This was a double-blind clinical trial, with a crossover design, to compare the efficacy of a non-steroidal anti-inflammatory drug, diclofenac sodium, intramuscularly administered, and placebo in the treatment of migraine attacks. The drug was administered to 40 patients once a day in three consecutive migraine attacks. If pain still remained after 6 h following administration the patient was given a 100 mg diclofenac sodium suppository, in open condition. Evaluation was by a complete medical examination performed by the physician and by the patient completing a specially designed self-assessment card. A total of eight patients dropped out of the trial (all during placebo administration): three due to poor compliance, four for refusal to continue and one because no further migraine attacks developed. Results were analysed after having checked the absence of both period and carry-over effects. In all cases diclofenac sodium was more effective than placebo (P less than 0.01). This was also confirmed by data obtained from the patient self-assessment cards (P less than 0.001) and by preferences expressed by patients at the end of the trial (P less than 0.001). Tolerance to the drug was similar to that of placebo.