blindrecalc
Blinded Sample Size Recalculation
Computation of key characteristics and plots for blinded sample size recalculation.
Continuous as well as binary endpoints are supported in superiority and non-inferiority trials.
See Baumann, Pilz, Kieser (2022)
blindrecalc
blindrecalc facilitates the planning of a clinical trial with an internal pilot study and blinded sample size recalculation.
Installation
Install the current CRAN version of blindrecalc with:
install.packages("blindrecalc")
Or install the development version from GitHub with:
# install.packages("devtools")
devtools::install_github("imbi-heidelberg/blindrecalc")
Usage
blindrecalc currently supports continuous and binary endpoints for
superiority and non-inferiority test problems. Continuous endpoints are
analyzed using Student’s t-test, binary endpoints are analyzed using the
Chi-squared test for superiority trials and the Farrington-Manning test
for non-inferiority trials. Each design can be defined using a
setup-function: setupStudent, setupChiSquare and
setupFarringtonManning. For example, to setup a superiority trial with
a continuous endpoint:
library(blindrecalc)
design <- setupStudent(alpha = 0.025, beta = 0.2, r = 1, delta = 5)
alpha and beta refer to the type 1 and type 2 error rate, r is the
sample size allocation ratio and deltais the effect size between the
null and the alternative hypothesis. For a non-inferiority trial with a
shifted t-test, additionally the argument delta_NI must be specified.
To calculate the sample size for a fixed design, use n_fix:
n_fix(design, nuisance = c(5, 10, 15))
#> [1] 31.39552 125.58208 282.55967
nuisance refers to the nuisance parameter of the design, which in the
case of the t-test is the common variance of the outcome variable.
To calculate the type 1 error rate of the design using blinded sample
size recalculation, use toer:
toer(design, n1 = c(30, 60, 90), nuisance = 10, recalculation = TRUE)
#> [1] 0.0265 0.0265 0.0232
n1 refers to the sample size of the internal pilot study
recalculation = TRUE specifices that the type 1 error rate for a
design with blinded sample size recalculation should be computed.
To compute the power of the design, use pow:
pow(design, n1 = c(30, 60, 90), nuisance = 10, recalculation = TRUE)
#> [1] 0.7923 0.8015 0.7946
To calculate the distribution of the total sample sizes use n_dist:
n_dist(design, n1 = c(30, 60, 90), nuisance = 10)
#> n_1 = 30 n_1 = 60 n_1 = 90
#> Min. : 41 Min. : 63.0 Min. : 90
#> 1st Qu.:109 1st Qu.:117.0 1st Qu.:120
#> Median :131 Median :133.0 Median :133
#> Mean :134 Mean :134.1 Mean :134
#> 3rd Qu.:156 3rd Qu.:149.0 3rd Qu.:147
#> Max. :293 Max. :244.0 Max. :230
Reference
A paper describing blindrecalc can be found here.