This paper is from Session 6D: Medical statistics

which comes under Topic 6: Statistics education, training and the workplace

Other papers under session 6D ... 6D1: Continuous variables: to categorise or to model? 6D3: The applied statistical scientist in a high-profile academic...

(Tuesday 13th, 16:30-18:00)

Control in clinical trials

Presenter

• Stephen Senn (University of Glasgow, United Kingdom)

Abstract

Amongst the many types of medical scientific investigation that are possible, the randomized double-blind controlled clinical trial has a very high reputation. Without control there can be no randomization and without randomization no convincing blinding, It seems, therefore, that control is the key feature of such trials. Yet the way in which such trials are analyzed, including the way in which they are presented, shows that many trialists do not understand the value of what they have done. I illustrate the problem with various examples. One possible reason that trialists may underestimate the value of concurrent control is that they do not understand what a powerful source of bias regression to the mean constitutes. I consider how physicians can be taught to understand this difficult phenomenon. I also present a striking example of differences between populations, which can be used to teach care in using control information appropriately.

Other papers under session 6D ... 6D1: Continuous variables: to categorise or to model? 6D3: The applied statistical scientist in a high-profile academic...