Add-on Studies / Sequential Design [Design Issues]

posted by Helmut Homepage – Vienna, Austria, 2006-08-16 16:05 (6257 d 04:23 ago) – Posting: # 219
Views: 9,147

Dear Joy,

sorry for the late reply ;-)

❝ 1. Are these acceptable in bioequivalence studies if specified a priori in the protocol? Would you know if US FDA, EMEA, Canada, ASEAN accept such design?

❝ 2. When an add-on study is conducted (and combined with the original study), what would be the statistical method of analysis?

❝ 3. Can we still apply the 90% Confidence Interval criteria for bioequivalence or would it have to be more restrictive i.e. 95% CI?

First a little review of international guidelines (maybe answering you questions #1-#3)

If BE is not shown, additional subjects are included. If an F-test (equality of variances) shows no significance, you may go for a pooled analysis. No alpha-adjustment needed.
If BE is not shown, a second part with sample size of at least half of the 1st part is initiated. Pooled analysis without alpha-adjustment.
South Africa
Maximum sample size must be stated a-priori; for details see Ahmed's post and the followings.
New Zealand
Group Sequential Design with alpha-adjustment (Gould's* method?).
No way :-(

Second some notes on the European practice.

Add-On and Group Sequential Designs are not covered in the NfG on BA/BE.
Group sequential designs are standard in clinical research.
Although discussed at BioInternationals '89 to '96, no consensus about their application in BE was reached.

Personal Experience:
A proposed method* was not accepted in the planning phase (three cases by the German BfArM).
Gould's method is an adaption of 'classical' procedures in clinical research (e.g., P. Armitage, 1975; J. Whitehead, 1992).
With such a method you are penalized for 'looking' at the data, i.e., the combined sample size is always larger than the one of an optimal fixed-sized study.
Unfortunatelly both the deviation from the reference and the variability are not known with certainty; therefore such an optimal study may be an illusion (i.e., in the long run less subjects are needed with Group Sequential Designs).

The following Add-On design was accepted in the EU (two cases Germany, one case France):
Evaluation of first part by an independent statistician (calculation of CV only!), performance of a second part, evaluation of pooled data without alpha-adjustment (by another statistician). The maximum additional sample size (stop criterion) was stated in the protocol. Drawback: since no point estimate must be calculated for the first part, such a study may fail even after pooling (if your assumptions are not met).

❝ 4. Could the first study be discarded if the second study passed the 90% CI BE criteria on its own?

At the first glance I saw no practical application. But it may be possible (as always if stated a-priori) if the F-test (Canada!) rejects pooling, and the second part is large enough...

❝ 5. More importantly, what are its pros and cons with regards to consumer protection or risk.

With the exception of US-FDA, all other regulators are answering with a 'Positive Maybe' ;-)
I think it's hypocritical accepting Group Sequential Designs (or even Adaptive Designs) in phase III, but rejecting them in BE...

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz

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

Complete thread:

UA Flag
 Admin contact
22,771 posts in 4,777 threads, 1,627 registered users;
12 visitors (0 registered, 12 guests [including 5 identified bots]).
Forum time: 20:28 CEST (Europe/Vienna)

The real struggle is not between the right and the left
but between the party of the thoughtful
and the party of the jerks.    Jimmy Wales

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz