Hi, I need help in performing simultaneous fitting in Phoenix WNL using a user-defined model. A simple example is fitting mean conc.-time data following IV bolus administration at two dose levels to a one-cmpt model. In Phoenix WNL, I am able to fit each data separately but not simultaneously. In the old WNL, I would use the “Function Variable” to designate the data to the different functions (dose levels). I would appreciate if someone could provide a solution (I do not have NLME option and so cannot analyze this data as a PopPK data set). Thank you in advance Anjaneya
Hi Anjaneya, The Function Variable option in WinNonlin 5.3 is for cases in which the data are truly from two different models. This would be the case if the data are from two different compartments (e.g., plasma and urine data). If the data are all from the same compartment, and therefore can be modeled with the same function, then Function Variable has no effect. That is, using Function Variable is the same as only specifying the X variable and Y variable for the modeling, and having the data pooled during the analysis. If you look at the plots when you use Function Variable, you should see that the model fit looks like it fits through the average of the data, which is the same as doing a naïve pooled fit. I even tried several examples to make sure the algorithms are working exactly the same, and I got answers that were identical in all 6 decimal places in every case. That is, when running the models, the results were identical when Function Variable was used and when it was not. So Function Variable in 5.3 was not actually doing the simultaneous fit as you wanted. Note: in Phoenix, you can indicate that urine data is being included in the analysis by selecting “Elim. Cpt.” in the Structural Model Tab. If you decide to try Phoenix NLME, although you could simultaneously model the average profiles of two different dose groups, I would recommend including all of the individual profiles. You can input the dose levels and also sort by ID, to optimally leverage all of the information in your data. If you have any questions about which type of modeling might work best for the case you’re considering, we could discuss that further. Best regards, Helen Moore
Hi Helen, Thanks for your response. Following some additional discussions with Pharsight tech. support, I think there is no “easy way” using Phoenix WNL to simultaneously fit plasma PK data from multiple dose levels to the same model (apparently there is a tricky workaround which tech support can help on an individual basis). This issue will apparently be fixed in the next release. With regard to the use of urine data by selecting the “Elim. Cpt.” option, I originally wanted to fit only plasma conc. time data from multiple dose levels to the same model. I dont have urine data for fitting (I apologize, I should have framed my question more clearly) With regard to using NLME option, I agree that using the individual data would be more meaningful. However, as I mentioned in my original post, I dont have NLME option. Also I intend to use this for preclinical PK data sets wherein the data is relatively tight and so using a Pop. PK approach may be an overkill. Thank you Anjaneya
Hi Anjaneya, You can fit the DME model in a similar manner as you did for 5.3. Simply create a column for each “function” (using the pivot toll for example). Then assign separate observation blocks for each fit. eg: deriv(A1 = -(ClCiv)-(Cl2(Civ-C2))) deriv(A2 = (Cl2*(Civ-C2))) deriv(Aa = -(AaKa)) Civ = A1/V C2 = A2/V2 dosepoint(A1) error(CEps=0.176814) observe(CObsIV = Civ+CEps) dosepoint(Aa,bioavail=BA) deriv(A3 = -(ClCpo)-(Cl2*(Cpo-C4))+(AaKa)) deriv(A4 = (Cl2(Cpo-C4))) Cpo = A3/V C4 = A4/V2 observe(CObsPO = Cpo+CEps) This is the same as would have been coded in 5.3 ASCII (see example PK 19 in Dan’s Book), with functions being replaced by observation blocks. there is no need for the functions to be from different conceptual models. I agree that it’s not as user friendly as it could be, but is is on equal footing as the 5.3 ASCII coding IMHO. There is an even simpler method taking advantage of the new “reset” function in DME, but i wont get into that here for now. martin