mult_sampsize

Description

R package for determining sample size required in studies using composite endpoints with mixed continuous and discrete components analysed using the augmented binary method.

Getting started

Install from Github using devtools::install_github("martinamcm/mult_sampsize")

Implementation as a Shiny app with further documentation on functionality and examples available at MultSampSize

Details

Composite Endpoints

Function sampsizecomp() provides the required sample size using both the augmented binary method based on a latent variable model and a standard binary method based on a logistic regression model with the following arguments:

• augmean mean risk difference treatment effect estimated using augmented binary method
• binmean mean risk difference treatment effect estimated using standard binary method
• augvar variance of risk difference treatment effect estimated using augmented binary method
• binvar variance of risk difference treatment effect estimated using standard binary method
• alpha one-sided alpha level
• beta beta level: 1-desired power

Estimates of these quantities can be obtained from existing data using the augbin_rheum package, as shown below.

Example

Assuming the endpoint of interest is a composite endpoint comprised of two continuous and one binary component, the sample size required in each arm when the dichotomisation thresholds are equal to 18 and 6 is obtained as below.

More details and Egdata21 can be obtained from MultSampSize.

devtools::install_github("martinamcm/augbin_rheum")

data_fit <- augbinrheum(Egdata21,2,1,c(18,6))

augmean_est <- data_fit$risk_diff$est[1]
binmean_est <- data_fit$risk_diff$est[2]

augvar_est <- 0.5*dim(Egdata21)[1]*((data_fit$risk_diff$ci_upper[1]-augmean_est)/1.96)^2
binvar_est <- 0.5*dim(Egdata21)[1]*((data_fit$risk_diff$ci_upper[2]-binmean_est)/1.96)^2

sampsizecomp(augmean_est,binmean_est,augvar_est,binvar_est,0.05,0.2)

References

McMenamin M, Barrett JK, Berglind A, Wason JMS. Sample Size Estimation using a Latent Variable Model for Mixed Outcome Co-Primary, Multiple Primary and Composite Endpoints. arXiv. 2019. arXiv:1912.05258.

McMenamin M, Grayling MJ, Berglind A, Wason JMS. Increasing power in the analysis of responder endpoints in rheumatology: a software tutorial. medRxiv. 2020. doi: 10.1101/2020.07.28.20163378

McMenamin M, Barrett JK, Berglind A, Wason JM. Employing a latent variable framework to improve efficiency in composite endpoint analysis. Statistical Methods in Medical Research. 2021;30(3):702-716. doi: 10.1177/0962280220970986