In this study the human glyceryltrinitrate (GTN) model of migraine was for the first time used to test the effect of a prophylactic drug. We chose to test valproate due to its well documented effect as a migraine prophylactic drug. Efficacy of this compound would support the usefulness of the model in prophylactic antimigraine drug development. Twelve patients with migraine without aura were included in a randomized double blind crossover study. Valproate 1000 mg or placebo was given daily, each for a minimum of 13 days. On the last treatment day of each arm a 20 min intravenous infusion of GTN (0.25 μg/kg/min) was given. Headache was registered for 12 h after the infusion and headache intensity was scored on a scale from 0 to 10. Fulfillment of IHS criteria was recorded for 24 h. The middle cerebral arteries were evaluated by transcranial Doppler and the diameter of the superficial temporal and radial arteries were measured with high frequency ultrasound. GTN evoked migraine fulfilling IHS criteria 1.1 in 6 patients after placebo and in 2 patients after valproate (P = 0.125). Including additionally 3 patients on placebo and 1 patient on valproate who felt they had suffered a migraine attack, but who had as associated symptoms only photophobia or phonophobia, a significant reduction in the number of patients with induced migraine after valproate was seen (P = 0.031). Median peak headache intensity was 1 (range 0–9) after valproate compared to 4.5 (range 0–8) after placebo (P = 0.120). Pretreatment with valproate as compared to placebo reduced the velocity in both middle cerebral arteries after GTN (left P = 0.021, right P = 0.031). No effect of valproate was seen in the diameter of the superficial temporal artery (P = 0.781) or the radial artery (P = 0.367) before or after GTN. The study indicates that a prophylactic effect of valproate may be demonstrated using the GTN human migraine model. Although, all headache parameters were reduced after valproate compared to placebo, only one parameter was statistically significantly reduced probably because of the small number of patients. The size of the effect was similar to that of valproate in clinical trials. The GTN model may therefore be a valid tool for testing new prophylactic antimigraine drugs.