[Newbie Help] Simulated terminal elimination phase is too high and fails to decline, showing a significant discrepancy with observed data - seeking guidance

[zhou-xl1202]

Hello everyone,
I am a newcomer to PK-Sim and currently teaching myself how to use the software. I've encountered some difficulties while trying to build/reproduce a PBPK model and am hoping to get some guidance from the experts and community members here.
The main issue I'm facing is a significant discrepancy between my simulated concentration-time profile and the observed data. Specifically, the simulated terminal elimination phase remains too high; the drug doesn't seem to be effectively cleared, and the blood concentration declines very slowly.
I have initially checked the compound's physicochemical properties and clearance parameters, but due to my limited understanding of the software's underlying mechanisms, I haven't been able to identify the root cause of this slow elimination.
To make it easier for you to diagnose the issue, I have attached the following files:
My observed data
My PK-Sim project file (.pksim5 format, containing all my current settings)
Could the experts here advise on which parameters or model settings should be prioritized when troubleshooting this typical "high/slow elimination phase" scenario? (e.g., clearance definition, enzyme kinetics, distribution model selection, etc.)
help.zip

Thank you very much for taking the time to look into my issue! I look forward to your replies and guidance.

[github-actions[bot]]

🤖 Discussion Assistant: I found some similar discussions that might be helpful:

• PBPK Model Problem: Extremely Fast Elimination (81% similar)
• The concentration of the elimination phase is constant. (77% similar)
• PK-Simbegineer (76% similar)
• elimination pathways (75% similar)
• IV furosemide PBPK model: Cmax underprediction and transporter (OAT1/OAT3/MRP4) parameterization in PK-Sim (75% similar)

[tobiasK2001]

Hi,

in case you haven't found the solutions in the answers from our bot, just some quick thoughts:
Your rat individual misses the CYP3A3 enzyme (without that the metabolisation of midazolam via it's main metabolic process will not work

Solution: add CYP3A4 in the individual BB

update your rat in the simulation form BB

now you can match the processes in the configuration (should be already automatically matched)
window

rerun the updated simulation. It looks a tiny bit better but not enough CYP3A4 CL (assumption: in vitro process form publication does not translate correctly to in vivo?)
Fit Kcat CYP3A4 Process in PI, Remove all the other processes you tried as they are not relevant here

rerun PI,

It gets better, but all results run to very high Kcat, and the profile looks better, but not really great

So maybe the assumption that Midazolam is cleared in the rat via CYP3A4 only, is a bit far fetched, but at least you know know how to connect metabolism to enzymes in the individual and how to optimize. Transporters and Protein binding partners work the same way. You can figure out the relevant ones from literature and test the impact in your model.

Good luck, Tobias
attached is the updated PK-sim file
midazolam-rat-TK.zip

P.s. check out the tutorials hands on workshops at the OSP-CC e.g. here: https://www.eventleaf.com/Attendee/Attendee/EventPage?eId=O3Aw6hfGQb%2FOfI3ULSNaAw%3D%3D#ss

[zhou-xl1202]

Thank you for the detailed explanation. I gained a lot of insight from it.

[tobiasK2001]

Glad to hear :-)