Toxicity Probability Matrices Generation for Trial Simulation

Description

A function to generate toxicity probability matrices employing proportional odds model, used as input for trial simulation. The generated matrices are a four-dimension array containing the probabilities of the occurrences of certain grades for certain types of toxicities, at each dose level and cycle under consideration. Dimension 1 refers to doses; dimension 2 corresponds to cycles of the treatment; dimension 3 regards the types of toxicities while dimenion 4 relates to grades.

Usage

GenToxProb(toxtype = c("Neurological", "Renal", "Hematological"), 
           intercept.alpha = c(2, 3, 4.2, 5.7), 
           coef.beta = c(-0.2, -0.4, -0.7), 
           cycle.gamma = 0, MaxCycle = 6, Numdose = 6)

Arguments

Details

The probability matrices of the occurrences of certain grades for certain types of toxicities, at each dose level and cycle are simulated from a proportional odds model that takes toxtype, intercept.alpha, coef.beta and cycle.gamma as the inputs. The model is as follows:

logit(c_j) = \alpha_j + \beta * dose + \gamma * cycle, j=0,1,2,3

where given a dose and cycle, c_j is the cumulative probabilities of occurrences of toxicities grades up to j; \alpha_0, \alpha_1, \alpha_2, \alpha_3 are the intercepts, making up intercept.alpha; coef.beta stores \beta for different types of toxicities while cycle.gamma acts as \gamma, controlling the cycle effect. The probability for the occurrence of a certain grade of a certain toxicity type at a given dose and cycle is obtained from taking the corresponding differences between cumulative probabilities.

Value

A four-dimension array where dimension 1 refers to doses; dimension 2 corresponds to cycles of the treatment; dimension 3 regards the types of toxicities while dimenion 4 relates to grades.

Examples

#tox.matrix <- GenToxProb()

Toxicity-based Dose estimation for the next patient cohort.

Description

RunRMD estimate the dosage for the next patient cohort using TTP measured multidimensional toxicity data collected from multiple treatemnt cycles of the existing patients.

Usage

  RunRMD(data, control, trlSize = 36, tox.target = 0.28, 
  sdose = 1:6, strDose = 1, iter=10000, burnin=4000, thin=1, chains=1)

Arguments

Details

The RunRMD function uses the patient data structure patdata to fit the linear mixed model as specified by formula based on the prior distributions as specified by control. The parameters of the MCMC simulation are specified by iter: the number of iterations, burnin: burn-ins, thin: thinning parameter and chains: number of chains in the MCMC simulation. The target nTTP score is defined by tox.target, default = 0.28.

Value

A list containing the following elements:

Examples

# Setup the prior distributions for the phase I dose-finding model:
control <- list(
 beta.dose = parm("normal", mean = 0, var = 1000),
 beta.other = parm("normal", mean = 0, var = 1000 ),
 gamma = parm("normal", mean = 0, var = 100 ),
 s2.gamma = parm("invgamma", shape = 0.001, scale = 0.001),
 s2.epsilon = parm("invgamma", shape = 0.001, scale = 0.001)
 )
 
# Load the example patient toxicity data:
# data(patdata)

# Estimate the dose-recommendation of the next patient cohort based on the toxicity data:
# fit <- RunRMD(data = patdata, control = control, trlSize = 36, tox.target = 0.28, 
#  sdose = 1:6, strDose = 1, iter = 2000, burnin = 500, thin = 1, chains = 1)
  

Interactive Implementation for an Adaptive, Multi-Stage Phase I Dose-Finding Design

Description

A function to interactively implement an adaptive, multi-stage phase 1 dose-finding design incorporating continuous efficacy and toxicity data from multiple treatment cycles, proposed by Du et al(2017)

Usage

RunRMDEFF(efficacy.dat = NULL, toxicity.dat, trialSize = 36,
          seed = 624, chSize = 3, MaxCycle = 6, doses = 1:6, 
          tox.target = 0.28, p1 = 0.2, p2 = 0.2, ps1 = 0.2, 
          thrd1 = 0.28, thrd2 = 0.28, proxy.thrd = 0.1, 
          dose_flag = 0, param.ctrl = list())

Arguments

Details

p1,p2,thrd1 and thrd2 are used to define allowable (safe) doses through the two probability conditions:

P(tox1 < thrd1) > p1

P(tox2 < thrd2) > p2

, where tox1 and tox2 denote the estimated toxicity at cycle 1 of the treament and late cycles, respectively. In stage 1, ps1 acts as p1 and p2.

Note that our design allows patients dropout due to DLT events, so the data frames efficacy.dat and toxicity.dat only include the information for the available patients and available cycles.

Value

The function prints information regarding the posterior toxicity/efficacy profile for each dose under exploration, the recommended dose for the next cohort of patients, and return a list with the following components:

Examples

#load the patient efficacy data
#data(eff_dat)
#head(eff_dat)
#----------------------#
#load the patient toxicity data
#data(tox_dat)
#head(tox_dat)

#RunRMDEFF(efficacy.dat = eff_dat, toxicity.dat = tox_dat)

Simulate trial characters of the phase I dose-finding design.

Description

SimRMD uses the design parameters to simulate the trial characters of phase I dose-finding design.

Usage

SimRMD(seed=2014, strDose=1, chSize=3, trlSize=36,
   numTrials=1000, sdose=1:6, MaxCycle=6, tox.target=0.28,
   control, iter=10000, burnin=4000, thin=1, chains=1,    
   pathout='./', tox.matrix, wm = matrix(c(0, 0.5, 0.75, 1  , 1.5, 
                                        0, 0.5, 0.75, 1  , 1.5, 
                                        0, 0  , 0   , 0.5, 1  ), 
                                        byrow = T, ncol = 5), toxmax = 2.5)

Arguments

Details

The SimRMD function uses the patient data structure patdata to fit the linear mixed model as specified by formula based on the prior distributions as specified by control. The parameters of the MCMC simulation are specified by iter: the number of iterations, burnin: burn-ins, thin: thinning parameter and chains: number of chains in the MCMC simulation. The target nTTP score is defined by tox.target, default = 0.28.

Value

A list containing the following elements:

Examples

# Setup the prior distributions for the phase I dose-finding model:
control <- list(
 beta.dose = parm("normal", mean = 0, var = 1000),
 beta.other = parm("normal", mean = 0, var = 1000 ),
 gamma = parm("normal", mean = 0, var = 100 ),
 s2.gamma = parm("invgamma", shape = 0.001, scale = 0.001),
 s2.epsilon = parm("invgamma", shape = 0.001, scale = 0.001)
 )


# Generate the toxicity matrix
# tox.matrix <- GenToxProb(
#    toxtype = c("Renal", "Neuro", "Heme"), 
#	intercept.alpha = c(2, 3, 4.2, 5.7), 
#	coef.beta = c(-0.2, -0.4, -0.7), 
#	cycle.gamma = 0)

# Simulate the dose-recommendation based on the toxicity matrix
# simu <- SimRMD(seed=2014, strDose=1, chSize=3, trlSize=12,
#    numTrials=1, sdose=1:6, MaxCycle=5, tox.target=0.28,
#    control=control, iter=10, burnin=2, thin=1, chains=1,    
#    pathout='./', tox.matrix=tox.matrix)
	

Simulation for an Adaptive, Multi-Stage Phase I Dose-Finding Design

Description

A function to implement simulation for an adaptive, multi-stage phase 1 dose-finding design incorporating a continuous efficacy outcome and toxicity data from multiple treatment cycles, proposed by Du et al(2017)

Usage

SimRMDEFF(numTrials = 100, trialSize = 36, doses = 1:6, cycles = 1:6,  
          eff.structure = c(0.1, 0.2, 0.3, 0.4, 0.7, 0.9), 
          eff.sd = 0.2, tox.target = 0.28, p1 = 0.2, p2 = 0.2, 
          ps1 = 0.2, StrDose = 1, chSize = 3, 
          tox.matrix = NULL, 
          proxy.thrd = 0.1, thrd1 = 0.28, thrd2 = 0.28, 
          wm = matrix(c(0, 0.5, 0.75, 1  , 1.5, 
                        0, 0.5, 0.75, 1 , 1.5, 
                        0, 0  , 0   , 0.5, 1  ), 
                        byrow = T, ncol = 5), 
          toxmax = 2.5, toxtype = NULL, intercept.alpha = NULL, 
          coef.beta = NULL, cycle.gamma = NULL, 
          param.ctrl = list())

Arguments

Details

p1,p2,thrd1 and thrd2 are used to define allowable (safe) doses through the two probability conditions:

P(tox1 < thrd1) > p1

P(tox2 < thrd2) > p2

, where tox1 and tox2 denote the toxicity at cycle 1 of the treament and later cycles, respectively. In stage 1, ps1 acts as p1 and p2.

The user can explore different dose-efficacy patterns using the argument eff.structure. For example, eff.structure = c(0.5, 0.5, 0.5, 0.5, 0.5, 0.5) corresponds to a flat relationship where the efficacy does not change with dose.

When tox.matrix is null, the probability matrices of the occurrences of certain grades for certain types of toxicities, at each dose level and cycle will be simulated from a proportional odds model that takes toxtype, intercept.alpha, coef.beta and cycle.gamma as the inputs. The model is as follows:

logit(c_j) = \alpha_j + \beta * dose + \gamma * cycle, j=0,1,2,3

where given a dose and cycle, c_j is the cumulative probabilities of occurrences of toxicities grades up to j; \alpha_0, \alpha_1, \alpha_2, \alpha_3 are the intercepts, making up intercept.alpha; coef.beta stores \beta for different types of toxicities while cycle.gamma acts as \gamma, controlling the cycle effect. The probability for the occurrence of a certain grade of a certain toxicity type at a given dose and cycle is obtained from taking the corresponding differences between cumulative probabilities.

Value

A list with the following components:

Examples

#######Dose 5 target toxicity dose, flat cycle effect#####################

###############a flat dose-efficacy pattern###############################

# simul <- SimRMDEFF(toxtype = c("H", "L", "M"), 
                   # intercept.alpha = c(1.9, 2.3, 2.6, 3.1), 
                   # coef.beta = c(-0.3, -0.2, -0.25), 
                   # cycle.gamma = 0, tox.target = 0.23, 
                   # thrd1 = 0.23, thrd2 = 0.23,
                   # eff.structure = rep(0.5, 6),
                   # p1 = 0.1, p2 = 0.1, ps1 = 0.1, 
                   # numTrials = 1)

Compute the nTTP score baed on the observed toxicity.

Description

Tox2nTTP use the design parameters to compute the nTTP score baed on the observed toxicity.

Usage

Tox2nTTP(tox = c(1,2,0),
 wm = matrix(c(0, 0.5, 0.75, 1 , 1.5,
 0, 0.5, 0.75, 1 , 1.5,
 0, 0 , 0 , 0.5, 1 ),
 byrow = T, ncol = 5),
 toxmax = 2.5)

Arguments

Details

The Tox2nTTP function calculates the nTTP values based on the weight matrix.

Value

nTTP value.

Examples

#data(tox_matrix)

# Simulate the operating characteristics of clinical trials:
#Tox2nTTP(tox = c(1,2,0))
 

Patient efficacy data

Description

This data set provides the efficacy data from 22 patients.

Usage

eff_dat

Format

A matrix containing 22 observations.

Source

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

References

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

Create prior for the phase 1 dose-finding model

Description

parm creates an S3 object of class parm that contains all necessary prior information to estimate the phase I repeated measures model.

Usage

parm(prior = c("gamma", "invgamma", "normal"), 
   mean=0, var=100, shape=0.001, scale=0.001)

Arguments

Details

The function parm builds three types of priors: gamma, invgamma and normal.

Gamma parameters:

• shape : shape parameter for Gamma prior for sigma

• rate : rate parameter for Gamma prior for sigma

Inverse gamma parameters:

• shape : shape parameter for the inverse gamma prior on sigma.

• scale : scale parameter for the inverse gamma prior on sigma.

Normal parameters:

• mean : mean parameter for the normal prior.)

• var : variance parameter for the normal prior.)

Value

An object of class parm, containing information for the prior.

Author(s)

Jun Yin

Examples

# Setup the prior distributions for the phase I dose-finding model:
control <- list(
 beta.dose = parm("normal", mean = 0, var = 1000),
 beta.other = parm("normal", mean = 0, var = 1000 ),
 gamma = parm("normal", mean = 0, var = 100 ),
 s2.gamma = parm("invgamma", shape = 0.001, scale = 0.001),
 s2.epsilon = parm("invgamma", shape = 0.001, scale = 0.001)
 )

Patient toxicity data

Description

This data set provides the toxicity data from multiple treatment cycles. The toxicity data is defined by nTTP toxicity score.

Usage

patdata

Format

A data frame containing 9 observations.

Source

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

References

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

plot nTTP Toxicity-based in cycle 1 for all doses.

Description

plot.RunRMDVal plot the dose-nTTP estimation in cycle based on the RunRMD estimation.

Usage

## S3 method for class 'RunRMDVal'
plot(x, ...)

Arguments

Details

The plot.RunRMDVal function uses the patient data structure patdata to fit the linear mixed model as specified by formula based on the prior distributions as specified by control. The parameters of the MCMC simulation are specified by iter: the number of iterations, burnin: burn-ins, thin: thinning parameter and chains: number of chains in the MCMC simulation. The target nTTP score is defined by tox.target, default = 0.28.

Value

This function plots the dose-nTTP estimation in cycle based on the RunRMD estimation.

Examples

# Setup the prior distributions for the phase I dose-finding model:
control <- list(
 beta.dose = parm("normal", mean = 0, var = 1000),
 beta.other = parm("normal", mean = 0, var = 1000 ),
 gamma = parm("normal", mean = 0, var = 100 ),
 s2.gamma = parm("invgamma", shape = 0.001, scale = 0.001),
 s2.epsilon = parm("invgamma", shape = 0.001, scale = 0.001)
 )
 
# Load the example patient toxicity data:
#data(patdata)

# Estimate the dose-recommendation of the next patient cohort based on the toxicity data:
#res <- RunRMD(data=patdata, control=control, tox.target=0.28, 
#  iter=2000, burnin=500, thin=1, chains=1)

#plot(res);

The probability matrix of toxicity events

Description

This data set provides the toxicity matrix of 3 types of toxicity events over 6 doses and 5 treatment cycles.

Usage

probaT

Format

A list containing 3 matrices for the probability for 3 type of toxicity events, each matrix with 6 rows for 6 doses and 5 columns for 5 treatment cycles.

Source

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

References

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

Probability matrix for toxicity

Description

This data set is a four-dimension array (Dose, Cycle, Toxicity Type, Toxicity Grade) specifying the probabilities of the occurrences of certain grades for certain types of toxicities.

Usage

tox.matrix

Format

A 4-dimension array.

Source

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

References

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

Patient toxicity data

Description

This data set provides the toxicity data from 9 patients with multiple treatment cycles, matching with the efficacy data.

Usage

tox_dat

Format

A data frame containing 183 observations.

Source

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

References

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

Patient toxicity data

Description

This data set provides the toxicity data in three toxicity types from 33 patients with multiple treatment cycles, matching with the efficacy data.

Usage

tox_dat_bytype

Format

A data frame containing 125 observations.

Source

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.

References

Yin, Jun, et al. "A Bayesian dose finding design incorporating toxicity data from multiple treatment cycles." Statistics in Medicine 36.1 (2017): 67-80.