Dr_Dan
★★

2010-08-04 10:29
(4188 d 08:50 ago)

Posting: # 5713
Views: 15,913

## inter-batch variability? [Regulatives / Guidelines]

Dear all
I need your help regarding following problem: we performed 2 BA (pilot) and 2 BE (pivotal, one 2 way cross-over and one replicate design) studies with the same test formulation.
The batches we used were:
Study 1 (pilot): test 1 (lab scale), ref. 1, study centre 1
Study 2 (pilot): test 2, ref. 2, study centre 1
Study 3 (pivotal): test 2, ref. 1, study centre 2
Study 4 (pivotal): test 3, ref. 3, study centre 3
Both pivotal studies show bioequivalence, the pilot studies were not powered adequately to demonstrate bioequivalence but the results hint in this direction. The drug is highly variable.
The point estimator (AUC/Cmax) differ between the studies:
Study 1: 111/116
Study 2: 84/95
Study 3: 111/114
Study 4: 98/97
The assessor concludes that these different point estimators hint at inter batch variability of the test product (why not the reference?) and therefore a marketing authorisation can not be granted.
What should be our response, which argumentation could help to safe our product?
Your input is very much appreciated.
Kind regards
Dan

Kind regards and have a nice day
Dr_Dan
Pavidus
●

2010-08-04 11:57
(4188 d 07:21 ago)

@ Dr_Dan
Posting: # 5714
Views: 14,316

## inter-batch variability?

Dear Dr_Dan & all,

Allow me one general remark:
IMHO BE studies are definitely not suitable to assess inter-batch variability of any product.

I am not saying, that this remark will "save" your product .... discussion with 'assessors' is always difficult...

(btw. nice figures for a highly-variable drug)

Kind regards, P.
d_labes
★★★

Berlin, Germany,
2010-08-04 13:58
(4188 d 05:21 ago)

@ Dr_Dan
Posting: # 5715
Views: 14,414

## inter-batch variability?

Dear Dan!

» Study 1: 111/116
» Study 2: 84/95
» Study 3: 111/114
» Study 4: 98/97
» The assessor concludes that these different point estimators hint at inter batch variability of the test product [color:red](why not the reference?)/color:red] and therefore a marketing authorisation can not be granted.
(emphasis by me)

I never heard about a regulatory criterion requesting that the point estimators between BE studies must not varying. And that you must express articulately against the regulatory body.
I would wonder if there is no variability!

As you have told: the pivotal studies have shown bioequivalence. IMHO that was it. All other concerns are negligible (if not safety is concerned).

More over the assessors judgment is logical flawed:
Considering Study 1/Study 3 with the same reference 1 (batch?), neglecting the fact that study 1 is pilot and therefore not suitable for a reliable point estimator, there is no substantial difference in the point estimators although test 1 and test 2 (batches?) are used.
Since the other 2 studies use different test/reference (batches?) one cannot use them to judge. If judgment of inter-batch variability is ever possible from BE studies which I highly question.

But to cite myself: "Regulators ways are inscrutable!"

Eventually you can argue using not the point estimators but the mean/geometric mean values of the PK metrics themselves. Hopefully they show greater variability between the reference batches.
But IMHO this is clearly nonsense.

Regards,

Detlew
ElMaestro
★★★

Denmark,
2010-08-04 17:09
(4188 d 02:10 ago)

(edited by ElMaestro on 2010-08-04 17:22)
@ Dr_Dan
Posting: # 5716
Views: 14,450

## inter-batch variability?

Dear Dan,

interesting question; I am not sure I understand it all, but I am sure there's hope!
First of all, there is no law against variation of any kind.
Second, you do not write anything about but did you receive a deficiency question from module 3 regarding the product's release specifications? After all, this is where the inter-batch variability comes into the picture. If the assessor for module 3 did not comment, then there is no issue.
On the positive side, the assessor tried to use common sense rather than cookbook in his/her assessment. On the negative side, he/she got something wrong in the creative mental process. I strongly recommend you to play the game of regulatory chicken if this is an EU submission.

Feel free to PM me re. this.

Pass or fail!
ElMaestro
Helmut
★★★

Vienna, Austria,
2010-08-04 19:45
(4187 d 23:34 ago)

@ ElMaestro
Posting: # 5717
Views: 14,531

## Between study variability common for HVDs

Dear ElMaestro, Dan and all!

» […] did you receive a deficiency question from module 3 regarding the product's release specifications? After all, this is where the inter-batch variability comes into the picture.

Not necessarily. Dan stated in his post that the drug is highly variable. It’s a common property of HVDs that not only the variance is high, but also the location of the T/R-ratio may vary across studies. If would suggest to have a look at papers by the “Two Lászlós” for further explanations. Remember that the restriction of the point estimate of [0.8-1.25] was introduced by the FDA after long discussions for political reasons [sic!]* – and copypasted by the EMA. Such a restriction is questionable and not statistically justified. The pivotal studies demonstrated BE, but I would build an argument based on the natural property of the underlying distribution.

The concept of batch-to-batch in vivo BE is only of historical interest. It was discussed at the BioInternational 1994 in Munich and was dropped. Same participants even asked the heretic question whether a product would be still bioequivalent to itself at the end of the shelf-life (don’t ask me how we would store the reference batch for the final BE study: in liquid N2?).

BTW, I agree with D. Labes’ observations about ref. 1, which was employed in studies 1 & 3.

• To quote Leslie Benet’s presentation (FDA/CDER, Pharmaceutical Science Advisory Committee (ACPS) Meeting, October 06, 2006) on the point estimate restriction:
1. There is no scientific basis or rationale for the point
estimate recommendations
2. There is no belief that addition of the point estimate
criteria will improve the safety of approved generic
drugs
3. The point estimate recommendations are only
“political” to give greater assurance to clinicians and
patients who are not familiar (don’t understand) the
statistics of highly variable drugs

Dif-tor heh smusma 🖖
Helmut Schütz

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

Denmark,
2010-08-04 21:01
(4187 d 22:18 ago)

@ Helmut
Posting: # 5718
Views: 14,290

## Between study variability common for HVDs

Dear HS,

» Not necessarily. Dan stated in his post that the drug is highly variable. It’s a common property of HVDs that not only the variance is high, but also the location of the T/R-ratio may vary across studies.

I know what you mean, but I think we are talking two different phenomena here.
If there truly is b2b variation, then this is a module 3 issue. However, there could be an -if I may call it so- "apparent" b2b variation that reflects stochastic phenomena even though the batches are not varying (or there could be both, shrug). Variation is evil and... well... unpredictable. The burden of giving a proof that relates to the latter phenomenon is almost impossible, however, for the former the task is somewhat easier and more practical. So what Dan can do is to provide something quite concrete in the form of batch data here in stead of entering the discussion around the possibility of random noise accounting for something that could be interpreted as b2b variation.

In another field of equivalence for pharmaceuticals, the closely related topic of batch selection is slowly becoming a hot topic. I think it will become hot for this type of BE as well. Regulators know of the existence of batch selection, and the potential for variation between batches. In the EU guidance we have "Unless otherwise justified, the assayed content of the batch used as test product should not differ more than 5% from that of the batch used as reference product determined with the test procedure proposed for routine quality testing of the test product."

IOW, the choice of batches can affect a study's outcome. In that situation, a lot of companies would be interested in knowing in which way they can use this as an argument for saving a failed BE study: "We selected the wrong batch so our study failed. We will conduct a new one with another batch" And who can blame them? The EU guideline specifies that one cannot just neglect the presence of a failed study. I think regulators have actively avoided discussing this issue in depth, in part because the acknowledgment of batch selection being key to a succesful BE study conflicts with a common way of interpreting success in BE:
Product A is bioequivalent with product B means one batch of T has been shown to be BE to one batch of R. But it does not imply that all batches of T are (or would be in a study) BE by the regulatory standard to all batches of R; nevertheless the latter is a common way for lay people to view it.
Add to that some alpha and beta and you have the perfect legal storm brewing; most EU assessors would rather have plague than enter this discussion before a judge. And who can blame them?

Pass or fail!
ElMaestro
Helmut
★★★

Vienna, Austria,
2010-08-04 23:42
(4187 d 19:37 ago)

@ ElMaestro
Posting: # 5719
Views: 14,286

## Representative batches?

Dear ElMaestro,

nice summary; couldn’t agree more.
Section 4.1.2 of the BE-GL: “representative” appears three times and “justify/justification” five times.

In the past sponsors tried to select T & R batches which matched closely in vitro (content & dissolution). If BE could be demonstrated, everybody was happy (believing that the “selection” worked). If not, most people did rather not question the (lacking) discriminatory power of the in vitro method, but sighed and said “Well, we have to face that it did not work in in vivo – but it’s well known, that without IVIVC, etc. etc.”

I have strong doubts that it is possible to justify that a batch is representative – if we move away from pharmaceutical quality and start thinking about BE.

I changed the category; doesn’t look as a statistical issue to me (any more). [Helmut]

Dif-tor heh smusma 🖖
Helmut Schütz

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

Denmark,
2010-08-05 08:40
(4187 d 10:39 ago)

@ Helmut
Posting: # 5723
Views: 14,442

## Representative batches?

Completely agree, HS.
I also occasionally wondered what representative batch actually means, and was preparing to bring the question up during a discussion session at a conference later this year. I imagine it would be OK to think of a repr. batch as one that has some in vitro properties that are close to the means/medians from the distribution of said properties.
The distributions (let's say of dissolution aspects, content) though are never well known at the time of batch selection and if regulators tried to force selection of 'average batches' (pardon my French) as a means of ensuring repr. batches I would imagine many companies would have to give up.

Real life, as I know it works like this: We have X batches of Ref, and Y batches of Test. Now we measure whatever property we consider most likely to correlate with in vivo results and pick the pair that provide the closest match of Test and Ref for that property without in any way having regard to what's typical. I would be tremendously interested in hearing how you bright people out there handle it in practice.

Pass or fail!
ElMaestro
Helmut
★★★

Vienna, Austria,
2010-08-05 12:30
(4187 d 06:49 ago)

@ ElMaestro
Posting: # 5725
Views: 15,424

## Representative batches?

Dear ElMaestro!

» I also occasionally wondered what representative batch actually means, […]
» […] I imagine it would be OK to think of a repr. batch as one that has some in vitro properties that are close to the means/medians from the distribution of said properties.
» The distributions (let's say of dissolution aspects, content) though are never well known at the time of batch selection and if regulators tried to force selection of 'average batches' (pardon my French) as a means of ensuring repr. batches I would imagine many companies would have to give up.

Reference product

The selection of the reference product used in a bioequivalence study should be based on assay content and dissolution data and is the responsibility of the Applicant. Unless otherwise justified, the assayed content of the batch used as test product should not differ more than 5% from that of the batch used as reference product determined with the test procedure proposed for routine quality testing of the test product. The Applicant should document how a representative batch of the reference product with regards to dissolution and assay content has been selected. It is advisable to investigate more than one single batch of the reference product when selecting reference product batch for the bioequivalence study.

Test product

The test product* used in the study should be representative of the product to be marketed and this should be discussed and justified by the applicant.
1. The test product should usually originate from a batch of at least 1/10 of production scale or 100,000 units, whichever is greater, unless otherwise justified.
2. The production of batches used should provide a high level of assurance that the product and process will be feasible on an industrial scale.
In case of a production batch smaller than 100,000 units, a full production batch will be required.
3. The characterisation and specification of critical quality attributes of the drug product, such as dissolution, should be established from the test batch, i.e. the clinical batch for which bioequivalence has been demonstrated.
* Test product - not batch of the test product!

Note the subtle difference: representative batch of the reference product vs.(test) study product representative of the to be marketed test product. The requirement to select from a least three reference batches in the draft was dropped (almost impossible for non-blockbusters) and replaced by ‘advisable to investigate more than one single batch’.

» Real life, as I know it works like this: We have X batches of Ref, and Y batches of Test. Now we measure whatever property we consider most likely to correlate with in vivo results and pick the pair that provide the closest match of Test and Ref for that property without in any way having regard to what's typical.

Exactly. I would replace ‘consider’ by ‘hope’. The GL calls for content and dissolution.
Content: OK. Dissolution? Well…
Look at c): ‘critical quality attributes’ - side batches tested in vivo beforehand?

Even if we would know ‘critical quality attributes’ from pilot studies, we can only get this data of the test. Representative in a statistical sense – forget it. What about the reference? No way at all, IMHO.

Dif-tor heh smusma 🖖
Helmut Schütz

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

2010-08-05 08:58
(4187 d 10:21 ago)

@ Helmut
Posting: # 5724
Views: 14,199

## Representative batches?

Dear ElMaestro, dear Helmut and all who contribute to this fruitful discussion

» I changed the category; doesn’t look as a statistical issue to me (any more).

You are right Helmut, I made the same experience, at the beginning it looks like a statistical problem but in the course of discussion it turned out that it becomes more and more a regulatory issue. But can you argue with the assessor without statistics?

As other sponsors we selected T & R batches which matched closely. Comparative dissolution profiles showed similarity and the assay content differed less than 5%.
From the quality point of view we see no difference between the batches.

The logic consequence would be that for the release of any batch you produce for the market you have to perform a BE study as quality control.

I am still at a loss.

Kind regards

Dan

Kind regards and have a nice day
Dr_Dan
Helmut
★★★

Vienna, Austria,
2010-08-05 12:45
(4187 d 06:34 ago)

@ Dr_Dan
Posting: # 5726
Views: 14,302

## Representative batches?

Dear Dan!

» But can you argue with the assessor without statistics?

Statistics – A subject which most statisticians find difficult
but in which nearly all physicians are expert.
Stephen Senn

EMEA. The European Medicines Evaluation Agency.
The drug regulatory agency of the European Union.
A statistician-free zone.
Stephen Senn; Statistical Issues in Drug Development. Wiley, p386, 2004

» As other sponsors we selected T & R batches which matched closely. Comparative dissolution profiles showed similarity and the assay content differed less than 5%.
» From the quality point of view we see no difference between the batches.

Yes, everybody did/does it that way (see also the last paragraph in ElMaestro’s post). Of course that's not representative, but 'best'.

» The logic consequence would be that for the release of any batch you produce for the market you have to perform a BE study as quality control.

Yes, but would be applicable to the innovator as well. I would remind the assessor that this idea was already abandoned 15+ years ago and on the consequences.

Dif-tor heh smusma 🖖
Helmut Schütz

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

2010-08-06 09:55
(4186 d 09:24 ago)

@ Helmut
Posting: # 5727
Views: 14,236

## Confidence intervals vs. point estimators

Dear Helmut
I would prefer if you could change the category back to statistics in order to to proceed in discussion, because I do not think that speculations about regulatory affairs aspects would help solving my problem. To get into discussion with the assessor I really need some strong arguments.

Dear all

I am not an expert in statistics but as far as I know the 90% confidence interval for Cmax and AUC tells one that the true test/reference ratio lies within this interval with a probability of 90%. On the one side we will argue that for both primary bioequivalence parameters there is a overlapping of confidence intervals for all four studies. On the other side the assessor will argue that the point estimators of the second pilot study are not included by the confidence intervals of the other studies -> batch variability! Is this reasonable?

As Pavidus explained BE studies are definitely not suitable to assess inter-batch variability of any product. Which arguments can support this idea? To my understanding the only possibility to detect batch variability of the test product would be to perform a three way cross-over study with test 1, test 2 and reference). If you get different point estimators for your test batches (they can not be exactly the same) how far can they differ in order to claim similarity? Beside this nobody raised the question if the results we obtained are due to batch differences of the orginator.

Your input is very much appreciated.
Kind regards
Dan

Category changed back. [Helmut]

Kind regards and have a nice day
Dr_Dan
ElMaestro
★★★

Denmark,
2010-08-06 12:34
(4186 d 06:45 ago)

@ Dr_Dan
Posting: # 5728
Views: 14,250

## Confidence intervals vs. point estimators

Hi Dan,

» I am not an expert in statistics but as far as I know the 90% confidence interval for Cmax and AUC tells one that the true test/reference ratio lies within this interval with a probability of 90%.

If the experiment is repeated an infinite number of times, 90% of the point estimates would be within your bounds, that's what you'd expect, all other factors being equal and assumptions holding. This is interpreted to mean that the true PE with 90% certainty is within the bounds you first found.

» On the one side we will argue that for both primary bioequivalence parameters there is a overlapping of confidence intervals for all four studies. On the other side the assessor will argue that the point estimators of the second pilot study are not included by the confidence intervals of the other studies -> batch variability! Is this reasonable?

Yes, in principle. Do I get you correctly that there is no single point all four CI's have in common?
Here's a proposal, the guideline specifically reads: "The test product should usually originate from a batch of at least 1/10 of production scale or
100,000 units, whichever is greater, unless otherwise justified."
. This is because everybody knows that there is variability between batches and that performance of a pilot batch is not necessarily equal to a full scale batch. So, you will just with kind words remind the assessor that a bunch of his/her colleagues acknowledge that fact that pilot performance is not pivotal performance and cannot be expected to be. Yes, there is variability between batches. No, it is not a problem, you are handling it well by separating pivotal data from pilot data. This is what you can tell the assessor.

» As Pavidus explained BE studies are definitely not suitable to assess inter-batch variability of any product. Which arguments can support this idea?

Well... If you wanted to study everything you might do a mixed model with batch (within treatment some would say) as random factor. In that case one could take variability between batches into consideration. Noone does it, though, and the PK subgroup does not expect you to do this either ("The terms to be used in the ANOVA model are usually sequence, subject within sequence, period and formulation.").
So you can just explain back that batch is never a factor in ANOVAs according to recommendations published by the Efficacy Working Party, and that you see no compelling reasons to deviate from that document.
You might also speculate: The ref. product is probably 500 years old and chances are the production parameters and specifications for that process reflects to some extent reflects habits and traditions from those days, whereas your product accords with today's standards. Therefore, for all practical purposes, generics are often less variable. This is a generalisation and unfortunately impossible to know if it specifically holds for your product.

Honestly, I don't think you have wildly big a problem. I understand why the question is scary, but the task the assessor faces if he/she has to defend his/her views at CMD(h) is worse.

S.

Pass or fail!
ElMaestro
Helmut
★★★

Vienna, Austria,
2010-08-06 13:20
(4186 d 05:59 ago)

@ ElMaestro
Posting: # 5729
Views: 14,221

## Confidence intervals vs. point estimates

Dear ElMastro, Dan & all!

» […] the guideline specifically reads: "The test product should usually originate from a batch of at least 1/10 of production scale or 100,000 units, whichever is greater, unless otherwiseustified."
» This is because everybody knows that there is variability between batches and that performance of a pilot batch is not necessarily equal to a full scale batch. So, you will just with kind words remind the assessor that a bunch of his/her colleagues acknowledge that fact that pilot performance is not pivotal performance and cannot be expected to be. Yes, there is variability between batches. No, it is not a problem, you are handling it well by separating pivotal data from pilot data.

I would say, this argument holds exactly for Study 1 (lab batch). Study 2 (pilot) was performed with a full size batch (I assume), but due to power considerations (sample size!) and the property of HVDs (variable PE across studies) a direct comparison with pivotal studies is futile.

Agree with the mixed model – would be fun setting up (other factors: study and center).
But I would not dive too deep into the murky waters of statistics, but rather go with common sense. According to guidelines (not only BE) nobody expects that all studies in the file were performed with the same batches. Imagine a hypothetical situation: Two studies fasting/fed. The same batch of test, but two batches of the reference. There is a true food effect of -20% and a true ‘batch effect’ of +25%. What would we get? BE fasting and BE fed. As ElMaestro suggested I would remind the assessor in nice words, that following his/her arguments would call for a requirement which is not covered in GL(s).

» S.
  ^

Dif-tor heh smusma 🖖
Helmut Schütz

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

2010-08-06 14:44
(4186 d 04:35 ago)

@ ElMaestro
Posting: # 5731
Views: 14,268

## Confidence intervals vs. point estimators

Dear ElMaestro

» Honestly, I don't think you have wildly big a problem. I understand why the question is scary, but the task the assessor faces if he/she has to defend his/her views at CMD(h) is worse.

Thanks for these encouraging words!

» Do I get you correctly that there is no single point all four CI's have in common?

No, that's not correct:
AUC:  98.57-124.33%, 68.58-103.22%, 100.95-122.94%, 91.56-105.70%
Cmax: 99.69-134.01%, 78.49-116.04%, 102.67-125.98%, 89.06-106.49%

Helmut suggested to cite the assessor in detail:
"The applicant's response does not resolve the objections concerning the bioequivalence issue. The first formulation of the Test product in the study showed bioinequivalence in comparison to the Reference product (Cmax outside of the widened acceptance range of 75-133%) and furthermore the values for the point estimates for the ratio of the geometric means (AUC0-t, AUC0-inf and Cmax) were all over 100%. Afterward the formulation of the Test product was changed and a new BE-study was conducted which again showed a bioinequivalence (AUC0-t and AUC0-inf outside of the acceptable range of 80-125%), with the point estimates all under 100%. The same test batch and the same formulation was investigated in a third BE-study, in which Cmax was out of the normally accepted range of 80-125% and all the point estimates were again over 100%, excluding the value 100% itself. Another batch of the same formulation of the Test product was then investigated in a fourth Replicate-BE-study which this time showed acceptable results for AUC0-t, AUC0-inf and Cmax (all in the normally acceptable range of 80-125%) but again the point estimates were only closely to 100%.
Even if it should be assumed that the first two BE-studies were only pilot studies and hence not suitable for statistical inclusion and taking into consideration that the formulation of the Test product was changed after the first BE-study, it still remains unclear why the same formulation of the Test product in the following three performed BE-studies shows varying values concerning point estimates in these studies (all values were lower, then higher and then again lower than 100% successively). This fluctuation can not be explained through the "intra-subject variability" as the latter influences Cmax and hence justifies the widening of the acceptance range for Cmax but not the fluctuating point estimates which are independent of this variability. This fluctuation is also independent of the number of subjects in the studies.
The Applicant is stating that the differences in point estimates in the four conducted BE-studies might result among others from different test batches. This is exactly the major objection raised by DE that the bioequivalence of the future batches of this formulation is not warranted."

Regards
Dan

Kind regards and have a nice day
Dr_Dan
ElMaestro
★★★

Denmark,
2010-08-06 15:01
(4186 d 04:18 ago)

@ Dr_Dan
Posting: # 5732
Views: 15,032

## Confidence intervals vs. point estimators

Dear Dan,

» AUC:  98.57-124.33%, 68.58-103.22%, 100.95-122.94%, 91.56-105.70%
» Cmax: 99.69-134.01%, 78.49-116.04%, 102.67-125.98%, 89.06-106.49%

Thanks - now much more clear and improves your chances.

» "The applicant's response does not resolve the objections concerning the bioequivalence issue. The first formulation of the Test product in the study showed bioinequivalence in comparison to the Reference product (Cmax outside of the widened acceptance range of 75-133%) and furthermore the ...

No, here the assessor is outright wrong. The CI's for Cmax above do not show bioINequivalence. Bioinequivalence is when the the CI has no point in common with the acceptance range. You can even argue that with increased sample sizes the studies would all be expected to be acceptable.
It is often forgotten that bioequivalence studies have at least three outcome: Equivalent, not conclusive, inequivalent. Studies 1 and 3 are inconclusive but do not indicate inequivalence. Dr. Dan vs Regulator 1-0.

» ...values for the point estimates for the ratio of the geometric means (AUC0-t, AUC0-inf and Cmax) were all over 100%. Afterward the formulation of the Test product was changed and a new BE-study was conducted which again showed a bioinequivalence (AUC0-t and AUC0-inf outside of the acceptable range of 80-125%), with the point estimates all under 100%. The same test batch and the same formulation was investigated in a third BE-study, in which Cmax was out of the normally accepted range of 80-125% and all the point estimates were again over 100%, excluding the value 100%itself. Another batch of the same formulation of the Test product was then investigated in a fourth Replicate-BE-study which this time showed acceptable results for AUC0-t, AUC0-inf and Cmax (all in the normally acceptable range of 80-125%) but again the point estimates were only closely to 100%.
» Even if it should be assumed that the first two BE-studies were only pilot studies and hence not suitable for statistical inclusion and taking into consideration that the formulation of the Test product was changed after the first BE-study, it still remains unclear why the same formulation of the Test product in the following three performed BE-studies shows varying values concerning point estimates in these studies (all values were lower, then higher and then again lower than 100% successively). This fluctuation can not be explained through the "intra-subject variability" as the latter influences Cmax and hence justifies the widening of the acceptance range for Cmax but not the fluctuating point estimates which are independent of this variability
» This fluctuation is also independent of the number of subjects in the studies.

I think the assessor is implicitly making the assumption that the test product does not vary between batches. Dan vs. Regulator 2-0. The wording highlighted in red lacks qualification. Variability, whenever present, causes fluctuations.

» The Applicant is stating that the differences in point estimates in the four conducted BE-studies might result among others from different test batches. This is exactly the major objection raised by DE that the bioequivalence of the future batches of this formulation is not warranted."

This is certainly not a good comment. The only way to ensure this would be to do the mixed model with batch as random. Bioequivalence by today's definitions require two batches tested against each other and showing equivalence. If the rules change tomorrow, then ok, take that comment into consideration, otherwise just keep your cool and await the referral.

Pass or fail!
ElMaestro
martin
★★

Austria,
2010-08-06 17:25
(4186 d 01:54 ago)

@ ElMaestro
Posting: # 5733
Views: 14,038

## meta analysis?

Dear ElMaestro!

what do you think about a meta analysis?

best regards

martin
ElMaestro
★★★

Denmark,
2010-08-06 17:57
(4186 d 01:21 ago)

@ martin
Posting: # 5734
Views: 14,089

## meta analysis?

Dear Martin,

I am afraid I do not have a qualified opinion.
Meta analysis are not frequently used in the context of BE to the best of my knowledge. To me, pooling and meta analysis come close to one another and we all know the regulatory attitude todards the former...
Google does not seem to know a lot of useful documents combining BE and m.a. or m.a. and cmd(h).
Do you have a good idea about m.a.? I'd be happy to hear.

Pass or fail!
ElMaestro
Helmut
★★★

Vienna, Austria,
2010-08-06 18:31
(4186 d 00:48 ago)

@ ElMaestro
Posting: # 5735
Views: 14,474

## meta analysis?

Dear ElMaestro!

» I am afraid I do not have a qualified opinion.

Me not either, but I’m a little bit concerned about the (lacking) α-spending in meta analysis.

» Do you have a good idea about m.a.? I'd be happy to hear.

See Chapter 16 in Chow & Liu (3rd ed. 2009). They refer to two models (based on 2×2 cross-over studies only), where the first one is rather restrictive – assuming equal CVintra and CVinter across studies. Personal experience: nil.

1. Chow S-C, Liu J-p. Meta-analysis for bioequivalence review.
J Biopharm Stat. 1997:7(1):97–111. doi:10.1080/10543409708835172
2. Chow S-C, J Shao J. Bioequivalence review for drug interchangeability.
J Biopharm Stat. 1999;9(3):485–97. doi:10.1081/BIP-100101189

Dif-tor heh smusma 🖖
Helmut Schütz

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

France,
2010-08-06 23:21
(4185 d 19:58 ago)

@ martin
Posting: # 5738
Views: 14,424

## meta analysis?

Dear Martin, dear all,

I'm not a statistician. But as a "lay person" my question would be: what do we need to demonstrate ? The problem raised by the assessor is not that one study shows bioequivalence but a previous study was inconclusive (that's the situation where there were discussions about meta analysis at some point in time, if I'm not mistaken). The problem is rather that there are fluctuations in the point estimates between different studies, and I'm not sure what new information a meta analysis will bring.

When reading the objection from the assessor, it seems to me that he/she is considering the point estimate to be a "true value", not an experimental value affected by some level of uncertainty. This is reflected in the statement "This fluctuation can not be explained through the "intra-subject variability" as the latter influences Cmax and hence justifies the widening of the acceptance range for Cmax but not the fluctuating point estimates which are independent of this variability. This fluctuation is also independent of the number of subjects in the studies". This statement is not correct: the number of subjects in your study will directly influence the "reliability" of your experimental determination of the point estimate. Particularly with a high intra-CV. Actually, that level of uncertainty is precisely what the confidence interval represents, isn't it ?

I would see two possible ways out:
• disregard the pilot studies (insufficiently powered, poor estimate of the point estimate, test 1 was lab scale) and focus on the two pivotals. There is an overlap between the 90 % CI of the two studies, which, if I'm not mistaken, could be considered as meaning that the two point estimates are not statistically different;
• argue that there were also differences in the batch number of the reference product and that the difference in point estimates cannot be attributed a priori to the test product alone. Now let's first restrict the comparison to study 1 vs. study 3, which used different test batches but the same reference batch. OK, both studies fail to demonstrate BE, but that's not the question here: let's just focus on the point estimates. They're very close, right ? Now let's compare studies 2 and 3: same test batch, different reference batch. There we have much larger differences in point estimates. Conclusion: it seems that there is a batch-to-batch difference in the reference product, not in yours .
(And with a second thought, you can still use the end of the first option with the second option...)

Regards
Ohlbe

Regards
Ohlbe
Dr_Dan
★★

2010-08-10 12:27
(4182 d 06:52 ago)

@ Ohlbe
Posting: # 5759
Views: 14,268

## No chance against RMS?

Dear all,
First of all thank you very much for your invaluable input which I summarize in the following:
1. BE studies are not suitable to assess inter-batch variability. Batch is never a factor in ANOVAs according to recommendations published by the Efficacy Working Party
2. A batch to batch variation is a module 3 issue. The assessment of module 3 revealed no difference between test and reference batches. Comparative dissolution profiles showed similarity and the assay content differed less than 5%.
3. It's a common property of HVDs that not only the variance is high, but also the location of the T/R-ratio may vary across studies. There is no regulatory criterion requesting that the point estimators between BE studies must not varying.
4. There were also differences in the batch number of the reference product and thus the difference in point estimates cannot be attributed a priori to the test product alone.
5. None of the four studies demonstrated bioINequivalence. With increased sample sizes the studies would all be expected to be acceptable.
This was the first part of my problem...

The second is: To be right is one thing, that the assessor admits that we are right is the other thing.
My experience with this assessor is that he does not want to revise his opinion and he will not care about our arguments.

This assessor represents the view of the RMS (Germany). Even if one, two or all CMS disagree the procedure will come at an end and we will not get a marketing authorization (please correct me if I am wrong). A referral will not take place, so we do not have the possibility to argue against this assessor opinion in front of an expert committee.
What shall we do?
I am looking forward to your replies.
Kind regards
Dan

P.s.: Dear Helmut, please change the category to regulatory, thanks!

Done. [Helmut]

Kind regards and have a nice day
Dr_Dan
ElMaestro
★★★

Denmark,
2010-08-10 16:26
(4182 d 02:53 ago)

@ Dr_Dan
Posting: # 5765
Views: 14,057

## No chance against RMS?

Dear Dan,

Under the first point I think I would add that you have not (have you?) received questions for module 3 and specifications; thus, the potential variability from the studies reflects -at least in part- the natural variation within the specifications that the M3 assessor has accepted.

You are right, a referral will not be triggered if the RMS disagrees. This is a new thing and there is very little you can do. If the RMS enforces a principle that has never been defined in a guideline document and which is not in line with contemporary equivalence thinking then honestly I think this would be a case for the European Generics Association. This is a slow process and not one that can be expected to work here and now.
Having said that, I am pretty sure regulators from BfArM are seeing this thread and giving it some consideration. That would potentially be in your favour.

I actually still believe your product is alive.

Pass or fail!
ElMaestro
kumarnaidu
★

Mumbai, India,
2016-07-20 07:16
(2011 d 12:03 ago)

@ ElMaestro
Posting: # 16494
Views: 11,366

## Batch-to-Batch Pharmacokinetic Variability

Kumar Naidu
Helmut
★★★

Vienna, Austria,
2016-07-20 10:48
(2011 d 08:31 ago)

@ kumarnaidu
Posting: # 16495
Views: 11,253

## tlast (Common)

Hi Kumar,

THX for posting this  amazing  scary article ( free resource). Some observations:
• I checked the CVs; the study had enough power. Hence, the results are likely not by pure chance.
• See my remarks about the EMA’s GL in this post above. When we think about a “representative batch of the reference product” this study opens a can of worms.
• We discussed numerous times (f.i. here) the potential bias of AUC-ratios if tlast of T and R are different. Quote from the article:
• As previously reported,* within-subject differences in the time of the last quantifiable drug concentration (tlast) between PK profiles that differ in magnitude (and therefore fall below assay limit of quantitation [LOQ] at different times) can lead to bias in the AUC(0-t) geometric mean ratio. This is simply a consequence of comparing AUC(0-t) values that have been calculated using different time windows, and is remedied by use of a common tlast for all profiles for a given subject and analyte (the revised PK parameter is referred to as AUC(0-tcommon)). As expected, Table 4 displays this phenomenon, with all AUC(0-t) geometric mean ratios differing from 100% more than the corresponding AUC(0-tcommon) ratios.
I love it.

• Fisher D, Kramer W, Burmeister Getz E. Evaluation of a Scenario in Which Estimates of Bioequivalence Are Biased and a Proposed Solution: tlast (Common). J Clin Pharm. 2016; 56(7): 794–800. doi:10.1002/jcph.663. free resource.

Dif-tor heh smusma 🖖
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
nobody
nothing

2019-02-21 15:20
(1065 d 02:59 ago)

@ Helmut
Posting: # 19960
Views: 7,377

## tlast (Common)

Hi Hellmuth and Band!

So what is your solution, do you specify in the protocol/SOP that AUC0-tlast is calculated employing tlast common? Any outcries in EU? Did anybody (sponsor/reg) recognize/ask back?

Edit: Never ever address me with Hellmuth again! [Helmut]

Kindest regards, nobody
ElMaestro
★★★

Denmark,
2019-02-21 16:32
(1065 d 01:47 ago)

@ nobody
Posting: # 19961
Views: 7,411

## tlast (Common)

Hello nobody,

» So what is your solution, do you specify in the protocol/SOP that AUC0-tlast is calculated employing tlast common? Any outcries in EU? Did anybody (sponsor/reg) recognize/ask back?

As far as I know, regulators are well aware of the publications of Getz et al. They totally get the idea aired in their papers, but they don't change their views or ask applicants to change practices.
We also need to bear in mind that it is quite common to have time point deviations on late samples, especially if they are ambulatory. Some of those are missing, even.
There is, for practical purposes, not always a crystal clear way we can speak of a meaningful last common time point regardless of how well the trial is planned and executed.

Pass or fail!
ElMaestro
nobody
nothing

2019-02-21 17:02
(1065 d 01:17 ago)

@ ElMaestro
Posting: # 19962
Views: 7,265

## tlast (Common)

» ...but they don't change their views or ask applicants to change practices.

Let's play devel's advocate for a moment and assume, AUC tlast fails, but AUC common tlast is OK (besides the fact that in 80% Cmax should fail alone or in combination with AUC, but....). And you can show a high number of non-common tlast (equally distributed between T and R). No way forward, or?

Kindest regards, nobody
ElMaestro
★★★

Denmark,
2019-02-21 18:02
(1065 d 00:17 ago)

@ nobody
Posting: # 19963
Views: 7,380

## tlast (Common)

Hi again,

» Let's play devel's advocate for a moment and assume, AUC tlast fails, but AUC common tlast is OK (besides the fact that in 80% Cmax should fail alone or in combination with AUC, but....). And you can show a high number of non-common tlast (equally distributed between T and R). No way forward, or?

Would likely be accepted in Sweden

Pass or fail!
ElMaestro
nobody
nothing

2019-02-21 18:17
(1065 d 00:02 ago)

@ ElMaestro
Posting: # 19964
Views: 7,216

## tlast (Common)

» Would likely be accepted in Sweden

...you're so funny... NOT :-p

Kindest regards, nobody