## Intrauterine Device – Levo­nor­gestrel 52 mg [Design Issues]

Hi Marianna,

» We used data from the article: "Comparative pharmacokinetic analysis of levonorgestrel-releasing intrauterine systems and » levonorgestrel-containing contraceptives with oral or subdermal administration route" Hofmann et. al, 2020.
» But the point still the number of subjects in this BE/BA study, because in this case, parallel design needs an intra+inter CV%.

Actually you need CVtotal. That’s not the sum (or average) of both. You need to average the variances and back-convert them to a CV. In :

pooling <- function(CV.inter, CV.intra) {   CV2var <- function(CV)  return(log(CV^2 + 1))   var2CV <- function(var) return(sqrt(exp(var) - 1))   return(var2CV(mean(c(CV2var(CV.inter), CV2var(CV.intra))))) } CV.inter <- 0.80 CV.intra <- 0.40 CV.total <- pooling(CV.inter, CV.intra) cat("CV.inter =", sprintf("%.2f%%", 100 * CV.inter),     "\nCV.inter =", sprintf("%.2f%%", 100 * CV.intra),     "\nCV.total =", sprintf("%.2f%%", 100 * CV.total), "\n")

Gives

CV.inter = 80.00% CV.inter = 40.00% CV.total = 61.59%

» But in a scientific literature review, we don't find something like that.

I don’t have this article. If the study was performed in a parallel design, you have already CVtotal. If it was not reported, you can calculate it from the confidence interval and the sample size by the function CVfromCI() in PowerTOST. If it was a crossover (very, very unlikely), use the code above.

» A suggestion about mitigating the error in subject number determination?

Dif-tor heh smusma 🖖
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes