- On 22 Jul 2003 at 10:39:49, Rostam Namdari (chista90.-a-.yahoo.com) sent the message

Back to the Top

The following message was posted to: PharmPK

Dear Colleagues,

For comparing PK parameters between two cycles we

proposed to look at the 95% confidence intervals

associated with the least square geometric mean (LSGM)

for the Cl, AUC, MRT and Vss and if contained the

value of 1 then conclude no statistically significant

difference. Do we need to declare the width of the CI

first and if the CI falls within that declared

interval, then conclude that the two Cycles are

equivalent. What should be the width (80%-125% or

others)? Is it necessary to include 1.0 in the

interval? What about just looking at p-values?

Rostam - On 22 Jul 2003 at 15:48:08, "Bonate, Peter" (pbonate.-a-.ilexonc.com) sent the message

Back to the Top

The following message was posted to: PharmPK

Dear Rostam,

You asked the following:

For comparing PK parameters between two cycles we proposed to look at

the 95% confidence intervals associated with the least square geometric

mean (LSGM) for the Cl, AUC, MRT and Vss and if contained the value of

1 then conclude no statistically significant difference. Do we need to

declare the width of the CI first and if the CI falls within that

declared interval, then conclude that the two Cycles are equivalent.

What should be the width (80%-125% or others)? Is it necessary to

include 1.0 in the interval? What about just looking at p-values?

I would not look at p-values because these are meaningless. First,

they test the hypothesis that the means are different. This is a silly

null hypothesis because of course they will be different. The question

is, are they clinically or meaningfully different. For that you need a

confidence interval approach. That is the rationale for the whole CI

approach in bioequivalence. Looking only at whether the CI contains 1

is not very informative either and can be quite misleading. Suppose

your CI is 0.1 to 10. This interval does contain 1, but obviously the

standard error of the mean is so large that the estimate is very

imprecise. This might happen if the sample size is small. So, yes, I

would a priori define a acceptance interval that if my entire CI is

contained within I would declare the two means to be equivalent.

I would just use what some of the guidance's recommend: 90% CI

contained within 80 to 125% or 70 to 143%, depending on the therapeutic

window of the drug. If you don't know what the window is go with 80 to

125%.

Good luck,

Peter Bonate - On 23 Jul 2003 at 07:35:08, Priti Pandey (priti_pandey.at.yahoo.com) sent the message

Back to the Top

Dear Mr. Rostam Namdari,

If you are going for95% CI calculation then you need to write range(or

width)of your CI, means upper limit and lower limit of interval and if

interval include 1 mean there is no statistically significant

difference and if 1 comes in between your interval then definitely you

have to include that. I don’t have idea about what should be the width

of interval (it again depends upon regulatory agency) . If you look for

only "p" value means you will get only conclusion about, “Is there any

significant difference between two cycle or not?” But when you look for

CI means you will evaluate spread of your observation around your mean

value and can evaluate strength of significance also. From CI, you can

evaluate how much variation is there in your observation.

Regards

Priti

Biostatistician

Aurobindo Pharma, Hyderabad, India

Want to post a follow-up message on this topic? If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Difference between PK cycles" as the subject

PharmPK Discussion List Archive Index page

Copyright 1995-2010 David W. A. Bourne (david@boomer.org)