Repository files navigation

Optimal and near-optimal group-sequential designs for clinical trialswith continuous outcomes

OptGS is anR package that provides asuite of functions to assist with the design, analysis, andvisualization of randomized two-arm group-sequential clinical trialswith continuous outcome variables.

Specifically, support is provided to perform a sample size calculationfor popular applicable (non-optimal) designs. The unique focus, however,is on determining optimal and near-optimal designs, using the methodsfromWasonet al (2012) andWason(2015) respectively.

Additional functions then allow point estimates to be computed and pointestimators to be evaluated. Plotting functions also permit theinformative depiction of several important quantities.

You can install the released version ofOptGS fromCRAN with:

install.packages("OptGS")

Alternatively, the current development version fromGithub can be installed with:

devtools::install_github("mjg211/OptGS")

This is a basic example, which demonstrates how to determine anoptimized power-family design (a near-optimal design), plot its stoppingboundaries, determine its operating characteristics, and subsequentlyproduce a plot of the expected sample size curve.

First, determine the design (for the default parameters) with:

des<- des_nearopt()

We can then plot the stopping boundaries of this design with:

plot(des)

The operating characteristics of the design can also be determined with:

opchar<- opchar(des,tau= seq(-des$delta,2*des$delta,length.out=100))

Finally, we can then plot the expected sample size and power curves forthis design using:

plot(opchar)

Between v.1.1.1 (the latest released version onCRAN) andv.2.0.0 (which the current development version onGithub has built upon), several majorchanges were made toOptGS:

• Dependence on C++ code was replaced with equivalentR functionality for stability and easeof further development.
• Support for additional plots were added (e.g., median sample sizecurves).
• Functions to determine operating characteristics (opchar()),perform inference on trial conclusion (est()), simulategroup-sequential trials (sim()), and build bespoke designs(build()) were added.
• Arguments in, and names of, previously present functions have beenmodified (e.g.,optgs() is replaced bydes_nearopt()).

Consequently, if all that you require is the functionality presented inWason (2015), it will likely bequicker to use v.1.1.1 fromCRAN, which isa substantially simpler and also faster (in terms of execution time)package.

However, as time progresses, the additional support provided by v.2.0.0+should make them preferable with some small time investment tounderstand the purpose of the different functions.

An extensive guide to usingOptGS will soon be provided in the formof a package vignette. For v.1.1.1 and earlier,Wason(2015) also provides a detailedintroduction to the package.

If you cannot find the answer to a problem, or a function is returningan unexpected error for your inputs, please contact James Wason(james.wason@newcastle.ac.uk) or Michael Grayling(michael.grayling@newcastle.ac.uk) for assistance.

Wason JMS (2015) OptGS: An R package for finding near-optimalgroup-sequential designs.Journal of Statistical Software 66(2):1–13.DOI:%5B10.18637/jss.v066.i02](https://doi.org/10.18637/jss.v066.i02).

Wason JMS, Mander AP, Thompson SG (2012) Optimal multistage designs forrandomised clinical trials with continuous outcomes.Statistics inMedicine 31(4):301–312.DOI:%5B10.1002/sim.4421](https://doi.org/10.1002/sim.4421).