CRM (version 1.1.0)

crmsim: CRM Simulator

Description

crmsim, crmsiminc1 and crmsiminc2 implement the continued reassessment method (CRM) for dose finding in Phase I clinical trials. The operating characteristics of CRM are summarized through simulations.

crmsim allows users to select a variety of cohort sizes. A cohort of subjects are treated at the same dose.

The cohort size is fixed to 1 in crmsiminc1 and crmsiminc2. crmsiminc1 implements an algorithm that allows a clinical trial to proceed to the next subject's dose assignment before observing the last subject's toxicity data. crmsiminc2 allows a clinical trial to proceed to the next subject's dose assignment before observing the last two subject's toxicity data (see Iasonos et al. for details).

Usage

crmsim(target,prior,true,rate,cycle,cohort=1,nsubject=24,nsim=1000,
       model=1,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)
crmsiminc1(target,prior,true,rate,cycle,nsubject=24,nsim=1000,
           model=1,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)
crmsiminc2(target,prior,true,rate,cycle,nsubject=24,nsim=1000,
           model=1,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)

Arguments

target

Target probability of toxicity. The value must be in (0, 1).

prior

Prior probabilities of toxicity for each dose. The values must be in (0,1) and in an ascending order. For example, prior=c(0.05,0.1,0.2,0.3,0.5,0.7),which corresponds to dose levels 1, 2, 3, 4, 5 and 6,respectively.

true

True probabilities of toxicity. The values must be in (0,1) and in an ascending order. e.g. (0.1,0.2,0,3,0.4,0.5,0.8)

rate

Recruitment/accrual rate of subjects in a 30 window. For example,if 1 subject can be recruited per 30 days, then set rate = 1/30 = 0.033; if 2 patients per 30 day then rate = 2/30 = 0.0667.

cycle

The length of treatment cycle in days.

cohort

Cohort size of subjects entering into the trials. Default is 1. The value for cohort must be less than or equal to the value for nsubject.

nsubject

Total number of subjects in one simulation(or trial). Default is 24. nsubject should be equal to n*cohort,where n is positive integer.

nsim

Total number of simulations. Default is 1000.

model

Dose-toxicity model. The value must be 1 (hyperbolic tangent model) or 2 (one-parameter logistic model). Default is 1. Hyperbolic tangent model: $p(y=1|x,a) = ((tanh(x)+1)/2)^a$. One-parameter logistic model: $p(y=1|x,a,b) = exp(b+ax)/(1+exp

a0

Initial value for parameter a. Default is 1.0.

b

A fixed parameter for the one-parameter logistic model. Default is 3.0.

jump

jump=FALSE means NOT allowing that the proposed dose by the CRM program has an increase of more than one level than the previous level; jump=TRUE means allowing more-than-one-level increase of the proposed dose by the CRM program.

start.dose

Initial dose for each trial. Default is 1.

seed

Seed for the random number generator. Default is 777.

Value

• SimResult is a matrix that summarizes the operating characteristics of CRM. The column names are the dose levels. The row names are the operating characteristics.
• % Selection
  the percentage of selection of each dose as MTD
• % Subjects Treated
  the percentage of subjects treated at each dose
• # Subjects Treated
  the average number of subjects treated at each dose
• Average Toxicities
  the average toxicities per trial at each dose
• True Probabilities
  the true probability of toxicity of each dose

References

John O'Quigley, Margaret Pepe,and Lloyd Fisher. (1990). Continual Reassessment Method: A Practical Design for Phase 1 Clinical Trials in Cancer. Biometrics, 46, pp.33-48

Alexia Iasonos,Andrew S Wilton,Elyn R Riedel,Venkatraman E Seshan and David R Spriggs. A comprehensive comparison of the continual reassessment method to the standard 3+3 dose escalation scheme in Phase I dose-finding studies. Clinical Trials,2008,nil:1-12

See Also

crm

Examples

prior1 <- c(0.05,0.1,0.2,0.3,0.5,0.7)
true1 <- c(0.1,0.15,0.2,0.4,0.5,0.8)

# simulations using model 1 (hyperbolic tangent model)
crmsim(target=0.2,prior=prior1,true=true1,rate=0.1,cycle=21,cohort=1,nsubject=24,nsim=100,
       model=1,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)

crmsiminc1(target=0.2,prior=prior1,true=true1,rate=0.1,cycle=21,nsubject=24,nsim=100,
           model=1,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)

crmsiminc2(target=0.2,prior=prior1,true=true1,rate=0.1,cycle=21,nsubject=24,nsim=100,
           model=1,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)

# simulations using model 2 (one-parameter logistic model)
crmsim(target=0.2,prior=prior1,true=true1,rate=0.1,cycle=21,cohort=1,nsubject=24,nsim=100,
       model=2,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)

crmsiminc1(target=0.2,prior=prior1,true=true1,rate=0.1,cycle=21,nsubject=24,nsim=100,
           model=2,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)

crmsiminc2(target=0.2,prior=prior1,true=true1,rate=0.1,cycle=21,nsubject=24,nsim=100,
           model=2,a0=1,b=3,jump=FALSE,start.dose=1,seed=777)