martin
★★  

Austria,
2020-07-29 18:57
(1338 d 04:44 ago)

Posting: # 21798
Views: 2,952
 

 long half-life and large variability [Regulatives / Guidelines]

Dear all,

I would appreciate the forum members thoughts on the following situation.

In short:
  • pretty long half-life (i.e. replicate designs are out of scope)
  • huge total variablity (i.e. within + between variability)
  • total variability is expected to be explained to some degree by some covariates (i.e. including these covariates in the model is expected to lead to a feasable study size)
My interpretation of regulatoy guidelines: adding additional fixed effect factors to the factors mentioned in the guidelines is typically considered as not acceptable by regulators.

best regards & looking forward to some ideas to potentially address this situation

martin
ElMaestro
★★★

Denmark,
2020-07-29 23:24
(1338 d 00:17 ago)

@ martin
Posting: # 21800
Views: 2,464
 

 long half-life and large variability

Hi martin,

❝ My interpretation of regulatoy guidelines: adding additional fixed effect factors to the factors mentioned in the guidelines is typically considered as not acceptable by regulators.


There was not a lot of concrete info in your post, but in my humble experience you are totally right where ordinary BA/BE is concerned: regulators rarely allow deviations in pivotal trials.

However, milder winds are blowing when we talk BE for biologics in EU. Here covariates found their way into the melée because many companies suggested it (and they did so because FDA wanted covariates).

I don't from the top of my head remember a case where an applicant wanted a "less obvious" fixed effect in the proposed model and was outright successful.
The new/creative/funky/funny fixed factors I have come across recently were in crossovers where we always have subjects fit as fixed. In such cases, when the proposed new fixed factors are subject-specific they don't really do anything in terms of the residual or its df's. For example, if we specify CYP phenotype as fixed factor, columns for phenotype in the model matrix would add up exactly to an intercept, so we wouldn't be going anywhere.

Pass or fail!
ElMaestro
martin
★★  

Austria,
2020-07-30 00:49
(1337 d 22:52 ago)

@ ElMaestro
Posting: # 21801
Views: 2,438
 

 long half-life and large variability

dear ElMaestro,

Thank you for your swift reply and sharing your thoughts.

There are two types of covariates (type 1: continous = subject specfic values; and type 2: categorical variable with a few levels).

At the first glance a different regulatory pathway might be worth considering based on a PK or exposure similarity evalulation as in such a case there could be more flexibility regarding acceptance of statistical models used to test for equivalence in PK metrics.

Happy to get your thoughts on this idea.

best regards

Martin
ElMaestro
★★★

Denmark,
2020-07-30 10:17
(1337 d 13:24 ago)

@ martin
Posting: # 21802
Views: 2,427
 

 long half-life and large variability

Hi mrtin,

❝ There are two types of covariates (type 1: continous = subject specfic values; and type 2: categorical variable with a few levels).


Type 2 sounds like the very thing that I would call a fixed factor. :-)

❝ At the first glance a different regulatory pathway might be worth considering based on a PK or exposure similarity evalulation as in such a case there could be more flexibility regarding acceptance of statistical models used to test for equivalence in PK metrics.


❝ Happy to get your thoughts on this idea.


I would always recommend to approach regulators for a discussion. They will always listen but not always agree.

Injectable? Sometimes surface area or body weight is reasonable for exposure modeling. That's a covariate in the word's best sense :-) As in, a continuous variable. If multiple periods or longitudinal study, a covariates like bw may change over time.

Pass or fail!
ElMaestro
dshah
★★  

India/United Kingdom,
2020-08-10 08:18
(1326 d 15:23 ago)

@ ElMaestro
Posting: # 21846
Views: 2,274
 

 long half-life and large variability

Hi All!

I believe there are few MR drug with high variability and still USFDA recommends partial or fully replicate design. Mesalamine PSG or
Mesalamine PSG for 800 mg strength
Considering the Mechanism of Action, PK variability, non-Linearity- the study and formultion could indeed be challenging.

Regards,
Dshah
Achievwin
★    

US,
2020-08-12 17:22
(1324 d 06:19 ago)

@ martin
Posting: # 21864
Views: 1,740
 

 long half-life and large variability

Welcome to the home of funky and funny pharmacokinetics where "Garbage in Gospel out" is day of life.

1. allowing long washout periods have seen studies with 4 -5 months washout.
2. other choice we can imagine is Two stage adaptive study design and including ANCOA in place of ANOVA...

want to know what other experts think?

❝ In short:

  • pretty long half-life

  • huge total variablity (i.e. within + between variability)
  • total variability is expected to be explained to some degree by some covariates
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2020-08-12 17:53
(1324 d 05:48 ago)

@ Achievwin
Posting: # 21865
Views: 1,748
 

 long half-life and large variability

Hi Achievwin,

❝ Welcome to the home of funky and funny pharmacokinetics where "Garbage in Gospel out" is day of life.


:-D

❝ 1. allowing long washout periods have seen studies with 4 -5 months washout.


Seen a couple as well.

❝ 2. other choice we can imagine is Two stage adaptive study design and including ANCOA in place of ANOVA...


Slippery grounds – nothing published. You are on your own to demonstrate that the Type I Error is controlled. For the FDA and Health Canada likely simulations are sufficient. For the EMA no way.
I’m a fan of TSDs but if you proceed to the second stage it more than doubles the time of study compared to a fixed sample design. With fast to moderate half lives not an issue. But here?

Furthermore, reference-scaling for HVD(P)s in a TSD is not possible. Some tried, all studies were rejected due to potential inflation of the TIE.

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

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
78 visitors (0 registered, 78 guests [including 4 identified bots]).
Forum time: 22:41 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5