analysis of non-positive DV

Dear NONMEM users, I'm analyzing a dataset of 4 patients, there are 2 types of observations for each patient - vehicle (say baseline) and dose (treatment). We want to analyze our dataset in terms of (treatment-baseline). So I've modified my dataset, and now I have only one "column" of observations for each patient: DV= (treatment-baseline). But now I have some problems because I have some negitive values in DV column. In case of some models it results in problems, for example: TEFF = TBAS + (TSLP*CP)**TALP so TSLP should be positive but it's not that right... Maybe there are some common methods to deal with this case? Thanks a lot, Svetlana

Have a look at those two references: J Pharmacokinet Pharmacodyn. 2009 Oct;36(5):381-405. Epub 2009 Aug 21. Methods of utilizing baseline values for indirect response models. Woo S, Pawaskar D, Jusko WJ. J Pharmacokinet Pharmacodyn. 2008 Jun;35(3):269-83. Epub 2008 Apr 30. Approaches to handling pharmacodynamic baseline responses. Dansirikul C, Silber HE, Karlsson MO. Vinogradova Svetlana wrote: > Dear NONMEM users, > > I'm analyzing a dataset of 4 patients, there are 2 types of observations for each patient > - vehicle (say baseline) and dose (treatment). We want to analyze our dataset in terms of > (treatment-baseline). So I've modified my dataset, and now I have only one > "column" of observations for each patient: DV= (treatment-baseline). But now I > have some problems because I have some negitive values in DV column. In case of some > models it results in problems, for example: > > TEFF = TBAS + (TSLP*CP)**TALP > > so TSLP should be positive but it's not that right... > > Maybe there are some common methods to deal with this case? > > Thanks a lot, > Svetlana

You can do it that way, but you are throwing away data (the baseline effect). Better to model baseline and effect. May prevent negative values as well.

Svetlana, The answer depends on whether you would like to allow your effect to be negative or not. If yes, you need to rewrite the model so that the model allows negative predictions, for example, EFF = TBAS + TSLP*CP**TALP where TSLP is allowed to be negative. If not, then you can use the model that produces only the positive productions. In this case, negative values will be treated as a random noise (you need to have additive error component in this case), e.g.,: Y=EFF+EPS(1) or Y=EFF*(1+EPS(1))+EPS(2) Thanks Leonid -------------------------------------- Leonid Gibiansky, Ph.D. President, QuantPharm LLC web: www.quantpharm.com e-mail: LGibiansky at quantpharm.com tel: (301) 767 5566

Svetlana, Can you clarify your model a bit more in terms of what is a parameter and what is a covariate? You indicated that the DV is change from baseline. If so, why do you have TBAS in the model? Is DV = TEFF - TBAS? If so, it seems that your model should be written as: DEFF = (TSLP*CP)**TALP (where DEFF = TEFF - TBAS). Alternatively, you could define DV as the raw measurement following treatment and create another column on your dataset for the observed baseline response where TBAS is now a covariate rather than a parameter and then your model could be as defined below: TEFF = TBAS + (TSLP*CP)**TALP (where TSLP and TALP are parameters but TBAS is a covariate). Ken

try an additive IIV and/or RV, depending on your MOA and measurement error pattern. -------------------------------------------------------------------------- Confidentiality Notice: This message is private and may contain confidential and proprietary information. If you have received this message in error, please notify us and remove it from your system and note that you must not copy, distribute or take any action in reliance on it. Any unauthorized use or disclosure of the contents of this message is not permitted and may be unlawful.