Simultaneous Population and Dose Selection in Clinical Trials and Cluster ValidationLi, Siyu (2014) Simultaneous Population and Dose Selection in Clinical Trials and Cluster Validation. Doctoral Dissertation, University of Pittsburgh. (Unpublished)
AbstractIn clinical trials, the population of interest may be heterogeneous with regard to a subject's protein expression level, genotype, or other characteristics, e.g., age or initial disease severity. In particular, there can exist a subpopulation of subjects with certain characteristics that are more sensitive to the targeted agents. Wang et al. (2009) suggested a two-stage design involving subpopulation enrichment along with a sample size adaptation in the second stage when evaluating the treatment effects on the overall population and the subpopulations. An important component of drug development is to select the minimum effective dose (MED). Multiple comparisons and adaptive designs have been used for dose selection, typically in Phase 2 clinical trials. In this research, we consider Phase 2 clinical trials with multiple populations and multiple doses. We propose methodologies for both non-adaptive and adaptive designs to select the most desired dose and population to enter the Phase 3 confirmative clinical trials. A testing scheme is established under the closed testing principle to strongly protect the familywise type I error rate for the population and MED choice for both non-adaptive and adaptive designs. Flexible test orderings are considered in order to achieve the largest power for a variety of study goals. In related research for post-mortem tissue studies where we again study the heterogeneity of a population, we externally validate a previous subpopulation finding in a schizophrenia population. Previous research of ours had suggested a subpopulation of all individuals diagnosed with schizophrenia (Volk et al. 2012). This subpopulation was termed the low GABA marker (LGM) cluster. A new study was undertaken to validate these findings. In our research we first extend the classification approach proposed by Kapp and Tibshirani (2007) and apply it to the validating data set. Then we apply the clustering analysis, as used in the previous research, on the validating data set and the combination of the defining and validating data sets to again demonstrate that the LGM finding is valid. Share
Details
MetricsMonthly Views for the past 3 yearsPlum AnalyticsActions (login required)
|
![[feed]](/images/feed-icon-32x32.png){kind=icon}