Group Sequential Design (GSD)

Resources

Statistical software for design and monitoring

Some authors have reviewed existing resources for implementing GSDs 1 including other adaptive designs 2, 3, 4. Here, we provide an updated non-exhaustive list of several open-access and commercial software packages for deriving, monitoring, and analysing GSD trials. PANDA users may wish to search for new R packages on METACRAN. For example, using the terms “group sequential”.   Of note, when flexible stopping boundaries are used and the planned timing does not coincide with the actual analysis, stopping rules will need to be updated to match the observed information fraction. A number of these packages provide functionality for researchers to achieve this.

References 

1. Zhu et al. Group sequential methods and software applications. Am Stat. 2011;65(2):127–35.
2. Wassmer et al. Group sequential and confirmatory adaptive designs in clinical trials. Springer. 2016.
3. Grayling et al. A review of available software for adaptive clinical trial design. Clin Trials. 2020;17(3):323-331.
4. He et al. Practical considerations for adaptive trial design and implementation. Springer. 2014.

Open-access software


gsDesign (R package) and gsDesignExplorer (Shiny app version)
 The package is for design and inference is limited to repeated confidence intervals during monitoring and adjusted p-values.
  • derives group sequential design for binary, continuous, and time to event outcomes,
  • offers many stopping rules and spending functions,
  • accommodates equal and unequal spaced interim analyses, option for a non-binding futility boundary,
  • accommodates symmetric and non-symmetric designs, and one- and two-sided tests,
  • offers an option to account for overrun participants (enrolled but without outcome data at an interim analysis).,
  • accommodates user-specified stopping rules and compute probabilities of crossing the boundaries at each interim analysis as a function of the effect size. It also offers an option to reset the boundaries when the timing of interim analysis has changed from the initially planned at the design stage.
RCTdesign (R package)
Comprehensive package for the design, monitoring and analysis for various outcomes. Offers many stopping rules and spending functions. Accommodates user-defined stopping rules and non-binding futility stopping.

gsbDesign (R package)
Evaluates statistical characteristics of GSDs using Bayesian methods with normally distributed outcomes. See underlying methods.

rpact (R package)
Comprehensive package for the design, monitoring, and inference of group sequential designs and other confirmatory adaptive designs for several outcomes (methods detailed in this book). The application is illustrated in this tutorial paper.

AGSDest (R package)
Performs inference: estimation of the median unbiased estimate based on stagewise ordering and maximum likelihood estimate, calculation of repeated confidence intervals at interim analysis, and calculation of confidence intervals for final reporting based on the stagewise ordering.

multiarm (R package)
A shiny app that supports the design of GSDs with binary and continuous outcomes. 

Adoptr (R package)
This is for two-stage designs including GSDs with additional flexibilities such as sample size re-estimation based on conditional power (see underlying methods).

OptGS (R package)
 Searches for designs with continuous outcomes that minimise the expected sample size.
ldbounds (R package)
 Derives group sequential boundaries using Lan-DeMets spending functions and probabilities of stopping at interim analyses.

seqmon (R package)
Derives and monitors the GSD and computes probabilities for declaring efficacy at interim analyses. It also updates the design if the future looks need to be changed and tracks the history of the design.

PwrGSD (R package)
Derives GSDs for a time-to-event (survival) outcomes and evaluates the properties of alternative competing designs.

gsrsb (R package)
Derives group sequential boundaries for the primary and secondary outcome simultaneously using error spending functions and gatekeeping keeping procedure (see detailed methods).

gscounts (R package)
 Design of a group sequential trial with a count outcome that follows a negative binomial distribution (see detailed methods).

gset (R package)
Designs a group sequential trial with equivalence hypotheses, incorporates non-binding and binding futility boundaries, and explores operating characteristics of the design.

SurvGSD (R package)
Derives a GSD with censored survival data using information fraction and spending functions (see detailed methods).

seqDesign (R package)
 Simulates and designs group sequential trials with survival outcome and one interim analysis.

Closed source software

GROUPSEQ (Stata module)
Designs group sequential trials with normally distributed outcomes. Computes samples sizes and stopping boundaries for several statistical stopping rules. 

LANDEMETS (Stata module)
 Derives boundaries for a group sequential trial using Lan-DeMets alpha spending functions.

FACTS
Facilitates the design and simulation of Bayesian group sequential trials. See a detailed example here.

Commercial source software


SAS (Modules: PROC SEQDESIGN and PROC SEQTEST)
 Design, monitoring, and analysis of group sequential trials. See more details here.

East (SEQUENTIAL module)
Design, simulation, monitoring, and analysis of group sequential trials. See some details here.

ADDPLAN (BASE module)
Design, simulation, monitoring, and analysis of group sequential trials.

PASS
Design and monitoring of group sequential trials.

nQUERY
See brochure for details.

Tutorial papers

There are some useful papers that have illustrated the application of group sequential trials and discussed related issues:
 
  • Lakens et al. Group sequential designs : A tutorial. Preprint. 2021;1–13. 
  • Gerber et al. gsbDesign : An R package for evaluating the operating characteristics of a group sequential Bayesian design. J Stat Softw. 2016;69(11):1–23. 
  • Gsponer et al. A practical guide to Bayesian group sequential designs. Pharm Stat. 2014;13(1):71–80.  
  • Whitehead. Group sequential trials revisited: Simple implementation using SAS. Stat Methods Med Res. 2011;20(6):635–56. 

Statistical books

  •  Wassmer et al. Group sequential and confirmatory adaptive designs in clinical trials. Springer. 2016.