Ethics and statistical methodology in clinical trials.

摘要

Statisticians in medicine can disagree on appropriate methodology applicable to the design and analysis of clinical trials. So called Bayesians and frequentists both claim ethical superiority. This paper, by defining and then linking together various dichotomies, argues there is a place for both statistical camps. The choice between them depends on the phase of clinical trial, disease prevalence and severity, but supremely on the ethics underlying the particular trial. There is always a tension present between physicians primarily obligated to their own patients (the weight of 'individual ethics') and ethical committees responsible for the scientific merit of the trial and its long-term implications ('collective ethics'). Individual ethics, it is proposed, favour the Bayesian approach; collective ethics, the frequentist. Though in some situations the choice appears clear-cut, there remain other where both methodologies can be appropriate.