Evaluate BLRM using `BLRMeval`

By Fenguoerbian

August 12, 2024

The BLRMeval package

The goal of BLRMeval is to provide reports on BLRM specification in protocol development and evaluation when trial is ongoing.

Extension to crmPack

Accelerated titration (AT) to standard escalation transition

This package provides some extensions to crmPack about accelerated titration. When the first DLT event is observed and the transition from accelerated titration to standard dose escalation is triggered, usually we will continue to enroll subjects into the same dose level/cohort to the size of standard escalation and evaluate the model when these full cohort finish DLT observation.

But in crmPack’s frame work, the model will be evaluated when first DLT is observed, and enroll subjects into the next recommended dose with the size of standard escalation. See details at crmPack::CohortSizeDLT(), crmPack::NextBestNCRM() and related vignettes of crmPack.

This package provides new classes CohortSizeDLT_AT(), CohortSizeConst2() and NextBestNCRM_AT() to extend their existing counterpart in crmPack. These new new classes fit in the framework of crmPack, but take into consideration the accelerated titration to standard escalation transition and the hard threshold of full sample size

Please refer to the documentation for more details about these classes.

New classes Counterpart in crmPack Modifications
CohortSizeDLT_AT() CohortSizeDLT() Consider accelerated titration to standard dose escalation transition
CohortSizeConst2() CohortSizeConst() Consider hard thresholding of maximum sample size
NextBestNCRM_AT() NextBestNCRM() Consider accelerated titration to standard dose escalation transition. Also consider back-up propose of next dose level when the original recommendation is NA (None of the dose is recommended)

Note: This accelerated titration to standard dose escalation transition is normally taken care of from the operation side. But these extension can make the OC simulation closer to real world usage.

Better knit_print() output

knit_print() methods are very useful for generating user-friendly output in Rmd reports of different R objects. crmPack packs a lot of these methods for its classes but some of them can benefit from some improvement. This package provides these knit_print() methods.

Classes Improvements Status to upstream
StoppingAny(), StoppingAll(), StoppingList() Better indentation. Auto-adjusted preamble strings Merged into crmPack main branch already in this PR
CohortSizeMax(), CohortSizeMin(), CohortSizeOrdinal(), CohortSizeDLT(), CohortSizeRange() Better indentation Working on a PR

Essentially we are working to get these improvements merged into crmPack so one can enjoy the benefit without having to install this package.

Note: If PR is merged into crmPack, then the corresponding functions will be deleted in this package so that users will not encounter conflict/overwrite problems.

Note: As stated in crmPack’s documentation, these methods are mainly focused on html output. If additionally word document format is desired, one can save the html report first, then open it with word. Modern versions of word can handle this type of html output quite nicely.

BLRM specification and evaluation reports

Two Rmd report templates are included in the package. One for BLRM specification and another one for BLRM evaluation.

Rmd template Purpose
BLRM Evaluation Report Evaluate the toxicity based on BLRM and make recommendation (dose level, cohort size, whether to stop the trial) for the next step
BLRM Specification Report Specify details (priors, model settings) of a BLRM. Test-run the model under various hypothetical escalation settings and perform simulation to evaluate the operation characteristics.

If you are using RStudio, you can choose the template when creating Rmd document, shown in the following screenshot:

rmd template screenshot

Or if you are not using Rstudio, you can invoke the command

rmarkdown::draft(file = "your_filename", 
                 template = "blrm-evaluation-report",    # or "blrm-specification-report"
                 package = "BLRMeval")

The demo content of these report templates can be found in the package’s vignettes. If you don’t build the vignettes locally, you can find a preview content at the Articles section of the package’s website.

Posted on:
August 12, 2024
Length:
3 minute read, 600 words
Tags:
R BLRM EWOC MTD
See Also:
Making Go/No-go decision in proof-of-concept trials
Parallel Computing in R, from parallel to foreach and future
分期付款的内含收益率