Ratio calculation of fully replicate study [RSABE / ABEL]

posted by Helmut Homepage – Vienna, Austria, 2020-07-09 14:19 (1376 d 00:19 ago) – Posting: # 21662
Views: 4,028

Hi Mithu,

❝ ❝ Without knowing your SAS code it would be reading tea leaves.


❝ We have used Progesterone program.


… which nowhere gives the subject’s T/R-ratios.

❝ ❝ $$x/y=\exp (\log_{e}x-\log_{e}y)\;\left\{x,y \in \mathbb{R}^{+} \right\}\; {\color{red}{\neq}}\log_{e}x/\log_{e}y$$

❝ Still I could not understand.


Calculating the ratio of logs (2. of your OP) is nonsense.
The individual averaged log-differences \(I_{ij}\) (for subject \(i\) in sequence \(j\)) are computed in the FDA’s SAS macro by

ilat=0.5*(lat1t+lat2t-lat1r-lat2r);

Simply back-transform them.

Let’s have a look at the EMA’s full replicate reference data set:

SUBJECT   DATA  FORMULATION  PERIOD  SEQUENCE  logDATA 
   1    2285.96      R          1      BABA    7.734541
   1    1955.82      T          2      BABA    7.578565
   1    1345.94      R          3      BABA    7.204848
   1    2856.24      T          4      BABA    7.957261

Then$$\overline{I_{11}}=0.5\times(7.578565+7.957261-7.734541-7.204848)=0.2982185$$and$$\overline{T_{11}/R_{11}}=\exp \overline{I_{11}}=\exp 0.2982185=1.347456$$Confirmed by Phoenix WinNonlin’s RSABE-template and my pocket calculator. :-D

Which leaves my second question open:

❝ ❝ Why on earth do you want to use Excel?


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
22,983 posts in 4,822 threads, 1,648 registered users;
40 visitors (0 registered, 40 guests [including 6 identified bots]).
Forum time: 14:39 CEST (Europe/Vienna)

Complex, statistically improbable things are by their nature
more difficult to explain than
simple, statistically probable things.    Richard Dawkins

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