Next Dose Combination in a BOIN Drug-Combination Trial

Decide the dose combination for the next cohort of a trial run with the BOIN drug-combination design of Lin and Yin (2017), from the numbers of patients and DLTs observed so far at every combination and the combination that treated the last cohort. The rules are those of comb_boin_simulate, so the decisions taken during the trial are the ones whose operating characteristics sim_comb_boin reports.

Usage

comb_boin_next(
  target,
  n_pts,
  n_tox,
  current,
  n_earlystop = 18,
  p_saf = NULL,
  p_tox = NULL,
  cutoff_eli = 0.95,
  extrasafe = FALSE,
  offset = 0.05
)

Arguments

target: Numeric scalar. Target DLT probability, for example 0.30.
n_pts: Integer matrix. Number of patients treated so far at each dose combination, with one row per dose level of drug A and one column per dose level of drug B.
n_tox: Integer matrix of the same shape as n_pts. Number of patients with a DLT at each combination.
current: Integer vector of length two. Dose levels of drug A and drug B of the combination that treated the last cohort.
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. Use the value of the simulations that support the protocol.
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.

Value

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

decision: One of "escalate", "de-escalate", "stay" and "stop".
next_dose: Integer matrix with columns dose_a and dose_b: one row for the next combination, two rows when two candidates tie, and no row when the trial stops.
tie: Whether two candidates tie, in which case the design picks one of them at random.
stop_reason: NA, or why the trial stops: "lowest_dose_eliminated", "lowest_dose_too_toxic" or "n_earlystop".
note: A sentence explaining a decision to stay when the rule called for a move that is not available, otherwise NA.
candidates: Data frame of the two combinations considered for an escalation or a de-escalation, one level of drug A and one level of drug B away, with their data, the posterior probability that their DLT rate lies between lambda_e and lambda_d, the score used to choose between them and whether they are available; NULL when no move was considered.
eliminated: Logical matrix of the combinations eliminated by the data.
boundary: The escalation, de-escalation and elimination boundaries at the number of patients treated at the current combination.

together with the data, the current combination and the design parameters.

Details

The decision follows the same steps as a simulated trial after each cohort. A combination whose DLT count reaches its elimination boundary is eliminated together with every combination with at least the same level of both drugs, and the trial stops if the lowest combination is eliminated or, with extrasafe, judged too toxic. The trial also stops when n_earlystop patients have been treated at the current combination and the design would neither escalate nor de-escalate. Otherwise the observed DLT rate at the current combination is compared with the boundaries of boin_boundary, and an escalation or a de-escalation moves one level of drug A or of drug B to the candidate with the higher score: the posterior probability, under a Beta(0.5, 0.5) prior, that its DLT rate lies between lambda_e and lambda_d, plus 0.0005 per patient already treated there. An eliminated combination is not a candidate for escalation. When the two scores are equal the design chooses at random, and the function returns both.

During a simulated trial a combination is eliminated as soon as its data reach the boundary and is never treated again, so the combinations eliminated by the data are exactly those eliminated in the course of the trial.

BOIN::next.comb() applies somewhat different rules from the simulations of BOIN::get.oc.comb(): it stops as soon as n.earlystop patients have been treated at the current combination, it does not escalate drug A when a combination with the next level of drug A and at most the current level of drug B has an observed DLT rate at or above lambda_d (and likewise for drug B), and it draws a random number to break ties. This function follows the simulations instead, so that a trial is run with the rules whose operating characteristics were evaluated. Where the rules coincide, the two functions agree.

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.

Examples

# Three levels of drug A and four of drug B; the last cohort was treated at
# A2B2, where two of three patients had a DLT
n_pts <- matrix(c(6, 6, 0, 0,
                  3, 3, 0, 0,
                  0, 0, 0, 0), nrow = 3, byrow = TRUE)
n_tox <- matrix(c(0, 1, 0, 0,
                  1, 2, 0, 0,
                  0, 0, 0, 0), nrow = 3, byrow = TRUE)

nxt <- comb_boin_next(target = 0.25, n_pts = n_pts, n_tox = n_tox,
                      current = c(2, 2))
nxt
#> BOIN drug-combination design: dose for the next cohort
#>   target DLT rate : 0.25
#>   current         : A2B2, 2 DLTs out of 3 patients (0.667)
#>   boundaries      : escalate if DLTs <= 0, de-escalate if DLTs >= 1, eliminate if DLTs >= 3 (3 patients)
#> 
#> Decision: de-escalate to A1B2
#> 
#> Candidates (interval 0.1968 to 0.2984):
#>  Direction Combination Patients DLTs Pr(in interval)  Score Available
#>     drug A        A1B2        6    1           0.218 0.2211       yes
#>     drug B        A2B1        3    1           0.167 0.1687       yes
#> 
#> Eliminated combinations: none
nxt$next_dose
#>      dose_a dose_b
#> [1,]      1      2