Statistical theory of prophylactic and therapeutic trials. I. Limitations of the unique null hypothesis.

Abstract

validify the testimony of the patient or the clinical judgement of the author, a clinical trial so conceived violates any or all of three canons of scientific method, as is now becoming recognized widely in all branches of medicine, except perhaps psychiatry. A change of outlook is largely due to the impact of more exacting standards of evidence established at an earlier date in connexion with the assessment of prophylactic measures, partly as a consequence of public controversy over the merits of vaccination. In this context, the term prophylactic calls for no comment. We here employ the expression therapeutic measures in the widest sense, including administration of drugs or convalescent sera, operative and manipulative surgery, diathermy and radiation (deep X-ray, short wave, radium) treatments, occupational and physio therapy (including remedial gymnastics, faradization, massage), rehabilitation techniques. Bradford Hill (1951), who has himself directed a series of therapeutic trials on the now familiar pattern expounded by Greenwood (1935), has lately set forth in clear and simple language some of the essential safeguards of a scientific assessment of remedial measures; and there is no need to recapitulate them in this context. Our aim in what follows is to examine statistical procedures invoked to validate results within a framework of the precautions to which he has drawn attention; but it will simplify our task if we first specify the desiderata. Prophylactic or thera