Log in
Register|
Dirk ☆ Neu-Ulm, Germany, 2012-12-03 01:14 (4957 d 20:12 ago) Posting: # 9654 Views: 6,514 |
|
|
Dear all, According to the EMA BE guidance “…the decision to exclude a subject from the statistical analysis must be made before bioanalysis”. We currently discuss the interpretation of “before bioanalysis”, considering two options: Option 1: “Before bioanalysis” will be interpreted as “prior to availability of bioanalytical data” (assuming the processes will guarantee that the clinical and statistical teams and whoever will be involved in the allocation of subjects to analysis sets will not have any access to bioanalytical data prior to fixed and documented decisions on the allocation of subjects to analysis sets). Option 2: “Before bioanalysis” will be interpreted as “prior to start of any bioanalytical assessments”. We wonder whether anybody in this forum can report any negative experiences (which may in worst case include rejections or at delays of regulatory approvals) from applying option 1 rather than the more strict interpretation (option 2). Please note, that this discussion wants to fully recognize that the allocation of subjects to analysis sets (i.e. the potential exclusion of subjects from the PK set) has to happen at a time point when the clinical and statistical teams (and whoever else may be involved in this process) have not yet seen any bioanalytical data (and would not even have the chance to do so; Option 1 is to be discussed under these requirements only). I.e., the discussion does not wish to question the need for a high-quality allocation/decision process, at all. It is just a question about timelines and total study duration (sponsors prefer, of course, not to wait with the start of bioanalytical assessments longer than necessary). Option 2 means that the start of bioanalysis has to wait more or less until DB lock (clinical data). Option 1 would allow running assays earlier as long as the data and information flow can be clearly separated. For some compounds the start of bioanalysis cannot wait for technical / analytical reasons (stability etc.) that long as option 2 may request. In those cases only option 1 will be feasible. However, this will not be an issue, anyway, if option 1 is accepted in general. We can share with this forum that, so far, we are not aware of any negative experiences when applying option 1. However, this does not mean that anybody else might have run into such an experience. Any feedback / opinion / reporting of negative experiences (if any) is highly appreciated. Many thanks and best regards, Dirk Lehnick |
|
Dr_Dan ★★ Germany, 2012-12-03 13:56 (4957 d 07:30 ago) @ Dirk Posting: # 9659 Views: 5,399 |
|
|
Dear Dirk My experience is that if you interpret "Before bioanalysis” as “prior to availability of bioanalytical data” no assessor will care and ask for the evaluation of the study with the full data set. Of course you will then present a sensitivity analysis but it could be hard to convince the assessor that the exclusion of the subject was justified. I made several bad experiences especially with MHRA. Even if you proof that the allocation of subjects to analysis sets happened at a time point when the clinical and statistical teams had not seen the bioanalytical data, the assessor will insist on the full analysis. So if you think that a clinical event will influence the outcome of the study you have to exclude this subject before you start sample measurement. I do not think that bioanalysis has to wait more or less until DB lock. You could organize a data review meeting and decide which subject should be excluded due to clinical reasons and you have to document this properly. Therefore you should define that “Before bioanalysis” will be interpreted as “prior to start of any bioanalytical assessments”. Consequently you should not measure samples of an excluded subject as long as this is not necessary (e.g. AE suspected to be related to drug concentration level). CPMP/QWP/EWP/1401/98 Rev. 1 states "If available, concentration data and pharmacokinetic parameters from such subjects should be presented in the individual listings, but should not be included in the summary statistics." So itś better not to have these data and to avoid a sensitivity analysis. I hope this helps. Kind regards Dan — Kind regards and have a nice day Dr_Dan |
|
Helmut ★★★   Vienna, Austria, 2012-12-03 14:43 (4957 d 06:43 ago) @ Dr_Dan Posting: # 9661 Views: 5,406 |
|
|
Dear Dirk & Dan, I have a different position here. The clinical site should issue a “end of clinical phase” document (or however you prefer to name it) and provide it to the sponsor / analytical site. This year we had a minor finding in an inspection by the Austrian AGES because the analytical protocol stated “A samples from B subjects in C periods; i.e. D samples in total”. One subject dropped out after the first period. Samples of this subject were documented in the shipment record as “missing”. All samples were analysed and the deviation from the protocol (fewer samples than expected, but in agreement with the shipment record) given in the analytical report. The inspector mentioned that the fact that samples were missing should have been acknowledged by the sponsor beforehand and the analytical protocol amended. BTW, we always analyse all available samples. For a justification see this story. — 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 |
|
Dr_Dan ★★ Germany, 2012-12-03 17:39 (4957 d 03:48 ago) @ Helmut Posting: # 9663 Views: 5,408 |
|
|
Dear Helmut It always depends on the situation. Do you remember the following case? In 2 subjects of a BE study samples were missing and we decided that if the missing sample was adjacent to the highest concentration, it would be assumed that the missing sample could have been the Cmax. In this case, the subject would be excluded from the statistical and pharmacokinetic analyses for this period. This assessment was done once the bioanalytical data for this study became available and before the statistical and pharmacokinetic analyses. This was in agreement with the study protocol and this was not questioned by the assessor. The assessment of the study showed bioequivalence when the 2 subjects were excluded but when the subjects were included the upper limit of the 90% CI was very slightly outside of the acceptance range. Data used in a sensitivity analysis for missing values that would still lead to a conclusion of bioequivalence were within the range of Cmax values obtained in the study. However, the assessor insisted of regarding the full data set analysis as valid and consequently the study did not pass. BTW assessors of different regulatory authorities accepted the procedure for excluding the subjects. My lesson learned is that in doubtful cases you have to carefully evaluate if you really should analyze all subjects. Kind regards Dan — Kind regards and have a nice day Dr_Dan |
|
Dirk ☆ Neu-Ulm, Germany, 2012-12-04 03:52 (4956 d 17:34 ago) @ Dr_Dan Posting: # 9665 Views: 5,245 |
|
|
Dear Dan and Helmut, many thanks for your posts. Regarding bioanalysis of samples, we tend to agree to Helmut. We use to analyze all samples (maybe exceptions only in very clear cases; e. g. an early drop-out in period 1 of a 2*2 design). We feel this is in line with EMA statements (e.g. the statement within the comments to the Draft EMA guideline; also cited in the EGA Symposium document): All samples should be bio-analysed. This is useful if there is disagreement between Member States over an exclusion as it enables the statistical analysis to be repeated. We wish to be as transparent as possible and will, in borderline cases, not dread sensitivity analyses. We just do not wish to be punished for this approach of transparency (which is even supported by Option 1) by receiving negative regulatory feedback on option 1 just from the formal reason "date of start of bioanalytical assessments too early" although option 1 does not take any misuse from this circumstance. Dan, I agree to the conduct of a Data Review Meeting to fix the allocation of analysis sets. We always do this, irrespective of applying option 1 or 2. However, we prefer to perform this on data which is at least sufficiently clean and validated. Everybody who is long enough in this business knows that e.g. the probability of disagreements between Medical Monitors and Investigators on relationship of AEs will be higher than the corresponding AE incidence  .I.e. if you fix the exclusion of subjects before the sponsor's Medical Monitors have seen all the relevant cleaned data, this may lead into trouble. If you wait until clean data can be used as a base (which we prefer) then you will be in the situation I mentioned: i.e., option 2 will not allow starting the bioanalytical assessment more or less prior to DB lock. Thanks a lot and best regards, Dirk Edit: Document linked. [Helmut] |
|
d_labes ★★★ Berlin, Germany, 2012-12-04 09:33 (4956 d 11:53 ago) @ Dirk Posting: # 9666 Views: 5,158 |
|
|
Dear Dirk, dear All, I have always considered the sentence "…the decision to exclude a subject from the statistical analysis must be made before bioanalysis." at least as a curiosity (not to say something in-polite). How could we do that for such allowed reasons (EMA guidance, page 14) following the above sentence nearly direct in the text: "1) A subject with lack of any measurable concentrations or only very low plasma concentrations for reference medicinal product. A subject is considered to have very low plasma concentrations if its AUC is less than 5% of reference medicinal product geometric mean AUC (which should be calculated without inclusion of data from the outlying subject). The exclusion of data due to this reason will only be accepted in exceptional cases and may question the validity of the trial. 2) Subjects with non-zero baseline concentrations > 5% of Cmax. Such data should be excluded from bioequivalence calculation (see carry-over effects below)." Try this without knowledge of any bioanalytical data  .Ok, they handle that reasons under "exclusion of data", but in consequence it is the exclusion of subjects (as stated also in the text), at least if we talk 2x2 crossover. — Regards, Detlew |
|
lukamar ☆ Poland, 2012-12-04 11:11 (4956 d 10:15 ago) @ d_labes Posting: # 9669 Views: 5,110 |
|
|
Dear all, ❝ Try this without knowledge of any bioanalytical data That's why these two situations are called exceptions  In regards to discussion option 1 vs option 2 from first Dirk post I would agree with Dr_Dan and interpret guideline wording as "prior to start of any bioanalytical assessment". This wording is repeated and rephrased in EGA Q&A document as: "In any case, the decision to exclude data from this subject from the statistical analysis must be made before performing the bio-analysis, in order to avoid bias. Any reason to exclude data is acceptable provided it is pre-specified and a decision taken before the bio-analysis.". I think there was just too many cases in which it was evident that "unconvenient" subject was excluded just to get 90% CI within acceptance criteria. And it is very difficult to guarantee (IMHO it's not possible at all) that clinical and statistical team will not have access to bioanalytical data once bioanalysis has started. Based on my experience, in case of exclusion of subject after start of his/her samples bioanalysis, reanalysis with full set of data may be requested. Best regards L. |
|
Dirk ☆ Neu-Ulm, Germany, 2012-12-05 15:04 (4955 d 06:22 ago) @ lukamar Posting: # 9673 Views: 4,946 |
|
|
Dear Lukamar, ❝ ❝ And it is very difficult to guarantee (IMHO it's not possible at all) that clinical and statistical team will not have access to bioanalytical data once bioanalysis has started. I guess, one can "guarantee" the separation of data and information flow in terms of audited and well controlled processes - at least, in a way which may comparable to e.g. handling randomisation lists. Even most accurate processes can never totally exclude criminal energy. But I won't tend to a priori assess bioanalytical centers as being less trustworthy than randomisation units, pharmacies / trial supply units etc. who deal with randomisation lists … .In many pivotal Phase III studies one has to establish processes of separation of data and information flow which seem to be even more critical and complex to "guarantee". Let's think of unblinded Safety Monitoring Boards, unblinded interim evaluations even of primary efficacy (third party, independent statistician, and corresponding boards)... or settings in which one may need to control how to separate the data flow for parameters that have unblinding potential (e.g. if a test product modifies liver values in another way than the comparator, you need to separate the liver values from the data flow …). Let's think not at least about a situation in which non-inferiority might be the primary hypothesis in such a pivotal Phase III setting (where per protocol is the primary set and exclusion of subjects from the PP set can heavily affect the primary outcome). Just wish to say: at least the principle of separating data and information flow via accurate processes seems to be well accepted (even by authorities) also in settings which are even more critical amd more complex to control. The question is: Will regulatory reviewers assess us, while working on BE settings, as less trustworthy than acting in pivotal Phase III settings? To avoid this, I think, we can only contribute by being as transparent as possible (analyzing all samples also for excluded subjects etc.) when working on BE studies. Many thanks and best regards, Dirk |
Ing. Helmut Schütz