Achievwin
★

US,
2021-05-14 21:20
(1101 d 15:02 ago)

Posting: # 22354
Views: 2,364

## Two-stage replicate design studies for RSABE [Design Issues]

Does anyone have knowledge or experience in conducting two stage adaptive study for a four period full replicate design study for RSABE candidate?
1. Has anyone conducted this kind of study (due to Covid or other reasons)
2. What kind of statistical analysis is best for two stage design. Stage 1 or Stage 1+2 data presented separately what penalty we need to incorporate in Stage 1+2 pooled data for evaluating RSABE?
Thanks for all suggestions

Helmut
★★★

Vienna, Austria,
2021-05-14 22:14
(1101 d 14:07 ago)

@ Achievwin
Posting: # 22355
Views: 2,064

## Two-stage design ❌ RSABE

Hi Achievwin,

sorry. See there.

If you are – very – courageous, you can opt for Bonferroni (α = 0.025). All methods I know (for 2×2×2 crossover and parallel designs) require less adjustments, i.e., a higher α. No idea whether that’s acceptable. Ask the OGD.

library(PowerTOST) theta0 <- 0.90 target <- 0.80 design <- "2x2x4" CV     <- seq(0.3, 0.8, 0.025) res    <- data.frame(CV = sprintf("%.1f%%", 100*CV)) for (j in 1:nrow(res)) {   # sample size for α = 0.05   res$n1[j] <- sampleN.RSABE(alpha = 0.05, CV = CV[j], theta0 = theta0, targetpower = target, design = design, details = FALSE, print = FALSE)[["Sample size"]] # sample size for α = 0.025 res$n2[j] <- sampleN.RSABE(alpha = 0.025, CV = CV[j],                              theta0 = theta0, targetpower = target,                              design = design, details = FALSE,                              print = FALSE)[["Sample size"]] } res$penalty <- sprintf("%+.1f%%", 100*(res$n2-res$n1)/res$n1) names(res)[2:3] <- c("n (alpha=0.05)", "n (alpha=0.025)") print(res, row.names = FALSE)     CV n (alpha=0.05) n (alpha=0.025) penalty  30.0%             32              40  +25.0%  32.5%             30              38  +26.7%  35.0%             28              34  +21.4%  37.5%             26              32  +23.1%  40.0%             24              30  +25.0%  42.5%             24              30  +25.0%  45.0%             24              28  +16.7%  47.5%             22              28  +27.3%  50.0%             22              28  +27.3%  52.5%             22              26  +18.2%  55.0%             22              26  +18.2%  57.5%             22              26  +18.2%  60.0%             24              26   +8.3%  62.5%             24              26   +8.3%  65.0%             24              28  +16.7%  67.5%             24              28  +16.7%  70.0%             26              28   +7.7%  72.5%             26              28   +7.7%  75.0%             26              28   +7.7%  77.5%             28              30   +7.1%  80.0%             28              30   +7.1%

Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

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

US,
2021-05-19 21:08
(1096 d 15:14 ago)

@ Helmut
Posting: # 22356
Views: 1,805

## Two-stage design ❌ RSABE

Thanks Helmut:

My questions is two-fold -

1) is there a precedence of conducting two stage study for 4-period full replicate design? (four periods in stage 1 and stage 2)

and

2) what kind of penalty we can factor in?

Regards,
Helmut
★★★

Vienna, Austria,
2021-05-19 22:40
(1096 d 13:42 ago)

@ Achievwin
Posting: # 22357
Views: 1,759

## Two-stage design & RSABE: Forget it!

Hi Achievwin,

❝ 1) is there a precedence of conducting two stage study for 4-period full replicate design? (four periods in stage 1 and stage 2)

I have seen one since 2010. Not for RSABE but for the EMA’s ABEL. Asymmetric alphas – don’t remember which approach: Haybittle-Peto (0.001/0.049) or O’Brien/Fleming (0.005/0.048). Even in a 2×2×2 crossover TSD there is an inflated Type I Error. This study ended in a – vested – deficiency letter of the MHRA.

“The applicant has to demonstrate that the patient’s risk is controlled.”

Oops!

❝ 2) what kind of penalty we can factor in?

That’s not the point and should be the least of your worries. Unless you have a magic wand providing you with a suitable adjusted α, no way.

Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

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