Design efficiency in dose finding studies

Abstract An investigator about to undertake a dose finding study can choose from rival methods. In selecting any method he or she may have certain properties in mind. Our purpose here is to present an approach to evaluating rival methods which can help provide guidance to the investigator. The idea is to determine which of any given number of competing dose finding studies has a superior performance in given circumstances. The approach is centered around three components. The first of these is simply a very broad class of potential dose–toxicity relations, of which current examples in the literature can be seen to be special cases. The second component, for some given situation or class of situations, is an optimal method, not one that we can necessarily apply in practice, but one that provides a gold standard, beyond which improvements are not generally possible. The third component is a class of efficiency measures which indicate how close to the optimal method is the performance of any given method. We are then able to compare rival approaches both in a relative and absolute sense.