ENBIS: European Network for Business and Industrial Statistics

Actual and simulated trials in clinical research

23 September 2009, 09:00 – 09:45

Abstract

Submitted by

Thomas Svensson

Authors

Alessandra Giovagnoli, joint with Maroussa Zagoraiou

Abstract

Clinical trials are experiments on patients or healthy volunteers aimed at improving medical care. They are classified according to their purpose (Phase I is for safety of new drugs or devices, Phase II is for efficacy and Phase III for comparing new with standard-of-practice); statistical methods play a prominent role in the design as well as the analysis of the results.
Clinical trials are conducted according to well-defined rules laid out in the protocol of the trial, which includes the actual experimental plan, namely the sample size, the sampling rule, the stopping rule (for harm, efficacy or futility), the treatment allocation rule, etc. Recruitment of subjects is typically sequential so the attention of statisticians has recently focused on ways of designing trials that take place in several stages, using accumulated information to decide how to modify aspects of the study without undermining the validity of the study. A brief overview of several such designs (group sequential, sample size reassessment, covariate adjusted, response adaptive randomization, adaptive dose-finding) will be given.
The last decade has seen the advent of clinical trial simulation as a systematic tool to be employed in clinical drug development. This method of investigation is not aimed at replacing real life trials; rather, physical and computer experiments are two complementary sources of information with distinct roles and different degrees of cost, speed, and reliability. Recognizing the value of this new technology, the Food and Drug Administration have recommended that industries expand and accelerate development of simulated clinical trials. We present an attempt at formalizing an approach to clinical research that alternatively employs both physical and computer experiments. A strong emphasis is on the validating methodology of simulation, seen as a way of testing the agreement between two different sources of information. Similarities and differences with numerical experiments for product/process improvement will be highlighted.

Return to programme