Bayesian monitoring of clinical trials: Examples using conjugate priors

A clinical trial can save time and resources if it incorporates Bayesian monitoring. Generally speaking, conducting Bayesian analysis is a computationally intensive task. However, in the special case of hypotheses testing for clinical trials, and, moreover, when conjugate prior distributions of parameters are used, computational complexity is reduced remarkably. This thesis presents three examples where the Bayesian monitoring is achieved with a prior density of a parameter and the likelihood function of the data belonging to conjugate families of distributions. The first example studies a heart valve trial with a Poisson rate of adverse events and a gamma prior distribution of the rate. The second example focuses on testing certain drug efficacy for lowering high blood pressure, with self-conjugate normal family of distributions. In the third example, the probability of a false positive alarm produced by a heart defibrillator is modeled with beta prior distribution conjugate to binomial likelihood function.