STATISTICAL ASPECTS OF CONTROLLED SINGLE SUBJECT TRIALS

Randomized controlled trials in groups and single subjects differ in several statistical aspects. In group trials the experimental unit is a randomly selected subject from a predefined population and this subject is randomly assigned to a treatment. Outcome is confined to average effects which can be generalized to the specific population, but which do not necessarily apply to individual persons. In single subject trials the experimental unit is a treatment period and each treatment period is randomly allocated in a mutiple cross-over sequence of periods. The single subjects is only representative of itself, but similar responses in corresponding single subject trials may justify careful extrapolation of the results. Single subject trials have a high risk of Type ll errors. However, the randomization procedure chosen and the type of statistical test applied may enhance the statistical power of such trials. Internal validity depends on modelling the trial design to the clinical features, drug properties and statistical requirements, while reliability is determined by the reproducibility of the trial response.