Bioequivalence and Bioavailability Forum

Sathya
☆

India,
2008-09-06 08:46
(6090 d 03:09 ago)

Posting: # 2324
Views: 9,681

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 (6089 d 17:06 ago) @ Sathya Posting: # 2330 Views: 7,993	BE parallel design Post reply
	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 (6088 d 06:30 ago) @ Ohlbe Posting: # 2332 Views: 7,995	BE parallel design Post reply
	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
(6088 d 00:42 ago)

(edited on 2008-09-08 11:46)
@ Ohlbe
Posting: # 2335
Views: 7,969

BE parallel design

Post reply

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
(6087 d 23:08 ago)

@ Sathya
Posting: # 2336
Views: 8,016

a posteriori power...

Post reply

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 (6087 d 22:04 ago) @ Jaime_R Posting: # 2337 Views: 8,013	a posteriori power... Post reply
	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
(6087 d 21:27 ago)

(edited on 2008-09-08 19:23)
@ Sathya
Posting: # 2338
Views: 8,030

a posteriori power...

Post reply

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 (6086 d 00:05 ago) @ Ohlbe Posting: # 2348 Views: 7,877	a posteriori power... Post reply
	Dear Ohlbe, Thank you for your response. I help me a lot. Once again thanks — Sathya

BE parallel design [Study As­sess­ment]

BE parallel design

BE parallel design

BE parallel design

a posteriori power...

a posteriori power...

a posteriori power...

a posteriori power...

BE parallel design [Study Assessment]