Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


DissertationCorrectedSiyoenKil.pdf (3.59 MB)

ETD Abstract Container

Abstract Header

Finding a Targeted Subgroup with Efficacy for Binary Response with Application for Drug Development

Kil, Siyoen
http://rave.ohiolink.edu/etdc/view?acc_num=osu1386082319

Abstract Details

2013, Doctor of Philosophy, Ohio State University, Statistics.
Identifying subpopulations that benefit from treatment or whose benefit is enhanced over the population at large can improve healthcare for patients and targeted marketing for drug developers. The goal of subgroup analyses in clinical trials is to quantify such heterogeneity of treatment effects across subpopulations. One way of inferring an enhanced drug effect on the target population is to test the treatment- biomarker interaction. A significant treatment-biomarker interaction indicates that the biomarker is predictive and thus can be used to classify the target population with enhanced efficacy. On the other hand, if a single continuous biomarker is involved, we often try to find a threshold that divides the total population into two subsets, one of which is the target population with meaningful efficacy. Permutation tests have been proposed for both testing interactions and finding a threshold in subgroup analyses for personalized medicine. The focus of this dissertation is to both demonstrate the inadequacy of simple permutation testing in each of these applications and to propose valid alternatives. First, the dissertation precisely explores the validity of permutation-based reference distributions for testing treatment-biomarker interactions in randomized clinical trials with binary biomarkers and outcomes. It turns out that the most prevalent method of permutation does not produce a valid reference distribution as expected. Thus, we propose an atypical method of permutation to test for the existence of an interaction term that is equivalent to the underlying idea of an exact conditional logistic regression. Second, we show that proposed permutation-based methods to find a threshold in fact test an inadequate null hypothesis. Instead of using a permutation distribution, we propose a flexible parametric decision-making procedure that can solve the questions not only of how to test for the existence of a the target population but also of how to find the biggest target population that benefits from the treatment on average.
Kaizar Eloise (Advisor)
134 p.
Statistics
Subgroup Analysis; Permutation Testing; Personalized Medicine; Drug development; Exact Conditional Logistic Regression

Recommended Citations

Citations

  • Kil, S. (2013). Finding a Targeted Subgroup with Efficacy for Binary Response with Application for Drug Development [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1386082319

    APA Style (7th edition)

  • Kil, Siyoen. Finding a Targeted Subgroup with Efficacy for Binary Response with Application for Drug Development. 2013. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1386082319.

    MLA Style (8th edition)

  • Kil, Siyoen. "Finding a Targeted Subgroup with Efficacy for Binary Response with Application for Drug Development." Doctoral dissertation, Ohio State University, 2013. http://rave.ohiolink.edu/etdc/view?acc_num=osu1386082319

    Chicago Manual of Style (17th edition)

osu1386082319
684
© 2013, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.