The assessment of bias in randomized controlled clinical trials

Dissertation, RWTH Aachen University, 2018; Aachen 1 Online-Ressource : Illustrationen (2018). = Dissertation, RWTH Aachen University, 2018
In controlled clinical trials, randomization is the key technique to ensure the comparability of two or more treatment groups. However, selection bias and time trends can lead to covariate imbalances that can influence the results of a clinical trial despite randomization. The unified assessment of selection bias and time trends is the objective of this thesis.Within the scope of this thesis, a sequence based approach for the assessment of selection bias and time trend was developed, and an explicit form for the influence of selection bias on the type-I-error rate in multi-arm trials was derived. A bias adjusted test was proposed and its power was investigated via simulations. The main result is the randomizeR software package in R that unifies the assessment of randomization procedures with respect to their susceptibility to selection and chronological bias. The sequence based approach allows the graphical presentation of the differences of the randomization procedures based on the type I and II error probabilities. For certain randomization procedures, the selection bias adjusted test can detect the treatment effect without substantial loss of power. In total, researchers are supported in choosing a randomization procedure that isleast susceptible to bias given the clinical scenario of the specific trial.
Published by Aachen