Sathya
☆    

India,
2008-09-06 08:46
(5683 d 07:32 ago)

Posting: # 2324
Views: 8,210
 

 BE parallel design [Study As­sess­ment]

Dear pkpdpkpd, Jaime and all,

I am a beginner to this Bioequivalence Study.

Your discussion about Parallel bioequivalence is very helpful to me.

This is the result which i got for one parallel study

pointestimate Upper lower
91.1244 109.417 75.8897
88.1705 105.443 73.7278
95.8295 106.065 86.5814

Please Clarify,

in below two condition which one i used to conclude the bioequivalence.(Parallel BE)

80 < lower and upper < 125 and
80 < point estimate < 125 and also
lower < point estimate < upper then bioequivalent = yes.

or

80 < point estimate < 125 and also
lower < point estimate < upper then bioequivalent = yes.

Other wise please help me how to conclude bioequivalence of the above said results.

Please Help me.


Edit: Category changed. [Helmut]

Sathya
Ohlbe
★★★

France,
2008-09-06 18:48
(5682 d 21:30 ago)

@ Sathya
Posting: # 2330
Views: 6,800
 

 BE parallel design

Dear Sathya,

The 90% CI is built around the point estimate ! You will always have lower < point estimate < upper.

The only criterion to consider is 80 < lower and upper < 125.

Regards
Ohlbe
Sathya
☆    

India,
2008-09-08 05:24
(5681 d 10:53 ago)

@ Ohlbe
Posting: # 2332
Views: 6,805
 

 BE parallel design

Dear Ohlbe,

Thank you for your response.

So my results does not meet the criteria for bioequivalence.Because results has lower of both AUCtot and AUclast is less than 80. but lower of CMax is greater than 80.

In this case, what i have to conclude? please help.

Sathya
Sathya
☆    

India,
2008-09-08 11:12
(5681 d 05:06 ago)

(edited by Sathya on 2008-09-08 11:46)
@ Ohlbe
Posting: # 2335
Views: 6,797
 

 BE parallel design

Dear ohlbe,

Thank you for your response.

I got sample SAS report of Cross over study. I tried the same dataset. I got all the values exactly except power. I tried different Calculation and formula. But unable to get the exact answer. Please help.

Here i attached required result and also my power calculations.

Power Calculations Using Different Formulae:-

Required Output:

Parameter Lcmax
lsmref    361.94
lsmtest   346.85
DF        22
mse       0.021091
se        0.041924
diff      -15.095
ratio     95.8295
intra_cv  14.5997
power     99.8158
lower     89.1733
upper     102.983
smanp1    0.000143
smanp2    1.113E-06



Method:

data Power2;
Estimate = -0.04260;
StdErr  = 0.04192;
DF = 22;
delta=log(1.25);
test1=(estimate+delta)/stderr;
test2=(estimate-delta)/stderr;
probt1=PROBT(test1,df,0);
probt2=PROBT(test2,df,0);
p1=1-probt1;
p2=probt2;
ANSWER =probt1 - probt2;
Power = answer*100;
run;


output

Obs  Estimate  StdErr   DF  delta    test1    test2     probt1   probt2
1    0.0426    0.04192  22  0.22314  6.33930  -4.30686  1.00000  .000142565
p1          p2          ANSWER  Power
.000001112  .000142565  0.99986 99.9856

Sathya
Jaime_R
★★  

Barcelona,
2008-09-08 12:46
(5681 d 03:32 ago)

@ Sathya
Posting: # 2336
Views: 6,846
 

 a posteriori power...

Dear Sathya,

since you have posted similar questions in various threads, please search the forum first.
If you have seen a posteriori power in reports, simply ignore them - don't try to reproduce a meaningless result. As an entry point have a look at one of Helmut's posts.
I'm neither a statistician nor a SAS-user, but I guess there's a problem with the noncentrality-parameter you used (0 works with the normal t-distribution).

Regards, Jaime
Sathya
☆    

India,
2008-09-08 13:51
(5681 d 02:27 ago)

@ Jaime_R
Posting: # 2337
Views: 6,845
 

 a posteriori power...

Dear Jaime,

Will you give me the importance of power in bioequivalence analysis.

Shall we submit the report without power calculation

Whether the power calculation is same for both parallel and crossover study

Please help me.

Sathya
Ohlbe
★★★

France,
2008-09-08 14:27
(5681 d 01:50 ago)

(edited by Ohlbe on 2008-09-08 19:23)
@ Sathya
Posting: # 2338
Views: 6,822
 

 a posteriori power...

Dear Sathya,

To summarise: calculating post-hoc power makes no sense and has no interest in BE trials. Unless it is specifically requested by the guideline of the country in which you intend to submit your trial data, just forget about it and stop losing your time trying to calculate it.

The only question once your trial is completed is: is bioequivalence demonstrated or not ? Whether your 90 % CI is included within 80-125 with 60 % power or 95 % power makes no difference in the end: you demonstrated bioequivalence, that's all ! You were just more lucky not to fail in the first case. That's why we are calculating 90 % CI and not trying to show a difference between formulations. The alpha risk is fixed and is the patient's risk; the beta risk is for the sponsor.

If you failed to demonstrate bioequivalence, then you can start wondering whether you had enough power to demonstrate it. But here again calculating post-hoc power is of no interest: rather check whether your pre-study assumptions used to calculate the number of subjects were valid or not (difference between test and reference, intra-subject CV in a cross-over study).

Regards
Ohlbe
Sathya
☆    

India,
2008-09-10 11:49
(5679 d 04:29 ago)

@ Ohlbe
Posting: # 2348
Views: 6,686
 

 a posteriori power...

Dear Ohlbe,

Thank you for your response.

I help me a lot.

Once again thanks

Sathya
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
132 visitors (0 registered, 132 guests [including 5 identified bots]).
Forum time: 15:18 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