Skip to contents

1. Getting started with psborrow2

Matt Secrest and Isaac Gravestock

psborrow2.Rmd

In this vignette, you’ll learn about the scope of psborrow2 and where to find additional information on how to implement analyses in psborrow2.

Introduction

While the randomized controlled trial (RCT) comparing experimental and control arms remains the gold standard for evaluating the efficacy of a novel therapy, one may want to leverage relevant existing external control data to inform the study outcome. External control data can help increase study power and thereby shorten trial duration or reduce the number of subjects needed. However, analysis of external control data can also introduce bias. One method for incorporating external control data to mitigate bias is Bayesian dynamic borrowing (BDB), in which external control data is borrowed to the extent that the external and RCT control arms have similar outcomes (Viele et. al. 2014).

Implementing BDB is computationally involved and requires Markov chain Monte Carlo (MCMC) sampling methods, which in turn may require knowledge of MCMC sampling software. To overcome these technical barriers and we developed psborrow2 using cmdstanr, an R package which facilitates the use of the MCMC sampling program Stan (via CMD Stan).

psborrow2 has two main goals:

  1. to provide a user-friendly interface for applying BDB on study results that handles the MCMC sampling on behalf of the user

  2. to facilitate simulation studies that compare different borrowing parameters (e.g. full borrowing, no borrowing, BDB) and other trial and BDB characteristics in a unified way

Right now, psborrow2 supports time-to-event endpoints as well as binary endpoints.

Analyze your own data

psborrow2 can implement BDB in a scenario wherein a two-arm RCT is supplemented with external data on the control arm. Three arms are required to implement BDB in psborrow2. They are:

  • The internal control arm from within the RCT
  • The external control arm using observational data or experimental data from another trial with the same intervention/treatment and population as the internal control arm
  • The internal experimental arm from within the RCT

Such scenarios are common in drug development because the comparator arm for a novel therapy is often the standard of care, for which data exists from electronic health care records or from previous phase III registrational trials.

Refer to the “dataset” vignette for more information on how to implement BDB analyses on your own data: vignette('dataset', package = 'psborrow2')

Conduct a simulation study

Refer to the “simulation study” vignette for more information on how to create a simulation study involving BDB and other innovative trial designs: vignette('simulation_study', package = 'psborrow2')

Additional information

psborrow2 is the successor to psborrow. While the core functionality has been replicated in psborrow2 such that there is no longer any reason to prefer psborrow, we cite it here for posterity.

References

Viele, K., Berry, S., Neuenschwander, B., Amzal, B., Chen, F., Enas, N., Hobbs, B., Ibrahim, J.G., Kinnersley, N., Lindborg, S., Micallef, S., Roychoudhury, S. and Thompson, L. (2014), Use of historical control data for assessing treatment effects in clinical trials. Pharmaceut. Statist., 13: 41–54.