ABSTRACT
Clinical trial simulation (CTS) depends fundamentally on a set of models to simulate observations that might arise in a clinical trial. Three distinct categories of model have been proposed (1):
• Covariate distribution • Input-output • Execution
They are presented in this sequence because the first decision that must be made when designing a clinical trial is what kind of subjects will be enrolled. The covariate distribution model defines the population of subjects in terms of their characteristics, such as weight, renal function, sex, etc. Next, the input-output model can be developed to predict the observations expected in each subject using that individual’s characteristics defined by the covariate distribution model. Finally, deviations from the clinical trial protocol may arise during execution of the trial. These may be attributed to subject withdrawal, incomplete adherence to dosing, lost samples, etc. The execution model will modify the output of the input-output model to simulate these sources of variability in actual trial performance.
