Simulate BOIN Drug-Combination Trials

Run the BOIN drug-combination design of Lin and Yin (2017) over many simulated trials of two drugs given together, and return the patient and DLT counts at every dose combination. This is the simulation engine behind sim_comb_boin, exposed for users who want the raw trial data.

Usage

comb_boin_simulate(
  target,
  p_true,
  n_cohort,
  cohort_size,
  n_trials = 10000,
  start_dose = c(1, 1),
  n_earlystop = 18,
  p_saf = NULL,
  p_tox = NULL,
  cutoff_eli = 0.95,
  extrasafe = FALSE,
  offset = 0.05,
  titration = FALSE,
  seed = 123
)

Arguments

target: Numeric scalar. Target DLT probability, for example 0.30.
p_true: Numeric matrix. True DLT probability of each dose combination, with one row per dose level of drug A and one column per dose level of drug B, both in increasing dose order.
n_cohort: Integer scalar. Number of cohorts in a trial.
cohort_size: Integer scalar or vector. Number of patients per cohort. A scalar is used for every cohort. A vector shorter than n_cohort is padded with its last element and a longer one is truncated.
n_trials: Integer scalar. Number of trials to simulate. Defaults to 10000.
start_dose: Integer vector of length two. Dose levels of drug A and drug B for the first cohort. Defaults to c(1, 1). Titration, when requested, also starts here.
n_earlystop: Integer scalar. The trial stops once this many patients have been treated at the current combination and the design would neither escalate nor de-escalate, or cannot. Defaults to 18. Set it to a value above the maximum sample size to switch this rule off.
p_saf: Numeric scalar. Highest DLT probability deemed subtherapeutic. Defaults to 0.6 * target.
p_tox: Numeric scalar. Lowest DLT probability deemed overly toxic. Defaults to 1.4 * target.
cutoff_eli: Numeric scalar. Posterior probability cutoff for eliminating a combination. Defaults to 0.95.
extrasafe: Logical scalar. Apply the stricter safety stopping rule at the lowest combination. Defaults to FALSE.
offset: Numeric scalar between 0 and 0.5. Amount by which cutoff_eli is relaxed for the safety stopping rule. Defaults to 0.05.
titration: Logical scalar. Treat one patient per combination, moving up one level of either drug, until the first DLT, before switching to full cohorts. Ignored when the first cohort size is one. Defaults to FALSE.
seed: Integer scalar or NULL. Random seed. When NULL the current state of the random number generator is used. The state of the calling session is restored on exit. Defaults to 123.

Value

An object of class comb_boin_trials, which is a list with components

n_pts: Integer array of patients treated, with dimensions trial by level of drug A by level of drug B.
n_tox: Integer array of the same shape, of DLTs observed.
eliminated: Logical array of the same shape, marking combinations eliminated during the trial.
cohorts_used: Integer vector of cohorts completed in each trial.
stop_reason: Character vector giving why each trial ended.
boundary: The boin_boundary object used.
settings: List of the design parameters.

Details

The decision boundaries are those of the single-agent design, from boin_boundary. After each cohort the observed DLT rate at the current combination is compared with them. When the design escalates, the next cohort goes one level up of drug A or of drug B, whichever has the larger posterior probability, under a Beta(0.5, 0.5) prior, that its DLT probability lies between lambda_e and lambda_d; de-escalation chooses one level down in the same way. Eliminated combinations are not escalated to. When a combination is eliminated, so is every combination with at least the same level of both drugs, and the trial stops when the lowest combination is eliminated.

The rules, and the order in which they are applied, are those of BOIN::get.oc.comb() with mtd.contour = FALSE. That implementation adds two things to the article: a candidate's posterior probability is increased by 0.0005 for every patient already treated there, so that ties go to the better informed combination, and the trial can stop at n_earlystop. Random variates are drawn in the same order as there: one per patient, one for the direction of every titration step that can go either way, and one for every escalation or de-escalation that has a candidate, even when there is only one. With the same seed and matching arguments a run therefore reproduces the reference implementation trial by trial.

During titration the patients treated before the first DLT are added to the planned cohorts, as in the reference implementation: the first full cohort only tops up the combination reached. A trial can therefore treat a few more patients than n_cohort cohorts hold.

Possible values of stop_reason are "lowest_dose_eliminated", "lowest_dose_too_toxic", "n_earlystop" and "max_cohorts".

References

Lin, R. and Yin, G. (2017). Bayesian Optimal Interval Design for Dose Finding in Drug-Combination Trials. Statistical Methods in Medical Research, 26(5), 2155-2167.

Yan, F., Zhang, L., Zhou, Y., Pan, H., Liu, S. and Yuan, Y. (2020). BOIN: An R Package for Designing Single-Agent and Drug-Combination Dose-Finding Trials Using Bayesian Optimal Interval Designs. Journal of Statistical Software, 94(13), 1-32.

Examples

p_true <- matrix(c(0.02, 0.04, 0.08, 0.14,
                   0.08, 0.25, 0.42, 0.48,
                   0.25, 0.45, 0.50, 0.60), nrow = 3, byrow = TRUE)

trials <- comb_boin_simulate(
  target = 0.25,
  p_true = p_true,
  n_cohort = 16,
  cohort_size = 3,
  n_trials = 200,
  seed = 123
)
trials
#> Simulated BOIN drug-combination trials
#>   trials          : 200
#>   dose levels     : 3 of drug A, 4 of drug B
#>   target DLT rate : 0.25
#>   cohorts         : 16 of size 3
#>   max sample size : 48
#> 
#> Average patients treated per trial:
#>      B1   B2   B3   B4
#> A1 3.98 3.70 2.71 2.84
#> A2 5.80 9.03 3.93 1.50
#> A3 4.74 4.04 1.03 0.30
#> 
#> Average DLTs observed per trial:
#>      B1   B2   B3   B4
#> A1 0.07 0.18 0.18 0.41
#> A2 0.40 2.12 1.65 0.72
#> A3 1.22 1.73 0.50 0.20
#> 
#> Stopping reason (%):
#> 
#> max_cohorts n_earlystop 
#>        49.5        50.5 

# Patients treated at each combination in the first trial
trials$n_pts[1, , ]
#>    B1 B2 B3 B4
#> A1  3  3 15  0
#> A2  0  0 12  9
#> A3  0  0  0  6