[Controversial viewpoints on the use of historical controls: Is there an all-or-nothing law?].

In the field of clinical trials the choice of appropriate controls has often given rise to controversy. Historical controls rely too often on biased, incomplete, and inappropriate information. Historical controls may be suitable if the natural history of the illness is known and stable or for planning prospective studies. The minimal requirement for use of historical controls is to explain the methods by which the data were collected. When the goal of clinical investigation is to test for a true difference, a randomized clinical study should be performed.