DM (30 mg/d) by decreasing depressive symptoms in BD patients.