Yura Regular Belarus, 20170501 09:34 Posting: # 17285 Views: 1,907 

Dear All Do I understand correctly that using the TRT / RTR design it will be necessary to compare the RR and TT extension interval that are built on the data of subjects of different groups (50% of study subjects are used for RR and 50% of other subjects for TT), the comparison will be for different Subjects? Or will it be necessary to compare the extension interval RR, built on 50% of the subjects, and the confidence interval TR, built on the data of all subjects? The appropriate sample size must be. Alpha correction. Best regards 
mittyri Senior Russia, 20170502 18:38 @ Yura Posting: # 17295 Views: 1,668 

Dear Yura, you should use all data included in PK analysis (ie all subjects) for CI estimation. To extent the limits you need to calculate variability of RR, that is, 50% of subjects. A hint: even if you try to calculate variability of RR for overall dataset, the majority of software (at least SAS PHX, R) is smart enough not to include the subjects with one R. » The appropriate sample size must be. » Alpha correction. what did you mean here? — Kind regards, Mittyri 
Yura Regular Belarus, 20170502 21:19 @ mittyri Posting: # 17296 Views: 1,647 

Dear mittyri, The approach of referencescaled average BE (RSABE) requires that the reference formulation be measured twice in each subject. And therefore do not understand the legality of the calculation in this design. Best regards 
Helmut Hero Vienna, Austria, 20170502 22:04 @ Yura Posting: # 17297 Views: 1,649 

Hi Yura, » […] referencescaled average BE (RSABE) requires that the reference formulation be measured twice in each subject. Says who? » And therefore do not understand the legality of the calculation in this design. Keep in mind that the early work on RSABE (by the two Lászlós) was solely based on exactly this design. — Regards, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. ☼ Science Quotes 
Yura Regular Belarus, 20170503 07:41 (edited by Ohlbe on 20170503 10:28) @ Helmut Posting: # 17298 Views: 1,621 

Hi Helmut "The approach of referencescaled average BE (RSABE) requires that the reference formulation be measured twice in each subject." An Exact Procedure for the Evaluation of ReferenceScaled Average Bioequivalence Best regards Edit: two posts merged (remember you can edit your post within 24 hours) and link corrected [Ohlbe] 
Helmut Hero Vienna, Austria, 20170503 21:49 @ Yura Posting: # 17302 Views: 1,546 

Hi Yura, » An Exact Procedure for the Evaluation of ReferenceScaled Average Bioequivalence Oops! Incidentally I was one of the reviewers of this paper. The phrase … » "The approach of referencescaled average BE (RSABE) requires that the reference formulation be measured twice in each subject." … slipped through my intention. My fault. This statement is – despite the two great authors – wrong. — Regards, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. ☼ Science Quotes 
ElMaestro Hero Denmark, 20170503 22:04 @ Helmut Posting: # 17303 Views: 1,525 

Hi Hötzi, » » "The approach of referencescaled average BE (RSABE) requires that the reference formulation be measured twice in each subject." » … slipped through my intention. My fault. This statement is – despite the two great authors – wrong. Not your fault. As a reviewer you are not and will never be responsible for anything authored by others. But having said that I think the sentence is OK, only the meaning is a bit out of context. To calculate meaningfully the intrasubject variability for Ref we need subjects in whom it has been measured (at least) twice. That was presumably what the authors tried to remind readers. The entire issue of rsaBE cannot be approached in the absence of such subjects, and therefore the sentence makes quite some sense. Blumentopferde — I could be wrong, but… Best regards, ElMaestro No, I still don't believe much in the usefulness of IVIVCs for OIPs when it comes to picking candidate formulations for the next trial. This is not the same as saying I don't believe in IVIVCs. 
Helmut Hero Vienna, Austria, 20170506 18:33 @ ElMaestro Posting: # 17312 Views: 1,358 

Hi ElMaestro & Yura, » But having said that I think the sentence is OK, only the meaning is a bit out of context. Right. It appears in the introduction about the FDA’s RSABE. Nothing like that in the next paragraph about the EMA’s ABEL. The GL states only the periods (three or four) and no specific designs/sequences. In the Q&A the RTRTTRTR and the RRTRTRTRR are given as examples. In Rev. 12 of the Q&A the TRTRTR is discussed (requiring at least 12 subject in sequence RTR). » Blumentopferde Love it! — Regards, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. ☼ Science Quotes 
Yura Regular Belarus, 20170509 11:35 @ Helmut Posting: # 17332 Views: 1,298 

Hi Helmut What if the design of the TRR / RTT obtained by Cmax CI TR = 0.881.21; CV TR =48%; expansion [0.711.40] CV RR=47%; GMR TT=0.708 not belong [0.80; 1.25]; CI TT = 0.471.06; CV TT =77%. Is it possible to draw a conclusion about bioequivalence? Best regards Edit: Subject line changed; see also this post #2. [Helmut] 
Helmut Hero Vienna, Austria, 20170509 13:06 @ Yura Posting: # 17333 Views: 1,304 

Hi Yura, » What if the design of the TRR / RTT obtained by Cmax » CI TR = 0.881.21; » CV TR =48%; » expansion [0.711.40] CV RR=47%; Fulfills all requirements for ABEL: CV_{wR} >30%, CI within expanded limits, and PE within 0.8000–1.2500. Following the ‘logic’ of the EMA’s Q&A document at least 12 subjects should have finished the sequence TRR in order to get a ‘reliable’ estimate of CV_{wR}. I guess that the sample size was 34, right? If an assessor does not accept this design (not mentioned in the Q&A…) the study demonstrated even ABE. » GMR TT=0.708 not belong [0.80; 1.25]; » CI TT = 0.471.06; » CV TT =77%. » Is it possible to draw a conclusion about bioequivalence? Yes. By definition BE is the (desired) outcome of a comparison of T with R. The results of T vs. T are interesting but not relevant. BTW, how did you calculate the GMR and the CI? IMHO, if you use only the data of the sequence RTT (as for CV_{wT}) you can evaluate it only as a paired design (e.g., arbitrarily comparing data of the second administration to the first). Such an evaluation assumes no period effects – which might be false. — Regards, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. ☼ Science Quotes 
Yura Regular Belarus, 20170509 16:18 @ Helmut Posting: # 17335 Views: 1,254 

Hi Helmut GMR  the average difference T1T2 CI  Student's ttest for dependent samples (for 17) Results for TT does not take into account? Best regards 
Helmut Hero Vienna, Austria, 20170509 16:56 @ Yura Posting: # 17336 Views: 1,251 

Hi Yura, » GMR  the average difference T1T2 » CI  Student's ttest for dependent samples (for 17) I guess by T1 you mean T in period 2 and by T2 T in period 3. As said above such an evaluation assumes no period effects. Only if the period effect would be exactly zero, the difference in period means would be an unbiased estimate of the treatment effect. Your comparison is not correct. I assume when you run the same test on RR you will also get ‘strange’ results. » Results for TT does not take into account? Does not take what into account? — Regards, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. ☼ Science Quotes 
Yura Regular Belarus, 20170509 17:36 @ Helmut Posting: # 17337 Views: 1,234 

Hi Helmut I also conducted ANOVA with the factors subject and period. The same results are obtained as in the SAS code. For constructing CI TT MSE from this analysis? Best regards 
Helmut Hero Vienna, Austria, 20170509 17:53 @ Yura Posting: # 17338 Views: 1,217 

Hi Yura, » I also conducted ANOVA with the factors subject and period. » The same results are obtained as in the SAS code. Sure. » For constructing CI TT MSE from this analysis? You are taxing my patience. Heck, it is not possible – or better: any result is nonsense! — Regards, Helmut Schütz The quality of responses received is directly proportional to the quality of the question asked. ☼ Science Quotes 
Yura Regular Belarus, 20170509 18:07 @ Helmut Posting: # 17339 Views: 1,236 

HI Dear Helmut Got it I'll take a break Best regards 