10 replies to this topic

[Yulan Qi]

Hello, I have a question about sequential PK-PD modeling using NLME. I have modeled PK data by NLME and PK parameters have been estimated for each subject. I want to use the individual PK parameters as PK input and not sure how to do that. Any suggestions would be highly appreciated!

 

Kind regards,

 

Yulan

[Emily Colby]

Dear Yulan,

 

Currently, NLME can only take the fixed effects PK parameters as input when doing population PK/PD modeling with PK frozen. We have added a request for the ability to freeze post hoc PK parameters (QC 10647) for individual subjects. This request will be taken into consideration for future versions of Phoenix NLME.

 

Best regards,

 

Emily

[Emily Colby]

Dear Yulan,

 

A workaround is to take in the individual post-hoc PK parameter estimates as covariates. The structural parameter estimates can be requested using the "Add table" button in the Run Options tab and "Structural Parameter Estimates" button. Then, merge the original data with post-hoc estimates, sorting by ID and selecting "carry along data for like sort levels". In the PD model, bring in the results from the Merge worksheets as input. Add the covariates for the PK parameters. Set up the model so that PK is frozen to the covariate values (I've attached a project file in which I set the fixed effects to 0 and the covariate coefficients to 1-- based on the model structure, that fixes the PK parameters to the covariate values).

 

Emily [file name=pkpd3102_indPK.phxproj size=2174571]http://pharsight.com/extranet/media/kunena/attachments/legacy/files/pkpd3102_indPK.phxproj[/file]

[Yulan Qi]

Dear Emily,

 

Thank you very much for your reply. I will try the woraround method and may ask for your help if enconter any problems.

 

Thanks,

Yulan

[Charvi]

Thanj

 

Dear Yulan, A workaround is to take in the individual post-hoc PK parameter estimates as covariates. The structural parameter estimates can be requested using the "Add table" button in the Run Options tab and "Structural Parameter Estimates" button. Then, merge the original data with post-hoc estimates, sorting by ID and selecting "carry along data for like sort levels". In the PD model, bring in the results from the Merge worksheets as input. Add the covariates for the PK parameters. Set up the model so that PK is frozen to the covariate values (I've attached a project file in which I set the fixed effects to 0 and the covariate coefficients to 1-- based on the model structure, that fixes the PK parameters to the covariate values). Emily [file name=pkpd3102_indPK.phxproj size=2174571]http://pharsight.com/extranet/media/kunena/attachments/legacy/files/pkpd3102_indPK.phxproj[/file]

 

Thank you Emily and Yulan for discussing this. Do we still have to follow the same procedure for Pheonix 8.1? I wish to use the individual post-hoc estimates from Pop PK modeling to model the PD data. It will be great if you could share an example project. The above file does not seem to be downloadable. 

 

Thanks

Charvi

Dear Emily, Thank you very much for your reply. I will try the woraround method and may ask for your help if enconter any problems. Thanks, Yulan

[bwendt@certara.com]

Dear Charvi,

 

your question comes at the right point in time. It is indeed in Phoenix 8.1 where you can use the posthoc parameters for sequential PKPD modeling.

I will prepare an example project for you.

 

Bernd

 

[bwendt@certara.com]

Here is the example:

 

 

 example_seq_PKPD.phxproj   742.33KB   465 downloads

 

 

 

Bernd

[Charvi]

Hi Bernd,

 

Thank you for your response and for the example project as well. I followed a similar approach as you have shown in the example for naive pooled method, but in addition to mapping the Theta file to the parameters, I also mapped the Eta file from the fitted Pop PK model to the Random Effects tab in the set up for the sequential PK/PD model. I did get expected post-doc estimates (for PK) for the all the IDs. This allows for inclusion of population value and the ETAs - giving eventually the post hoc PK estimates to drive the PD. I think the approach works out while having the Population box checked (instead of naive pooled). Do you recommend or use a similar approach as well for Pheonix 8.1 for Pop PK/PD sequential modeling? I would appreciate your thoughts on this. 

 

Thanks,

Charvi

[bwendt@certara.com]

Hi Charvi,

 

depending on whether you want to run a sequential PKPD model in individual or population mode the approach differ:

 

1. individual mode: here you have to map your Thetas from an individual PK fit model to your Parameters panel:

 

 

Alternatively, you can use posthoc-estimates from a population PK fit. It will give the same results.

 

2. population mode: here you don't need to map thetas or posthoc estimates, rather you do:

 

- copy your PK Fit model and paste it back to workflow

- accept all fixed and random parameters under Parameters>>Fixed Effects Tab

- change the type of model to PKPD, e.g. PK/Emax

- check the box: sequential PKPD - this will freeze all fixed effects of your PK parameters

- map your etas from previous PK Fit to the Random Effects panel

 

 

In both cases, individual and population mode, the PK results will be the same, the predicted concentration time curves will be identical.

 

 

Bernd

[Charvi]

Thanks Bernd. I followed the procedure as outlined by you. Just wanted to confirm, I was on the right path.

This feature in Pheonix 8.1 is great !

 

Many Thanks

Charvi

[wanghaiqing2004@gmail.com]

Dear Emily and Bernd,

Does   example_seq_PKPD.phxproj   742.33KB  works in earlier version of Phoenix (Phoenix 64) ?  Or is it possible that you can post 

 

file name=pkpd3102_indPK.phxproj size=2174571]http://pharsight.com...2_indPK.phxproj[/file]

 

HQ