Back to search

University of Minnesota

Innovations in Dose-Finding Designs Using Bayesian Methods

Abstract

dc:description.abstract

Phase I clinical trials are a relatively short part of the research and development process for a new treatment, but are arguably one of the most important steps. The main goal of a phase I trial is to learn about the risk-benefit profile of different doses of the treatment. With this information, researchers select a dose to use for the subsequent phases of development, with the aim of selecting the truly optimal dose. Selecting a sub-optimal dose can have detrimental effects on future clinical trial phases and on the viability of the treatment. Selecting the optimal dose is done most consistently when the dose-finding design is tailored to the investigational treatment, which may necessitate some innovation to account for situations beyond what existing designs can handle. The proposed trial in Chapter 2 requires a concurrent control arm to assess the safety of the treatment via the relative increase in serum lactate on hospital admission. Six Bayesian models, which characterize expected serum lactate as a function of sodium nitroprusside dose and randomization group, circumvent a major restriction of most dose-finding designs by incorporating a concurrent control arm. These models are proposed and compared for guiding dose-escalation and each model makes a different assumption about the expected change in serum lactate across control cohorts concurrently randomized with each dose. Strategies for selecting a model, choosing a randomization ratio for the non-initial cohort and specifying a maximum sample size are outlined. Through simulation study, the model which uses spike and slab prior distributions on the cohort-specific effects in combination with a 2:1 randomization ratio in the non-initial cohorts is determined to be the best. A maximum sample size of 70 individuals is required to achieve the desired maximum number of dose-escalation errors. In Chapter 3, I propose a phase I/II trial design to support dose-finding when uncertainty exists about whether the optimal biological dose (OBD) differs in two pre-specified subgroups. The proposed design uses a utility function to quantify efficacy-toxicity trade-offs, and a Bayesian model with spike and slab prior distributions for the subgroup effect on toxicity and efficacy to guide dosing and to facilitate identifying either subgroup-specific OBDs or a common OBD depending on the resulting trial data. In a simulation study, I find the proposed design performs nearly as well as a design that ignores subgroups when the dose-toxicity and dose-efficacy relationships are the same in both subgroups, and nearly as well as a design with independent dose-finding within each subgroup when these relationships differ across subgroups. Chapter 4 concerns the subjectivity inherent in techniques that quantify the risk-benefit trade-off for dose-finding designs. I propose two order-based decision criteria that depend only on the ordering of the joint efficacy-toxicity outcomes as a more robust alternative to the standard utility-based or trade-off contour-based decision criteria. In simulation study, the proposed decision criteria performed approximately as well as the comparison decision criteria while reducing the complexity and improving the robustness of the design. I also present an alternative admissibility criteria which uses the desirability of the efficacy-toxicity trade-off rather than marginal rates. This is a more exclusive criteria, but performed well, in simulation study, when the OBD falls within the admissibility constraints.

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Porter, Sydney

Rights

Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Citation
Porter, Sydney. (2023). Innovations in Dose-Finding Designs Using Bayesian Methods. Retrieved from the University Digital Conservancy, https://hdl.handle.net/11299/282389.

Chain of custody

source
Harvested from
University of Minnesota
Base URL
conservancy.umn.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Porter, Sydney. Innovations in Dose-Finding Designs Using Bayesian Methods. 2023. https://hdl.handle.net/11299/282389