Tanacetum parthenium (feverfew), is a well-known herb for the prophylactic treatment of migraine. The primary objective was to show a dose–response of a new stable extract (MIG-99) reproducibly manufactured with supercritical CO2 from feverfew (T. parthenium). Furthermore, the study should provide data on the safety and tolerability of MIG-99. In a randomized, double-blind, multicentre, controlled trial with an adaptive design, the clinical efficacy and safety of three dosages of MIG-99 (2.08 mg; 6.25 mg; 18.75 mg t.i.d.) were compared with placebo. The patients (n=147) suffered from migraine with and without aura according to International Headache Society (IHS) criteria and were treated with one of the study medications for 12 weeks after a 4-week baseline period. The primary efficacy parameter was the number of migraine attacks during the last 28 days of the treatment period compared with baseline. Secondary endpoints were total and average duration and intensity of migraine attacks, mean duration of the single attack, number of days with accompanying migraine symptoms, number of days with inability to work due to migraine as well as type and amount of additionally taken medications for the treatment of migraine attacks. The design of the study included a pre-planned adaptive interim analysis for patients with at least four migraine attacks within the baseline period. With respect to the primary and secondary efficacy parameter, a statistically significant difference was not found between the overall and the confirmatory intention-to-treat (ITT) sample in the exploratorily analysed four treatment groups. The frequency of migraine attacks for the predefined confirmatory subgroup of patients (n=49) with at least four migraine attacks during the baseline period decreased in a dose-dependent manner (P=0.001). The highest absolute change of migraine attacks was observed under treatment with 6.25 mg t.i.d. (mean±SD=−1.8±1.5 per 28 days) compared with placebo (−0.3±1.9; P=0.02). Overall, 52 of 147 (35%) patients reported at least one adverse event (AE). The incidence of AEs in the active treatment groups was similar to that in the placebo group, and no dose-related effect was observed in any safety parameter. MIG-99 failed to show a significant migraine prophylactic effect in general. Accordingly, in the ITT analysis a dose–response relationship could not be observed. MIG-99 was shown to be effective only in a small predefined subgroup of patients with at least four attacks during the 28-day baseline period where the most favourable benefit–risk ratio was observed with a dosage of three capsules of 6.25 mg MIG-99 extract per day. Because of the low number of patients, these findings need to be verified in a larger sample. The incidence of AEs was similar for all treatment groups.