R Package

R is free of charge and there is increasing interest by the industry to use R. At the moment, no R package is available for performing confirmatory adaptive designs in a comprehensive sense (e.g., simulation and analysis for continuous, binary, and survival endpoint).

Nevertheless, for the design of group sequential tests there is the R package gsDesign, developed by Keaven Anderson (copyright Merck Research Laboratories), which is well established and covers all relevant designs. Among the over 10.000 available packages (May 2017) there are several packages that address the issue of adaptive designs, most of them with special reference to research results from the authors, but none covers the broad range of applications that is nowadays available.

In RPACT (R Package for Adaptive Clinical Trials) particularly, the methods described in the recent monograph of Wassmer and Brannath (published by Springer, 2016) will be implemented and made available for the public.

As a first step, until 06/2018 we plan to deliver a comprehensive package that enables the simulation and analysis of parallel group designs with continuous, binary, and survival endpoint. This package can be downloaded per CRAN and stands under the "GNU Lesser General Public License" version 3. We include a complete documentation, validation, and publication, e.g., in Journal of Statistical Software. Further developments (e.g., multi-armed and enrichment designs) are planned for near future.

The simulation based evaluation of operating characteristics of adaptive designs are becoming increasingly important, and the package will address this issue. We develop these simulations for the most relevant types of endpoints (continuous, binary, and survival) and will include the assessment of sample size reassessment strategies based on conditional power, of futility rules and other strategies. As adaptive strategies classical group sequential tests, combination tests (inverse normal, Fisher's combination test), and adaptive tests based conditional rejection probability (CRP) principle will be available.
A comprehensive output in form of graphs and tables will be provided. Additionally, a specific methodology for survival endpoints with adaptation based on surrogates will be available.

For the analysis and execution of an adaptive trial, all methods provided by the simulation will be available. Specific results of the adaptive methodology are also available, e.g., overall confidence intervals and p-values and conditional and predictive power assessments. The R package will be fully integrated in R (i.e., no "stand alone" package) such that R specific data entry, transformations, and summary statistics can be utilized.

Screenshots: take a look to the appearance and graphical output of the RPACT package.

Validation

The R package RPACT will be a fully documented and validated product, including

  • user requirements specification,
  • functional specification,
  • technical design specification,
  • test plan,
  • installation guides,
  • user guides, and
  • release notes.

The validation of the R package will be done compliant to FDA/GxP guidelines and to the validation process of “Base R” and “Recommended Packages” as described in: “R: Regulatory Compliance and Validation Issues, A Guidance Document for the Use of R in Regulated Clinical Trial Environments” (The R Foundation for Statistical Computing, December, 2014).

Please visit www.rpact.org to get more information about the package.