Ethical procedures in comparative evaluation of drugs

Abstract

The comparison of sampled distributions is one of the classical problems of statistics. Evaluation of any technique for doing so depends at least implicitly on an assumed cost function. In one of the most important applications, that of choosing treatments for the ailments of man, costs are measured in suffering and death, and must be accorded primary importance. A highly simplified but in fact prototypical situation of this kind is addressed, in which two drugs, A and B, are available for a specified illness, their actions are unequivocal or dichotomic: success or failure, and this action is on a time scale short compared to that on which patients receiving the drugs present themselves. The assumption is that the mode of treatment with each drug is sufficiently stylized that each drug can be fully represented for each patient independently by success probability p. The problem then is to design a sequential selection of treatments so that the drug with the higher success probability is used as often as possible.

Keywords

This publication has 0 references indexed in Scilit: