1 reply to this topic

[zhoumeng]

We are developing a 2-compartment PK model for a drug after subcutaneous injection in lactating goat using Phoenix NLME. The model predicted negative concentrations of this drug in the plasma at later time points? What does the negative concentration mean? The concentration should not be negative? How to avoid predicting negative concentrations?

[serge guzy]

We are developing a 2-compartment PK model for a drug after subcutaneous injection in lactating goat using Phoenix NLME. The model predicted negative concentrations of this drug in the plasma at later time points? What does the negative concentration mean? The concentration should not be negative? How to avoid predicting negative concentrations?

I believe first that if you get negative concentrations that you should put  0 lower boundaries for all PK parameters.

now are you talking about predicting the drug level with or without noise? If it is without noise then enough to juts put 0 lower boundaries for all PK parameters. If you want to predict what could be an observation rather than the predicted level only, then the problem can be with the error model you are using. If it is so, use multiplicative error model with small cv.

I need to know better how did you reach the simulation level. Is it first by fitting a model to data or you just start from scratch.

Are you using log-normal distribution for all PK?

best Regards

Serge