We aimed to compare the efficacy of valproic acid (VPA) on paresthesias and sleep in RLS to that of levodopa (LD). Twenty patients with idiopathic restless legs syndrome (RLS) were treated with 600 mg slow-release VPA and 200 mg slow-release LD+50mg benserazid in a randomized, placebo-controlled, cross-over, double-blind setting with polysomography (PSG) at the end of each 3-week treatment periods. There was no major difference between the efficacy of valproic acid or LD. Periodic leg movements in sleep (PLMS) and PLM arousal index (PLMAI) significantly decreased with LD (p < or= 0.005). However, LD, but not VPA, significantly increased arousals not associated with PLMS (p = 0.002). Decrease of intensity and duration of RLS symptoms were more pronounced with VPA (p < or= 0.022) than with LD (NS). We conclude that slow-release VPA provides a treatment alternative for RLS.