In martinlaw/curtailment: Finds Binary Outcome Designs Using Stochastic Curtailment

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  out.width = "100%"
)

curtailment

The goal of curtailment is to allow users to create single- or two-arm binary outcome clinical trial designs that use a form of early stopping known as stochastic curtailment. Using stochastic curtailment can result in trial designs with a lower sample size on average, compared to typical trial designs or even multi-stage designs. The package contains functions that will search for designs that are suitable for your trial, and draw a diagram showing all points in the trial that a decision may be made to stop early. The package can also find two-stage designs that stop for futility only (Simon designs) or for benefit or futility (Mander and Thompson designs). For all designs, a maximum conditional power for futility can be specified, above which a trial would not be permitted to stop early for futility.

Installation

You can install the released version of curtailment from CRAN with:

install.packages("curtailment")

And the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("martinlaw/curtailment")

Example

This is a basic example which shows you how to solve a common problem:

library(curtailment)
## Find a single-arm design with a maximum sample size in the range 10-25, allowing early stopping after every 5 results, with a type-I error-rate of 0.05 when the response rate equals 0.1 and a power of 0.8 when the response rate equals 0.4:
output <- singlearmDesign(nmin = 10,
                          nmax = 25,
                          C = 5,
                          p0 = 0.1,
                          p1 = 0.4,
                          power = 0.8,
                          alpha = 0.05)
# Obtain the stopping boundaries and a diagram of the above trial:
fig <- drawDiagram(output)

Future plans

Speed up Mander and Thompson and Simon design searches by reducing the number of combinations of n1/n2/r1/r/e1 search over, using Wald's sequential probability ratio test as in the multi-stage/continuous functions: