I’m delighted to share that our paper, Some Group Sequential Trials From Industry Over the Last 30 Years, has been published in Statistics in Biopharmaceutical Research.
You can download the accepted manuscript (PDF, 457 KB) here.
Key takeaways
The paper began as a conference talk presented by Keaven, our lead author, in 2022. Recognizing the value of these stories, we decided to develop them into a full-length paper. The peer review process significantly improved the quality of our work, and we are grateful for the reviewers’ constructive feedback.
Our paper discussed how group sequential design has been applied in industry trials over the past three decades. Besides a concise review of theory, it offers an informed perspective into the flexibility of group sequential design in solving challenges compared to more complex adaptive designs. We emphasized important considerations such as timing of interim analyses, boundary setting for interim and final analyses, and handling multiple hypotheses created by dose groups, populations, and endpoints.
Highlights
What sets this paper apart is its use of real-world examples to illustrate the intellectual challenges and practical solutions in trial design. Keaven carefully selected several representative trials that showcase the advantages of group sequential design. These include but are not limited to practical benefits like early futility analysis and facilitating fast decision-making in trials with significant efficacy or safety outcomes.
Naturally, the paper discusses the use of open source R packages, such as gsDesign, gsDesign2, and gMCPLite, for creating group sequential designs. These tools provide the flexibility needed to design and analyze group sequential trials, including features for flexible timing and spending, non-proportional hazards, and stratified populations. They were also used to generate the designs presented in the paper’s tables and figures.
Behind the scenes: write publication-quality papers using R Markdown
For those interested in the technical aspects, we used a plain text workflow to write the paper:
- gsDesign, gsDesign2, and gMCPLite for producing the designs, tabular summaries, and visualizations.
- R Markdown for technical writing and publishing.
- Git for version control and team collaboration.
In particular, we leveraged the following tools to produce high-quality PDF outputs. They helped us meet the style and formatting standards while keeping the Markdown simple so that we could focus on the content.
Template
We used the ASA article template
from the rticles package.
This template helped us meet the journal formatting requirements.
It also provided a convenient blinded option for producing
double-anonymized review PDFs.
Table
Most design summary tables were generated with R and piped into gt to produce nicely formatted LaTeX tables. For simpler, manually crafted Markdown tables, pipe tables offered precise column alignment control.
Figure look and feel
Key elements of our figure design include:
- Using
cowplot::theme_cowplot()for its sensible default appearance. - Using
cowplot::plot_grid()for arranging multi-panel figures. - Using ggrepel for physics-based text label placement to avoid overlap.
- Using
gMCPLite::hGraph()to create elliptical multiplicity graphs.
Figure production
We used the Cairo PDF device by setting dev = "cairo_pdf" in knitr chunk
options to correctly render Unicode characters (such as Greek letters) in plots.
LaTeX fine-tuning
We customized the knit button behavior
using the knit option in the YAML header. This trick allowed us to
programmatically parse and process the intermediate LaTeX output before
rendering it into PDF. It can help address issues in the original LaTeX
output, for example, making the gt table captions numbered and resetting
the table font sizes.
Looking ahead
I hope this paper serves as a valuable reference for statisticians and clinical researchers interested in implementing group sequential designs in their trials. It aims to contribute to ongoing discussions and statistical methodology research on efficient, effective clinical trial design.
If you find our paper useful, please cite it as follows (APA style):
Anderson, K. M., Zhao, Y., Xiao, N., Ge, J., & Weisman, H. F. (2024). Some group sequential trials from industry over the last 30 years. Statistics in Biopharmaceutical Research, 16(3), 281–293. doi:10.1080/19466315.2024.2334354
For LaTeX users, the BibTeX entry is:
@article{anderson2024some,
title = {Some group sequential trials from industry over the last 30 years},
author = {Anderson, Keaven M and Zhao, Yujie and Xiao, Nan and Ge, Joy and Weisman, Harlan F},
journal = {Statistics in Biopharmaceutical Research},
volume = {16},
number = {3},
pages = {281--293},
year = {2024},
doi = {10.1080/19466315.2024.2334354}
}