# Bioequivalence and Bioavailability Forum

Back to the forum  2018-07-23 00:43 CEST (UTC+2h)
verstc
Junior

China,
2018-02-26 10:15

Posting: # 18462
Views: 1,312

## intra-subject variability from CI [General Sta­tis­tics]

Dear All,

From the forum, I found the way to calculate intra-subject variability from confidence interval.

I am wondering how I cite this calculation or anyone can tell me where this calculation originate from?

Thanks,
Shawn
d_labes
Hero

Berlin, Germany,
2018-02-26 12:26

@ verstc
Posting: # 18464
Views: 1,154

## intra-subject variability from CI

Dear shawn,

» From the forum, I found the way to calculate intra-subject variability from confidence interval.

Fine for you. If you are interested in a software for doing this, habe a look into R package `PowerTOST` / function `CI2CV()`.

» I am wondering how I cite this calculation or anyone can tell me where this calculation originate from?

This is a little bit algebra based on the formula of the CI. I don't think that there is a paper dealing with this algebra. I further think there is no need to have a citation for that.

IIRC the little algebra was first mentioned in Helmut's lecture for the conference on "Bioequivalence and Bioavailability" Budapest 2009.

Regards,

Detlew
verstc
Junior

China,
2018-02-26 13:41

@ d_labes
Posting: # 18465
Views: 1,118

## intra-subject variability from CI

Dear Detlew

best regards,

Shawn

Edit: Full quote removed. Please delete everything from the text of the original poster which is not necessary in understanding your answer; see also this post #5! [Helmut]
Helmut
Hero

Vienna, Austria,
2018-02-26 13:51

@ verstc
Posting: # 18466
Views: 1,117

## intra-subject variability from CI

Hi Shawn,

no rocket science but basic algebra an 8th-grade pupil should be able to handle. I derived the formula about thirty years ago.  Hence, no reference…
For background see this post and followings. If you want a reference for such a trivial calculation see this post.

Cheers,
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes
Louis52
Junior

2018-03-09 14:27

@ Helmut
Posting: # 18510
Views: 688

## intra-subject variability from CI

On this subject of finding the CV intra based on info from a previous study/paper, is there a function in Power.TOST or other package that would be something like CVfromCI but based on a DDI rather than BE? Let's say a DDI with one sequence for which we know the PE and CLs for the GMR (test vs. reference).
Helmut
Hero

Vienna, Austria,
2018-03-09 14:39

@ Louis52
Posting: # 18511
Views: 682

## known.designs()

Hi Louis,

» […] something like CVfromCI but based on a DDI rather than BE? Let's say a DDI with one sequence for which we know the PE and CLs for the GMR (test vs. reference).

One sequence? Are you thinking about a paired design? IMHO, DDI studies should be performed as crossovers. In paired designs you have to assume (!) that there are no period effects. However, if you have data of such a study, use e.g.,

```library(PowerTOST) CVfromCI(lower=0.89, upper=1.15, n=24, design="paired") # [1] 0.2634199```

BTW, the function `CVfromCI()` (alias `CI2CV()`) supports all designs listed in `known.designs()`.

Cheers,
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes
Louis52
Junior

2018-03-09 15:13

@ Helmut
Posting: # 18512
Views: 669

## known.designs()

Hi Helmut,

I thought about the 'paired' design as I looked into known designs(), but I was not sure. Indeed the DDIs are COs but in several instances it happens that in period 1 a subject receives what is called the 'probe' while in period 2 it receives the "probe+investigational drug". I'm trying to find the CV intra associated with a particular "probe" based on DDI study in a published paper (no CV intra, just PE, CI, and N), as I need to use it in the sample size determination (precision only, no power) for a new DDI. Would 'paired' cover this situation? Of course, I can try to do it manually, but if it is already out there,...then why not use it.
Helmut
Hero

Vienna, Austria,
2018-03-09 15:42

@ Louis52
Posting: # 18513
Views: 688

## known.designs()

Hi Louis,

» I thought about the 'paired' design as I looked into known designs(), but I was not sure.

You were guessing right.

» […] in several instances it happens that in period 1 a subject receives what is called the 'probe' while in period 2 it receives the "probe+investigational drug".

Designed without consulting a statistician? Lacking period effects is a strong assumption.
The only designs where a paired design is common is assessing linear PK of a new drug. Such a study starts with a single dose followed by saturation and another profile in (pseudo) steady state. Then we compare AUC0–τ (MD) with AUC0–∞ (SD). Performing such a study as a crossover would be a logistic nightmare. Never ever have seen one.

» I'm trying to find the CV intra associated with a particular "probe" based on DDI study in a published paper (no CV intra, just PE, CI, and N), as I need to use it in the sample size determination (precision only, no power) for a new DDI. Would 'paired' cover this situation?

Yes.

Cheers,
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. ☼
Science Quotes
Bioequivalence and Bioavailability Forum |  Admin contact
18,547 posts in 3,941 threads, 1,190 registered users;
online 14 (0 registered, 14 guests [including 13 identified bots]).

[The] impatience with ambiguity can be criticized in the phrase:
absence of evidence is not evidence of absence.    Carl Sagan

The BIOEQUIVALENCE / BIOAVAILABILITY FORUM is hosted by
Ing. Helmut Schütz