Package 'SubTite'

Determines Prior ESS for fixed values of sigma_alpha^2 and sigmabeta^2

Description

Uses the prior means for the intercept and slope parameters and the number of doses to obtain an approximate prior ESS for the given prior variances. The user should calibrate varint and varbeta with varint>varbeta such that the ESS value is 1.

Usage

GetESS(
  Dose,
  meanmu,
  meanslope,
  MeanInts,
  MeanSlopes,
  VarInt,
  VarSlope,
  phetero
)

Arguments

Dose	Vector containing standardized doses.
meanmu	Prior mean for baseline intercept.
meanslope	Prior mean for baseline slope.
MeanInts	Vector of prior means for the group specific intercept parameters.
MeanSlopes	Vector of prior means for the group specific slope parameters.
VarInt	Prior variance for the intercept parameters.
VarSlope	Prior variance for the slope parameters.
phetero	Prior probability of clustering

Value

Returns the nonlinear regression model whos parameter estimates will be used as prior means for the SubTITE Design.

References

[1] Chapple and Thall (2017), Subgroup-specific dose finding in phase I clinical trials based on time to toxicity allowing adaptive subgroup combination.

Examples

###Specify the prior hypermeans
meanmu=-.5
meanslope=-.05
MeanInts = c(0,-.5,-.1)
MeanSlopes = c(0,.1,0)
Dose=sort(rnorm(5))
VarInt=5
VarSlope=1
phetero=.9
GetESS(Dose,meanmu,meanslope,MeanInts,MeanSlopes,VarInt,VarSlope,phetero)

---

Obtains true simulation parameters for each supported distribution function to correspond to a probability of the truth by time T1.

Description

Obtains true simulation parameters for each supported distribution function to correspond to a probability of the truth by time T1.

Usage

GetParams(Family, ParamNum, Param, GroupProb, T1)

Arguments

Family	What distribution Family to simulate from. Options include: Exponential,Gamma, Lognormal, Uniform, Weibull.
ParamNum	Parameter index for user set value. For example, ParamNum=1 for a Gamma distribution means that the user will supply the shape parameters in the param matrix. If ParamNum=2, the user will supply the rate parameters in the param matrix.
Param	#Groups X #Doses Matrix containing one parameter for each subgroup and dose.
GroupProb	#Groups X #Doses Matrix containing the true toxicity probability by time T1.
T1	Toxicity observation window.

Value

A list containing the hyperparameter matrices to input into the SimTrial function. Also plots the hazard of toxicity for each subgroup and dose.

Examples

GroupProb =matrix(c(.05,.3,.6,.7,.8,.01,.02,.13,.27,.5),nrow=2,byrow=TRUE)
##True Simulation distribution
Family="Weibull"
T1=6
Param = GroupProb*0 + 4 ##Late onset weibull
SimTruth = GetParams("Weibull",1,Param,GroupProb,T1)

---

Calibrates prior means for Dose Finding Trial

Description

Uses the clinician elicited prior reference probabilities for each subgroup and dose to obtain prior means for the Bayesian logistic regression model used in the SubTite trial design.

Usage

GetPriorMeans(Prior, Dose)

Arguments

Prior	#Groups X #Doses matrix containing the elicited prior toxicity probabilities at the reference time for each dose and subgroup.
Dose	Vector containing standardized doses.

Value

Returns the a list containing the nonlinear regression model whos parameter estimates will be used as prior means for the SubTITE Design.

References

[1] Chapple and Thall (2017), Subgroup-specific dose finding in phase I clinical trials based on time to toxicity allowing adaptive subgroup combination

Examples

##Specify elicited reference toxicity probabilities
Prior = matrix(c(.2,.3,.4,.5,.6,.1,.2,.3,.4,.5,.05,.1,.15,.2,.3),byrow=TRUE,nrow=3)
Dose=sort(rnorm(5))
GetPriorMeans(Prior,Dose)

---

Gives the subgroup specific optimal dose vector.

Description

Returns a list containing the optimal doses to enroll each subgroup at and the subgroups that should have their accrual suspended temporarily.

Usage

GetSubTite(
  Y,
  I,
  Doses,
  Groups,
  Include = rep(1, length(Y)),
  ID,
  cohort,
  Conservative,
  T1,
  Target,
  Upper,
  Dose,
  meanmu,
  meanslope,
  MeanInts,
  MeanSlopes,
  VarInt,
  VarSlope,
  phetero,
  Borrow,
  B
)

Arguments

Y	Vector containing observed event or censoring times.
I	Vector containing event indicators (1 if patient experiences an event for a patient).
Doses	Vector containing numerical doses assigned to patients in the trial.
Groups	Vector containing group assignment of patients, 1 is baseline group.
Include	Binary vector indicating whether each patient record should be included in the decision making process.
ID	Vector of patient IDs. Can be numeric or character valued.
cohort	Number of patients needed to be assigned at a dose level prior to escalation.
Conservative	Binary Indicator of Whether conservative escalation, i.e. not allowing escalation until cohort patients have been fully evaluated at the highest tried dose level.
T1	Reference time for toxicity.
Target	Target cumulative toxicity probability vector at time T1.
Upper	Cutoff values used to determine if accrual in a subgroup should be suspended.
Dose	Vector containing the standardized doses considered.
meanmu	Prior mean for baseline intercept.
meanslope	Prior mean for baseline slope.
MeanInts	Vector of prior means for the group specific intercept parameters.
MeanSlopes	Vector of prior means for the group specific slope parameters.
VarInt	Prior variance for the intercept parameters.
VarSlope	Prior variance for the slope parameters.
phetero	Prior probability of heterogeneous subgroups.
Borrow	Parameter to specify subgroup borrowing/clustering. 0=No borrowing, 1=Borrowing but no clustering, 2=Borrowing and clustering.
B	Number of Iterations to run for MCMC

Value

Returns a list with two objects, a vector of optimal doses for each subgroup and matrix of posterior toxicity probabilities at each dose level within each subgroup.

References

[1] Chapple and Thall (2017), Subgroup Specific Dose Finding in Phase I Clinical Trials Based on Time to Toxicity Within a Fixed Follow Up Period.

Examples

T1=28 ##Reference time for toxicity
Target=rep(.3,2) ##Target toxicity probability
Upper=rep(.95,2) ##Upper cutoffs for excessive toxicity
##How many patients in each subgroup have been assigned at each dose level?
cohort=3 ##Cohort size required for escalation
Conservative = 1 ##Conservative escalation
##Only can escalate with a fully evaluated cohort at the highest dose level.
##Matrix of umber of patients tried or fully evaluated at each dose level.
##Hyperparameters
meanmu=-0.4467184 ##Common Intercept hypermean
meanslope= 0.8861634 ##Common slope hypermean
MeanInts =c(0, -0.5205379) ##Group Intercept hypermeans
MeanSlopes = c(0, 0.1888923) ##Group slope hyperneabs
VarInt=5 #Prior Variance of the intercept betas
VarSlope=1 ##Prior Variance of slope betas
phetero=.9 ##Prior Probability of hetergeneity
Borrow=0 ##Borrowing specification, 0=none, 1=some, 2=clustering.
B=5000 ##Number of iterations
Borrow=2
Y=c(28,26,29,28,29,5,1)
RawDose=c(350,420,530,660,825)
Dose=(RawDose-mean(RawDose))/sd(RawDose)
I <- c(0,0,0,0,0,0,0)
Doses <- rep(2,7)
Groups <- c(0,1,1,0,0,1,1)
Include <- rep(1,7)
ID=1:length(Y)
Z=GetSubTite(Y, I,Doses, Groups, Include,ID,cohort, Conservative,
T1,Target,  Upper, Dose,  meanmu, meanslope,
 MeanInts,  MeanSlopes ,VarInt,VarSlope,phetero, Borrow,B)
Z

---

Performs MCMC and returns needed values for dose-finding in a list.

Description

Performs MCMC and returns needed values for dose-finding in a list.

Usage

MCMC(
  Y,
  I,
  Doses,
  Groups,
  T1,
  Target,
  Upper,
  Dose,
  meanmu,
  meanslope,
  MeanInts,
  MeanSlopes,
  varint,
  varbeta,
  phetero,
  Stopped,
  NumPat,
  SubRout,
  B
)

Arguments

Y	Vector containing observed event or censoring times.
I	Vector containing event indicators (1 if patient experiences an event for a patient).
Doses	Vector containing Doses of patients in trial.
Groups	Vector containing group assignment of patients, 0 is baseline group.
T1	Reference time for toxicity.
Target	Target cumulative toxicity probability vector at time T1.
Upper	Cutoff values used to determine if accrual in a subgroup should be suspended.
Dose	Vector containing the standardized doses considered.
meanmu	Prior mean for baseline intercept.
meanslope	Prior mean for baseline slope.
MeanInts	Vector of prior means for the group specific intercept parameters.
MeanSlopes	Vector of prior means for the group specific slope parameters.
varint	Prior variance for the intercept parameters.
varbeta	Prior variance for the slope parameters.
phetero	Prior probability of heterogeneous subgroups.
Stopped	Current vector of STOPPED groups
NumPat	Number of patients
SubRout	Parameter to specify subgroup borrowing/clustering. 0=No borrowing, 1=Borrowing but no clustering, 2=Borrowing and clustering.
B	Number of Iterations to run for MCMC

Value

A list of quantities needed for determining the next dose to enroll each subgroup.

---

Performs MCMC and returns needed values for dose-finding in a list.

Description

Performs MCMC and returns needed values for dose-finding in a list.

Usage

MCMCSIM(
  Y,
  I,
  Doses,
  Groups,
  T1,
  Target,
  Upper,
  Dose,
  meanmu,
  meanslope,
  MeanInts,
  MeanSlopes,
  varint,
  varbeta,
  phetero,
  Stopped,
  NumPat,
  SubRout,
  B
)

Arguments

Y	Vector containing observed event or censoring times.
I	Vector containing event indicators (1 if patient experiences an event for a patient).
Doses	Vector containing Doses of patients in trial.
Groups	Vector containing group assignment of patients, 0 is baseline group.
T1	Reference time for toxicity.
Target	Target cumulative toxicity probability vector at time T1.
Upper	Cutoff values used to determine if accrual in a subgroup should be suspended.
Dose	Vector containing the standardized doses considered.
meanmu	Prior mean for baseline intercept.
meanslope	Prior mean for baseline slope.
MeanInts	Vector of prior means for the group specific intercept parameters.
MeanSlopes	Vector of prior means for the group specific slope parameters.
varint	Prior variance for the intercept parameters.
varbeta	Prior variance for the slope parameters.
phetero	Prior probability of heterogeneous subgroups.
Stopped	Current vector of STOPPED groups
NumPat	Number of patients
SubRout	Parameter to specify subgroup borrowing/clustering. 0=No borrowing, 1=Borrowing but no clustering, 2=Borrowing and clustering.
B	Number of Iterations to run for MCMC

Value

A matrix of quantities needed for determining the next dose to enroll each subgroup while using the SimTrial function.

---

Gives summaries of GetSubTite Objects.

Description

Gives summaries of GetSubTite Objects.

Usage

Print_SubTite(Z)

Arguments

Z	List produced by GetSubTite.

---

Simulates a Sub-TITE trial design

Description

Simulates replicates from a Sub-TITE trial with user specified true toxicity time distributions for different doses and subgroups and returns average summary statistics of the trial.

Usage

SimTrial(
  nSims,
  Nmax,
  T1,
  Target,
  Dose,
  DoseStart,
  Upper,
  Accrue,
  groupprob,
  meanmu,
  meanslope,
  MeanInts,
  MeanSlopes,
  VarInt,
  VarSlope,
  phetero,
  Family,
  SimTruth,
  NSep,
  NBorrow,
  cohort,
  FULL
)

Arguments

nSims	Number of Trials to Simulate.
Nmax	Maximum Number of Patients to enroll in the trial.
T1	Reference time for toxicity.
Target	Target cumulative toxicity probability (or subgroup specific vector) at time T1.
Dose	Standardized vector of doses to try.
DoseStart	Dose (or vector of Doses) to enroll the first patient in each subgroup at.
Upper	Cutoff values used to determine if accrual in a subgroup should be suspended.
Accrue	Expected montly patient accrual rate.
groupprob	Probability vector of subgroup assignment.
meanmu	Prior mean of the baseline intercept parameter.
meanslope	Prior mean of the baseline slope parameter.
MeanInts	G-1 length vector of subgroup specific prior intercept means.
MeanSlopes	G-1 length vector of subgroup specific prior slope means.
VarInt	Prior Variance of Intercept Parameters.
VarSlope	Prior Variance of Slope Parameters.
phetero	Prior probability of clustering
Family	What distribution Family to simulate from. Options include: Exponential,Gamma, Lognormal, Uniform, Weibull.
SimTruth	List of 2 #Groups by #Doses matrices containing the true parameter values needed for simulating from different true time to toxicity distributions. When a Uniform distribution is chosen, the user will instead supply the true toxicity probabilities for each dose/subgroup combination in both list entries. For a gamma distribution, the user will supply a matrix for the shape parameters for each dose and subgroup, and a second matrix for the rate parameters of each dose and subgroup.
NSep	Number of patients to assign based on no borrowing.
NBorrow	Number of patients to assign based on no clustering
cohort	Number of patients to enroll before escalating.
FULL	Do we have to fully evaluate a cohort before escalating?

Value

A list with first entry corresponding to summaries of the operating characteristics of the design including

Examples

##Note: nSims  should be set larger than the example below.
nSims=1
###TRIAL PARAMETERS###
##Specify reference toxicity time and target
T1=6
Target=.3
##Number of Groups
##Specify upper bound for determining if the lowest dose is too toxic in a subgroup
Upper=c(.95,.95)
#' ##Standardized Dose Values and starting dose index
Dose=sort(rnorm(5))
DoseStart=1
##Maximum Sample Size
Nmax=25
##Number of patients to run separately
NSep=0
##Number of patients to borrow, but NOT cluster
NBorrow=0
##Number of patients to fully evaluate or TREAT before ESCALATING
cohort=3
##Do we fully evaluate a cohort before escalating?
FULL=0
#HYPERPARAMETERS#
##Hypermeans for baseline terms
meanmu=2.21
meanslope=-.57
##Hypervectors for subgroup specific terms
MeanInts = c(0,.46)
MeanSlopes = c(0,.04)
##Hypervariances
VarInt=5
VarSlope=1
######SIMULATION TRUTH####
##True Accrual Rate
Accrue=2
##True Distribution of subgroups
groupprob=c(.5,.5)
##True Group Toxicity probabilities at each dose level
GroupProb =matrix(c(.05,.3,.6,.7,.8,.01,.02,.13,.27,.5),nrow=2,byrow=TRUE)
##True Simulation distribution
Family="Uniform"
SimTruth = as.list(c(0,0))
SimTruth[[1]]=GroupProb
SimTruth[[2]]=GroupProb
phetero=.9
RESULTS=SimTrial(nSims,Nmax,T1,Target,Dose,DoseStart,
              Upper,Accrue,groupprob,meanmu,meanslope,
              MeanInts,MeanSlopes,VarInt,VarSlope,phetero,
              Family,SimTruth,NSep,NBorrow,cohort,FULL)
              RESULTS[[1]]

---

Simulates a Sub-TITE trial design

Description

Simulates replicates from a Sub-TITE trial with user specified true toxicity time distributions for different doses and subgroups and returns average summary statistics of the trial.

Usage

SimTrial1(
  nSims,
  Nmax,
  T1,
  Target,
  Dose,
  DoseStart,
  Upper,
  Accrue,
  groupprob,
  Family,
  Param1,
  Param2,
  meanmu,
  meanslope,
  MeanInts,
  MeanSlopes,
  varint,
  varbeta,
  phetero,
  NSep,
  NBorrow,
  cohort,
  FULLY
)

Arguments

nSims	Number of Trials to Simulate.
Nmax	Maximum Number of Patients to enroll in the trial.
T1	Reference time for toxicity.
Target	Target cumulative toxicity probability (or subgroup specific vector) at time T1.
Dose	Standardized vector of doses to try.
DoseStart	Dose (or vector of Doses) to enroll the first patient in each subgroup at.
Upper	Cutoff values used to determine if accrual in a subgroup should be suspended.
Accrue	Expected montly patient accrual rate.
groupprob	Probability vector of subgroup assignment.
Family	What distribution Family to simulate from. Options include: Exponential,Gamma, Lognormal, Uniform, Weibull.
Param1	nGroups X nDose matrix of first parameter values.
Param2	NGroups X nDose matrix of second parameter values.
meanmu	Prior mean of the baseline intercept parameter.
meanslope	Prior mean of the baseline slope parameter.
MeanInts	G-1 length vector of subgroup specific prior intercept means.
MeanSlopes	G-1 length vector of subgroup specific prior slope means.
varint	Prior Variance of Intercept Parameters.
varbeta	Prior Variance of Slope Parameters.
phetero	Prior prob of heterogeneity.
NSep	Number of patients to assign based on no borrowing.
NBorrow	Number of patients to assign based on no clustering
cohort	Number of patients to enroll before escalating.
FULLY	Do we have to fully evaluate a cohort before escalating?

Value

A list of simulation outputs to be processed in R.

---