Repository files navigation

The goal of tout is to optimise the design ofthree-outcomeclinical trials by determining their sample size and progressioncriteria.

Install the released version of tout from CRAN:

install.packages("tout")

Or you can install the development version of tout fromGitHub with:

# install.packages("devtools")devtools::install_github("DTWilson/tout")

Consider a pilot trial which will measure the proportion of patients, inthe intervention arm, who adhere to the intervention. We want to usethis estimate of the probability of adherence (denoted$\rho$) in astatistical test with three possible outcomes: stop, go, or pause (wherea pause outcome allows us to make the progression decision outwith thetesting framework, based on other information and/or stakeholders).

Let the null and alternative hypotheses be$\rho_0 = 0.5$ and$\rho_1 = 0.7$, and suppose we wish to control error rates at$\alpha \leq 0.05, \beta \leq 0.2$ and$\gamma \leq 0.6$. We can findthe smallest possible design satisfying these constraints using thetout_design() function:

library(tout)#> Warning: package 'tout' was built under R version 4.5.2design<- tout_design(rho_0=0.5,rho_1=0.7,alpha_nom=0.05,beta_nom=0.2,gamma_nom=0.6)design#> Three-outcome design#>#> Sample size: 57#> Decision thresholds: 33 38#>#> alpha = 0.04808966#> beta = 0.186012#> gamma = 0.5433195#>#> Hypotheses: 0.5 (null), 0.7 (alternative)#> Modification effect range: 0 0#> Error probability following an intermediate result: 0.5 0.5

Plotting the returned object gives an illustration of the samplingdistribution of the estimated adherence probability under eachhypothesis, alongside the optimal progression criteria and the decisionsthey will lead to:

plot(design)

The same function can be used when we anticipate making some adjustmentsfollowing a pause outcome (see thetau argument); when assumingparticular probabilities of making correct decisions following a pauseoutcome (see theeta argument); and when the outcome is continuousrather than binary (see thesigma argument). Further details can befound in the vignette and in the associated manuscript.