Log in
Register|
Dr_Dan ★★ Germany, 2010-06-02 16:57 (5871 d 01:09 ago) Posting: # 5412 Views: 8,283 |
|
|
Dear all In a BE study you have a certain statistical risk to fail even if test and reference product are bioequivalent. Additionally you could have an statistical outlier who bias the study results. According to the new BE guideline these outliers can not be excluded. The draft version of the guideline contained the following text which is now missing in the final: Evaluation of data from several bioequivalence studies If the application contains some studies which demonstrate bioequivalence and others that do not, the documentation must be considered as a whole. The existence of a positive study does not mean that negative studies can be ignored. In this situation the interpretation of the overall documentation is not straightforward but there are three distinct situations which can be considered:
Can you suggest a strategy to present additional data/studies? I am looking forward to your replies. Kind regards Dan — Kind regards and have a nice day Dr_Dan |
|
ElMaestro ★★★ Denmark, 2010-06-03 00:36 (5870 d 17:31 ago) @ Dr_Dan Posting: # 5418 Views: 7,705 |
|
|
Hi Dan, ❝ Would it be reasonable after implementation of the new guideline just to repeat the study (if I am convinced that both formulations are bioequivalent) and to discuss the results with the assessors? I would not dare to generalise an answer here, but you are touching something that may very well be assessed differently across EU countries, and it could even come down to whether or not the product is submitted in a dcp or nationally. Perhaps you can tell a bit more about how your study failed - was the entire CI outside the acceptance region? And could you elaborate on how you feel convinced the formulations are equivalent? If you for some reason feel very convinced of equivalence then I would ask an auditor to take a look at the material generated by your clinical/analysis lab/statistician. In the absence of a good reason to distrust the first data, you will with another study in the best case end up with potentially discrepant study results or in the worst case just be plainly wasting money. Best regards EM. — Pass or fail! ElMaestro |
|
Dr_Dan ★★ Germany, 2010-06-03 11:21 (5870 d 06:45 ago) @ ElMaestro Posting: # 5422 Views: 7,618 |
|
|
Dear ElMaestro Thanks for your reply. I agree that this issue will be assessed differently across EU countriesit and surely it will be a great difference whether or not the product is submitted in a dcp or nationally. Actually I have the following problem: 3 studies, US fasted and fed bioequivalent, EU failed (upper CI for Cmax 125.1%). US and EU reference have a different formulation, test was the same. Failed EU study: Point estimator for Cmax 117%, for AUC 87% => test and reference different formulations with different PK. Honestly I see no chance. Do you agree that a second BE study would not convince the assessor even if bioequivalence is shown? Kind regrads Dan — Kind regards and have a nice day Dr_Dan |
|
ElMaestro ★★★ Denmark, 2010-06-04 03:48 (5869 d 14:18 ago) @ Dr_Dan Posting: # 5432 Views: 7,618 |
|
|
Dear Dan, ❝ I have the following problem: 3 studies, US fasted and fed bioequivalent, EU failed (upper CI for Cmax 125.1%). US and EU reference have a different formulation, test was the same. Failed EU study: Point estimator for Cmax 117%, for AUC 87% => test and reference different formulations with different PK. Honestly I see no chance. Do you agree that a second BE study would not convince the assessor even if bioequivalence is shown? In your situation, if I get you right the PE's are within the standard acceptance range. This to me would indicate that at least one of the following points is true:
A true showstopper is when the entire CI is outside the acc. range. Best regards EM. — Pass or fail! ElMaestro |
|
Dr_Dan ★★ Germany, 2010-06-04 15:13 (5869 d 02:54 ago) @ ElMaestro Posting: # 5435 Views: 7,598 |
|
|
Dear ElMaestro Thanks for your reply, but I disagree regarding hope. ❝ 1. The products are more dissimilar than anticipated (actually, I have no idea what you expected but my guess is you powered your study on basis of a max 5%-10% difference). Lack of power. ❝ 2. The intrasubj. variability is higher than anticipated. Lack of power. ❝ 3. You were unlucky. I agree that a true showstopper is when the entire CI is outside the acc. range, but this is not the question. Cmax: 117% [109-125] AUC: 87% [83-92] As an assessor I would argue that the formulations are not similar to such a degree that their effects regarding safety and efficacy would be the same. In fact the study shows that similarity can be excluded since both CI do not include 100%. The PK of test and reference are different. How could you defend the study even if the upper CI for Cmax would be within the acc. range? Kind regards Dan — Kind regards and have a nice day Dr_Dan |
|
ElMaestro ★★★ Denmark, 2010-06-04 17:26 (5869 d 00:40 ago) @ Dr_Dan Posting: # 5436 Views: 7,492 |
|
|
Dear Dr Dan, ❝ In fact the products are more dissimilar than anticipated (17% instead of max. 5%), nevertheless the study was well powered. Increasing the power by including more subjects will not convince the regulatory authority if the PE is 117%, right? Otherwise you can make nearly every formulation bioequivalent if you only use enough subjects. As long as the CI is within the acceptance range there should be no issue. The decision is not based on the PE. ❝ As an assessor I would argue that the formulations are not similar to such a degree that their effects regarding safety and efficacy would be the same. In fact the study shows that similarity can be excluded since both CI do not include 100%. The PK of test and reference are different. No assessor can do that on basis of your data. They measure no such thing as effect and safety directly. By convention the effects and safety aspects are anticipated to bad as good or as bad as that of the originator if the product is considered bioequivalent. The fact that 100% is not part of the CI only shows they are different; of course they are because they are manufactured different places by different methods by different people etc etc. But they might still be similar enough. Best regards EM. — Pass or fail! ElMaestro |
|
Ohlbe ★★★ France, 2010-06-05 02:39 (5868 d 15:28 ago) @ Dr_Dan Posting: # 5438 Views: 7,508 |
|
|
Dear Dan, ❝ Cmax: 117% [109-125] AUC: 87% [83-92] ❝ As an assessor I would argue that the formulations are not similar to such a degree that their effects regarding safety and efficacy would be the same. In fact the study shows that similarity can be excluded since both CI do not include 100%. The PK of test and reference are different. No: I would not say that the study shows that similarity can be excluded, just because the CI does not include 100 %. What this shows is that you have a statistically significant difference between your formulations. Whether this is a real difference or whether it is due to a high power (low variability, and your CI are not very wide) remains to be discussed. Yes: it looks like the two formulations have a different PK profile: lower AUC but higher Cmax. If both parameters were on the low side or both on the high side, you could argue it is just by chance (or rather bad luck) and repeat the study. But with these contrasted results it looks more difficult to justify. Regards Ohlbe — Regards Ohlbe |
|
Helmut ★★★   Vienna, Austria, 2010-06-06 01:31 (5867 d 16:36 ago) @ Ohlbe Posting: # 5445 Views: 7,544 |
|
|
Dear Dan, Ohlbe, and ElMaestro! ❝ ❝ Cmax: 117% [109-125] AUC: 87% [83-92] ![[image]](img/uploaded/image48.png) Let's assume that the upper CL of Cmax was 125.00 instead of 125.1...The formal requirements for BE are fulfilled (except maybe in Denmark, my Prince). But: If PE of AUC and Cmax are far away from unity and in opposite direction, a true difference in PK is likely. Let's assume that 87% is a true lower F of the test, D=100, V=5, kel=0.1. If kin=0.5 for the reference, it needs a kin=3.434 (almost 7times faster absorption) in order to get a PE of 117% for Cmax. Profiles are very different; tmax is 1.06 for test and 4.02 for reference. I don't know how assessors would handle such a situation. Maybe on a case-by-case basis. Remember IR nifedipine? — Dif-tor heh smusma 🖖🏼 Довге життя Україна! ![[image]](https://static.bebac.at/pics/Blue_and_yellow_ribbon_UA.png) Helmut Schütz ![[image]](https://static.bebac.at/img/CC by.png) The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
ElMaestro ★★★ Denmark, 2010-06-06 22:31 (5866 d 19:35 ago) @ Helmut Posting: # 5446 Views: 7,486 |
|
|
Dear Helmut, ❝ But: If PE of AUC and Cmax are far away from unity and in opposite ❝ direction, a true difference in PK is likely. Let's assume that 87% is a true lower F of the test, D=100, V=5, kel=0.1. If kin=0.5 for the reference, it needs a kin=3.434 (almost 7 times faster absorption) in order to get a PE of 117% for Cmax. Profiles are very different; tmax is 1.06 for test and 4.02 for reference. Hehe, although I can follow your point and agree different PK characteristics could be a reason I think this is speculation; lower extent, higher rate. I do agree that this could be an explanation, but when you say a true difference in PK is likely, do you intend to say more likely than other explanations? I don't think this would be reflecting in the true spirit of the way CIs are usually used (there is a decent chance for example that the true PE is not 87% but 91% etc), and I would not be surprised if it would be possible to come up with a numerical example that shows the opposite (?). ❝ Remember IR nifedipine?  Finally, violation of underlying assumptions seems often to be of no practical concern here in Europe (non-parametric statistics would serve as a good example  ).Best regards, EM. — Pass or fail! ElMaestro |
|
Helmut ★★★ Vienna, Austria, 2010-06-06 23:43 (5866 d 18:24 ago) @ ElMaestro Posting: # 5447 Views: 7,682 |
|
|
Dear ElMaestro! ❝ Hehe, although I can follow your point and agree different PK characteristics could be a reason I think this is speculation; lower extent, higher rate. I do agree that this could be an explanation, but when you say a true difference in PK is likely, do you intend to say more likely than other explanations? Well, luckily this is not one of our studies - so I played Devil’s advocate. What looked strange to me was the large deviation of both metrics from unity in opposite directions (PE of Cmax is almost 1/PE of AUC). For sure it could have been only chance (the non-PK ‘explanation’), but on the other hand Dan said that the CV was <10% – so I have some doubts… We know that Cmax is a lousy metric of the rate of absorption (because influenced by the extent as well). This was the reason why some colleagues (Lacey, Midha, Endrényi,…) advocated Cmax/AUC in he past. ❝ I don't think this would be reflecting in the true spirit of the way CIs are usually used (there is a decent chance for example that the true PE is not 7% but 91% etc),… Well, right. ❝ …and I would not be surprised if it would be possible to come up with a numerical example that shows the opposite (?). Don’t think so (untested – only a gut feeling).  ❝ ❝ Remember IR nifedipine? ❝ No, please enlighten me IR nifedipine lead to nasty adverse events due to the fast initial ramp. MR formulations were developed afterwards; I’m not sure whether IR nifedipine is still on the market. ❝ Finally, violation of underlying assumptions seems often to be of no practical concern here in Europe (non-parametric statistics would serve as a good example Yes. If the study would have been BE (with 125.1% it was a quite close miss), formal requirements would have been satisfied. On the other hand the difference in tmax shows a sunstantial difference in the rate of absorption. Of course we do not test for that (nonparametrics  ), only:However, if rapid release is claimed to be clinically relevant and of importance for onset of action or is related to adverse events, there should be no apparent difference in median tmax and its variability between test and reference product. I had this section in mind when I referred to IR nifedipine. BTW, since nonparametrics are not acceptable, how will we compare formulations after August 1st, 2010? Boxplots?— Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
|
ElMaestro ★★★ Denmark, 2010-06-07 00:44 (5866 d 17:23 ago) @ Helmut Posting: # 5448 Views: 7,476 |
|
|
Ahoy Helmut, ❝ (...) Of course we do not test for that (nonparametrics ❝ However, if rapid release is claimed to be clinically relevant and of importance for onset of action or is related to adverse events, there should be no apparent difference in median tmax and its variability between test and reference product. I had this section in mind when I referred to IR nifedipine. BTW, since nonparametrics are not acceptable, how will we compare formulations after August 1st, 2010? Boxplots? That is one helluva good question. Hey wait.... I know the elmaestrolophystic solution. We just transform tmax so that it becomes log-normal. Easy. Of course, it is below my dignity to give details on how to tranform tmax, since it is standard statistical technique. And who decides in this context what "clinically relevant and of importance for onset of action or is related to adverse events" means? The assessor? The ref. SPC? I could be afraid this is going to be "The applicant should document that the rate of absorption of Schützomycin is not linked to onset of action or adverse events" galore... Best regards EM. — Pass or fail! ElMaestro |
|
statistician ☆ India, 2010-06-05 10:43 (5868 d 07:24 ago) @ Dr_Dan Posting: # 5440 Views: 7,646 |
|
|
dear all, i would like to know the assignable cause for BA/BE failure. also how to interpret if CI falls below 80% and above 125% what will be the reason for the wider range of CI value falling below lower limit and above upper limit. thank you, with warm regards RK |
|
Jaime_R ★★ Barcelona, 2010-06-05 14:07 (5868 d 04:00 ago) @ statistician Posting: # 5441 Views: 7,563 |
|
|
Hi RK! ❝ i would like to know the assignable cause for BA/BE failure. See ElMaestro’s post. ❝ also how to interpret if CI falls below 80% and above 125% what will be the reason for the wider range of CI value falling below lower limit and above upper limit. Don’t mix up CL (confidence limit) and CI (confidence interval). I guess you are talking about something like this? ├────┼─────┤ (AR: 80%–125%)├──────┼───────┤ (CI: 75%–135%)Obviously the variability was higher than expected in designing the study. A manifold of potential reasons:
— Regards, Jaime |
|
Helmut ★★★ Vienna, Austria, 2010-06-05 21:27 (5867 d 20:40 ago) @ Jaime_R Posting: # 5444 Views: 7,561 |
|
|
Dear Jaime! ❝ β-error hit you. Remember that if you planed the study for 80% power, you have a 20% chance of failure of a ‘true’ bioequivalent product (β=1–power).
— Dif-tor heh smusma 🖖🏼 Довге життя Україна! Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. 🚮 Science Quotes |
![[image]](img/uploaded/image12.jpg)
|
Dr_Dan ★★ Germany, 2010-06-08 13:32 (5865 d 04:35 ago) @ Jaime_R Posting: # 5465 Views: 7,365 |
|
|
Dear all Thank you very much for this discussion. As you see the problem is more complicated as you would expect on the first glance. To my experience the assessors are more strikt regarding the guidelines and the worst thing is that they will not discuss with you. Either you fullfil the requirements or not and in this case you will have no chance regardless of clinical significance. My intention was to make you sensitive to a changed assessment by the regulatory authorities. Dear Jamie We did not use an unreliable CV to plan the study and the intra-subject variability (ANOVA CV) was very low. Kind regards Dan — Kind regards and have a nice day Dr_Dan |
|
Jaime_R ★★ Barcelona, 2010-06-08 14:05 (5865 d 04:02 ago) @ Dr_Dan Posting: # 5466 Views: 7,377 |
|
|
Dear Dan! Agree with your POVs about current regulatory practices. BTW, Jaime # Jamie ❝ We did not use an unreliable CV to plan the study and the intra-subject variability (ANOVA CV) was very low. I didn't say "unreliable" in my answer to RK's post; "uncertain" is the better term (as I have learned from the statistical experts here). Seeing your study's results must have given you nightmares, I guess. — Regards, Jaime |
|
ElMaestro ★★★ Denmark, 2010-06-08 22:42 (5864 d 19:24 ago) @ Dr_Dan Posting: # 5481 Views: 7,368 |
|
|
Dear Dan, ❝ Thank you very much for this discussion. As you see the problem is more complicated as you would expect on the first glance. To my experience the assessors are more strikt regarding the guidelines and the worst thing is that they will not discuss with you. Either you fullfil the requirements or not and in this case you will have no chance regardless of clinical significance. My intention was to make you sensitive to a changed assessment by the regulatory authorities. This was a very good comment; I completely understand what makes you say what you do, however, try for a second to put yourself in the shoes of a regulator. You have a boss to please, and if there is something bosses do not like then it is when their assessors do something that ends up in a referral. Even worse, in some countries the reg.authority can be dragged into court for approving something the originator doesn't like and that actually does happen nowadays when parts of AR's are public. Being a young assessor questioned by a team of overpaid lawyers in suits plus a judge and asked to justify a deviation from a guideline is not exactly a picnic, so I've been told. Therefore, if the guideline says you must be within 0.8000 to 1.2500 or whatever and you have 0.7991 to 1.0363 etc then although the assessor is convinced there might not be any particular problem with that, his/her boss my not allow a positive assessment. The risk of a resource-draining referral or court case may simply outweigh any other benefit in the mind of the boss. Best regards EM. — Pass or fail! ElMaestro |
Ing. Helmut Schütz