Jump to content


Photo

Additive error being bound to minimize the effect of the BLQ imputation


  • Please log in to reply
6 replies to this topic

#1 Davidq8

Davidq8

    Member

  • Members
  • PipPip
  • 29 posts

Posted 16 June 2022 - 10:43 PM

Hi,
I am working on a model with LLOQ concentrations being imputed. I found the following in a published PopPk analysis:
 
" Residual variability was assumed to have an additive and a proportional element, with the additive error being bound to be at least 20% of the LLOQ to minimize the effect of the BLQ imputation".
 
I question how to put such a bound on additive error using the PML or built-in faction?
 
Thanks  


#2 cradhakr

cradhakr

    Advanced Member

  • Members
  • PipPipPip
  • 54 posts

Posted 16 June 2022 - 11:19 PM

Hi David,

 

Thanks for contacting Certara. In regards to your question, I am not sure which publication you are referring to.

 

I have attached a link (Presentation on handling LLOQ) that may be of use to you.

 

 

https://www.youtube....h?v=xX-yCO5Rzag

 

Thanks and regards

Chnadramouli



#3 Davidq8

Davidq8

    Member

  • Members
  • PipPip
  • 29 posts

Posted 17 June 2022 - 03:08 AM

Thanks, Chnadramouli, for sharing the link. It is very helpful. I am quite familiar with handling LLOQ using NLME, but I still do not know how they applied the bound on the additive error to be at least 20% of the LLOQ .
I am attaching the study and NONMEM codes.
Thanks 

Attached Files



#4 cradhakr

cradhakr

    Advanced Member

  • Members
  • PipPipPip
  • 54 posts

Posted 05 July 2022 - 05:43 AM

Hi David,

 

Apologies thought I had answered your question. With Phoenix you will be using the M3 method as mentioned in the link below

 

 

 

Thanks

Mouli

 

https://www.youtube....h?v=xX-yCO5Rzag



#5 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,230 posts

Posted 05 July 2022 - 10:51 AM

The model seems somewhat complex but this can be done in Phoenix too.

 

AddErr and PropErr are the fixed effects and they are combined into W which is used as a coefficient for the sigma.

That could be used in PML too (create fixefs AddErr and PropErr and use them to build W).

Since AddErr has a low limit as 0.00626, I assume it is the constraint you/the authors are talking about.

 

Prop=C*PropErr

Add=AddErr

W= sqrt(Add^2 + Prop^2))

error(CEps = 1)

observe(CObs = C + W*CEps)

fixef(AddErr = c(0.0138181,0.00626,))

fixef(PropErr = c(,0.171622,))


Edited by Simon Davis, 05 July 2022 - 11:39 AM.


#6 Davidq8

Davidq8

    Member

  • Members
  • PipPip
  • 29 posts

Posted 06 July 2022 - 07:42 PM

Hi David,

 

Apologies thought I had answered your question. With Phoenix you will be using the M3 method as mentioned in the link below

 

 

 

Thanks

Mouli

 

https://www.youtube....h?v=xX-yCO5Rzag

Thanks Mouli for your explanation. 



#7 Davidq8

Davidq8

    Member

  • Members
  • PipPip
  • 29 posts

Posted 06 July 2022 - 07:44 PM

The model seems somewhat complex but this can be done in Phoenix too.

 

AddErr and PropErr are the fixed effects and they are combined into W which is used as a coefficient for the sigma.

That could be used in PML too (create fixefs AddErr and PropErr and use them to build W).

Since AddErr has a low limit as 0.00626, I assume it is the constraint you/the authors are talking about.

 

Prop=C*PropErr

Add=AddErr

W= sqrt(Add^2 + Prop^2))

error(CEps = 1)

observe(CObs = C + W*CEps)

fixef(AddErr = c(0.0138181,0.00626,))

fixef(PropErr = c(,0.171622,))

Thanks, Simon, for providing me with text codes for PML, this is really helpful!

Best  






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users