Methods for Blinded Sample Size Recalculation in Adaptive Enrichment Designs

by Marius Placzek
Cumulative thesis
Date of Examination:2022-10-26
Date of issue:2022-12-21
Advisor:Prof. Dr. Tim Friede
Referee:Prof. Dr. Tim Friede
Referee:Prof. Dr. Thomas Kneib
Persistent Address: http://dx.doi.org/10.53846/goediss-9638

 

 

Files in this item

Name:diss_summary_pub.pdf
Size:2.45Mb
Format:PDF

The following license files are associated with this item:

Abstract

English

In precision medicine the potential heterogeneity of patient populations becomes more accessible through technical and medical innovations. A population might be heterogeneous and different strata may in turn respond heterogeneously to a certain treatment. A certain proportion of the population might respond not at all or even negatively to the investigated treatment. Personalized medicine and targeted therapies aim to use this information to find tailored treatment for subjects depending on subgroup affiliation, increasing the success rate of treated subjects. Therefore, clinical trials with the ability to reveal an increased treatment benefit in particular subgroups compared to the whole population are in great demand. An efficient and flexible option promises an adaptive enrichment designs with blinded sample size recalculation in an internal pilot study. I split the design into its two main components which I first analyzed separately: On the one hand there is the analysis, sample size determination and blinded sample size recalculation in a multiple subgroups design. On the other hand there is the adaptive enrichment design with subgroup selection at an interim analysis. Both topics individually pose several challenges, e.g. concerning variability across the (sub)populations, blinded reestimation of parameters or distributional properties of the test statistics. Having analyzed those issues step by step I finally put them together once again presenting a combined method for blinded sample size recalculation in adaptive enrichment designs. In all three parts the performance of the proposed methods were examined with simulations in R.
Keywords: adaptive design; BSSR; interim analysis; enrichment; subgroup analysis
 
Statistik