InferenceBEAGSD: Inference for Binary Endpoint Adaptive Group-Sequential Designs

Description Usage Arguments Details Value Author(s) References See Also Examples

adjustMet1 calculates the multiplicative ajustment factor f to be applied to Phase II efficacy estimate, and the factor ρ to be applied to Phase III sample size estimate using Method 1 proposed by Nhacolo and Brannath (in press).

1
2
3

adjustMet1(p2d, p2r, p2e, p2p0 = NULL, p2p1 = NULL, p2a = NULL,
  p2b = NULL, p3p0 = NULL, p3p1 = NULL, p3a = NULL, p3b = NULL,
  nsimul = 5000, seed = NULL)

`p2d`	Dataframe with Phase II design, with similar as in `EKOptAdaptDesigns`.
`p2r`	Dataframe containing results of Phase II trials following the design `p2d`. It is the output of the function `AnalyzeEKOAD`.
`p2e`	Phase II estimate to consider among the estimates used by codeAnalyzeEKOAD. It can be `"pip"` (naive MLE) or one of the four estimates from methods proposed by Nhacolo and Brannath (2018): `"pim1"`, `"pim2"`, `"pim2v2"` or `"pim3"`.
`p2p0`	Phase II response rate under H_0. If `NULL` (default), the value is taken `p2d`.
`p2p1`	Phase II response rate under H_1. If `NULL` (default), the value is taken `p2d`.
`p2a`	Phase II type I error rate. If `NULL` (default), the value is taken `p2d`.
`p2b`	Phase II type II error rate. If `NULL` (default), the value is taken `p2d`.
`p3p0`	Phase III response rate of the control group. If `NULL` (default), the value is set to `p2p0`.
`p3p1`	Phase III response rate of the treatment group. If `NULL` (default), the value is set to `p2p1`.
`p3a`	Phase III type I error rate. If `NULL` (default), the value is set `p2a`.
`p3b`	hase III type II error rate. If `NULL` (default), the value is set `p2b`.
`nsimul`	Number of (parametric) bootstrap samples (default 5000).
`seed`	Seed for random number generator. If `NULL` (default), no seed is set.

The aim of the adjustment is to get an adequately powered Phase III trial based on Phase II data. See the documentation of the function AnIItoIIIRe for more details about the designs.

A list containing two dataframes final and intermed. final contains the final measures for the adjustment factors (f and ρ) and power. intermed holds the intermediate results (of each bootstrap sample).

Arsenio Nhacolo

Nhacolo, A. and Brannath, W. Using Estimates from Adaptive Phase II Oncology Trials to Plan Phase III Trials. In press.

Nhacolo, A. and Brannath, W. Interval and point estimation in adaptive Phase II trials with binary endpoint. Stat Methods Med Res, 2018.

Ahn, C., Heo, M. and Zhang, S. Sample Size Calculations for Clustered and Longitudinal Outcomes in Clinical Research. CRC Press, 2014.

adjustMet1, SimulateEKOAD, AnalyzeEKOAD.

## Not run: 
vdid <- c(6,10) # design ids
vp2est <- c("pip","pim1","pim2","pim2v2","pim3")
nse <- 1000#number of simulations for each phase
cur <- 1; tot <- length(vdid)*length(vp2est)
for (did in vdid){
 for (p2est in vp2est){
   cat('Processing ',cur,' of ',tot,' (',100*round(cur/tot,1),'%)\n',sep = '')
    load(paste0("p2r",did,".rdata")) # output of the function AnalyzeEKOAD
    out <- adjustMet1(p2d = EKOADwn[EKOADwn$id==did,], p2r = rslt[1:nse,], p2e = p2est, nsimul = nse, seed = 3343)
    write.csv(out$final,file = paste0("final",did,p2est,".csv"),row.names = FALSE)
   write.csv(out$intermed,file = paste0("intermed",did,p2est,".csv"),row.names = FALSE)
   cur <- cur+1
 }
}


vdid <- c(6,10)
vp2est <- c("pip","pim1","pim2","pim2v2","pim3")
fa <- data.frame()
for (did in vdid)
{
 for (p2est in vp2est){
    f <- read.csv(paste0("final",did,p2est,".csv"))
    fn <- names(f)
    f$dsgn <- did
    f <- f[,c('dsgn',fn)]
    fa <- rbind(fa,f)
  }
}
write.csv(fa,file = "final_all.csv",row.names = FALSE)

## End(Not run)