Maximizing Statistical Efficiency in Clinical Trials: Ordinal Longitudinal Models and Two-Phase Designs
| dc.contributor.advisor | Harrell, Frank E | |
| dc.contributor.committeeChair | Schildcrout, Jonathan S | |
| dc.creator | Rohde, Maximilian Dimitrios | |
| dc.creator.orcid | 0000-0002-3731-3372 | |
| dc.date.accessioned | 2025-06-06T09:42:03Z | |
| dc.date.created | 2025-05 | |
| dc.date.issued | 2025-03-25 | |
| dc.date.submitted | May 2025 | |
| dc.description.abstract | Clinical trials are the primary method used to establish the safety and efficacy of medicines and interventions. Results from clinical trials form the foundation of medical knowledge and improve the lives of millions of patients. However, clinical trials are often expensive to run, slow to enroll participants, and subject to unforeseen delays -- leading to trials that are underpowered to detect a clinically relevant effect. Optimizing statistical efficiency through the use of modern study designs and analysis methods is therefore an important goal. In this work, we advance two Bayesian methods to improve the efficiency of clinical trials. First, we show how Bayesian ordinal transition models (OTMs) can be used to analyze ordinal longitudinal data, an information-rich type of outcome data that has become common in COVID-19 clinical trials, and develop the statistical theory for these models from first principles. We illustrate OTMs using data from ACTT-1, a clinical trial evaluating the use of remdesivir in hospitalized patients with COVID-19. We then compare OTMs to other commonly used models and approaches for analyzing ordinal longitudinal data through a comprehensive simulation study, and demonstrate that OTMs can greatly improve statistical power. Second, we describe a Bayesian factored likelihood approach to account for missing exposure data in two-phase designs using outcome-dependent sampling and BLUP-dependent sampling. We demonstrate how this approach can improve statistical efficiency compared to simple random sampling through simulations, and provide a case study using data from the Lung Health Study clinical trial. Both of these methods can inform clinical trial design to improve statistical efficiency and advance medical knowledge more quickly. | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.uri | https://hdl.handle.net/1803/19734 | |
| dc.language.iso | en | |
| dc.subject | Ordinal longitudinal outcomes, Two-phase designs, Bayesian methods, Statistical efficiency | |
| dc.title | Maximizing Statistical Efficiency in Clinical Trials: Ordinal Longitudinal Models and Two-Phase Designs | |
| dc.type | Thesis | |
| dc.type.material | text | |
| local.embargo.lift | 2027-05-01 | |
| local.embargo.terms | 2027-05-01 | |
| thesis.degree.discipline | Biostatistics | |
| thesis.degree.grantor | Vanderbilt University Graduate School | |
| thesis.degree.level | Doctoral | |
| thesis.degree.name | PhD |
Files
License bundle
1 - 1 of 1