The occurrence, severity, and duration of patient adverse events are routinely recorded during randomised clinical trials. This data is used by a trial's Data Monitoring Committee to make decisions regarding the safety of a treatment and may lead to the alteration or discontinuation of a trial if real safety issues are detected. There are many different types of adverse event and the statistical analysis of this data, particularly with regard to hypothesis testing, must take into account potential multiple comparison issues.Unadjusted hypothesis tests may lead to large numbers of false positive results, but simple adjustments are generally too conservative. In addition, the anticipated effect sizes of adverse events in clinical trials are generally small and consequently the power to detect such effects is low.A number of recent classical and Bayesian methods, which use groupings of adverse events, have been proposed to address this problem. We illustrate and compare a number of these approaches, and investigate if their use of a common underlying model, which involves groupings of adverse events by body-system or System Organ Class, is useful in detecting adverse events associated with treatments.For data where this type of grouped approach is appropriate, the methods considered are shown to correctly flag more adverse event effects than standard approaches, while maintaining control of the overall error rate.While controlling for multiple types of adverse event, these proposed methods do not take into account event timings or patient exposure time, and are more suited to end of trial analysis. In order to address the desire for the early detection of safety issues in clinical trials a number of Bayesian methods are introduced to analyse the accumulation of adverse events as the trial progresses, taking into account event timing, patient time in study, and body-system.These methods are suitable for use at interim trial safety analyses. The models which performed best were those that had a common body-system dependence over the duration of the trial.
|Date of Award||19 Oct 2017|
- University Of Strathclyde
|Sponsors||EPSRC (Engineering and Physical Sciences Research Council)|
|Supervisor||Chris Robertson (Supervisor) & David Young (Supervisor)|