Helmut
★★★
avatar
Homepage
Vienna, Austria,
2014-07-31 20:06
(3528 d 05:06 ago)

Posting: # 13327
Views: 6,487
 

 EMA: GL on quality of oral MR products [BE/BA News]

Dear all,

on 31 July EMA’s CHMP published the final “Guideline on quality of oral modified release products” (dated 20 March 2014). The GL will come into effect six months after publication.
Funny enough I received the notification today through EMA’s RSS-feed (lag-time?). Overview of comments here. Only 89 pages.

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
luvblooms
★★  

India,
2014-08-06 14:03
(3522 d 11:09 ago)

@ Helmut
Posting: # 13344
Views: 5,335
 

 EMA: GL on quality of oral MR products

Dear Helmut


Now I am going through the guidance and one point got my attention

on page 9/16
Although some methods of IVIVC analysis quantify biological variability (and allow prediction of confidence intervals), most methods predict mean concentration-time data only.

If I am right, WinNonlin too works on the mean data only, so what are the other methods (highlighted in blue and bold) discussed over there?

Can anyone please explain?

Looks like its time to upgrade my knowledge core!!!


P.S: Have you checked the comments section
For comment 10: Outcone is See coments 11 and 12
For comment 11: Outcone is See coments 10 and 12
For comment 12: Outcone is See coments 10 and 11 :-D


More to come

~A happy Soul~
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2014-08-06 16:13
(3522 d 08:59 ago)

@ luvblooms
Posting: # 13346
Views: 5,293
 

 EMA: GL on quality of oral MR products

Hi Luv,

[…] most methods predict mean concentration-time data only.


❝ If I am right, WinNonlin too works on the mean data only,


Correct.

❝ so what are the other methods […] discussed over there?


Maybe this one (and/or the references within):

Cardot J-M, Davit BM.
In vitro–In Vivo Correlations: Tricks and Traps.
AAPS J. 2012;14(3):491–9
doi 10.1208/s12248-012-9359-0

Although not dealing with individual in vivo data per se but with CVintra observed in biostudies this one is interesting as well:

Rodier B, Davit BM, Beyssac E, Cardot J-M.[/b]
In Vitro- In Vivo Correlation’s Dissolution Limits Setting.
Pharm Res (pre-print 28 March 2014)
doi 10.1007/s11095-014-1349-8


In both papers at least part of the calculations were done in Phoenix/WinNonlin. If I recall it correctly they presented an extract of the first one at an AAPS meeting. In reviewing the submission the AAPS told Jean-Michel that geometric mean curves are “too complicated”. :-D
Jean-Michel is a member of the forum; I will send him an e-mail – maybe he has some spare time giving an overview.

❝ P.S: Have you checked the comments section


10 ⇒ 11 & 12
11 ⇒ 10 & 12
12 ⇒ 10 & 11
Nitpicking, but precise. ;-)

❝ More to come


Go ahead. I always find the ”Overview of comments” very useful in understanding why some­thing ended up in the final version of any guideline.

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
JMCardot
☆    

France,
2014-08-10 18:02
(3518 d 07:10 ago)

@ Helmut
Posting: # 13356
Views: 5,261
 

 EMA: GL on quality of oral MR products

Hello,

First we are speaking here about SR/PR formulations in fasting state.

Working on individual data can be tried in population approach but that is rather complicated and not easy at all to validate (in my point of view). On classical 2 steps approach a lot of question arises to know if you mean before or after deconvolution etc… that is a little casuistic question!

I think that the main question is: if all the subjects behave similarly in rate of absorption and shape of curves (even is not at the same extent) or not? The initial test to assess this point is the Tmax. If all the Tmax are close together it is a good starting point. After the comparison of the mean vs individual results is of interest (is the mean Cmax close or far from the Cmax of the mean curve, etc…) and so on one test after the other, for example are all the absorption curves in %FD close together or not? If all the data goes in the same direction of a mean curve reflecting a mean subject then no major problem. Similarly you never use the individual dissolution data but the mean of them as you assume that variability is low (low means max 20% first time then <10% other times) and all tablets dissolutions are similar! In the same way you do not know which subject was dosed with which tablet => which dissolution profile. That is quite funny when you see dissolution data which had passed only with step 2!


Now imagine that all the subjects behaves differently with the same tablets: not the same shape at all, with Tmax ranging from 2 to 12h in fasting state. A lot of questions, among them only two: is that linked with absence of robustness of your formulation and incorrect design of it or with purely physiological factors. In case of physiology IVIVC cannot help to predict as it is subject dependent and (except if you know which marker is responsible of that) you will be unable to predict a new subject Lambda. If it is linked with your formulation that means that you have not found the key factors responsible of the quality of it => you are back to formulation step and to the dissolution method setting up (variability was not anticipated).


Do not forget that at the end the predictability is done in comparing one AUC and one Cmax simulated vs observed (use similar data source mean curve for example and not mean of observed individual vs simulated mean curve) and not the full comparison time by time of the curves. That is logic as you want to assess the in-equivalence risks.

As your data come from a pilot or pivotal bioavailability study you have the intra (and inter subject variability). So based on the results of the IVIVC you can try to estimate the variability of your simulated results … reconstructing for example a 90%CI base on the simulated results (I apologize for the purist statisticians) and see the risk to be not bioequivalent with your new formulation that being the ultimate goal of IVIVC. Do not forget that somehow differences between subject will always exist. What you want to predict is: how a subject Lambda treated with your form A is going to react if you switch from formulation A to B (safety and efficacy).


Helmut I apologize as usual I am long in my answers!


Best regards,

JM


Edit: No worries – I’m notoriously verbose as well. You post has 3,267 characters and the forum’s limit is 15,000. Still some headroom. [Helmut]
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
73 visitors (0 registered, 73 guests [including 7 identified bots]).
Forum time: 00:13 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

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