Skip to main content

Group Sequential Design Web Interface

1 Introduction

We begin with background, including an overview of group sequential designs for clinical trials, the gsDesign web interface, and the organization of this book.

Group sequential design was first developed in the 1960’s by Peter Armitage, CK McPherson, and BC Rowe1, but became much more popular in the 1980’s and 1990’s as more methods became available.2 The group sequential designs presented here are all based on spending functions as developed by KK Gordon Lan and David L DeMets3. This provides a great deal of flexibility for the designs you may wish to derive. An excellent book on examples of interim monitoring applied to group sequential trials has been compiled by David L DeMets, Curt D Furberg, and Lawrence M Friedman, eds.4

Here we describe the gsDesign web interface that will allow you to generate group sequential designs simply without software other than an internet browser. The software underlying this is the gsDesign R package. The web interface has been developed using the shiny R package developed by Posit. The group sequential design web interface allows great flexibility in designing group sequential trials, including:

  1. time-to-event outcome trials, binomial outcome trials and normal outcome trials;
  2. creating a group sequential design based on a fixed design sample size;
  3. flexible spending functions to derive design boundaries.

The web interface provides text and tabular output as well as a variety of graphs summarizing a design.

With the second edition, we have added a variety of features:

  1. Saving and re-loading designs.
  2. Generation of R markdown reports to document design; these can be saved in order to reproduce designs using the gsDesign R package.
  3. Updating bounds at the time of analysis.

In addition to the web browser, Chapter 8 also demonstrates code for group sequential designs in R. Chapter 8 also demonstrates several concepts and capabilities in the R package that have not been incorporated into the web interface.

  1. Computing probability of success, conditional and predictive power and conditional probability of success.
  2. Computing repeated confidence intervals.
  3. How to update bounds if actual sample sizes at analyses differ from planned.
  4. Adapting sample size using conditional power or information-based design.

Note that we have tried to keep any mathematical detail at a minimal level. Here we focus on the concepts of group sequential design, providing references for those interested in more technical details. A classic reference for this is the book by Christopher Jennison and Bruce W Turnbull5. A brief set of references is provided at the end of the book for those wishing to read further.

1.1 Citation

To cite this manual, use:

Anderson, Keaven M, Lueders John, and Nan Xiao. “Group Sequential Design Web Interface.” https://keaven.github.io/gsd-shiny/, 2023.

BibTeX entry:

@misc{gsDesignShiny,
  title        = {Group sequential design web interface},
  author       = {Anderson, Keaven M and Lueders, John, and Xiao, Nan},
  howpublished = {\url{https://keaven.github.io/gsd-shiny/}},
  year         = 2023,
  note         = {Accessed: 2023-04-21}
}