Jump to content


Photo

Strange behaviour of partial AUCs

partial AUC

  • Please log in to reply
5 replies to this topic

#1 rosame

rosame

    Newbie

  • Members
  • Pip
  • 3 posts

Posted 19 January 2023 - 04:05 PM

Dear All,

 

During an analysis I encountered a case when AUC0-24, which was extrapolated since the last measurable observation was around 4 h, was (very slightly) larger than AUCinf_pred:

 

- AUC024:        1.314578765489

- AUCinf_pred: 1.314034376517

- AUCinf_obs:   1.314654951761

 

I played around with the partial AUCs to figure out what could have happened and I think it is related to the difference between Clast and Clast_pred. 

 

Interestingly however, during my investigation, when I tried to approximate AUCinf with partial areas using extremely big "End Time" values (100000000 h), I started to get very strange results. The AUC suddenly became much larger, which I cannot explain. Is this an expected behaviour somehow?

 

Thanks,

Marta

Attached Thumbnails

  • wnl_p1.PNG

Attached Files


Edited by rosame, 19 January 2023 - 04:18 PM.


#2 cradhakr

cradhakr

    Advanced Member

  • Members
  • PipPipPip
  • 78 posts

Posted 20 January 2023 - 12:55 AM

Hi,

 

The predicted AUC concentration is not computed to the expected value. Since there is no observed concentration at 1000000, Phoenix is prediction 0 concentration at that time. This is more of an computational issue and we need further clarification from our developers.

 

Thanks 

Mouli

 

Note: this is how partial ares are calculated in Phoenix (Follows NCA rule)

 

https://onlinehelp.c... Areas&rhsyns=



#3 Linda Hughes

Linda Hughes

    Advanced Member

  • Members
  • PipPipPip
  • 73 posts

Posted 20 January 2023 - 01:53 PM

That is correct, Mouli.  The predicted concentration at 100000000, although in theory is not exactly zero, can only be represented in double precision as zero.  Since an endpoint of the partial area has a zero concentration value, NCA uses the linear trapezoidal rule:

 

(100000000 - 24) * 0.00000641378771090087 / 2 = 320.689308579591



#4 rosame

rosame

    Newbie

  • Members
  • Pip
  • 3 posts

Posted 26 January 2023 - 12:18 PM

Thank you for the clarification! So if I understand it correctly, it is indeed an expected behaviour, although the result itself is not correct.

Regarding the AUC024 > AUCinf problem, I always use AUCINF_pred, but seeing this result, I am not sure anymore. Would it be better to use AUCINF_obs instead of _pred?

Attached Thumbnails

  • capt.PNG


#5 Helmut Schütz

Helmut Schütz

    Advanced Member

  • Members
  • PipPipPip
  • 316 posts
  • LocationVienna, Austria

Posted 28 January 2023 - 10:03 AM

Hi rosame,

 

So if I understand it correctly, it is indeed an expected behaviour, although the result itself is not correct.

 

Define ‘correct’. :D

 

Regarding the AUC024 > AUCinf problem, I always use AUCINF_pred, but seeing this result, I am not sure anymore. Would it be better to use AUCINF_obs instead of _pred?

 

I prefer AUCINF_pred as well. With AUCINF_obs you base the extrapolated AUC on a single concentration which is likely the most inaccurate of the entire profile (close to the LLOQ).

Some background in this article. IMHO, more important is the trapezoidal rule you apply (see another article).


 Best regards,
Helmut

https://forum.bebac.at/


#6 rosame

rosame

    Newbie

  • Members
  • Pip
  • 3 posts

Posted 30 January 2023 - 09:18 AM

Define ‘correct’.  :D

 

AUC after 24h cannot be 320.7 h*ng/mL, this value is way too large.

I understand that my "AUC24lot" partial area has no practical use, but still, I think it would be better if Phoenix dropped an error message rather than giving a large AUC value.

 

Some background in this article. IMHO, more important is the trapezoidal rule you apply (see another article).

 

Thank you for the links! As per our SOP, we always use the linear-log trapezoidal rule for extravascular administration with rich sampling, but I see that the linear-up / log-down trapezoidal rule is advised in the article. Thank you for pointing this out, I will look into it more. Nevertheless, AUC024 is slightly greater than AUCinf_pred with both methods.


Edited by rosame, 30 January 2023 - 09:18 AM.






Also tagged with one or more of these keywords: partial AUC

0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users