From: Kristel Crommentuyn, pharmD Apkcr@SLZ.NL Subject: [NMusers] problem with acquiring post hoc estimates Date: Tue, May 18, 2004 11:21 am Dear all, We are experiencing problems with a pk model, in acquiring post hoc estimates and hope someone will be able to help us The model describes the pharmacokinetics of an orally dosed compound and its metabolite in children. The data consists of complete pharmacokinetic curves as well as sparse samples for 38 subjects at different occasions (1-28). Both parent and metabolite data are best described with a one-compartment model with first-order elimination. Cl and V are allometrically scaled for body weight. The frequency of administration is a covariate on Cl1. We would like to generate posthoc estimates for the Cmin of the parent compound and metabolite to use in a pharmacodynamic analysis. When we use the controlstream (below) and dataset (enclosed test1.csv) the Cmin posthocs for both compounds are equal, whereas at the other time-points they differ. $PK TVCL1=THETA(1)*THETA(6)**FREQ CL1=TVCL1*(WT/18)**0.75*EXP(ETA(1)) TVV1=THETA(2)*(WT/18)**1.0 V1=TVV1*EXP(ETA(2)) K12=CL1/V1 D1=THETA(3)*EXP(ETA(3)) TVCL2=THETA(4)*(WT/18)**0.75 CL2=TVCL2*EXP(ETA(4)) TVV2=THETA(5)*(WT/18)**1.0 V2=TVV2*EXP(ETA(5)) K20=CL2/V2 F1=1*EXP(VF) S1=V1 S2=V2 I hope that someone can help us with this. Best regards, Kristel ___________________________________ Kristel Crommentuyn, pharmD AIDS research pharmacist Department of Pharmacy and Pharmacology Slotervaart Hospital Louwesweg 6 1066 EC Amsterdam The Netherlands apkcr@slz.nl 020-5124737 _______________________________________________________ From: Leonid Gibiansky lgibiansky@emmes.com Subject: RE: [NMusers] problem with acquiring post hoc estimates Date: Tue, May 18, 2004 12:01 pm Kristel Few questions: It would help to see a full control stream, not only the PK block. What is VF ? Why CL1 depends on frequency of administration (THETA(6)**FREQ) ? How Cmin is defined? Why oral dose is modeled as zero-order bolus dose (infusion with duration D1) rather than via absorption KA? I am not 100% sure but I would use EVID=1 rather than EVID=4 at the first line of your data set. Thanks Leonid _______________________________________________________ From: Leonid Gibiansky lgibiansky@emmes.com Subject: RE: [NMusers] problem with acquiring post hoc estimates Date: Tue, May 18, 2004 1:41 pm Kristel I do not know why, but with your control stream and data, NONMEM was using F in compartment 1 for EVID =2 (only!) even when CMT was equal to 2. You may use values that correspond to EVID=0 (3d and 4th lines of your data file) or modify the code as below and use C1, C2 or IPRED/PRED (that are computed explicitly using A1 and A2 rather than F). Leonid $ERROR C1 = A(1)/S1 C2 = A(2)/S2 IPRED=A(1)/S1 IF(CMT.EQ.2) IPRED=A(2)/S2 IF(CMT.EQ.1)THEN Y=IPRED+IPRED*THETA(7)*ERR(1)+THETA(8)*ERR(2) W=SQRT(IPRED*IPRED*THETA(7)**2+THETA(8)**2) ELSE Y=IPRED+IPRED*THETA(9)*ERR(3)+THETA(10)*ERR(4) W=SQRT(IPRED*IPRED*THETA(9)**2+THETA(10)**2) ENDIF IRES=DV-IPRED IWRES=IRES/W $TABLE .... CMT C1 C2 IPRED .. _______________________________________________________