10 replies to this topic

[marianamefaparicio@gmail.com]

Hi all,

 

I have a subject with unknown concentration at 72h and I want to calculate the AUC0-72.

 

Which method I should to employ in this subject?

 

Method 1) Partial area calculation where t0=0h and t72=72

 

Method 2) Based on topics that I've seen in this forum: Estimate the concentration at 72h by the expression in attached image "conc" and then apply the formula to calculate AUC given by the expression in attached image "auc".

 

Method 3) There is another method?

 

The results obtained by method 1 and 2 are a little different....it is normal?

 

Thanks for the support,

 

Best regards,

 

Mariana

 

 

 

Attached Thumbnails

• 
• 

[Helmut Schütz]

Hi Mariana,

 

I have a subject with unknown concentration at 72h and I want to calculate the AUC0-72.

 

Which method I should to employ in this subject?

 

The one which is specified in the protocol. SCNR.

 

 

Method 1) Partial area calculation where t0=0h and t72=72

Method 2) Based on topics that I've seen in this forum: Estimate the concentration at 72h by the expression in attached image "conc" and then apply the formula to calculate AUC given by the expression in attached image "auc".

Method 3) There is another method?

 

The results obtained by method 1 and 2 are a little different....it is normal?

 

Since I posted #2… I guess there will be a difference if you calculate AUClast by the linear trapezoidal method. Can you try to change your options to [Linear Up Log Down] and check whether there is still a difference?

 

 

[marianamefaparicio@gmail.com]

Hi Mariana,

 

 

The one which is specified in the protocol. SCNR.

 

 

Since I posted #2… I guess there will be a difference if you calculate AUClast by the linear trapezoidal method. Can you try to change your options to [Linear Up Log Down] and check whether there is still a difference?

Hi Helmut,

 

First of all, thanks for your quickly reply.

 

Yes, it must be specified in the protocol...but sometimes it is not.

 

With Linear Up Log Down the results of AUC0-72 calculated by method 1 and method 2 for this subject are closer than with linear trapezoidal method.

 

Another questions (sorry to bother you!):

 

1) Regarding the expressions already presented in the attached pictures what is the difference if I replace AUCINF_pred by AUCINF_obs? Which is the most suitable approach?

 

2) As an example, with method 2, I'm obtaining the values for AUC0-72 in column AUC72, however another value was reported to me (column AUC72_OTHER). I don't know if my result is the correct or not   (attached picture)

 

 

Thanks,

 

Best regards,

 

Mariana

Attached Thumbnails

• 

Edited by marianamefaparicio@gmail.com, 11 November 2016 - 10:40 AM.

[mittyright]

Dear Mariana and Helmut,

 

I've found what is the difference here.

WNL computes AUCpartial as:

partialAUC = AUClast+AUC(Clast, C72_pred)

 

But Helmut's code will compute it as

partialAUC = AUClast+AUC(Clast_pred, C72_pred)

 

So the code posted by Helmut is nice for conc prediction, but should be used with care

 

I hope it helps,

Mittyright

Edited by mittyright, 11 November 2016 - 11:08 AM.

[mittyright]

Dear Mariana,

 

1) Regarding the expressions already presented in the attached pictures what is the difference if I replace AUCINF_pred by AUCINF_obs? Which is the most suitable approach?

 

Could you please enlighten me what is the purpose of such a substitution? The Helmut's derived formula has strong PK/math background.

 

2) As an example, with method 2, I'm obtaining the values for AUC0-72 in column AUC72, however another value was reported to me (column AUC72_OTHER). I don't know if my result is the correct or not    (attached picture)

 

Sorry, it is impossible to answer without reference dataset. I suppose that the AUC calculation rules were different, because AUC72_Other is faraway from yours.

 

BR,

Mittyright

[marianamefaparicio@gmail.com]

Dear Mittyright,

 

Thanks for your answers. My comments are presented below.

 

Dear Mariana,

 

Could you please enlighten me what is the purpose of such a substitution? The Helmut's derived formula has strong PK/math background.

 

Ok...so, the substitution does not make sense....I understand now.

 

Sorry, it is impossible to answer without reference dataset. I suppose that the AUC calculation rules were different, because AUC72_Other is faraway from yours.

 

Yes, probably the method used for the AUC72_Other calculation was different...I just want to know if method 2 (Helmut's code) that I employed is the correct one or if there is a better approach to estimate AUC72h that I'm not seeing...

 

Thanks a lot!

 

Best regards,

 

Mariana

[mittyright]

Dear Mariana,

 

I just want to know if method 2 (Helmut's code) that I employed is the correct one or if there is a better approach to estimate AUC72h that I'm not seeing...

 

The best way for me is to use embedded in WNL partial AUC method calculation.

Otherwise(method 2) you need to make a note to your report that in case when Tlast<72h then AUC0_72 was computed as

partialAUC = AUClast+AUC(Clast_pred, C72_pred)

with log rule of AUC() calculation

 

In this case everyone can reproduce your results.

 

BR,

Mittyright

[Helmut Schütz]

Hi Mariana,

I agree with Mittyright. Is it possible to upload a project-file (containing just the time/concentrations of these two subjects?



I’m curious how AUC72_OTHER was calculated.

[marianamefaparicio@gmail.com]

Hi Helmut,

 

Thanks for your reply.

 

I'm sending in csv and excel format because I'm not being able to upload the project file.

 

Thanks in advance,

 

Best regards,

 

Mariana

 

 

Hi Mariana,

I agree with Mittyright. Is it possible to upload a project-file (containing just the time/concentrations of these two subjects?



I’m curious how AUC72_OTHER was calculated.

 

Attached Files

•  test.csv   540bytes   423 downloads
•  test.xlsx   9.17KB   396 downloads

Edited by marianamefaparicio@gmail.com, 14 November 2016 - 02:17 PM.

[Helmut Schütz]

Hi Mariana,

AUC72_OTHER unveiled. ;-)

First C72 is calculated according to my Method 2:

(AUCINF_pred - AUClast) * Lambda_z / exp(-Lambda_z * Tlast) * exp(-Lambda_z * 72)

Then a trapezoidal (Tlast to 72) is calculated and added to AUClast by

AUClast + 0.5 * (72 - Tlast) * (C72 + Clast)

for the linear method and similarly

AUClast + (72 - Tlast) * (C72 - Clast) / ln(C72 / Clast)

for the linear up / logarithmic down method.



As you can see AUC72_OTHER agrees with pAUC0-72 if (!) calculated by the lin-up/log-down (which is better than the linear, IMHO*). I would follow Mittyright’s suggestion and simply use PHX’ native method for the partial AUC.

* See this presentation (slides 21–23) for the pros and cons of the two trapezoidal methods.

Attached Files

•  Mariana 64.phxproj   762.89KB   467 downloads

[marianamefaparicio@gmail.com]

Hi Mariana,

AUC72_OTHER unveiled. ;-)

First C72 is calculated according to my Method 2:

(AUCINF_pred - AUClast) * Lambda_z / exp(-Lambda_z * Tlast) * exp(-Lambda_z * 72)

Then a trapezoidal (Tlast to 72) is calculated and added to AUClast by

AUClast + 0.5 * (72 - Tlast) * (C72 + Clast)

for the linear method and similarly

AUClast + (72 - Tlast) * (C72 - Clast) / ln(C72 / Clast)

for the linear up / logarithmic down method.



As you can see AUC72_OTHER agrees with pAUC0-72 if (!) calculated by the lin-up/log-down (which is better than the linear, IMHO*). I would follow Mittyright’s suggestion and simply use PHX’ native method for the partial AUC.

* See this presentation (slides 21–23) for the pros and cons of the two trapezoidal methods.

 

Hi Helmut,

 

Many thanks for your excellent clarification and all support.

 

Now I understand why different values were obtained.

 

Best regards,

 

Mariana