Jump to content


Photo

Simulation results


  • Please log in to reply
9 replies to this topic

#1 sundar

sundar

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 27 August 2019 - 03:30 PM

Hi 

 

 

I am simulating PD response from lower dose to higher dose using eta values, different occasion of simulation under same dose & condition i am getting different responses. Kindly let me know the possible reasons,

 

Procedure adopted while simulation;

  • Copied & accepted final estimates from final PD model
  • Dose units are same 

 

Thanks

Sundar. M 

 



#2 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 28 August 2019 - 05:33 AM

Are you able to share your model or better yet project file so we can look at your other settings?



#3 sundar

sundar

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 28 August 2019 - 07:09 AM

Hi, 

 

Please find uploaded project file. 

 

Now i am getting different kind of issue, i.e getting almost similar simulated profile even double the dose. 

 

Thanks

Sundar. M 

Attached Files



#4 sundar

sundar

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 04 September 2019 - 04:27 AM

Hi

 

Have you had a chance to review my project file. 

 

I have checked all possible mapping and its correct as per my knowledge. However, i am not able to understand the results, why its same for different dose level simulation. 

 

Thanks

Sundar. M 



#5 sundar

sundar

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 04 September 2019 - 05:33 AM

Hi,

 

In continuation to above post, while reviewing the individual subject's simulated profile, simulated profile of 3 subjects observed with -1.#IND (for all time points) instead of numerical value for one of the replicate (total replicates n=100).

 

Is it because of any error?

 

Thanks

Sundar. M 



#6 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 06 September 2019 - 01:15 PM

Hi Sundar, sorry for  the delay. I would be interested to see your original fits since the random error (omegas) are very large.

 

to Demonstrate, try this  in R

1.69304 * exp(rnorm(100, 0, sqrt(2.5395776)))

 

This is structural parameter gam simulation.

You will see that some gamma values are very large (more than 100) when normally beween 0.5 and 8 might be expected.

Using that values as a power leads to overflow.

I’d suggest to reduce ngam

 

  Simon



#7 sundar

sundar

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 12 September 2019 - 11:32 AM

Hi Simon, 

 

Thanks for your information, due to confidentiality reasons, i am not able to share the original fits project file. 

 

I will try to reduce ngam and post the outcome. 

 

Thanks
Sundar. M 



#8 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 12 September 2019 - 12:53 PM

Sundar, it's more a case of being sure that your original fit is good and you have confidence in *all* the model parameters before you start uisng them for simulation.



#9 sundar

sundar

    Advanced Member

  • Members
  • PipPipPip
  • 38 posts

Posted 13 September 2019 - 10:23 AM

Simon, I agree your point, however still i have following uncertainty,

 

---even with worst model - doubling the dose will not differentiate the simulation results? 

 

---simulated results are comparable with observed results (for eg., 30mg observed vs simulated) - how to buy then the validity of developed model?

 

 

Thanks

Sundar. M 



#10 Simon Davis

Simon Davis

    Advanced Member

  • Administrators
  • 1,316 posts

Posted 13 September 2019 - 10:38 AM

Without seeing your data, bur certainly in biological systems a saturation effect can be expected and the model should include that e.g. heart rate cannot increase beyond perhaps 200BPM for even a young healthy adult.

if you model matches well at observed doses that's great but again with PKPD models it's good to assess a minimum of three dose levels to better understand the dose response curve, is it sigmoid, or simple emax or a more complicated indirect response? etc.

you may also want to consider if there is a placebo effect you need to first build as your baseline model if your PD varies over time e..g cortisol levels within a day, tumour growth, other disease progression.

this course maybe interesting to you, it's available in person and online;

https://www.certarau...hoenixWinNonlin


Edited by Simon Davis, 13 September 2019 - 10:41 AM.





0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users