dfped (version 1.1)

simu: A simulation of a single dose-finding trials in paediatrics.

Description

Simulate a single dose-finding clinical trial with the given scenarios of toxicity and efficacy.

Usage

simu(targetTox, targetEff, skeletonTox, skeletonEff, startingDose, 
        nbSubjects, crmModel, cohortSize, scenarioTox, scenarioEff, 
        nbDesign, mu, sd = NULL, lesb, sigmaLI, sigmaHI, adaptivePrior)

Arguments

targetTox

Target/threshold of toxicity; must be a integer/double.

targetEff

Target/threshold of efficacy; must be a integer/double.

skeletonTox

Skeleton of toxicity for the BMA bivariate CRM, or the bivariate CRM. Must be a dataframe with the number of row corresponding to the number of doses and the number of columns corresponding to the number of working models for toxicity.

skeletonEff

Skeleton of efficacy for the BMA bivariate CRM, or the bivariate CRM. Must be a dataframe with the number of row corresponding to the number of doses and the number of columns corresponding to the number of working models for efficacy.

startingDose

First dose to be assigned; must be an integer.

nbSubjects

Maximum number of allocated patients; must be an integer.

crmModel

A model for STAN in C++.

cohortSize

The size of the cohorts for the 3+3 based algorithm before kickoff of the CRM; must be an integer.

scenarioTox

Toxicity scenario for the simulations with the probability of toxicity for each dose; must be a vector of length the number of doses.

scenarioEff

Efficacy scenario for the simulations; must be a vector of length the number of doses.

nbDesign

The number of different designs for the model selection using the Watanabe-Akaike information criteria (WAIC); must be an integer.

mu

The mean value which the model is using.

sd

The standard deviation.

lesb

A vector consisting of the variables b.

sigmaLI

The standard deviation when the model using non-informative prior.

sigmaHI

The standard deviation when the model using informative prior.

adaptivePrior

TRUE if you want to use as a prior an adaptive prior; FALSE otherwise.

References

Petit, C., et al, (2016) Unified approach for extrapolation and bridging of adult information in early phase dose-finding paediatric studies, Statistical Methods in Medical Research, <doi:10.1177/0962280216671348>.

See Also

simulation

Examples

# NOT RUN {
    library(rstan)
    adaptivePrior <- TRUE

    ####### Targets  ###########

    targetTox <- 0.25   # target of toxicity
    targetEff <- 0.20   # target of efficacy 

    ####### Skeleton ###########

    skeleton_tox1 <- c(0.10, 0.21, 0.33, 0.55, 0.76) 
    skeleton_tox2 <- c(0.21, 0.33, 0.55, 0.76, 0.88) 
    skeleton_tox3 <- c(0.05, 0.10, 0.21, 0.33, 0.55) 
    skeleton_tox4 <- c(0.025, 0.05,0.1, 0.21, 0.33) 
    skeleton_tox5 <- c(0.0125, 0.025, 0.05,0.1, 0.21) 

    skeleton_eff <- c(0.04937516, 0.20496890, 0.43388003, 0.64409781, 0.79313693) 
    skeleton_tox <-  data.frame(skeleton_tox1, skeleton_tox2, skeleton_tox3,
                        skeleton_tox4, skeleton_tox5)
    skeleton_eff <-  data.frame(skeleton_eff, skeleton_eff, skeleton_eff,
                        skeleton_eff, skeleton_eff)  

    ############# Priors ############
    priorModel  <- list(rep(1/5,5), 0.001)
    sd <- 0.65  
    mu <- -0.34   

    ####### Trial settings  #############

    startingDose <- 1 
    nbSubjects <- 15   
    cohortSize <- 3       

    nbDesign <- length(skeleton_tox[1,])  
    nbDoses <- length(scenario_tox)       

    lesb <- calcul.bi(skeleton_tox[,1], mu, a = NULL, "power_log", targetTox) 
    sigmaLI <- sigmaLI(skeleton_tox[,1],  mu, a = NULL, "power_log", targetTox)
    sigmaHI <- sigmaHI(skeleton_tox[,1],  mu, a = NULL, "power_log", targetTox, 0.80)

    ################## Scenarios ##############

    scenario_tox <- c(0.1301477, 0.2774171, 0.4184642, 0.6486846, 0.8257219)  
    scenario_eff <- c(0.07945205, 0.20000000, 0.33686856, 0.59537737, 0.80996173)  

    stancode <- 'data {
        int <lower = 0> J; //nb of patients
        int <lower = 0> K; // nb of doses and dose reference
        real r[K]; // skeleton for tox - K doses
        real q[K]; // skeleton for efficacy - K doses
        int y[J]; // toxicity of patient j
        int v[J]; // efficacy of patient j
        int d[J]; // dose received by patient j
        real moy; // mean for the normal prior of toxicity
        real standardError; //standard error of the normal prior of toxicity
    }
    parameters {
        real <lower = 0> alpha;
        real  <lower = 0> beta;
    }
    transformed parameters{
        real <lower = 0, upper = 1> varphi[K]; // marginal probability of toxicity for dose k
        real <lower = 0, upper = 1> psi[K]; // marginal probability of efficacy for dose k
        // defining the marginal probabilities for each value of a and b for each dose
        real p01[K]; // tox = 0, eff = 1
        real p10[K]; // tox = 1, eff = 0
        real p11[K]; // tox = 1, eff = 1
        real p00[K]; // tox = 0, eff = 0

        vector[J] logLike;

        for (k in 1:K){
            varphi[k] = exp(alpha*log(r[k]));
            psi[k] = exp(beta*log(q[k]));
        }
        
        
        // computing the marginal probabilities for each dose
            for (k in 1:K){
                p01[k] = (1-varphi[k])*psi[k]; 
                p10[k] = varphi[k]*(1-psi[k]);
                p00[k] = (1-varphi[k])*(1-psi[k]);
                p11[k] = varphi[k]*psi[k];
            }
        // Computing the log-likelihood
            for (j in 1:J){
                logLike[j] = y[j]*v[j]*log(p11[d[j]]) + y[j]*(1-v[j])*log(p10[d[j]]) 
                + (1-y[j])*v[j]*log(p01[d[j]]) + (1-y[j])*(1-v[j])*log(p00[d[j]]);
            }
        
    }
    model {
        // priors
        alpha ~lognormal(moy, standardError);
        beta ~ lognormal(0,sqrt(1.34));              
        increment_log_prob(sum(logLike));
    }'
    
    crm_model <- stan_model(model_code = stancode) 

    ################## Simulation ##############

    simu(targetTox, targetEff, skeleton_tox, skeleton_eff,
        startingDose, nbSubjects, crm_model, cohortSize, scenario_tox,
        scenario_eff, nbDesign, mu, sd = sd, lesb,
        sigmaLI, sigmaHI, adaptivePrior)
# }