Extensions of Methods for Clinical Trials: (1) Randomization-based ANCOVA for Longitudinal Clinical Trials with Missing Data, Particularly the Sequential Parallel Comparison Design, and (2) Methods for Clarifying Criteria for Study Continuation at Interim Analysis
Public Deposited- Creator
-
Wiener, Laura Elizabeth
- Affiliation: Gillings School of Global Public Health, Department of Biostatistics
- Abstract
- Randomization-based analysis of covariance (ANCOVA) is a useful analysis for randomized clinical trials when covariate-adjusted estimates of treatment effects are needed, but statistical assumptions for parametric methods cannot be verified. Li (2017) and Wiener et al. (2019) present methods to apply randomization-based ANCOVA to studies with the sequential parallel comparison design (SPCD) in settings where hypothesis testing or confidence interval estimation is needed. SPCD is a two-period crossover design, typically adopted by trials with high placebo response rates, wherein the primary analysis includes all subjects from the first period and only first period placebo nonresponders from the second period. This research expands the methods for randomization-based ANCOVA for SPCD studies to account for multiple visits per period and assesses how the methods perform for different endpoints and under various missing data imputations. Missing data is a ubiquitous problem in longitudinal studies and can have substantial impacts on drawn inferences. However, many longitudinal studies restrict their analyses to subjects with complete baseline data, thereby underutilizing follow-up data and possibly introducing bias. This research applies a hybridized mixed model for repeated measurements with randomization-based ANCOVA to longitudinal studies when there are missing data at baseline and/or post-baseline. Two case studies highlight the methods with parametric and nonparametric applications when the outcome can be presumed to follow a multivariate normal distribution, and when it cannot. Interim analyses are an important tool in monitoring the progress of a clinical trial, and often assess futility via conditional power. However, the opposite viewpoint of whether the study is demonstrating proof of concept and should continue is a valuable consideration and needs to be addressed with high power. This research interconnects the idea of proof of concept with conditional power, while highlighting the importance of type I and type II errors for study continuation or not at the interim analysis. Methods for analyzing subgroups motivate the interim analyses, and two conditional power modifications based on the assumed effect size or the observed interim effect size are considered. Graphical displays illustrate the connection of the interim type I and type II errors with conditional power.
- Date of publication
- 2019
- Keyword
- DOI
- Resource type
- Advisor
- Zink, Richard C
- Ivanova, Anastasia
- Herring, Amy H
- Koch, Gary G
- Brookhart, Maurice A
- Degree
- Doctor of Public Health
- Degree granting institution
- University of North Carolina at Chapel Hill Graduate School
- Graduation year
- 2019
- Language
Relations
- Parents:
This work has no parents.
Items
| Thumbnail | Title | Date Uploaded | Visibility | Actions |
|---|---|---|---|---|
|
|
Wiener_unc_0153D_18725.pdf | 2020-01-08 | Public | Download |
|
Wiener_unc_0153D_18725_DATA.xml | 2020-01-08 | Public | Download |