Use this URL to cite or link to this record in EThOS:
Title: Statistical issues in the design and analysis of early phase proof of concept clinical trials in rheumatoid arthritis
Author: Liu, Feng
Awarding Body: University of Sheffield
Current Institution: University of Sheffield
Date of Award: 2017
Availability of Full Text:
Access from EThOS:
Access from Institution:
Introduction: Rheumatoid Arthritis (RA) is a chronic inflammatory disorder that typically affects joints of the body such as knees, hands and feet. The prevalence of RA is around 1% in the general population and the incidence rate is about 41 per 100,000 persons per year. The use of adaptive designs in drug development is of increasing interests to medical researchers since adaptive design potentially delivers faster and better decisions compared to the traditional fixed sample-size designs. In RA clinical trials, after treatment starts, follow-up and outcome assessments usually take a number of months, with some outcomes collected up to two or more years after the start of study treatment. For this reason, adaptive designs are rarely undertaken in RA clinical trials as the timing of the endpoints make it difficult to make decisions from the interim analyses. Aims: The overall aim of this thesis is to assess whether or not we can apply an adaptive dose finding design, using Bayesian methods, to a Phase 2a Proof of concept (PoC) trials for the development of a new drug or pharmaceutical treatments for patients with RA disease when the dose response is likely to be non-monotonic. Data: The primary data was a PoC study which was re-analysed retrospectively using an adaptive design procedure to investigate the aspects of the study design. Method: In this dissertation, a Bayesian dose response model was applied to the adaptive design of a Phase 2a PoC clinical trial in RA patients. The first part of the design was to “learn” the dose response in multiple dose cohorts and the second part was to “confirm” the selected dose in a group sequential design with an O’Brien Fleming alpha spending correction. A delayed response predictive model using outcomes collected short-term (Day 14 post-randomisation) for the prediction of longer term outcomes (Day 56 post-randomisation time points) is proposed, based on retrospective analysis and a literature based meta-analysis which was undertaken to investigate the use of early time points to predict Day 56 responses with the intent of reducing the time taken to make decisions at interim analyses of efficient design. Simulations are used to evaluate the models in the desired settings. Results: It was shown that the two-part adaptive design with “learn” and “confirm” could be implemented in the Phase 2a PoC study with two potential improvements: 1) The delayed response predictive model can be used to predict Disease Activity Score (DAS28) Day 56 outcome based on Day 14 outcome to expedite the time to interim decisions in the context of a Phase 2a PoC design; 2) The Bayesian Normal Dynamic Linear Model (NDLM) can be used in the dose response analysis to handle both monotonic and non-monotonic dose responses without sacrificing statistical power or design performance. Conclusion: This thesis demonstrates that it is possible to apply an adaptive design to a PoC study in the treatment of RA. It is recommended that the dose response design with Bayesian NDLM model using predicted DAS28 outcomes at Day 56 based on Day 14 data can expedite the interim decision making. In most cases the Bayesian Emax model works effectively and efficiently, with low bias and a good probability of success in the case of a monotonic dose response. However, if there is a belief that the dose response could be non-monotonic based on prior knowledge then the NDLM is the superior model to assess the dose response. Based on the trial design proposed if the predictive model can be applied to a future adaptive trial, there is potential for a significant time-saving in Phase 2a study.
Supervisor: Julious, Steven ; Walters, Stephen Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available