In the context of simulating multiple dose profiles, from single dose data we would like to ask your advice about pros and cons of the following approaches: 1- execute descriptive statistics of plasma concentration-time data by group → fit the mean plasma concentration-time data with model from PK library model → use initial estimates from previous step to simulate multiple dose regimen 2- fit individual plasma concentration-time data with model from PK library model → calculate mean by group of individual initial estimates from previous step → use mean initial estimates from previous step to simulate multiple dose regimen Which of those would you rather recommend? Or also, would you have a different recommendation about how to best simulate multiple dose profiles from single dose data?
HI I suggest the following: Fit individual plasma concentration-time data with model from Phoenix Modeling using a population approach. Use the estimated PK parameters and assoicated variability to simulate the median and prediction intervals ond compare to the observed data this is what we call visual predictive check and let you verify that the developed model can simulate the obsered data well. Next step simulate multiple dose data using parameters and asosciated variability and compute median and prediction intervals this will give you the range of possible Multiple dose concentrations in your population. The population approach might not be feasible of you don’t have a lot of subjects. What you descirbe is akind of naive pooling compute a mean first then fit the mean profile versus two stage approach fit individual curves compute mean pk parameters and then simulate from estiamted parameters. All approaches might give similar answers for the mean but the variability is not well captured unless you do the population approach. To me the standard two stage is better than naive pooling since it can give you some variability estimate on your pk paramters. Is there any reason why you are not considering using a mixed effects moedling approach ?
Option 2 is preferred over option 1 as Samir mentions in his post. Pooling data (i.e., fitting mean concentration-time curve) can be deceptive, especially if there is a lot of variability in the data. I would recommend individual fits, then taking the mean (or median if N is small) for each PK parameter and run simulations for multiple doses.
Dear LLLi
Please find attached an example that shows you the strategy I believe is adequate.
What you need to remember is that when you just simulate based on population information you gathered from your SAD fit, you nee only the dosing information in your input data set and when simulating you put the times you want the results of simulations in the add/sim table feature we provide to you in the interface.
I am not sur what you know and what you do not know but the example attached shows you input data from a SAD. I fit the model to the data.
Then I copied the model to the workflow, accept all fixed and random, shifted to last run options (sim/Pred)and map the template data set called format 1,format2 and format3 (3 different format to simulate the same dosing regimen).
Please review what I did and come back if any clarifications are needed
SAD_fit_MAD_simulations.phxproj (1.92 MB)
Dear Serge guzy,
Thank you so much! Your example is so helpful! I still haves some questions. Any input will be appreciated.
The Rawsimtbl01.csv in your attachment contains data for 100 replicates. If I want to see the concentration-time curve after MAD, can I calculate the mean C in each time point and then plot the mean C vs time points? This mean C is the final result that we need consider to decide the MAD?
Do we need set N Iter to 0 when we run simulation?
Thank you again for your help!
LLLi
Dear LLLi
<The Rawsimtbl01.csv in your attachment contains data for 100 replicates. If I want to see the concentration-time curve after MAD, can I calculate the mean C in each time point and then plot the mean C vs time points? This mean C is the final result that we need consider to decide the MAD?>
If you want only the average, you can either do what you suggest or in the simulation, there is an option to remove all the random effects.
You do not need to put NIter to 0 if you have version 14 which I believe you have.
The program automatically set Niter to 0 whatever number is showing up.
Best Regards
Serge
Thank you Dear Serge!
About removing all the random effects, is it to clear the Ran checkbox in the structural in Parameters tab? If I did so, I can only get one concentration no matter how many replicate # we have?
If I don’t use the mean C, how can I decide the MAD from the simulated data (concentration)?
Thank you!
LLLi
Yes , you uncheck the ran check box.
I am not sure what you are trying to do when you say choosing the MAD.
I will respond later tonight(Have to go) once you clarify what you are trying to do.
MAD is multiple ascending dose, I guess.
Serge
Dear Serge,
Yes, MAD is multiple ascending dose. I want to simulate the initial multiple dose from several cohorts of single dose (single ascending dose). For my project, I should choose individual simulation or population simulation?
LLLi
Dear Serge,
I removed all the random effects in a pop model and ran the simulation. The result the same for each replicate. And also the simulation result is the same as the simulation result by using individual model and pop parameters. I think that removing random effect changed a pop model to an individual model. So which method is more reasonable, averaging the simulated concentrations from a pop model or simulating concentrations using an individual model?
If I simulated using a pop model (with random effect) and got the simulated concentration for each replicate, how to decide the multiple dose in further study besides using the average simulated concentration?
Thank you!
LLLi
Of course that if remove the random effects you will get same levels for each replicate.
Just ask for one replicate and use that as your average response.
Mathematically the average response using the average parameters or using the random effects and then averaging the concentrations will not be the same but similar only.
However I would suggest using the average PK parameters and 1 replicate only to get the average response from which you would decide the multiple dose in further studies. This is assuming that your reference for the right decision is base don the average response only.
Sometime the reference is different and can be for example.
I want to design my study in such way I am pretty sure at least 90% of the population will not have Cmax above x.
For that you would need to simulate many patients by turning on the random effects, calculating Cmax for each simulated patients and then see the percent that exceed x.
Best
Serge
Dear all,
I’m not sure if its the right place, but i want to simulated a multimple dose regimen from a single dose data that fits to a two compartamental model, but when i try to simulate multiple doses appears the same curve of a single dose without get steady state. I wanted to do something like PKS. Thank u so much.
PK VAN.phxproj (659 KB)
Dear Rosy,
could you please reformulate your question?
I see some units errors along the models causing strange plots output.
What do you want to achieve? I think Phoenix model is more suitable object for steady state simulation.
Bests,
Mittyright
Dear Mittyright,
Sorry if i didn ´t explain well. Thanks for suggest a Phoenix model, but i don’t pretend to do a population model. I just want to simulate how a subject reach the steady state after 5 or 6 doses using it’s individual parameters calculated from one dose.
Best regards.
Rosy
PK VAN.phxproj (644 KB)
¡Hola Rosy!
Please find the project attached.
I tried to show my best to you and all PHX users.
You can find in the Project 2 workflows: the first includes WNL PK solution and the second includes PHX Models solution (with 6 doses, you’ll see how to do it very fast and simple using ADDL input option; you can also find the example where I put the subject into the steadystate just after beginning)
Why the curves after 1 st dose are so similar to the results after 6th dose? I think you are going to understand that after looking at the results of Accumulation Estimation object (Accumulation index and Time to reach 99.9% of steady state level)
BR, Mittyright
PS: I kindly suggest to visit some Certara PKPD modelling courses, there are much more things which are doable in PHX
PK VAN_Rosy.phxproj (2.38 MB)
Dear Mittyright,
Thank you so much for everything, I’m sure you always do your best and that makes you a great person. I’m really grateful for your patience, I’m new to this and it’s still difficult for me to relate concepts to the program’s functions. Finally I apologize that maybe I will continue to bother you… just kidding 