(edited by Jay on 2018-03-26 13:50)
Posting: # 18595
In one of the full replicate study of NTID the results are as per below,
The ABE results are as below,
The point estimate is near to 100 and ABE is also within 80% to 125%.
But in SABE, the Cmax criteria is more than 0 i.e. 0.0001 which leads to bioinequivalence.
So can it be scientifically justified the 95%CI of Cmax value as its very borderline to 0 and T/R shows less difference between test and reference.
Edit: Category and subject line changed; see also this post #1 and #2. [Helmut]
Posting: # 18596
Did you check individual data? Any strange ones?
If FDA, "no" As in good luck in trying to convince them.
Posting: # 18628
Thanks, John for the reply...
I would like to understand that how come the study passing in ABE within 80% to 125%, low T/R and low ISCV would not meet the BE criteria of scaled average bioequivalence.
As we generally proceed with scaled approach when the ISCV is high (more than 30%) and would not meet BE criteria with ABE approach.
Please let me know your views on the same.
(edited by pjs on 2018-04-02 15:17)
Posting: # 18633
» I would like to understand that how come the study passing in ABE within 80% to 125%, low T/R and low ISCV would not meet the BE criteria of scaled average bioequivalence.
As per the Swr data shared scaled BE limits would be 95.87-104.30%. Now as per the ABE data you shared lower limit is mentioned to be 95%.
Now please note that variability in ABE and SABE would be different as there will be different models for the estimation. Also for SABE only subjects who had completed all four Periods would be included and in ABE any subject with atleast one reference and one test product would be included.
You may refer the lower limit for the ilat output in SABE approach which would be outside 95.87%. This could be reason behind the upper bound value marginally more than 0. Just one quick question in the outputs for upper bound, value to be rounded off till which decimal? If you consider three decimals study is passing and with four decimals same is failing.
Also with variability as low as 4%, you also need to check test product variability to comply with variability comparison criteria.
Just to check with forum members is there any method for outlier detection in such full replicate studies? In one of the earlier posts there was discussion for use of method proposed by Lazlo for the same. Any further update for the acceptability of this approach.
Can you pls confirm how many subjects were included in the study and how sample size was defined.
Posting: # 18639
Thanks for the reply.
In any two way ABE study, if the 90%CI is not within the range of 80-125%, most of time the same is reflected in the T/R and also the dissolution graphs.
For highly variable product, if the 95% CI is not ≤0, the T/R ratio would be very high and beyond the range of 80 to 125%.
For NTID as per USFDAs GL below three criteria should be followed,
- Upper 95% CI ≤0 (SABE)
- 90% CI should be within 80-125%
- UL of 90% CI for σWT/σWR should be ≤ 2.5 (ABE)
As per the case mentioned in above post, the ISCV of test and ref is very low (around 6-7), T/R is near to 100 and also meeting BE criteria in ABE. So if only 95% CI in SABE does not meet criteria (0.0001) so how the reason for the difference between test and reference can be interpreted.
Thanks in advance.
Posting: # 18640
» In any two way ABE study, if the 90%CI is not within the range of 80-125%, most of time the same is reflected in the T/R …
Well, the CI is constructed around the GMR.
» … and also the dissolution graphs.
Nope. Only if dissolution would be the rate-limiting step. Otherwise in vitro cannot (!) be predictive of in vivo. There a tons of studies with completely dissimilar dissolution profiles which passed BE easily.
» For NTID as per USFDAs GL below three criteria should be followed,
» - Upper 95% CI ≤0 (SABE)
» - 90% CI should be within 80-125%
» - UL of 90% CI for σWT/σWR should be ≤ 2.5 (ABE)
You are mixing something up. Actually:
- Upper 95% CI ≤0 (SABE)
- 90% CI should be within 80-125% (ABE)
- UL of 90% CI for σwT/σwR should be ≤ 2.5 (comparison of variabilities)
» As per the case mentioned in above post, the ISCV of test and ref is very low (around 6-7), T/R is near to 100 and also meeting BE criteria in ABE. So if only 95% CI in SABE does not meet criteria (0.0001) so how the reason for the difference between test and reference can be interpreted.
The FDA’s “implied BE limits” can be calculated by ℯ∓log(1.11111) · √log(CV²+1)/0.1 which are in your case:
Does the conventional ABE (90% CI) pass these narrower limits?
These three tests are not directly related. Fine if you pass the latter two. Bad luck if you fail the first by a small margin. How was the study powered? If you targeted a power of 80–90%, the chance of failing for a product which is BE would be 10–20%. Daily life. See John’s reply above.
Counterquestion: Did you ever succeed in convincing the FDA to accept a conventional BE study which showed a 90% CI of 79.99–125.01%?
The quality of responses received is directly proportional to the quality of the question asked. ☼