Wang, Yuanyuan
(2011)
Simulation Experiment Platform for Evaluating Clinical Trial Designs, with Applications to Phase 1 Dose-Finding Clinical Trials.
Doctoral Dissertation, University of Pittsburgh.
(Unpublished)
Abstract
Clinical Trial (CT) simulation is used by academic research centers and pharmaceutical companies to improve the efficiency and accuracy of drug development. Sophisticated commercial software for CT simulations is available for those with resources to cover fees and with design challenges that happen to match the software's capabilities. Academic research centers usually use locally developed or shared software for study design, mainly due to cost and flexibility considerations. Inspired by the success and immense influence of open-source software development projects, we are building an open-source simulation experiment platform with the intention of utilizing the power of distributed study design expertise, development talent, and peer review of code. The code base relies on S4 classes and methods within R. Design, baseline characteristic model, population model, outcome model, and evaluation criterion are five key object types. An action queue-based approach allows for complex decision making at the patient or CT level. Name matching mechanism is used to check interoperability among the objects. Extensibility, reuse and sharing come from the class/method architecture, together with automatic object and documentation discovery mechanisms.An extensive literature review of existing design evaluation criteria did not reveal the use of criteria based on utility functions. In this dissertation, we propose flexible criteria for evaluating Phase I trial designs by assessing through CT simulation the expected total personal utility, societal utility and total utility. To illustrate the application, we present several examples using the platform to investigate important questions in clinical trial designs. Specifically, we look at the logit model in the continual reassessment method (CRM), choices of parameterization and prior distribution for its model parameters, and the effect of patient heterogeneity on the performance of the standard "3+3" design and the CRM.This work creates an open-source highly flexible and extensible platform for evaluating CT designs via simulation, and promotes collaborative statistical software development. Its impact on public health will manifest itself in greatly speeding and expanding thorough and thoughtful design evaluations when developing clinical trials, for a community of CT designers.
Share
| Citation/Export: |
|
| Social Networking: |
|
Details
| Item Type: |
University of Pittsburgh ETD
|
| Status: |
Unpublished |
| Creators/Authors: |
|
| ETD Committee: |
|
| Date: |
31 January 2011 |
| Date Type: |
Completion |
| Defense Date: |
9 September 2010 |
| Approval Date: |
31 January 2011 |
| Submission Date: |
27 October 2010 |
| Access Restriction: |
5 year -- Restrict access to University of Pittsburgh for a period of 5 years. |
| Institution: |
University of Pittsburgh |
| Schools and Programs: |
School of Public Health > Biostatistics |
| Degree: |
PhD - Doctor of Philosophy |
| Thesis Type: |
Doctoral Dissertation |
| Refereed: |
Yes |
| Uncontrolled Keywords: |
clinical trial designs; open-source software; phase I dose-finding clinical trials; S4 classes and methods |
| Other ID: |
http://etd.library.pitt.edu/ETD/available/etd-10272010-204135/, etd-10272010-204135 |
| Date Deposited: |
10 Nov 2011 20:03 |
| Last Modified: |
15 Nov 2016 13:50 |
| URI: |
http://d-scholarship.pitt.edu/id/eprint/9525 |
Metrics
Monthly Views for the past 3 years
Plum Analytics
Actions (login required)
 |
View Item |
![[feed]](/images/feed-icon-32x32.png){kind=icon}