Global Statistical Tests for Multiple Endpoints

Design and evaluate global statistical tests (GST) that combine several endpoints into one test, including O'Brien's ordinary and generalized least squares tests (O'Brien (1984) ), power and sample size calculations, and an O'Brien-based truncated closed testing procedure (Luo et al. (2025) ). Two 'shiny' applications support interactive power calculation and the creation of forest plots for multiple endpoints from summary statistics.

gst

Assessing sample size and power, and presenting consolidated evidence

Originally developed at Sarepta Therapeutics (Sarepta-Therapeutics/gst). This fork is maintained by Xiao Ni.

Description

The {gst} package provides a toolbox for assessing sample size and power for Global Statistical Tests across a variety of scenarios, and presenting consolidated evidence using a forest plot.

Talk

R/Pharma 2026: Totality of Evidence in Minutes: Interactive Apps for GST Power and Publication-Ready Forest Plot

• Video (9.5 min): https://www.youtube.com/watch?v=UmHAQsYu50o
• Slides: https://xni7.github.io/talks/rpharma2026-gst/

Installation

The package is not yet available on CRAN. To install the development version of the package from Github run:

pak::pak("xni7/gst")

Shiny Apps

The {gst} package includes two shiny applications. For details on how to use the apps, please review the vignettes.

GST Power App

Hosted app: https://xni7-gstpower.share.connect.posit.cloud

library(gst)
gst_power_app()

Make Forestplot App

Hosted app: https://xni7-gstmakeforestplot.share.connect.posit.cloud

library(gst)
make_forestplot_app()

Acknowledgement

Lerong Li implemented the O'Brien GST test and the power calculations in this package. Anthony Pileggi built the interfaces for both Shiny apps. Xiaolong Luo led the methodological development of the truncated closed test in Luo et al. (2025). Stephen Knapp and Preston Burns contributed to the package.

The forest plot is drawn with forestploter by Alimu Dayimu.