In eniclas/FAIRsimulator: Simulation Platform for the FAIR Trial

First example: A three cohort, three cycle study

library(FAIRsimulator)
set.seed(3243)

Create the study

The first step is to create an initial study object.

StudyObjIni <- createStudy(
  recruitmentAges = list(c(0,1)*30,c(6,7)*30,c(12,13)*30),
  nSubjects = c(300,300,300),
  Recruitmentfunction=function(...) {return(5000)},
  samplingDesign = list(0:6*30, c(0,3,6)*30, c(0,3,6)*30),
  studyStopTime = 18*30+3,
  latestTimeForNewBirthCohorts=0*30,
  treatments =list(
    c("SoC-1","TRT-1","TRT-2","TRT-3","TRT-4"),
    c("SoC-2","TRT-5","TRT-6","TRT-7","TRT-8"),
    c("SoC-3","TRT-9","TRT-10","TRT-11","TRT-12")),
  effSizes = list(c(0,0.05,0.1,0.15,0.25),
                  c(0,0.05,0.1,0.15,0.25),
                  c(0,0.05,0.1,0.15,0.25)),
  randomizationProbabilities = list(rep(0.20,5),rep(0.20,5),rep(0.20,5)),
  strCovariates = c("BIRTHWT","MAGE", "MHTCM", "SEXN", "SANITATN"),
  minAllocationProbabilities = list(c(0.2,rep(0,4)),c(0.2,rep(0,4)),c(0.2,rep(0,4)))
  )

• Three groups, 0-6 months, 6-12 months and 12-18 months.
• Five possible treatments, including standard of care (SoC), in each age group.
• Initial randomization will assign 60 subjects to each treatment in an equiprobable fashion. In total 300 subjects per age group.
• Subjects between 0-1 month, 6-7 months and 12-13 months can be recruited to each of the three age groups, respectively.
• Recruitment is assumed instantaneous.
• Treatment effects are 0, 0.05, 0.1, 0.15 and 0.25 HAZ improvement over 6 months (the first treatment is assumed to be Soc).
• The youngest age group will have monthly HAZ observations while the two older groups will have bi-monthly HAZ observations.
• Once all observation in a cohort have been made, the HAZ data from that cohort is analysed using an LME model with treatment group, birth weight, maternal age, maternal height, sex and sanitation facilities as covariates.
• The estimated treatment effects are used to update the randomization probabilities for new subjects being randomized to the treatments. These probabilities are subject to constraints:
  • Randomization probabilities < 0.1 are explicitly set to 0 (for futility)
  • SoC has a minimum allocation probability and the probabilities are modified to accommodate this.

The above design choices can be changed by the user.

Run the study

The study is simulated with the AdaptiveStudy function.

StudyObj<-AdaptiveStudy(StudyObjIni)

Visualise the results

The design of the trial

The plotStudyCohorts function provides a high level picture of the study.

plotStudyCohorts(StudyObj,plotAnaTimes = T)

The number of subjects remaining

plotActiveSubjects(StudyObj)

The simulated HAZ data

The simulated HAZ values can be inspected using the plotHAZ function.

plotHAZ(StudyObj)

The randomization probabilities

The randomization probabilities before any modifications can be visualized with the plotProbs function while specifying strProb="UnWeightedUpdateProbabilities.

plotProbs(StudyObj,strProb="UnWeightedUpdateProbabilities")

The probabilities after modifications can also be plotted. The SoC probabilities are reset to 0.2 and TRT-9 and TRT-10 are deemed futile (set to 0). TRT-8 falls below a probability of 0.1. This is because the futility evaluation happens before the probabilities are adjusted to accommodate minimum allocation levels.

plotProbs(StudyObj,strProb="UpdateProbabilities")

Second example:

Create the study

This study is the same as the one in the first example except that new cohorts are included in the youngest age group. The randomization probabilities for the new cohorts are based on the analys of the cohort preceeding it.

set.seed(32423)

probTemperation <- function(probs) {
  probs <- sqrt(probs)/sum(sqrt(probs))
  return(probs)
}

StudyObjIni <- createStudy(
  recruitmentAges = list(c(0,1)*30,c(6,7)*30,c(12,13)*30),
  nSubjects = c(300,300,300),
  Recruitmentfunction=function(...) {return(5000)},
  samplingDesign = list(0:6*30, c(0,3,6)*30, c(0,3,6)*30),
  studyStopTime = 32*30,
  latestTimeForNewBirthCohorts=18*30,
  treatments =list(
    c("SoC-1","TRT-1","TRT-2","TRT-3","TRT-4"),
    c("SoC-2","TRT-5","TRT-6","TRT-7","TRT-8"),
    c("SoC-3","TRT-9","TRT-10","TRT-11","TRT-12")),
  effSizes = list(c(0,0.05,0.1,0.15,0.25),
                  c(0,0.05,0.1,0.15,0.25),
                  c(0,0.05,0.1,0.15,0.25)),
  randomizationProbabilities = list(rep(0.20,5),rep(0.20,5),rep(0.20,5)),
  strCovariates = c("BIRTHWT","MAGE", "MHTCM", "SEXN", "SANITATN"),
  minAllocationProbabilities = list(c(0.2,rep(0,4)),c(0.2,rep(0,4)),c(0.2,rep(0,4)))
  )

Run the study and look at the design

StudyObj <- AdaptiveStudy(StudyObjIni)

plotStudyCohorts(StudyObj)

Plot the probabilities

The unadjusted probabilities:

plotProbs(StudyObj,strProb="UnWeightedUpdateProbabilities")

The adjusted probabilities:

plotProbs(StudyObj)
#plotProbs(StudyObj,strProb="UnWeightedRandomizationProbabilities")

eniclas/FAIRsimulator documentation built on May 16, 2019, 5:12 a.m.