felipeberlinski
☆    

Brazil,
2014-10-22 16:20
(3445 d 21:47 ago)

Posting: # 13763
Views: 14,575
 

 3x3 or 2x4? [RSABE / ABEL]

Dear friends

Reading a BE reccomendation of FDA I would like to know about your experience using two different designs.

Since the drug is a HVD:
Their reccomendation is to use a 3-sequence 3-period or a 2-sequence 4 period design.

Could anybody explain the main differences between them and when should I use each one? Go easy, I'm not a biostatistician :cool:

Thanks in advance


Edit: Category changed. [Helmut]
ElMaestro
★★★

Denmark,
2014-10-22 16:42
(3445 d 21:26 ago)

@ felipeberlinski
Posting: # 13764
Views: 13,204
 

 3x3 or 2x4?

Hi Felipe,

❝ Since the drug is a HVD:

❝ Their reccomendation is to use a 3-sequence 3-period or a 2-sequence 4 period design.


❝ Could anybody explain the main differences between them and when should I use each one? Go easy, I'm not a biostatistician :cool:


When you do a replicated study for the purpose of proving BE by means of reference scaling it is of primary interest to you to have two administrations on Ref. and one administration of Test within each subjects.
Thus it isn't necessary to have two administrations of Test within each subject. Therefore I would personally prefer a 3-period, 3-sequence study. That designs is also lighter on the subjects - they undergo treatment (exposure) only in three periods and not four and that is arguably an ethical advantage.

I'll say this a little differently (principle applies to both the EU and US version, although there are subtle other difference in how it works under the hood, especially since the CI is implicit in the US). The three period design gives you:
  1. The treatment effects of Test and Ref. You use this to derive the point estimate.
  2. The variability of the point estimate. This determines the width of your CI.
  3. The within-subject variability associated with the reference product. This one determines the acceptance range for the confidence interval.
    The four-period design gives you all of the above plus
  4. The within-subject variability for the Test product. No-one asks for this and to the best of my knowledge there is no way to use that estimate towards proving BE.

Pass or fail!
ElMaestro
nobody
nothing

2014-10-22 17:06
(3445 d 21:01 ago)

@ ElMaestro
Posting: # 13765
Views: 13,053
 

 3x3 or 2x4?

Don't forget: find someone to calculate that kind of stuff. FDA differntiates between "normal" HV drugs (progesterone guidance document) and narrow therapeutic index Hv dugs (warfarin document) iirc...

Kindest regards, nobody
felipeberlinski
☆    

Brazil,
2014-10-22 17:36
(3445 d 20:31 ago)

@ nobody
Posting: # 13766
Views: 13,060
 

 3x3 or 2x4?

Thank you both for the help and explanation.

I fully understood your point on exposing subjects, and agree with you regarding the ethical question.

But talking a little about sample size, does a 3x3 is advantageous
on this matter? Which one will require a smaller number of subjects involved?


Thank you again
nobody
nothing

2014-10-22 17:41
(3445 d 20:26 ago)

@ felipeberlinski
Posting: # 13767
Views: 13,046
 

 3x3 or 2x4?

...as a starter:

http://bebac.at/lectures/Mumbai2013WS1.pdf

page 17 ;-)

...but drop outs may increase if you have to dose one individual FOUR times, with all the wash-out and potential adverse events...

Kindest regards, nobody
d_labes
★★★

Berlin, Germany,
2014-10-22 18:12
(3445 d 19:55 ago)

@ felipeberlinski
Posting: # 13768
Views: 13,431
 

 2x3x3 or 2x2x3 or 2x2x4?

Dear Felipe,

❝ But talking a little about sample size, ...


Since your original question refers to FDA lets see:
(using R with package PowerTOST, true GMR=0.95, target power=80%, sample size for reference scaled ABE using scABE criterion according to progesterone guidance)

> library(PowerTOST)
> sampleN.RSABE(CV=0.4, design="2x3x3")

++++++++ Reference scaled ABE crit. +++++++++
           Sample size estimation
---------------------------------------------
Study design:  2x3x3 (partial replicate)
log-transformed data (multiplicative model)
1e+05 studies for each step simulated.

alpha  = 0.05, target power = 0.8
CVw(T) = 0.4; CVw(R) = 0.4
Null (true) ratio = 0.95
ABE limits / PE constraints = 0.8 ... 1.25
Regulatory settings: FDA
- CVswitch =  0.3
- Regulatory constant = 0.8925742

Sample size search
 n     power
21   0.73975
24   0.80864

> sampleN.RSABE(CV=0.4, design="2x2x3")

++++++++ Reference scaled ABE crit. +++++++++
           Sample size estimation
---------------------------------------------
Study design:  2x2x3 (TRT|RTR)
log-transformed data (multiplicative model)
1e+05 studies for each step simulated.

alpha  = 0.05, target power = 0.8
CVw(T) = 0.4; CVw(R) = 0.4
Null (true) ratio = 0.95
ABE limits / PE constraints = 0.8 ... 1.25
Regulatory settings: FDA
- CVswitch =  0.3
- Regulatory constant = 0.8925742

Sample size search
 n     power
20   0.66022
22   0.71427
24   0.76072
26   0.79722
28   0.83092

> sampleN.RSABE(CV=0.4, design="2x2x4")

++++++++ Reference scaled ABE crit. +++++++++
           Sample size estimation
---------------------------------------------
Study design:  2x2x4 (full replicate)
log-transformed data (multiplicative model)
1e+05 studies for each step simulated.

alpha  = 0.05, target power = 0.8
CVw(T) = 0.4; CVw(R) = 0.4
Null (true) ratio = 0.95
ABE limits / PE constraints = 0.8 ... 1.25
Regulatory settings: FDA
- CVswitch =  0.3
- Regulatory constant = 0.8925742

Sample size search
 n     power
14   0.68849
16   0.76590
18   0.82274


The number of measurements (concentration samples) are
2x3x3 = 72 * number of timepoints
2x2x3 = 84 * number of timepoints
2x2x4 = 72 * number of timepoints

The design 2x2x3 needs somewhat higher sample size than the partial replicate but has the advantage of yielding also an intra-subject variability for Test. But this is a "nice to have".

Hope this helps.

Regards,

Detlew
MGR
★    

India,
2015-04-08 15:35
(3277 d 22:32 ago)

@ d_labes
Posting: # 14675
Views: 11,762
 

 2x3x3 or 2x2x3 or 2x2x4?

Dear d_Labes,

❝ (using R with package PowerTOST, true GMR=0.95, target power=80%, sample size for reference scaled ABE using scABE criterion according to progesterone guidance)


As per the above, can we do the sample size in SAS software, if so can we use pairedmeans option in procpower or anything?

Could you please help me in this regards?

Thanks in advance.

Regards,
MGR
d_labes
★★★

Berlin, Germany,
2015-04-08 18:06
(3277 d 20:02 ago)

@ MGR
Posting: # 14676
Views: 11,837
 

 Sample size for scABE in SAS?

Dear MGR,

❝ As per the above, can we do the sample size in SAS software, if so can we use pairedmeans option in procpower or anything?


Sorry :no:. You have to do the power calculations via simulations because the methods underlying the scaled ABE have no (simple or complicated) analytical solution. And there is no Proc in SAS available to do this out of the box.

In principle the simulations could be programmed in SAS, but IMHO this is a horrible task. Moreover the run-time expected is very slow.
For that reasons (ease of programming, run-time) I had chosen the language R to implement the power/sample size estimation for scaled ABE, also my daily job is struggling with the beast SAS.

R is free available at CRAN, package PowerTOST also.
Both are relative easy to install. See this post.

Thus I recommend you to use that way to be able to do the sample size estimation for scaled ABE.

Regards,

Detlew
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-04-08 18:19
(3277 d 19:48 ago)

@ MGR
Posting: # 14677
Views: 11,720
 

 Re-inventing the wheel

Hi MGR,

adding to what Detlew said: See the paper* by the two Lászlós about simulations (done in MatLab). When asked for their code they recommended PowerTOST instead. ;-)


  • Endrényi L, Tóthfalusi L. Sample Sizes for Designing Bioequivalence Studies for Highly Variable Drugs. J Pharm Pharmaceut Sci. 2011;15(1):73–84. [image] free resource.

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
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2014-10-22 18:36
(3445 d 19:31 ago)

@ felipeberlinski
Posting: # 13769
Views: 13,167
 

 RTR|TRT

Hi Felipe,

I oppose our Master when he wrote:

Thus it isn't necessary to have two administrations of Test within each subject.

Correct, if you have only CVWR needed for scaling in mind. But – if CVWR<30% – FDA’s mixed-effects code fails sometimes in partial replicated studies (TRR|RRT|RTR) since the model is over-specifed. Either parameterize the covariance structure as FA0(1) – instead of TYPE=FA0(2) given in the guidance – or perform a fully replicated three-period design (RTR|TRT). This design seemingly con­verges always and additionally gives you information about your product.

Sample sizes for FDA’s RSABE (T/R 90%, 80% power):
CV%          20  30  40  50  80
───────────────────────────────
RTRT|TRTR    32  24  24  22  28
TRR|RRT|RTR  30  45  33  30  42
RTR|TRT      28  46  38  34  44


I would not recommend a four-period study (more blood draws / subjects, higher drop-out rate).

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
Mauricio Sampaio
★    

Brazil,
2014-10-28 13:12
(3439 d 23:56 ago)

@ felipeberlinski
Posting: # 13810
Views: 12,973
 

 3x3 or 2x4?

Hi Felipe! How are you my friend?

In my opinion, as the my focus is generics drugs products, I suggest the implementation of a reference scaled average bioequivalence using a 4x2 design. Because with test and reference formulations replicated the statistical adjustment is more consistent. Because you have more information from two medicines. In addition, the 4x2 design gives you a real information about your product that can be used in an internal discussion to improve your formulation in case of bioequivalence study fail.

More information read:

Implementation of a Reference Scaled Average Bioequivalence Approach for Highly Variable Generic Drug Products by the US Food and Drug Administration.
The AAPS Journal, Vol.14, No.4, December2012
doi 10.1208/s12248-012-9406-x

Best regards
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
98 visitors (0 registered, 98 guests [including 7 identified bots]).
Forum time: 13:08 CET (Europe/Vienna)

With four parameters I can fit an elephant,
and with five I can make him wiggle his trunk.    John von Neumann

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