Skip to contents

gsdmvn: A package for group sequential design under non-proportional hazards (NPH)

Source: R/gsdmvn.r
gsdmvn.Rd

The gsdmvn package will eventually be incorporated into the gsDesign2 package. This version is for the Regulatory/ASA Biopharmaceutical Subsection training course in September, 2020 The package computes the asymptotic normal distribution for group sequential designs, generalizing the theory presented by Jennison and Turnbull (2000) to cases with non-homogeneous treatment effect over time. The primary application for this is group sequential design under the assumption of non-proportional hazards. The gsdmvn package has 4 types of functions

  1. support for asymptotic normal distribution computation,

  2. support for group sequential bound derivation, and

  3. support for design and power calculations.

  4. applications to designs for survival analysis under non-proportional hazards assumptions.

Details

In addition to the above function categeories, vignettes show how to implement

  1. design for a binomial endpoint as an example of how to extend the package to other endpoint types,

  2. the extensive capabilities around group sequential boundary calculations, including enabling capabilities not in the gsDesign package.

gsdmvn functions

The primary functions supporting non-proportional hazards in the short course are:

  • gs_design_ahr - derive group sequential design under non-proportional hazards (NPH) for logrank test

  • gs_power_ahr - compute power for a group sequential design under non-proportional hazards for logrank test

Key supportive functions specify bound derivation for designs:

  • gs_b - directly provide bounds for designs

  • gs_spending_bound - provide bounds based on a spending function (e.g., Lan-DeMets O'Brien-Fleming)

Underlying functions to support numerical integration that should not be directly needed by typical users are

  • gridpts - set up grid points and weights for numerical integration for a normal distribution

  • h1 - initialize numerical integration grid points and weights for NPH for first analysis

  • hupdate - update numerical integration grid points and weights for NPH from one interim to the next

  • gs_power_npe - general non-constant-effect size boundary crossing probability calculation for group sequential design

References

Jennison C and Turnbull BW (2000), Group Sequential Methods with Applications to Clinical Trials. Boca Raton: Chapman and Hall.

Author

Keaven Anderson keaven_anderson@merck.com