There is increasing emphasis on the importance of practising evidence-based medicine. Randomized controlled trials are the standard way to assess the benefits of an intervention, and observational studies are not usually accorded much weight; the results are likely to be considered misleading. For rare diseases, there are great difficulties in obtaining adequate evidence for interventions or for the benefits of early diagnosis. This is because the disorders are not only very rare but also have variable expression, may have very long courses, and have incompletely known late effects; and surrogate end-points often have to be used. Randomized controlled trials are usually impossible because of inadequate power, and because there are preconceived notions of the effects of treatments already in use. The adoption of the best possible design for observational trials, formation of a central registry of such trials, and a greater general appreciation of the problems that rare diseases pose will help in obtaining the best possible evidence for the effects of interventions.