Jump to content


Photo

Phoenix 6.3 Inflation of AUC in NonParametric Superposition


  • Please log in to reply
7 replies to this topic

#1 Ralph-Steven Wedemeyer

Ralph-Steven Wedemeyer

    Member

  • Members
  • PipPip
  • 16 posts

Posted 07 September 2016 - 12:52 PM

Dear all.

 

I have to admit that we are still running 6.3, waiting for 7 as our next step for implementation (never touch a running system, kind of).

Due to some miscalculation from my side I had a look into the result I got from a quick NPS today and observed something I was not able to understand (or resolve). Fortunately the values in the end still somehow match with my (corrected) calculations by hand, but still

I would like to ask, whether someone had observed something similar and might give some hints what I did wrong here.

 

For demonstration I provide one of the data sets: 1 subject, two products (I changed them to R and T naturally).

This was subject to the NPS object (sorted by treatment, regular dosing, 100 points, log-calculation, doses 1/1, Tau 24 (note: I did change the "display", number of points (24, 1441...) and Tau to several other values without a change in the dubious output; Lz was estimated from 8-72 for Reference and from 36-72 for Test (based on NCA)).

 

What I got when I compared the original data with the 1st dose was the profiles also attached.

In this (and all other subjects, some to a lesser degree, some even more clearly) the predicted concentrations failed to match the original data starting at 2 hours and showed a sharp decline at about 8 hours (there are even subjects with multipeaks until 2 hours that are nicely matched, but afterwards ...).

As I expected, that for "Display 1st dose" the profiles should be superimposable, I found this interesting. However, at least the sharp decrease at 8 hours is also clearly visible for displays at steady-state (the start of mismatching at 2 h not quite as clearly (higher concentrations/accumulation etc.) and here I started to worry, whether my calculations/values I would like to use for further assessments are correct.

 

As said, I would highly appreciate any input on what went wrong here and I will be really happy, if this is something that I did wrong (because that would naturally be the most comfortable issue to correct).

 

Please do not hesitate to ask for any further information I forgot to provide and thanks for reading!.

 

Best regards,

 

Steven.

                                                                                                                                                                                                         

 

Attached Thumbnails

  • Sample profile NPS.png

Attached Files



#2 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,210 posts

Posted 07 September 2016 - 03:11 PM



What I got when I compared the original data with the 1st dose was the profiles also attached.
In this (and all other subjects, some to a lesser degree, some even more clearly) the predicted concentrations failed to match the original data starting at 2 hours and showed a sharp decline at about 8 hours (there are even subjects with multipeaks until 2 hours that are nicely matched, but afterwards ...).
As I expected, that for "Display 1st dose" the profiles should be superimposable, I found this interesting. However, at least the sharp decrease at 8 hours is also clearly visible for displays at steady-state (the start of mismatching at 2 h not quite as clearly (higher concentrations/accumulation etc.) and here I started to worry, whether my calculations/values I would like to use for further assessments are correct.

Steven I am not sure I understand your question, your profiles at Steady-state show some accumulation, (if you compere NPS at '1st dose' there is no accumulation and therefore the lines overlap exactly.

NPS itself is a somewhat crude method since it is only using an Lz to extrapolate the data, (and you can see your data exhibits probably two phases, however your original profile is well sampled so, assuming linear kinetics, it is probably not bad.

What is it you expected to see from this exercise, why do you think whay you have now is wrong?
(I also ran it just now in 6.4 as I don't expect anythign to have changed in this function between versions)
Simon.

Attached Thumbnails

  • Steven_NPS.jpg

Edited by Simon Davis, 07 September 2016 - 03:12 PM.


#3 Ralph-Steven Wedemeyer

Ralph-Steven Wedemeyer

    Member

  • Members
  • PipPip
  • 16 posts

Posted 07 September 2016 - 03:45 PM

Dear Simon.

 

Thanks for the rapid reply and sorry for this post to be brief, but I don't want to postpone it, but have some time pressure also.

(if you compere NPS at '1st dose' there is no accumulation and therefore the lines overlap exactly.

What is it you expected to see from this exercise, why do you think whay you have now is wrong?

I expected exactly that, an exact overlap. If you look closely at the graph I attached, the "superpositioned" concentrations (violett circle and orange triangle) overlap well prior to 2 h and after about 8 h (after the sharp decline) with the black circle and the black triangle, respectively.

But I would have expected them to be nearly perfect matches, as after the first dose, there is no accumulation.

But if these profiles do not overlap, although the documentation and the procedures we all apply will expect only a simple interpolation at these phases, I am simply afraid, that these "artificial AUCs" will have an unrealistic impact on the extrapolation into steady state.

From my rough assessment of the provided subject, I would say, that the AUC from NPS between 2 and 8 hours might be about 1 third larger than the real AUC of the measured profile.

 

And actually, I have never seen such a drop as present at 8 hours before and I can also observe this in the "steady-state" profiles. But I cannot see such a "drop" in your profiles, so it seems to me that it is an error from my side.

I have no idea what I am doing wrong, but I will keep on trying.

 

Best regards,

 

Steven.

*corrected writing error


Edited by Ralph-Steven Wedemeyer, 07 September 2016 - 03:46 PM.


#4 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,210 posts

Posted 07 September 2016 - 03:51 PM

OK Steven, I see what you mean in your original plot now, it's somewhat similar to the difference you might see in observed vs predicted AUC, i.e. what Clast it would extrapolate from, but why you are seeing it at 8h I do not know.

 

Could you perhaps post a v simple project of just this data so I can look at your settings?

(and perhaps I can try re-running in 6.3 and 6.4 to see if there is indeed  a change)

   Simon.



#5 Ralph-Steven Wedemeyer

Ralph-Steven Wedemeyer

    Member

  • Members
  • PipPip
  • 16 posts

Posted 07 September 2016 - 04:14 PM

*Problem solved*

 

Dear Simon, hello everybody.

 

Well, sometimes...

From an optimistic perspective, this might be a mistake that now is documented and, because it is quite easy to avoid, my failure might help others :rolleyes: .

 

There is a longer story, but in short: I checked some settings and observed that there was no impact on my data at all (which it actually should have :blink: ).

Reason was, that time and concentrations were defined as "text" rather than "numeric". Set them to "numeric" and everything works like a charm :rolleyes: .

 

Admittedly, the values are larger than I would like, but that's another discussion.

 

Lesson learned: always keep the format correct and "text" is not always the superior option!

 

Simon, thanks again for the support!

 

Best regards,

 

Steven.

 

 



#6 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,210 posts

Posted 08 September 2016 - 05:46 AM

Hi Steven, I am glad it is solved however I would be interested to receive the file you had that exhibited the issues,since at least in 6.4 I can change time and conc to be formatted as text and still I get the same plots as above in my post 2.

 

 Simon.



#7 Ralph-Steven Wedemeyer

Ralph-Steven Wedemeyer

    Member

  • Members
  • PipPip
  • 16 posts

Posted 08 September 2016 - 07:53 AM

Hi Simon.

 

I think I can do this. I had to create a new project in order to get rid of the history and found that normally the import will recognize the data as being numeric.

Might be the reason that I did not see this issue before.

However, I checked that the "additional AUC" is present here after setting time and concentration to text.

Long story short: you should see the wrong profiles in this project and if this is not an issue in 6.4 anymore, the profiles should be come correct once you re-execute.

We will see.

 

Best regards,

 

Steven.

Hi Steven, I am glad it is solved however I would be interested to receive the file you had that exhibited the issues,since at least in 6.4 I can change time and conc to be formatted as text and still I get the same plots as above in my post 2.

 

 Simon.

 

Attached Files



#8 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,210 posts

Posted 08 September 2016 - 08:13 AM

Steven, you may or may not be able to open this v64 project in 6.3, but I opened it and saw the erroneous plot you had,

 

I double checked this by making a new plot (without needing join), remember the trick of 'add graphs' on the tab Graphs in Plot

 

Then I copied your whole workflow and re-executed in v64 without making any changes and it works fine, both with 1st dose and SS.

 

I had a quick look in the release notes for v64 and I can't see one directly related to what you are observing, so not sure why it's fixed but rest assured it is.

 

   Simon.

Attached Files






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users