GM
★

India,
2018-09-12 07:17
(1382 d 21:00 ago)

Posting: # 19267
Views: 6,310

## Group Effect in Full Replicate Design [RSABE / ABEL]

Dear All,

Hope all are doing well.

We are conducting a BE study of Highly variable drug in patient population. This study will be conducted in 3 groups, which are seperated by one month from each other. There is no clarity in the guidance to check the group effect for this design.

Our doubt is, whether the group effect should be checked in the full replicate design or not?

If so, what are the fixed/random effets need to be checked in RSABE and ABE?

Kindly explain.

Best Regards,
GM
Helmut
★★★

Vienna, Austria,
2018-09-12 11:20
(1382 d 16:58 ago)

@ GM
Posting: # 19270
Views: 5,727

## Group Effect in Full Replicate Design

Hi Gm,

» […] BE study of Highly variable drug in […] 3 groups, which are seperated by one month from each other. There is no clarity in the guidance to check the group effect for this design.

Which agency’s guidance?

» Our doubt is, whether the group effect should be checked in the full replicate design or not?

What do you mean by “checked”? Have group – together with some nested terms – as part of the model (only report the MSEs, p-values) or do you additionally want to base any decision on it? That’s not a good idea, since power might be compromised and the Type I Error inflated (see this presentation).

» If so, what are the fixed/random effets need to be checked in RSABE and ABE?
• EMA: all effects fixed
• FDA: all effects fixed except subject(group × sequence) random

Dif-tor heh smusma 🖖
Helmut Schütz

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

India,
2018-09-12 12:32
(1382 d 15:46 ago)

@ Helmut
Posting: # 19271
Views: 5,757

## Group Effect in Full Replicate Design

Dear Helmut,

Thank you for the quick reply.

» Which agency’s guidance?

FDA Guidance only.

» What do you mean by “checked”? Have group – together with some nested terms – as part of the model (only report the MSEs, p-values) or do you additionally want to base any decision on it? That’s not a good idea, since power might be compromised and the Type I Error inflated (see this presentation).

Sorry for poor english.

» If so, what are the fixed/random effets need to be checked in RSABE and ABE?
»
• EMA: all effects fixed
»
• FDA: all effects fixed except subject(group × sequence) random

We understand the above and this is as same as 2x2 design. But we thought this is only for ABE approach.

For RSABE approch, only sequence effect will be sufficient in the model? or any group terms need to be include here.

Best Regards,
GM
Helmut
★★★

Vienna, Austria,
2018-09-12 13:08
(1382 d 15:10 ago)

@ GM
Posting: # 19273
Views: 5,894

## Group Effect in Full Replicate Design

Hi GM,

» » Which agency’s guidance?
»
» FDA Guidance only.

Sorry that I didn’t ask for the specific guidance. I guess you are referring to this one stating in Section VII.A.:

If a crossover study is carried out in two or more groups of subjects (e.g., if for logistical reasons only a limited number of subjects can be studied at one time), the statistical model should be modified to reflect the multigroup nature of the study. In particular, the model should reflect the fact that the periods for the first group are different from the periods for the second group. This applies to all of the approaches (average, population, and individual BE) described in this guidance.

Since studies for PBE and IBE are also performed in replicate designs, one can conclude that the same is applicable to RSABE.

» » What do you mean by “checked”? Have group – together with some nested terms – as part of the model (only report the MSEs, p-values) or do you additionally want to base any decision on it? That’s not a good idea, since power might be compromised and the Type I Error inflated (see this presentation).
»
» Sorry for poor english.

» For RSABE approch, only sequence effect will be sufficient in the model? or any group terms need to be include here.

Here you lost me. You will have a lot of effects… Honestly, I have no clue how to modify the models given in the progesterone-guidance.

Dif-tor heh smusma 🖖
Helmut Schütz

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

India,
2018-09-12 13:23
(1382 d 14:55 ago)

@ Helmut
Posting: # 19275
Views: 5,726

## Group Effect in Full Replicate Design

» » » What do you mean by “checked”? Have group – together with some nested terms – as part of the model (only report the MSEs, p-values) or do you additionally want to base any decision on it? That’s not a good idea, since power might be compromised and the Type I Error inflated (see this presentation).
» »
» » Sorry for poor english.
»
» No worries. Mine is hardly better. But please answer my question.

We don't want to take any decisions on the basis of model terms.

Our thought is, we need to check group effect for both ABE and RSABE also.

If this correct, how the model is to be designed for RSABE?

Best Regards,
GM
Helmut
★★★

Vienna, Austria,
2018-09-12 15:09
(1382 d 13:08 ago)

@ GM
Posting: # 19276
Views: 5,790

## Group Effect in Full Replicate Design

Hi Gm,

» We don't want to take any decisions on the basis of model terms.

Very good.

» Our thought is, we need to check group effect for both ABE and RSABE also.

Again: you have to include “group” as a factor in various ways (see below).

» If this correct, how the model is to be designed for RSABE?

As I wrote above

» » Honestly, I have no clue how to modify the models given in the progesterone-guidance.

Specifically the scaling-part…
For the ABE part (if swR <0.294) like the FDA’s Model II:
• fixed:
Group, Sequence, Treatment, Period (nested within Group), Group-by-Sequence Interaction
• random:
Subject (nested within Group × Sequence)

Dif-tor heh smusma 🖖
Helmut Schütz

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

India,
2018-09-12 15:19
(1382 d 12:58 ago)

@ Helmut
Posting: # 19277
Views: 5,789

## Group Effect in Full Replicate Design

Dear Helmut,

Thank you very much for the clarification. One more last question...

» » » Honestly, I have no clue how to modify the models given in the progesterone-guidance.
»
» Specifically the scaling-part…
» For the ABE part (if swR <0.294) like the FDA’s Model II:
• fixed:
» Group, Sequence, Treatment, Period (nested within Group), Group-by-Sequence Interaction
»
• random:
» Subject (nested within Group × Sequence)

But as per Progesterone Guidance, Treatment as random term in Average BE.

Please see model SAS code from guidance.

MODEL LAUCT = SEQ PER TRT/ DDFM=SATTERTH;
RANDOM TRT/TYPE=FA0(2) SUB=SUBJ G;

As per your version, Subject (nested within Group × Sequence) is the random term.

Now am Confused,Which is the correct one?

Best Regards,
GM
Helmut
★★★

Vienna, Austria,
2018-09-12 15:31
(1382 d 12:47 ago)

@ GM
Posting: # 19278
Views: 5,767

## Group Effect in Full Replicate Design

Hi GM,

» » For the ABE part (if swR <0.294) like the FDA’s Model II:
• fixed:
» » Group, Sequence, Treatment, Period (nested within Group), Group-by-Sequence Interaction
» »
• random:
» » Subject (nested within Group × Sequence)

»
» But as per Progesterone Guidance, Treatment as random term in Average BE.

No way.

» Please see model SAS code from guidance.
»
» MODEL LAUCT = SEQ PER TRT/ DDFM=SATTERTH;
» RANDOM TRT/TYPE=FA0(2) SUB=SUBJ G;

Please make yourself familiar with the syntax of SAS’ PROC MIXED. Terms in the MODEL argument are fixed.

» As per your version, Subject (nested within Group × Sequence) is the random term.
»
» Now am Confused,Which is the correct one?

Mine. For the exact setup ask a qualified SAS-coder.

Dif-tor heh smusma 🖖
Helmut Schütz

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

Plzeň, Czech Republic,
2018-09-12 23:30
(1382 d 04:47 ago)

@ Helmut
Posting: # 19282
Views: 5,649

## Group Effect in Replicate Design

Dear GM and Helmut,

many years ago there is an old PDF ... only maybe interesting for you.
But bear in mind it is very old as data for the analyses were provided on computer diskette.

And it wasn't full replicate design but only partial replicate design. Nevertheless the subjects were treated in groups (day after day) and the model was modified accordingly.

Best regards,
zizou
Helmut
★★★

Vienna, Austria,
2018-09-13 11:34
(1381 d 16:43 ago)

@ zizou
Posting: # 19286
Views: 5,679

## Group Effect in Replicate Design

Hi zizou,

» many years ago there is an old PDF ... only maybe interesting for you.

THX – very interesting indeed!

Dif-tor heh smusma 🖖
Helmut Schütz

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