john john
☆    

Spain,
2015-05-13 11:36
(2030 d 07:28 ago)

Posting: # 14803
Views: 12,074
 

 Extrapolation [Power / Sample Size]

hi

Knowing the intrasubject variability is there any way to extrapolate the intersubject variability? Thanks.


Edit: Category changed; see also this post #1[Helmut]
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-05-13 12:33
(2030 d 06:31 ago)

@ john john
Posting: # 14804
Views: 11,026
 

 CVinter (or CVtotal) from CVintra

Hi John John,

» Knowing the intrasubject variability is there any way to extrapolate the intersubject variability?

No. There is no relationship between those. For examples see the end of this post. Please forget CVinter. In planning a crossover study you need CVintra and for a parallel study CVtotal.

[image]In 2012 I reviewed a manuscript claiming such a relationship (well: The other way ’round – estimating CVintra from CVtotal). All [sic] reviewers rejected the manuscript.
Obviously the authors were more successful in a journal with a lower impact factor… I think that it’s crap, but if you want to have a look:

Ramírez E, Abraira V, Guerra P, Borobia AM, Duque B, López JL, Mosquera B, Lubomirov R, Carcas AJ, Frías J. A Preliminary Model to Avoid the Overestimation of Sample Size in Bioequivalence Studies. Drug Res. 2013;63(2):98–103. doi:10.1055/s-0032-1333296.


Dif-tor heh smusma 🖖
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
nobody
nothing

2015-05-13 12:53
(2030 d 06:11 ago)

@ Helmut
Posting: # 14805
Views: 10,903
 

 CVinter (or CVtotal) from CVintra

OT: Again Drug Res. ? :-O notorious...

Kindest regards, nobody
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-05-13 13:11
(2030 d 05:53 ago)

@ nobody
Posting: # 14806
Views: 10,908
 

 OT: Drug Res

Hi nobody,

» OT: Again Drug Res. ? :-O notorious...

Yessir. Decades ago “Arzneimittelforschung” (founded in 1951!) had a good reputation. Gone with the wind (IF 0.507).
The authors of the MS tried it first at a journal with an IF of 2.037. ;-)

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
nobody
nothing

2015-05-13 13:18
(2030 d 05:46 ago)

@ Helmut
Posting: # 14807
Views: 10,837
 

 OT: Drug Res

When I was young Arznei-Frosch (sic!) was the standard for some things (at least in Germany, but also internationally...) Times change (quickly).

Kindest regards, nobody
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-05-13 13:29
(2030 d 05:35 ago)

@ nobody
Posting: # 14808
Views: 10,867
 

 completely OT

Hi nobody,

» Arznei-Frosch (sic!)

:-D

When I was young sex was safe and I recorded my favorite songs off the radio onto Compact Cassettes.

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
nobody
nothing

2015-05-13 13:37
(2030 d 05:27 ago)

@ Helmut
Posting: # 14809
Views: 10,915
 

 completely OT

Sex was never really safe but...

[image]

:-D

Kindest regards, nobody
john john
☆    

Spain,
2015-05-14 10:16
(2029 d 08:48 ago)

@ Helmut
Posting: # 14815
Views: 10,799
 

 CVinter (or CVtotal) from CVintra

Hi Helmut

In a bioavailability study with six healthy volunteers:

...(Cmax; range 135–422 ng ml-') and time to maximum (tmax; range 1.5–7.0 h) for the three tablet doses showed significant differences between subjects (P <0.009; between subject coefficients of variation (CVs) 34% and 27%, respectively). There were no within subject differences in C, (P = 0.32; mean within subject CV 11%

To calculate the sample size for a parallel study I use the total CV= 45?

or do I need the pooled variability? Which is?

Thanks


Edit: Merged with a later post. Don’t fork; you can edit your post with 24 hours. [Helmut]
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-05-15 11:05
(2028 d 07:59 ago)

@ john john
Posting: # 14827
Views: 10,792
 

 CVtotal ≠ CVintra + CVinter

Hi John John,

» In a bioavailability study with six healthy volunteers:

I would never plan a pivotal study based on results of such a small pilot – especially if the variability is high. Let’s assume a CV of 30%. From six subjects we can calculate a confidence interval of the CV with 18–102%. Not meaningful.

» ...(Cmax; range 135–422 ng ml-') and time to maximum (tmax; range 1.5–7.0 h) for the three tablet doses showed significant differences between subjects (P <0.009; between subject coefficients of variation (CVs) 34% and 27%, respectively). There were no within subject differences in C, (P = 0.32; mean within subject CV 11%

The quoted text leaves many questions open. I guess it was a crossover. Significant differences between subjects… Why report this at all? Since we know that results come from different sub­jects, this effect will  always  likely be significant.
[image]Two values for CVinter? How were those calculated? From the ANOVA you get only one. Some people use the (arithmetic!) SD/mean of treatments (like in the crappy paper above). Has nothing to do with the ANOVA on log-transfomed data.

» To calculate the sample size for a parallel study I use the total CV= 45?

No way! CVs are not additive (CVpooled  CVintra + CVinter). See the methylphenidate example from this post.

  MSe      MSs   CVintra CVinter CVpooled
0.004882 0.07675 7.00%   19.1%   20.4%


» or do I need the pooled variability?

Yes.

» Which is?

Not the slightest idea. Not accessible from what you posted. You need either the complete ANOVA or the raw data.

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

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

Spain,
2015-05-15 11:13
(2028 d 07:51 ago)

@ Helmut
Posting: # 14828
Views: 10,716
 

 CVtotal ≠ CVintra + CVinter

Hi Helmut

But how I find MSe and MSs? Thanks
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-05-15 11:14
(2028 d 07:49 ago)

@ john john
Posting: # 14829
Views: 10,666
 

 CVtotal ≠ CVintra + CVinter

Hi John John,

» But how I find MSe and MSs?

They are given in the ANOVA of a crossover study.

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

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

Spain,
2015-05-15 11:50
(2028 d 07:13 ago)

@ Helmut
Posting: # 14830
Views: 10,661
 

 CVtotal ≠ CVintra + CVinter

» They are given in the ANOVA of a crossover study.

But as hydroxychloroquine has a half-life of elimination of about 50 days, studies are made in parallel, so I don’t have that data.


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
★★★
avatar
Homepage
Vienna, Austria,
2015-05-15 12:07
(2028 d 06:57 ago)

@ john john
Posting: # 14831
Views: 10,709
 

 CVtotal ≠ CVintra + CVinter

Hi John John,

» But as hydroxychloroquine has a half-life of elimination of about 50 days, studies are made in parallel, so I don’t have that data.

Heck, this is not the topic of this thread. I gave you some examples in the other thread. CVtotal was reported with ~30–35%, both for AUC and Cmax.

From the ANOVA of a parallel design use \(CV_{total}(\%)=100\sqrt{e^{MSE}-1}\).

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

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

Spain,
2015-05-15 13:07
(2028 d 05:57 ago)

@ Helmut
Posting: # 14833
Views: 10,576
 

 CVtotal ≠ CVintra + CVinter

Hi Helmut

For complicate things further: if I do not have the ANOVA, it would be justified to use the sum of CVintra and CVinter, or only CVinter? Thanks
Helmut
★★★
avatar
Homepage
Vienna, Austria,
2015-05-15 13:10
(2028 d 05:54 ago)

@ john john
Posting: # 14834
Views: 10,681
 

 CVtotal ≠ CVintra + CVinter

Hi John John,

» […] it would be justified to use the sum of CVintra and CVinter, or only CVinter?

Did you read my posts – or at least their subjects lines? Are you a Troll?

EOD from my side.

Dif-tor heh smusma 🖖
Helmut Schütz
[image]

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes
Activity
 Admin contact
21,232 posts in 4,427 threads, 1,481 registered users;
online 6 (1 registered, 5 guests [including 3 identified bots]).
Forum time: Wednesday 19:04 UTC (Europe/Vienna)

A big data-analyst is an expert
in producing misleading conclusions from huge datasets.
It is much more efficient to use a statistician,
who can do the same with small ones.    Stephen Senn

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5