The EMA’s ‘Data set I’ is incomplete [Regulatives / Guidelines]

posted by Helmut Homepage – Vienna, Austria, 2021-09-24 11:36 (29 d 17:35 ago) – Posting: # 22592
Views: 413

Hi Loky do,

» is it mentioned in guidelines?

No. However, inspect the EMA’s ‘Data set I’ (TRTR|RTRT with 77 subjects) published in the Q&A document. Although it was fabricated by David Brown of the MHRA (at that time member of the Biostatistics Working Party) in collaboration with the Pharmacokinetics Working Party, likely it reflects what regulators have seen in the past and obviously is acceptable.
The data set is incomplete with missings not only in the last period. Eight subjects ‘returned’ after missed periods:$$\small{\begin{array}{ccccccccc}
\hline
\text{Subject} & \text{Missed period(s)} & \text{Missed treatment(s)} & \text{`Returned' period} & n_\textrm{R} & n_\textrm{T} & \text{1. BE} & 2.\;CV_\textrm{wR} & 3.\;CV_\textrm{wT}\\
\hline
11 & 3 & \text{T} & 4 & 2 & 1 & + & + & -\\
20 & 3 & \text{T} & 4 & 2 & 1 & + & + & -\\
24 & 2 & \text{R} & 3 & 1 & 2 & + & - & +\\
31 & 3 & \text{R} & 4 & 1 & 2 & + & - & +\\
42 & 3 & \text{T} & 4 & 2 & 1 & + & + & -\\
67 & \text{3, 4} & \text{R, T} & - & 1 & 1 & + & - & -\\
69 & 3 & \text{T} & 4 & 2 & 1 & + & + & -\\
71 & \text{3, 4} & \text{T, R} & - & 1 & 1 & + & - & -\\\hline
& \text{2: 1, 3: 7, 4: 2} & \text{R: 4, T: 6} & &\text{2: 4, 1: 4} & \text{2: 2, 1: 6} & & & \\
\hline
\end{array}}$$I agree with Dshah’s post.
  1. For ABE – or ABEL if applicable – all subjects with at least one T and R treatment.
  2. For the estimation of \(\small{CV_\textrm{wR}}\) (required to decide whether ABEL can be applied and, if yes, calculation of the expanded limits) subjects 24, 31, 67, and 71 excluded.
  3. Since this is a fully replicated design: For the estimation of \(\small{CV_\textrm{wT}}\) (not required by agencies but useful information) subjects 11, 20, 42, 67, 69, and 71 excluded.
In the Phoenix-template #1 & #2 and in the [image]-package replicateBE #1 – #3 is performed automatically.
Example using the package’s first internal reference data set evaluated by ‘Method A’ (all effects fixed):

library(replicateBE)
method.A(data = rds01)


Relevant part of the file DS01_ABEL_MethodA.txt written to tempdir():
Sequences (design) : TRTR|RTRT (4-period full replicate)
Subjects / sequence: 39|38   (unbalanced)
Missings / sequence: 7|3     (incomplete)
Missings / period  : 0|1|7|2 (incomplete)
Subjects (total)   :  77
Subj’s with T and R:  77 (calculation of the CI)
Subj’s with two Ts :  71
Subj’s with two Rs :  73
Degrees of freedom : 217
CVwT               :  35.16%
swT                :   0.34138
CVwR               :  46.96% (reference-scaling applicable)
swR                :   0.44645
Expanded limits    :  71.23% ... 140.40% [100exp(±0.760·swR)]
Confidence interval: 107.11% ... 124.89%  pass
Point estimate     : 115.66%              pass
Mixed (CI & PE)    :                      pass


Without writing to a file:

library(replicateBE)
x           <- method.A(data = rds01, print = FALSE, details = TRUE)
# round full precision results
x[c(11:14)] <- signif(x[c(11:14)], 4)
x[c(19:21)] <- round(x[c(19:21)], 2)
print(x[c(1, 6:8, 3:5, 11, 13, 12, 14, 19:21)], row.names = FALSE) # relevant stuff reordered


   Design Sub/seq Miss/seq Miss/per  n nTT nRR CVwT(%)    swT CVwR(%)    swR CL.lo(%) CL.hi(%)  PE(%)
TRTR|RTRT   39|38      7|3  0|1|7|2 77  71  73   35.16 0.3414   46.96 0.4464   107.11   124.89 115.66

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

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

Complete thread:

Activity
 Admin contact
21,753 posts in 4,548 threads, 1,544 registered users;
online 7 (0 registered, 7 guests [including 4 identified bots]).
Forum time: Sunday 05:12 CEST (Europe/Vienna)

They were “so intent of making everything numerical”
that they frequently missed seeing
what was there to be seen.    Barbara McClintock

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