Bayesian Clinical Trial Simulation (BCTS)

The Problem

In planning clinical trials, whether conventional efficacy-only trials or ones where economic data is gathered alongside efficacy, a key question is, "If I run the trial according to this design, what is the chance that we will obtain the results we are hoping for?" The desired result might for instance be achieving a statistically significant improvement (in efficacy or cost-effectiveness) for the new treatment compared with the comparator treatment.

Traditional trial design has only addressed this question by designing the trial to reach a requisite power for the significance test. When working with complex trial designs, it is often impossible to obtain analytical expressions for the power, and a common device for computing it is clinical trial simulation.

However, the conventional power calculation only answers the original question obliquely. That is, the power can only equate to the probability of a successful result if the true treatment difference equals the clinically meaningful difference at which the power has been defined (and if the test is one-sided). In reality, the true effect will not exactly equal that clinically meaningful difference. It may be greater or less, and changes in its value change the power calculation. Often, power calculations fix all sorts of other parameters whose values are not really known, such as between-patient variability in response, or baseline incidence.

The question can only be meaningfully answered by including the uncertainty in the value of the true treatment difference, and in any other uncertain parameters. This necessitates a Bayesian perspective, and the relevant computational tool is Bayesian clinical trial simulation (BCTS).

Research at CHEBS

BCTS was introduced in O´Hagan, Stevens and Campbell (2005), following earlier work by O´Hagan and Stevens (2001) on choosing a sample size to achieve a desired probability of a successful outcome. Recent research in CHEBS has been developing a substantial application of BCTS methods for AstraZeneca. The Rheumatoid Arthritis Drug Development Model is a tool based on BCTS to assist AstraZeneca in planning trials for new drugs to treat rheumatoid arthritis. Several papers are in preparation to report this work.

Ongoing and future research

  • Developing practical experience with BCTS through further substantive applications.

  • Methods to improve efficiency, particularly when BCTS is linked to economic modelling to evaluate probabilities of cost-effectiveness.

  • Developing formal methods to update a BCTS system as new information becomes available, e.g. as early trial results are obtained.

Published papers

  • O'Hagan, A., Stevens, J. W. and Campbell, M. J. (2005). Assurance in clinical trial design. Pharmaceutical Statistics 4, 187-201.


  • O'Hagan, A. and Stevens, J. W. (2001). Bayesian assessment of sample size for clinical trials of cost-effectiveness. Medical Decision Making 21, 219-230.