Skip to contents

Finding the next best dose

nextBest-NextBestNCRM_AT-numeric-Samples-GeneralModel-Data-method.Rd

This function computes the next best dose based on the NCRM method. Also the AT to standard dose escalation transition is considered.

Usage

# S4 method for class 'NextBestNCRM_AT,numeric,Samples,GeneralModel,Data'
nextBest(nextBest, doselimit = Inf, samples, model, data, ...)

Arguments

nextBest

the rule for the next best dose, a NextBestNCRM_AT() object

doselimit

number, the maximum allowed next dose. If this is infinite (the default), then essentially no dose limit will be applied in the course of dose recommendation.

samples

posterior samples from model parameters given data

model

The DLT model

data

crmPack::Data() object that was used to generate the samples

...

additional arguments without method dispatch

Value

the same object and type as crmPack::nextBest() for crmPack::NextBestNCRM() with the logic modification as specified in NextBestNCRM_AT().

Details

Basically this function works the same for nextBest() for crmPack::NextBestNCRM() object, with the following modification:

  • This function will consider whether currently the trial is at a AT to standard dose escalation transition. If current data indicates the transition, then the last used dose will be the recommendation.

  • When the recommended dose is NA_real_, which indicates the model thinks all doses are too toxic and can not a recommendation, nextBest@na_method will be used to propose a back-up plan:

    • na_method = 0: no back-up dose is proposed.

    • na_method = 1: the 1st dose in the dose grid is proposed.

    • na_method = 2: the last used dose is proposed.

    This is mostly for the very early stage of the trial when maybe only 1 or 2 subjects are enrolled and DLT event is observed. Sometimes the model will make estimation that all doses are too toxic while actually this is very early stage and we still want to make some exploration.

Note

The crmPack::DataParts() can achieve similar results by setting the trials into two parts and setup a crmPack::IncrementsRelativeParts() rule. But it requires the user to provided whether the next part of the trial has begun or not when updating the data and make next step recommendation.

See also

crmPack::nextBest() and crmPack::NextBestNCRM()