biomkrAccrual

Simulating recruitment at time of randomisation to adaptive trials with arm eligibility determined by biomarker status.

The {biomkrAccrual} package uses a Poisson-Gamma-Dirichlet model to simulate
trial recruitment for multi-site, multi-region, multi-arm trials. Recruitment per
site is modelled with the Poisson-Gamma model (Anisimov and Federov, 2007).
A hierarchical Dirichlet model is used to model biomarker proportions for sites
within regions. Recruitment to a given site in a given week is then randomised
to biomarker status using the prevalences drawn from the Dirichlet model for that
site.

Running a single simulation

biomkrAccrual()

The default settings will use the configuration files in the extdata directory, and will
keep the resulting data files and recruitment plots. The location can be specified with
output_path and figs_path.

Running a set of simulations

biomkrAccrualSim(n = 250)

The datafiles and recruitment plots from the individual runs will not be kept (this
can be changed with quietly = FALSE) but will preserve the summary datafiles and
distribution plots.

Practical notes

There are a very large number of arguments to both commands, and three configuation files,
one of which (the relationship of treatment arms to biomarker recruitment arms) is a JSON.
This is because flexibility is required, and they are intended to be driven by a dashboard
in future.

This package will not pass R CMD Check because it is written using the new object orientation
system for R, {S7}, and CMD Check does not yet understand the syntax.

Anisimov, V.V., Fedorov, V.V., 2007. Modelling, prediction and adaptive adjustment of
recruitment in multicentre trials. Statistics in Medicine 26, 4958–4975.
https://doi.org/10.1002/sim.2956