Combining PK and PD data during population PK/PD analysis
Janet R. Wade(1) and Mats O. Karlsson(2)
1 Medical Products Agency, Box 26, S-751 03 Uppsala, Sweden; 2 Uppsala University, Dept. of Pharmacy, BMC, Box 580, S-751 23 Uppsala, Sweden.
There are two general alternatives for combining pharmacokinetic (PK) and pharmacodynamic (PD) data during a population analysis. First, the data may be analysed simultaneously, this probably represents the purist approach. Second, the data may be analysed sequentially; the PK data are analysed separately, the results from that analysis are then included in the subsequent PD analysis.
A literature review has been performed to look at which of these two different approaches are used. There was no clear indication from that review that one approach was used more than another. A further breakdown of the methodology used in the published papers examined additionally found that the choice of algorithm (with respect to the NONMEM analyses) or the sparse/richness of the data to be analysed also seemed to have no implications for the PK/PD analysis approach applied.
A simulation study has been performed using NONMEM, Version V. Data were generated using a well designed PK/PD study from which sparse data were collected. The data were analysed in four different ways:
- A simultaneous PK/PD fit
- Fit PK data, fix individual PK parameters, fit PD data
- Fit PK data, fix population PK parameters (and take PK data out of data file for PD analysis), fit PD data
- Fit PK data, fix population PK parameters (leave PK data in data file for PD analysis), fit PD data
and where numbers 2 to 4 are variations of a sequential approach. Further, each of the four cases described above were analysed using both the FO and FOCE algorithms. The effect of PD model misspecification on estimation of PK parameters was also looked at.
Some general conclusions can be made from the results of this work, but much more remains to be done.