Using a PK-PD linked model for horse data, I have obtained PD estimates to describe drug effects with a sigmoidal Emax model (baseline) for six subjects. The graphical output looks promising in that the predicted concentrations follow the measured concentrations well. However, for at least 4 subjects, some paramter estimates , such as the Emax do not potray the graphical output . The subject numbers on the graphs correlate to the numbers in the table. I have tried adjusting the link paramters, and changing the PK model from a 2 to 3 comp and there is minimal difference. Does anyone know what accounts for these dicrepancies and how I can correct them? I appreciate any help. Best, Marc 
Sorry I’ve tried to upload my files, but was unable. 
[file name=mrpkpd.phxproj size=2724940]/extranet/media/kunena/attachments/legacy/files/mrpkpd.phxproj[/file]
Marc - I attached the image files for you, however I am not sure I follow your question. The graphical PD output looks to have the Correct E0, e.g. baseline per subject. At these dose levels it is possible you have not reached Emax predicted by the model, perhaps you can attach the project to help us visualise your model better - if you keep the file name below 8 characters it tends to be happier, we are going to be upgrading to a new forum platform in part to solve these attachment issues. Simon
Marc, I think your simpler 2cmpt PK model is the better way to go, the AICs are generally lower, the visual fits look good and increasing to 3 compartments overparameterises the model as can be seen by the much larger CV% of the parameters. 
Other comments you know that you can link the outputs from PK models directly into the PKPD models e.g. Final Parameters and Dosing used; 
Lastly I’d mention that i personally would not swap between giving the dose in ug/kg for PK and then total ng for the PKPD model - i would normalise to your units of conc initially and then let it follow through. Probably I would use the absolute dose but that’s a personal preference probably to keep units simpler. Simon.
