Exclusion of subjects [Design Issues]

posted by Helmut Homepage – Vienna, Austria, 2015-10-14 16:44 (3409 d 15:11 ago) – Posting: # 15561
Views: 6,907

[image]Hi Compliance,

❝ Does FDA raise any concern …


I don’t have a crystal ball. Substitute “does the FDA” by “will the FDA likely”.

❝ … if we get around 20 to 25% positive pre-dose concentration in subsequent period of the study?


I don’t think so. See the current NDA-draft (Section Subjects with predose plasma drug concentrations):

If the predose concentration is ≤ 5 per­cent of Cmax value in a subject with predose plasma con­cen­tra­tion, you can include the subject’s data without any adjustments in all pharmacokinetic measure­ments and calculations. We recommend that if the predose value is greater than 5 per­cent of Cmax, you drop the subject from all BE study evaluations.

If you exclude subjects you decrease the chance to demonstrate BE (= increase the producer’s risk). The patient’s risk (Type I Error) is of regulatory concern – not the Type  II Error (β = 1 – π). If you are able to show BE despite the lower sample size, fine.
The loss in power is not substantial in most cases:

library(PowerTOST)
CV     <- seq(0.2, 0.6, 0.1)
GMR    <- 0.95
pi     <- 0.85
design <- "2x2x2"
pct.do <- c(0, 20, 25)
min    <- 12
n      <- vector("integer", length(CV))
for (j in seq_along(CV)) {
  n[j] <- sampleN.TOST(CV=CV[j], theta0=GMR, targetpower=pi,
            design=design, print=FALSE)[["Sample size"]]
  if (n[j] < 12) n[j] <- 12
}
res    <- matrix(nrow=length(CV)*length(pct.do), ncol=6, byrow=TRUE,
            dimnames=list(NULL,
              c("pct.CV", "pct.do", "n", "power",
                "loss.abs", "loss.rel.pct")))
r <- 0
for (j in seq_along(CV)) {
  for (k in seq_along(pct.do)) {
    r         <- r+1
    res[r, 1] <- 100*CV[j]
    res[r, 2] <- pct.do[k]
    res[r, 3] <- trunc(n[j]*(100-pct.do[k])/100)
    res[r, 4] <- power.TOST(CV=CV[j], theta0=GMR,
                   n=res[r, 3], design=design)
    if (res[r, 3] == n[j]) {
      pwr.max <- res[r, 4]
    } else {
      res[r, 5] <- round(pwr.max - res[r, 4], 5)
      res[r, 6] <- round(100*(1-res[r, 4]/pwr.max), 3)
    }
  }
}
res    <- data.frame(res)
print(res, row.names=FALSE)

 pct.CV pct.do   n   power loss.abs loss.rel.pct
     20      0  22 0.86889       NA           NA
     20     20  17 0.76365  0.10524       12.112
     20     25  16 0.73541  0.13347       15.361
     30      0  44 0.85076       NA           NA
     30     20  35 0.75955  0.09121       10.721
     30     25  33 0.73220  0.11855       13.935
     40      0  76 0.85644       NA           NA
     40     20  60 0.76574  0.09070       10.590
     40     25  57 0.74270  0.11374       13.281
     50      0 112 0.85187       NA           NA
     50     20  89 0.76333  0.08854       10.394
     50     25  84 0.73758  0.11429       13.416
     60      0 154 0.85229       NA           NA
     60     20 123 0.76649  0.08580       10.067
     60     25 115 0.73669  0.11561       13.564


❝ […] wash out duration fixed based on the data available on FDA site under SBOA for reference product.


In the worst case the FDA might ask why you didn’t perform a pilot study.

If the applicant chooses, a pilot study in a small number of subjects can be carried out before pro­ceed­ing with a full BE study. This pilot study can be used to validate analytical methodology, assess vari­abi­lity, optimize sample collection time intervals, and provide other information.

But that’s not mandatory. A depot injection, well… IMHO, the reference product’s data are sufficient to justify the chosen design.

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

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

Complete thread:

UA Flag
Activity
 Admin contact
23,380 posts in 4,914 threads, 1,661 registered users;
40 visitors (0 registered, 40 guests [including 5 identified bots]).
Forum time: 06:55 CET (Europe/Vienna)

Everything is trivial, if you know the answer.    Thomas Jaki

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5