Events Calendar

26 Jul
New Statistical Insights to Precision Medicine, from Targeted Treatment Development to Individualized Tailoring Recommendation
Event Type

Defenses

Topic

Research

Target Audience

Faculty, Graduate Students

University Unit
Department of Biostatistics
Subscribe
Google Calendar iCal Outlook

New Statistical Insights to Precision Medicine, from Targeted Treatment Development to Individualized Tailoring Recommendation

Doctoral Candidate Yue Wei defends her dissertation on "New Statistical Insights to Precision Medicine, from Targeted Treatment Development to Individualized Tailoring Recommendation”

Advisors: Ying Ding, PhD & Chaeryon Kang, PhD, Department of Biostatistics

Committee Members: 

  • Joyce Chang, PhD, Department of Medicine
  • Jong Jeong, PhD, Department of Biostatistics
  • Yu Cheng, PhD, Department of Statistics

ABSTRACT:

ABSTRACT:
There has been increasing interest in discovering precision medicine in current pharmaceutical drug development and medical research.  One aspect of precision medicine research is to develop new therapies that target a subgroup of patients who exhibit enhanced treatment efficacy through clinical trials.  Another aspect is to tailor existing therapies to individual patient so that each patient can get the most “suitable” treatment.  Motivated by analyzing the Age-Related Eye Disease Study (AREDS) data, a large randomized controlled trial (RCT) to study the efficacy of nutritional supplements in delaying the progression of an eye disease, age-related macular degeneration (AMD), this dissertation proposes new statistical methods to address key issues in both aspects of precision medicine research.

The first part of my dissertation concerns with confidently identifying subgroups with differential treatment efficacy in RCTs.  I propose a novel multiple-testing-based approach to simultaneously identify and infer subgroups with enhanced treatment efficacy in RCTs. Specifically, I formulate the null hypotheses through contrasts and construct their simultaneous confidence intervals, which control both within- and across-marker multiplicity appropriately. Two types of outcomes are considered:  survival endpoint (e.g., progression time to late-AMD) and binary endpoint (e.g., 10-year progression status). I demonstrate the use of a logic-respecting efficacy measure and develop the subgroup mixable estimation procedure for each outcome.  Extensive simulations are conducted to evaluate the method performance and to provide practical guidance.  The method is then applied to AREDS data to assess the efficacy of antioxidants and zinc combination in delaying AMD progression.  Multiple gene regions including ESRRB-VASH1 on chromosome 14 and CHST3-SPOCK2 on chromosome 10 have been identified with subgroups showing differential efficacy in different genotype groups. I further validate our findings in an independent subsequent RCT, AREDS2, by discovering consistent differential treatment responses in the targeted and non-targeted sub-groups identified from AREDS.

The second part of my dissertation concerns with detecting and estimating heterogeneous treatment effect (HTE) so that individualized tailoring recommendation can be provided. Although various machine-learning-based methods have been proposed in the literature to estimate individualized treatment effects (ITE), existing approaches for estimating ITEs with survival outcomes are limited.  I implement three machine-learning methods, random survival forest (RSF), Bayesian accelerated failure time models (BAFT), and Cox-based DNN survival prediction model (DNNSurv) into the framework of two meta-algorithms: T-learner and X-learner, to accurately estimate ITEs with survival outcomes, measured by the difference of survival probabilities between the two treatments. Treatment recommendation is made based on patients’ ITE estimates and evaluated by various prediction performance metrics. The merits of the proposed methods are investigated with comprehensive simulation studies. Finally, I apply the proposed approach on AREDS data to estimate ITEs for tailoring recommendation. Boruta algorithm is used to identify the top variables that contribute to HTEs.

Public health significance: This dissertation addresses two fundamental questions that are in the heart of precision medicine research: (1) in targeted treatment development, whether there exists subgroup of patients with beneficial treatment efficacy and how to correctly infer efficacy in targeted, untargeted and overall mixture patient population; (2) in tailoring existing therapies, how to accurately estimate ITEs and determine the important features that lead to the heterogeneity. It has the potential to fundamentally improve the current practice in analyzing treatment effects from RCTs, and thus to increase the success of modern drug development and precision medicine research.
 

Dial-In Information

Join Zoom Meeting

https://pitt.zoom.us/j/97230380206

Passcode: 285132

Monday, July 26 at 2:00 p.m. to 4:00 p.m.

Virtual Event

New Statistical Insights to Precision Medicine, from Targeted Treatment Development to Individualized Tailoring Recommendation

Doctoral Candidate Yue Wei defends her dissertation on "New Statistical Insights to Precision Medicine, from Targeted Treatment Development to Individualized Tailoring Recommendation”

Advisors: Ying Ding, PhD & Chaeryon Kang, PhD, Department of Biostatistics

Committee Members: 

  • Joyce Chang, PhD, Department of Medicine
  • Jong Jeong, PhD, Department of Biostatistics
  • Yu Cheng, PhD, Department of Statistics

ABSTRACT:

ABSTRACT:
There has been increasing interest in discovering precision medicine in current pharmaceutical drug development and medical research.  One aspect of precision medicine research is to develop new therapies that target a subgroup of patients who exhibit enhanced treatment efficacy through clinical trials.  Another aspect is to tailor existing therapies to individual patient so that each patient can get the most “suitable” treatment.  Motivated by analyzing the Age-Related Eye Disease Study (AREDS) data, a large randomized controlled trial (RCT) to study the efficacy of nutritional supplements in delaying the progression of an eye disease, age-related macular degeneration (AMD), this dissertation proposes new statistical methods to address key issues in both aspects of precision medicine research.

The first part of my dissertation concerns with confidently identifying subgroups with differential treatment efficacy in RCTs.  I propose a novel multiple-testing-based approach to simultaneously identify and infer subgroups with enhanced treatment efficacy in RCTs. Specifically, I formulate the null hypotheses through contrasts and construct their simultaneous confidence intervals, which control both within- and across-marker multiplicity appropriately. Two types of outcomes are considered:  survival endpoint (e.g., progression time to late-AMD) and binary endpoint (e.g., 10-year progression status). I demonstrate the use of a logic-respecting efficacy measure and develop the subgroup mixable estimation procedure for each outcome.  Extensive simulations are conducted to evaluate the method performance and to provide practical guidance.  The method is then applied to AREDS data to assess the efficacy of antioxidants and zinc combination in delaying AMD progression.  Multiple gene regions including ESRRB-VASH1 on chromosome 14 and CHST3-SPOCK2 on chromosome 10 have been identified with subgroups showing differential efficacy in different genotype groups. I further validate our findings in an independent subsequent RCT, AREDS2, by discovering consistent differential treatment responses in the targeted and non-targeted sub-groups identified from AREDS.

The second part of my dissertation concerns with detecting and estimating heterogeneous treatment effect (HTE) so that individualized tailoring recommendation can be provided. Although various machine-learning-based methods have been proposed in the literature to estimate individualized treatment effects (ITE), existing approaches for estimating ITEs with survival outcomes are limited.  I implement three machine-learning methods, random survival forest (RSF), Bayesian accelerated failure time models (BAFT), and Cox-based DNN survival prediction model (DNNSurv) into the framework of two meta-algorithms: T-learner and X-learner, to accurately estimate ITEs with survival outcomes, measured by the difference of survival probabilities between the two treatments. Treatment recommendation is made based on patients’ ITE estimates and evaluated by various prediction performance metrics. The merits of the proposed methods are investigated with comprehensive simulation studies. Finally, I apply the proposed approach on AREDS data to estimate ITEs for tailoring recommendation. Boruta algorithm is used to identify the top variables that contribute to HTEs.

Public health significance: This dissertation addresses two fundamental questions that are in the heart of precision medicine research: (1) in targeted treatment development, whether there exists subgroup of patients with beneficial treatment efficacy and how to correctly infer efficacy in targeted, untargeted and overall mixture patient population; (2) in tailoring existing therapies, how to accurately estimate ITEs and determine the important features that lead to the heterogeneity. It has the potential to fundamentally improve the current practice in analyzing treatment effects from RCTs, and thus to increase the success of modern drug development and precision medicine research.
 

Dial-In Information

Join Zoom Meeting

https://pitt.zoom.us/j/97230380206

Passcode: 285132

Monday, July 26 at 2:00 p.m. to 4:00 p.m.

Virtual Event

Event Type

Defenses

Topic

Research

Target Audience

Faculty, Graduate Students

University Unit
Department of Biostatistics

Powered by the Localist Community Events Calendar ©