a question about unstable baseline for a soluble antibody target

Dear NMusers, I am trying to develop a TMDD model for an antibody against its soluble target in a monkey study. I have observations of the target before a single antibody IV bolus dose. My model has two compartments for antibody and target respectively. The challenging part is that the baseline levels of the target before the antibody dose increased instead of being stable. So I am trying to model the changing target baseline level. In my data file, the antibody dose is on Time=4 (d) instead of Time 0 (d). I have observations of target from day 0 to day 4. My question is that if it is possible to let NONMEM compute only the differential equation for target from day 0 to day 4 and compute both differential equations for antibody and target after day 4. If yes, how should I set up initial values for the two differential equations? Any suggestion and help will be appreciated. Thank you. Best, Xiaofeng

Hi Xiaofeng, There is no problem allowing the differential equation for the drug to be running prior to the first dose. The initial condition for DADT(drug) is zero and remains so until the dose is introduced. Modelling time-varying levels of soluble target, on the other hand, may not be so straightforward. Is it random, or is there a distinct pattern such that a function could be applied? If there is a pattern, then placebo data (i.e. not just the 4 days prior to dose) would be very useful to develop a model for the soluble target. Could it be that the target you are studying is responding to other factors, such as stress as the monkeys settle into the experimental routine? How could it change over time? Talk to the biological scientists to understand the possibilities. It is possible to model changes in the rate of target production or its clearance. One just needs to be inventive in thinking up suitable functions to feed in to the differentials, perhaps even linking these to other biomarkers. Playing with simulations prior to starting estimations in NONMEM really helps in this. I tend to use the Berkeley Madonna - it is far faster to explore model structures and come up with something that looks approximately like the data, before committing to a NONMEM run. Best regards, Phil. Physiologist, biochemist; Master Modeller Novartis Pharma AG, WSJ-027.6.25, CH-4056 Basel, Switzerland Phone: +41 61 324 4676; Mobile: +41 79 349 7806; [email protected]<mailto:[email protected]>