Jump to content


Photo

using the reset option in multiple dosing modelling


  • Please log in to reply
10 replies to this topic

#1 Omamah

Omamah

    Member

  • Members
  • PipPip
  • 15 posts

Posted 17 September 2018 - 11:35 PM

Dear Certara Team,

I was trying to model some data in which the patients have to take their drug daily but the PK sampling was done on sparse visits (2 visits only and in each PK visit there are average of 4 time points plus the predose drug level). there is an average of one month between each PK visit. it is Not crossover as the patients is taking the same drug daily. also the patients are on steady state because when they come to the first PK visit, they were almost on the drug for about 3 weeks.

so when i prepared the data, I flagged the steady state condition and i also used the ADDL to let the software know that there are additional doses until the second PK visit.

however, I don't know if i should use the reset option to overlay the data of the first and the second visits.

when i did that, i got higher AIC but the shape of the CWRES vs IVAR was improved. 

so is it recommended here to use the reset option?

 

also, many times when i run the model i get failure of execution message for a given id. I used to get it very frequent before I sort the steady state and the ADDL. however, now i get it sometimes (less frequent) and I am not sure why. most of the time when change the way the dose variable is displayed or the way the ADDL is displayed, i can get some results but this is confusing because i don't know why. and does the initial estimates values (if they are not precise) can lead to model excution failure?

I am attaching the excel file example to show how did i sort the time, the steady state and the ADDL. I apologize that there are many columns for each but i tried different ways of sorting. all of them are based on the id and the PK visit number. 

note that the patients are using different doses for the same drug (do i need to sort by the dose when i model?)

thank you very much

best regards

Omamah


  • veenu.bala likes this

#2 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 18 September 2018 - 09:35 AM

Omamah,

   Just a reminder that the forum is not the Support site; anyone can respond on here so there is no need to address it to the Certara team.  People answer questions in their own time so please try to make your question as clear as possible and try to make sure any provided data is complete but still easy to follow, eg labelled workflow or  name the specific model you want people to look at.

 

  Specific to your post the Excel sheet has not attached so if you can't edit your post to attach the file please send it to me and I'll add it for you.  I suspect thought that it would be more useful to attach the Phoenix project file so we can see how you've mapped the various columns as it depends onthe mappings how the data will be read from any file.

 

  Without seeing your data my initial thought is I would not use RESET since it sounds like you have continuous dosing and information between the two days so better to let the engine see that as one continuous time stream.  It may be that you need to code for some time-varying element e.g clearance if you think it's still changing over time, perhaps handle it as an occasion covariate?

 

  Don't forget to look at the NLME user manual and PML language guide forun under Help>Documents.

 

  Simon

from p44.

 

Check the Reset? checkbox if the main input dataset contains a reset column.
This option sets the state variables of differential equations to zero and restarts any sequence
statements in the model. The low value and high value fields for the Reset option allow a range of
values to be used to trigger a reset. The default is to reset when a value exactly equal to one is
found in the Reset column. See “Simple versus complex time” on page 120 for additional information.



#3 Omamah

Omamah

    Member

  • Members
  • PipPip
  • 15 posts

Posted 21 September 2018 - 12:34 PM

Simon

thank you for the response. I think I understand the concept of reset. it seems that is can be specially useful if there is a change over time in one of the parameters or the covariates



#4 veenu.bala

veenu.bala

    Member

  • Members
  • PipPip
  • 11 posts

Posted 02 July 2020 - 01:27 PM

Dear Certara Team,

I was trying to model some data in which the patients have to take their drug daily but the PK sampling was done on sparse visits (2 visits only and in each PK visit there are average of 4 time points plus the predose drug level). there is an average of one month between each PK visit. it is Not crossover as the patients is taking the same drug daily. also the patients are on steady state because when they come to the first PK visit, they were almost on the drug for about 3 weeks.

so when i prepared the data, I flagged the steady state condition and i also used the ADDL to let the software know that there are additional doses until the second PK visit.

however, I don't know if i should use the reset option to overlay the data of the first and the second visits.

when i did that, i got higher AIC but the shape of the CWRES vs IVAR was improved. 

so is it recommended here to use the reset option?

 

also, many times when i run the model i get failure of execution message for a given id. I used to get it very frequent before I sort the steady state and the ADDL. however, now i get it sometimes (less frequent) and I am not sure why. most of the time when change the way the dose variable is displayed or the way the ADDL is displayed, i can get some results but this is confusing because i don't know why. and does the initial estimates values (if they are not precise) can lead to model excution failure?

I am attaching the excel file example to show how did i sort the time, the steady state and the ADDL. I apologize that there are many columns for each but i tried different ways of sorting. all of them are based on the id and the PK visit number. 

note that the patients are using different doses for the same drug (do i need to sort by the dose when i model?)

thank you very much

best regards

Omamah

Hello,

I am facing the same problem. Could you please tell, how you did that. My data is also a steady state with additional dosing.



#5 veenu.bala

veenu.bala

    Member

  • Members
  • PipPip
  • 11 posts

Posted 02 July 2020 - 01:58 PM

Hello, 

 

I was trying to model some data in which the patients have to take their drug daily but the PK sampling was done on sparse visits (one time point plus the predose drug level). there is an average of one month between each PK visit. it is Not crossover as the patients is taking the same drug daily. also the patients are on steady state because when they come to the first PK visit.

so when i prepared the data, I flagged the steady state condition and i also used the ADDL to let the software know that there are additional doses until the second PK visit.

many times when i run the model i get failure of execution message for a given id. I used to get it very frequent before I sort the steady state and the ADDL. My data also contain some BLQ values and missing values so I censored that and use it's output. 

I am attaching my excel sheet here. I tried to make it two type in one sheet SS and ADDL are altogether and in one sheet they are in separate line. Please help

Thanks

 

Veenu

 

Attached Files



#6 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 13 July 2020 - 11:06 AM

Veenu - to save us some time in setting up, please post your project andthen we cna advise you what went wrong in your settings.  Simon.



#7 veenu2bala

veenu2bala

    Newbie

  • Members
  • Pip
  • 1 posts

Posted 13 July 2020 - 01:21 PM

Veenu - to save us some time in setting up, please post your project andthen we cna advise you what went wrong in your settings.  Simon.

Here I am attaching my project file. I would like to elaborate about that.

 

The data is from HIV infected patients already taking anti-HIV drug  (one dose daily) at the time of enrollment. So, we suppose to take it as a steady state when they come to the first PK visit. On the first day of visit predose conc was taken then after dosed and 2 hr and 6 hr. sample was taken.

The patients have to take their drug daily but the PK sampling was done on sparse visits (one time point only). There is 4 week gap between each PK visit.

So, when I prepared the data, I flagged the steady state condition and I also used the ADDL to let the software know that there are additional doses until the second PK visit. I flagged additional dosing at once for the whole study. In my data I make the excelsheets taking Addl and SS in same row for every IDs (SS and Addl altogether).

In another sheet I put ADDL and SS in separate row coz at first visit we took predose conc and then dosed the patient and take 2 hr and 6 hr conc. So, I put addl row after taking predose conc. (I am not sure, which is correct measure). My data also contain some BLQ values and missing values so I censored that and use it’s output for model input.

I took initial estimates from literature and run models for both excel files. Most of the times when I run model I get failure of execution message for a given Id that it is unable to reach steady state.  Both excel data behave differently for different models. So, I am not sure which excel file is correct in this situation or there is any other method to make input for this kind of data. Like taking Steady state in consideration is required or not. Or I need to use reset while starting study.

For SS and Addl altogether file only mix ratio models are working but Retcode is 2 and 3. Also, Omega table shows very high eta values for volume and clearance.

For SS and Addl separate file multiplicative error, additive+multiplicative and mix ratio model is working but Retcode is 3 for all. Also, Omega table shows very high eta values for volume and clearance.  

Why eta values are getting high in omega sheet?

 

One more thing we want to find out interoccassion variability in this study (like adherence of patients to the dosing schedule). Want to include it later as covariate assuming each data point as separate occasion.

 

 

 

Please help me in this regard. I will be highly thankful to you.

Attached Files



#8 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 14 July 2020 - 10:12 AM

Hi Veenu, sorry i haven't had a chance to look at your project in detail, however "unable to reach steady state. "

Is often an issue with the model fitting, when I looked at your initial estimates tab, they were giving predicted concentrations several fold higher than your observations. Please check your units (is dose administered really meant to be in ng/mL)

I just did a rough NCA (check the doses when you re-eexcute it) to ananlys the first 'complete' profile 24-48h. got some better estimates to start with,however the program still struggled to converge, see how all the parameter estimates keep increasing.


I also noted your initial estimates for random effects were all 1 i dropped these to .1 and then the first model of your project converged, take a look at that and update your others as necessary. Remember you can map in final estimates from a previous model if that helps. see your second model inthe first workflow?

remember also the model comparer tool and if you need help with using these models we have online training available. e.g.

https://www.certarau...-september-2020

Attached Thumbnails

  • can_not_converge.jpg

Attached Files


Edited by Simon Davis, 14 July 2020 - 11:14 AM.


#9 veenu.bala

veenu.bala

    Member

  • Members
  • PipPip
  • 11 posts

Posted 15 July 2020 - 01:02 PM

Thank you so much Simon! I understand that. Can we use initial estimate from only two data point slope (or three data points required)? Models are running after taking initial estimate from NCA. How to use reset option while starting study if patients already on medications but not sure whether they are following the dose as suggested so I just wanted to remove steady state and do a fresh multiple dosing PK.

Thanks
Veenu

Edited by Simon Davis, 15 July 2020 - 02:24 PM.
remove full quote


#10 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 15 July 2020 - 02:24 PM

Can we use initial estimate from only two data point slope (or three data points required)?
>yes it's just to get an initial estimate inthe right order of magnitude.

 

How to use reset option while starting study if patients already on medications but not sure whether they are following the dose as suggested so I just wanted to remove steady state and do a fresh multiple dosing PK?

 

I am not really sure I follow your question, even if they are not totally compliant then they are probably close to SS, missing the odd dose won't affect C compared to Css much in most cases.

If you do want them to be started 'fresh' then probably easier to just not use the SS flag isn't it?
 



#11 veenu.bala

veenu.bala

    Member

  • Members
  • PipPip
  • 11 posts

Posted 15 July 2020 - 02:26 PM

Can we use initial estimate from only two data point slope (or three data points required)?
>yes it's just to get an initial estimate inthe right order of magnitude.

 

How to use reset option while starting study if patients already on medications but not sure whether they are following the dose as suggested so I just wanted to remove steady state and do a fresh multiple dosing PK?

 

I am not really sure I follow your question, even if they are not totally compliant then they are probably close to SS, missing the odd dose won't affect C compared to Css much in most cases.

If you do want them to be started 'fresh' then probably easier to just not use the SS flag isn't it?
 

Thank You So much Simon for your quick response. Now I got that. it is really a great help. I appreciate.

 

Veenu






2 user(s) are reading this topic

0 members, 2 guests, 0 anonymous users