Memantine is a non-competitive N-methyl-d-asparate (NMDA) receptor antagonist with a mood-stabilizing effect. We investigated whether using valproic acid (VPA) plus add-on memantine to treat bipolar II disorder (BP-II) is more effective than using VPA alone (VPA + Pbo). We also evaluated, in BP-II patients, the association between the brain-derived neurotrophic factor (BDNF) Val66Met polymorphism with treatment response to VPA + add-on memantine and to VPA + Pbo. In this randomized, double-blind, controlled 12 wk study, BP-II patients undergoing regular VPA treatments were randomly assigned to a group: VPA + Memantine (5 mg/day) (n = 115) or VPA + Pbo (n = 117). The Hamilton Depression Rating Scale (HDRS) and Young Mania Rating Scale (YMRS) were used to evaluate clinical response during week 0, 1, 2, 4, 8 and 12. The genotypes of the BDNF Val66Met polymorphisms were determined using polymerase chain reactions plus restriction fragment length polymorphism analysis. To adjust within-subject dependence over repeated assessments, multiple linear regression with generalized estimating equation methods was used to analyze the effects of the BDNF Val66Met polymorphism on the clinical performance of memantine. Both groups showed significantly decreased YMRS and HDRS scores after 12 wk of treatment; the differences between groups were non-significant. When stratified by the BDNF Val66Met genotypes, significantly greater decreases in HDRS scores were found in the VPA + memantine group in patients with the Val Met genotype (p = 0.004). We conclude that the BDNF Val66Met polymorphism influenced responses to add-on memantine by decreasing depressive symptoms in patients with BP-II.