Methods for the Sequential Parallel Comparison Design Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 20, 2019
Creator
  • Silverman, Rachel
    • Affiliation: Gillings School of Global Public Health, Department of Biostatistics
Abstract
  • Sequential parallel comparison design (SPCD) has been proposed to increase the likelihood of success of clinical trials, especially trials with a possibly high placebo effect. SPCD is conducted with two stages, and subjects are randomized into three groups: (1) placebo in both periods, (2) placebo in the first period and active therapy in the second period, and (3) active therapy in both periods. Efficacy analysis of the study data includes all data from stage 1 and all placebo non-responding subjects from stage 2. Each stage is analyzed separately then the data are pooled to yield a single p-value. We first describe methods to use in a trial where we combine SPCD with the group sequential approach. We examine how to increase the sample size and adjust the design parameters during an interim analysis to increase power; these design parameters include allocation proportion to placebo in stage 1 of SPCD and weight of stage 1 data in the overall efficacy test statistic. Next, we develop new methods for SPCD with binary and time-to-event outcomes. These methods allow us to analyze SPCD stage-wise using the model of interest with adjustment for covariates. We show that under certain conditions the covariance between the estimated treatment effects in the two periods of SPCD is 0 under both null and alternative hypotheses. We also show that the stage-wise test statistics are uncorrelated under the null hypothesis. As a result, we can omit covariance in the construction of the overall test statistic and the confidence interval for the weighted sum of treatment effects. We develop framework and implementation of SPCD using permutation tests and bootstrap hypothesis testing. This approach allows the flexibility to use SPCD with any outcome. We examine two variations of permutation tests and three variations of the bootstrap. We show that the overall permutation as well as the stage-wise permutation test preserve type I error. Additionally, the bootstrap that maintains the original stage 1 group sample sizes and the stage-wise bootstrap also preserve type I error. The stage-wise permutation test and bootstrap make it easy to evaluate SPCD data with popular software.
Date of publication
Keyword
DOI
Resource type
Rights statement
  • In Copyright
Advisor
  • Fine, Jason
  • Zink, Richard
  • Baron, John
  • Koch, Gary
  • Ivanova, Anastasia
Degree
  • Doctor of Philosophy
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2017
Language
Parents:

This work has no parents.

Items