Several problems confront trials of drugs for migraine: no clinical or biological markers for objective diagnosis are available, severity of the attack cannot be readily evaluated, the placebo effect is significant, and both compliance with therapy and patient follow-up may be difficult to ensure. No statistical method is capable of solving these problems and it is the clinicians' role to minimize them by consistent efforts at establishing good communication with their patients during the trial. Because of these significant intrinsic difficulties, trials of drugs for migraine require especially stringent data collection and analysis protocols in order to minimize the risk of errors which exist in any drug trial. The protocol used have not always been satisfactory with this respect. A rating system for helping non-specialists to evaluate the reliability of migraine drug trials is proposed.