The escalation package

escalation is an R package for dose-escalation clinical trials, providing a consistent, extensible, modular approach.

It largely uses existing R implementations of popular dose-finding designs, including the continual reassessment method (CRM) design from the dfcrm package, the Bayesian optimal interval (BOIN) design from the BOIN package, and the perennial 3+3.

escalation bestows these model-fitting functions with a common interface. This allows the provision of add-on behaviours that affect the way that doses are selected, or trial stopping decisions are made. For example, it is simple to compose a CRM design that will not skip doses and will stop once a recommended dose has been given to $n$ patients; or a BOIN design that will demand that a certain dose is tested sufficiently before stopping is countenanced; and so on. Functions are provided to run simulations and enumerate all possible future dose paths. See related blog posts or the package documentation site for many illustrative examples.

CRAN listing

• https://cran.r-project.org/package=escalation

Source code

• https://github.com/brockk/escalation

Documentation

• https://brockk.github.io/escalation/