blrm_trial: Dose-Escalation Trials guided by Bayesian Logistic Regression...

In OncoBayes2: Bayesian Logistic Regression for Oncology Dose-Escalation Trials

Dose-Escalation Trials guided by Bayesian Logistic Regression Model

Description

blrm_trial facilitates the conduct of dose escalation studies guided by Bayesian Logistic Regression Models (BLRM). While the blrm_exnex only fits the BLRM model to data, the blrm_trial function standardizes the specification of the entire trial design and provides various standardized functions for trial data accrual and derivation of model summaries needed for dose-escalation decisions.

Usage

blrm_trial(
  data,
  dose_info,
  drug_info,
  simplified_prior = FALSE,
  EXNEX_comp = FALSE,
  EX_prob_comp_hist = 1,
  EX_prob_comp_new = 0.8,
  EXNEX_inter = FALSE,
  EX_prob_inter = 1,
  formula_generator = blrm_formula_saturating,
  interval_prob = c(0, 0.16, 0.33, 1),
  interval_max_mass = c(prob_underdose = 1, prob_target = 1, prob_overdose = 0.25),
  ...
)

## S3 method for class 'blrm_trial'
print(x, ...)

Arguments

data	dose-toxicity data available at design stage of trial
dose_info	specificaion of the dose levels as planned for the ongoing trial arms.
drug_info	specification of drugs used in trial arms.
simplified_prior	logical (defaults to FALSE) indicating whether a simplified prior should be employed based on the reference_p_dlt values provided in drug_info. Warning: The simplified prior will change between releases. Please read instructions below in the respective section for the simplified prior.
EXNEX_comp	logical (default to TRUE) indicating whether EXchangeable-NonEXchangeable priors should be employed for all component parameters
EX_prob_comp_hist	prior weight ([0,1], default to 1) on exchangeability for the component parameters in groups representing historical data
EX_prob_comp_new	prior weight ([0,1], default to 0.8) on exchangeability for the component parameters in groups representing new or concurrent data
EXNEX_inter	logical (default to FALSE) indicating whether EXchangeable-NonEXchangeable priors should be employed for all interaction parameters
EX_prob_inter	prior weight ([0,1], defaults to 0.8) on exchangeability for the interaction parameters
formula_generator	formula generation function (see for example blrm_formula_linear or blrm_formula_saturating). The formula generator defines the employed interaction model.
interval_prob	defines the interval probabilities reported in the standard outputs. Defaults to c(0, 0.16, 0.33, 1).
interval_max_mass	named vector defining for each interval of the interval_prob vector a maxmimal admissable probability mass for a given dose level. Whenever the posterior probability mass in a given interval exceeds the threshold, then the Escalation With Overdose Control (EWOC) criterion is considered to be not fullfilled. Dose levels not fullfilling EWOC are ineligible for the next cohort of patients. The default restricts the overdose probability to less than 0.25.
...	Additional arguments are forwarded to blrm_exnex, i.e. for the purpose of prior specification.
x	blrm_trial object to print

Details

blrm_trial constructs an object of class blrm_trial which stores the compelte information about the study design of a dose-escalation trial. The study design is defined through the data sets (see sections below for a definition of the columns):

data (historical data)

The data argument defines available dose-toxicity data at the design stage of the trial. Together with the prior of model (without any data) this defines the prior used for the trial conduct.

dose_info

Definition of the pre-specified dose levels explored in the ongoing trial arms. Thus, all dose-toxcitiy trial data added to the object is expected correspond to one of the dose levels in the pre-defined set of dose_info.

drug_info

Determines the drugs used in the trial, their units, reference dose level and optionally defines the expected probability for a toxicity at the reference dose.

Once the blrm_trial object is setup the complete trial design is specified and the model is fitted to the given data. This allows evaluation of the pre-specified dose levels of the trial design wrt. to safety, i.e. whether the starting dose of the trial fullfills the escalate with overdose criterion (EWOC) condition.

The blrm_trial trial can also be constructed in a 2-step process which allows for a more convenient specification of the prior since meta data like number of drugs and the like can be used. See the example section for details.

After setup of the initial blrm_trial object additional data is added through the use of the update method which has a add_data argument intended to add data from the ongoing trial. The summary function finally allows to extract various model summaries. In particular, the EWOC criterion can be calculated for the pre-defined dose levels of the trial.

Value

The function returns an object of class blrm_trial.

Methods (by generic)

• print: print function.

Simplified prior

As a convenience for the user, a simplified prior can be specifed whenever the reference_p_dlt column is present in the drug_info data set. However, the user is warned that the simplified prior will change in future releases of the package and thus we strongly discourage the use of the simplified prior for setting up trial designs. The functionality is intended to provide the user a quick start and as a starting point. The actually instantiated prior can be seen as demonstrated below in the examples.

Input data

The data given to the data argument of blrm_trial is considered as the available at design stage of the trial. The collected input data thus does not necessarily need to have the same dose levels as the pre-specified dose_info for the ongoing trial(s). It's data columns must include, but are not limited to:

group_id

study

stratum_id

optional, only required for differential discounting of groups

num_patients

number of patients

num_toxicities

number of toxicities

drug_A

Columns for the dose of each treatment component, with column names matching the drug_name values specified in the drug_info argument of blrm_trial

Drug info data

The drug information data-set defines drug properties. The fields included are:

drug_name

name of drug which is also used as column name for the dose

dose_ref

reference dose

dose_unit

units used for drug amounts

reference_p_dlt

optional; if provided, allows setup of a simplified prior

Dose info data

The drug_info data-set pre-specifies the dose levels of the ongoing trial. Thus, all data added to the blrm_trial through the update command must be consistent with the pre-defined dose levels as no other than those pre-specified ones can be explored in an ongoing trial.

dose_id

optional column which assigns a unique id to each group_id/dose combination. If not specified the column is internally generated.

group_id

study

drug_A

Columns for the dose of each treatment component, with column names matching the drug_name values specified in the drug_info argument of blrm_trial

References

Babb, J., Rogatko, A., & Zacks, S. (1998). Cancer phase I clinical trials: efficient dose escalation with overdose control. Statistics in medicine, 17(10), 1103-1120.

Neuenschwander, B., Roychoudhury, S., & Schmidli, H. (2016). On the use of co-data in clinical trials. Statistics in Biopharmaceutical Research, 8(3), 345-354.

See Also

Other blrm_trial combo2 example: dose_info_combo2, drug_info_combo2, example-combo2_trial

Examples

## Setting up dummy sampling for fast execution of example
## Please use 4 chains and 100x more warmup & iter in practice
.user_mc_options <- options(OncoBayes2.MC.warmup=10, OncoBayes2.MC.iter=20, OncoBayes2.MC.chains=1,
                            OncoBayes2.MC.save_warmup=FALSE)


# construct initial blrm_trial object from built-in example datasets
combo2_trial_setup <- blrm_trial(
  data = hist_combo2,
  dose_info = dose_info_combo2,
  drug_info = drug_info_combo2,
  simplified_prior = TRUE
)

# extract blrm_call to see setup of the prior as passed to blrm_exnex
summary(combo2_trial_setup, "blrm_exnex_call")

# Warning: The simplified prior will change between releases!
# please refer to the combo2_trial example for a complete
# example. You can obtain this example with
# ?example-combo2_trial
# or by running
# example_model("combo2_trial")

## Recover user set sampling defaults
options(.user_mc_options)

OncoBayes2 documentation built on July 26, 2023, 5:30 p.m.