SUMMARYPurpose: The frequency of valproate (VPA)-induced thrombocytopenia varied widely in previous studies, due to methodological differences. Our objective was to evaluate the relationship between trough VPA plasma levels and platelet counts and assess risk factors for the development of thrombocytopenia. Methods: Patients with refractory partial epilepsy were enrolled in this double-blind, multicenter, concentration-response trial that evaluated the efficacy and safety of high versus low trough plasma VPA concentrations following administration of divalproex sodium as monotherapy. Trough VPA concentrations and concomitant platelet counts were drawn at baseline and intermittently throughout the 24-week trial. Bivariate correlations and multivariate stepwise regression analysis were performed between platelet counts and multiple variables. A logistic regression analysis was done to determine the probability of developing thrombocytopenia at various VPA levels.Results: A total of 851 VPA levels and concomitant platelet counts were analyzed in 265 patients. Of these, 17.7% of patients experienced at least one episode of thrombocytopenia (platelet count ≤ 100,000/μl) after exposure to divalproex sodium. A significant negative correlation was found between VPA levels and platelet counts. Women were significantly more likely to develop thrombocytopenia. The probability of developing thrombocytopenia substantially increased at trough VPA levels above 100 μg/ml in women and above 130 μg/ml in men. Discussion: Our data strongly support a causal relationship between rising plasma VPA levels and reduced platelet counts, with additional risk factors including female gender and lower baseline platelet counts.