3 replies to this topic

[yangjinh]

Dear community,

 

I have a question about the prediction in popPK model in Phoenix 8.0. And due to I am new to Phoenix, I am not sure what exactly "options" or "buttons" I need to do next.

 

So what I did: i copied a popPK model after successfull parameter estimation and simulation. Then I pasted it to the workflow again.

 

Now, I am not sure what should do next if I want to get a prediction.

 

The dataset I used to fit the model is the data about clearance and there are 24 subjects. All subjects are adults. The parameters in that model include age, weight, the amount of drug, time and gender, subjectNum and concentration.

 

Now I want to do a prediction that if the subject is a child, with the sepcific amount of drug, how the concentration will be with time goes by, under the model I got before.

 

I think I need to give a "fake" data that just includes all variable values except concentration to the software. But here I dont know how to do that.

 

 

Thank you in advance and

best regards,

Yangjinh

 

[smouksassi1]

Hi Yangjinh,

Suppose you have a pop pk model and you copy pasted it and you accepted all final estimates as initial so your are really starting from your previous fitted model.

 

You want to predict a new subject based on this model using the following information

Subject specific dose

Subject specific age

Subject specific Weight

 

so you have a data like

ID      Time  DOSE     Weight Age Sex   conc

99999   0     dose1         15     4      1     empty

99999    12      dose2         15    4      1 empty

you can just tell the software that you do not want any iteration maxiter=0 and ask for a table

time : seq( 0,24,1)# make a prediction from time 0 to time 24 every 1 hr.

since your conc are empty all your eta will be zero and what you are really doing is prediciting this subject that had contributed no data i.e a population prediction given his or her covariate values.

 

if you are extrapolating beyond data range i.e predicitng for a weight much smaller than in your current population this can be dangerous and there is no guarantee that the equation holds for very low or very high weights

SM

[yangjinh]

Thanks for your help, SM.

 

Now I could get your idea about how to do that. I think your answer means that I dont need to have iteration to get better estimation, just use the parameter values I got before as intial estimate and get prediction.

 

But I am still not sure where should I have this operation in the software. I add a new data table in the "Data" category that just contain what I want to predict. And I think I need to choose Sim/Pred.Check in the Run Options? Any other options should I choose?

 

Maybe my question is too entry level. If you could offer some guidances/examples/videos to explain it, I have the willing learn it by myself.

 

Thank you in advance and

best regards,

Yangjinh

 

 

 

Hi Yangjinh,

Suppose you have a pop pk model and you copy pasted it and you accepted all final estimates as initial so your are really starting from your previous fitted model.

 

You want to predict a new subject based on this model using the following information

Subject specific dose

Subject specific age

Subject specific Weight

 

so you have a data like

ID      Time  DOSE     Weight Age Sex   conc

99999   0     dose1         15     4      1     empty

99999    12      dose2         15    4      1 empty

you can just tell the software that you do not want any iteration maxiter=0 and ask for a table

time : seq( 0,24,1)# make a prediction from time 0 to time 24 every 1 hr.

since your conc are empty all your eta will be zero and what you are really doing is prediciting this subject that had contributed no data i.e a population prediction given his or her covariate values.

 

if you are extrapolating beyond data range i.e predicitng for a weight much smaller than in your current population this can be dangerous and there is no guarantee that the equation holds for very low or very high weights

SM

 

[smouksassi1]

Hi Yangjinh,

To be able to help I need a reproducible example on my end like a phoenix project with some dummy data in it

If the subject has no data even if you ask for iteration there is no data to inform the model alternatively you can use the data of the subject and perform maximum a posterior step where the subject individual parameters will be computed based on population parameters and the data

 

it is a simple run that asks for the prediction for this subject no need for simu/PredCheck

 

can you share a project ?