jag009
★★★

NJ,
2022-03-08 00:47
(751 d 23:24 ago)

Posting: # 22824
Views: 2,067
 

 Significant figures in Certara Phoenix [Software]

Hi,

I have a stupid question. When I set the # of significant figures for my PK parameters as "12" and export the file as SAS XPT file, how come some of the parameter values do not contain 12 sig figs? I know why Tmax and Cmax are not because they are not numbers computed by Phoenix, but how come observed AUClast and AUCall only have 6 significant figures (unless the last digit is 0) while predicted AUClast/AUCinf, observed AUCinf have full 12 significant figures?

Thx
J
dshah
★★  

India/United Kingdom,
2022-03-08 18:07
(751 d 06:05 ago)

@ jag009
Posting: # 22826
Views: 1,785
 

 Significant figures in Certara Phoenix

Hello J!
I believe that the input significance figure might be causing to have 6 significance figure for observed while for predicted- as calculation is considered- it is 12.
I have found that even for observed, we can have 12 significance figure.
Observed AUClast is given below:
[image]
Predicted AUCinf is given below:
[image]
Regards,
Divyen Shah
SDavis
★★  
Homepage
UK,
2022-03-09 17:04
(750 d 07:07 ago)

@ jag009
Posting: # 22828
Views: 1,670
 

 Significant figures in Certara Phoenix

Hi John,

I'd like to see the original Phoenix project exhibiting this to investigae further. Could you please post it to [email protected] so we can take a look and advise you accordingly please?

Addtionally it would be helpful to have a copy of your SAS transport settings that you will find in the documents folder if you Save them.[image]

Simon

Simon
Senior Scientific Trainer, Certara™
[link=https://www.youtube.com/watch?v=xX-yCO5Rzag[/link]
https://www.certarauniversity.com/dashboard
https://support.certara.com/forums/
Ohlbe
★★★

France,
2022-03-10 15:10
(749 d 09:01 ago)

@ jag009
Posting: # 22831
Views: 1,709
 

 Significant figures in Certara Phoenix

Hi John,

❝ how come observed AUClast and AUCall only have 6 significant figures (unless the last digit is 0) while predicted AUClast/AUCinf, observed AUCinf have full 12 significant figures?


It all depends on the number of significant figures you have for concentrations, the blood sampling times, the method used to calculate AUC, and how Phoenix displays zeros. I assume you're using the linear trapezoidal method (which will make Helmut scream, but that's another topic, just like why you want to have 12 significant figures).

Let's imagine that you have 4 significant figures for concentrations. If you have sampling times at exact hours: each AUC "slice" will have 4 or 5 significant digits (average of the two concentrations x time between the two points). If you have half-hours or quarter hours: possibly 1 or 2 extra figures. If you have a sampling time which does not result in a round decimal value (e.g. 20 minutes = 0.33333[...]33 hours): then you can get 12 significant figures.

For instance, it the concentration at 1h is 1234 and the concentration at 2h is 2345: that "slice" of AUC will be ((1234 + 2345)/2) x (2-1) = 1789.5. That's 5 significant figures. If you want more, all you'll get is extra zeros (1789.50000000). These zeros are significant figures, granted. But I don't know whether they will be displayed in Phoenix.

Predicted values and AUCinf rely on the calculated elimination rate constant, and you can get 12 decimal figures.

Regards
Ohlbe
ElMaestro
★★★

Denmark,
2022-03-10 17:51
(749 d 06:20 ago)

@ Ohlbe
Posting: # 22832
Views: 1,641
 

 ...which naturally leads to another question....

...and all this naturally leads to the next big question, slightly OT but to me a much more important related matter :-)

Let us just say we work with a producer of ref standards located in... say we pick a completely random place... Toronto, Ontario, Canada.
Said producer renders CoA's where the purity of the ref material is given with two significant digits. Most often 99% or 98%. If you are very lucky you'll get something like 98.7% (three significant digits) but that's an exception and you cannot request a CoA with a given number of digits.
Now, how (or rather why) the [expletive] does one report concentrations with 5 significant digits from std curves produced from such standards? And how (why), in the grand scheme of things, does one meaningfully and reasonably get from such concentrations where two digits are the most you can squeeze out, to confidence intervals reported with four significant digits on the Western side of 100% and five digits on the Eastern side?

Does anyone, anywhere have anything better than "those are the rules" and "we just follow the standards"? :-D

Pass or fail!
ElMaestro
Ohlbe
★★★

France,
2022-03-10 18:34
(749 d 05:37 ago)

@ ElMaestro
Posting: # 22833
Views: 1,609
 

 ...which naturally leads to another question....

Hi ElMaestro,

❝ Now, how (or rather why) the [expletive] does one report concentrations with 5 significant digits from std curves produced from such standards?


Measured with a method for which the tolerance limits for precision and accuracy are 15% (20% at the LLOQ)...

(Well, actually I've seen worse than 5: sometimes 6 or even 7. And only very occasionally people using a fixed number of significant digits rather than a fixed number of decimals).

❝ Does anyone, anywhere have anything better than "those are the rules" and "we just follow the standards"? :-D


"I left my brain in a jar at the door" ? "That's what the FDA requires" ?
The basic rule I was taught in my 1st year at University: you can't give a result with more significant digits than the lowest you have in any of the figures you use to calculate it. Use full precision for calculations, but always round to the lowest number of significant figures to give the end result. We actually lost points if we gave too many significant figures at an exam.
This being said, if you give a result with only 2 significant digits and your LLOQ is 10 pg/ml, you would get a first step of 10% between 10 and 11 pg/ml, which is not great regarding precision and accuracy. I would not blame the use of 3 digits in such circumstances.

Regards
Ohlbe
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
82 visitors (0 registered, 82 guests [including 7 identified bots]).
Forum time: 00:12 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

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