This repository contains code for a paper that introduces a meta-analytic approach for adjusting external control studies for additional bias and variability due to their non-randomized design. Code for the advanced non-small cell lung cancer (aNSCLC) example and the simulation study are available:
-
Example: Code is in the
ecmeta.nsclcdirectory with a corresponding website with results available here. -
Simulation: Code is in the
simulationdirectory with results from the simulations available here.
The entire analysis can be replicated by running the file replicate.R, which runs code to replicate the simulation study and the aNSCLC example, and retrieves all results (figures, tables, and inline text) that are reported in the manuscript.
The aNSCLC analyses can only be reproduced with access to the Flatiron Health database, since it was used to construct the external control cohorts. Unfortunately, we are not permitted to share this data.
R package dependencies---including two internal Roche packages used to access the Flatiron data---are managed through the renv package. You can view all packages and their versions in the lockfile renv.lock. All required packages and the appropriate versions can be installed with renv::restore().
Note that an R package was created for the aNSCLC example, which can be installed with:
system("R CMD INSTALL --build --no-multiarch --with-keep.source ecmeta.nsclc")