Hi Patricia, If the stopping of pump is an artifact and you are interested in getting parent-metabolite parameters without bias, I would approach in a progressive manner: 1. I would model parent IV data alone with an eta on Duration and then fix duration parameter with EBE (for doses that have this problem). 2. I would combine parent IV and oral data with fixed EBE of duration to see if other parameters are comparable to explain combined data. You may need to have oral bioavailability here. 3. I would extend the model to include metabolite compartments and estimate all parameters with duration of EBEs continued to be fixed. 4. Your current model may also be tried with fixed duration EBEs for VPC. You may get similar model from steps 1-3 but given complex model going on , I would check in different ways to be confident.
I also welcome comments/suggestions from the experts on this approach. Regards, Ayyappa > On Aug 6, 2020, at 1:43 AM, Patricia Kleiner <pkle...@uni-bonn.de> wrote: > > Dear all, > > first of all, thanks a lot for your fast and helpful replies and efforts. I > am currently running the model with your suggested expressions to describe > variability on infusion duration. > > To answer you question, Ayyappa, I intend to simulate population from my > model and I see that including variability on infusion duration would not > reasonable. > Using an individual modeling approach to estimate duration and fix in > population model is an interesting suggestion, but unfortunately I think > observations next to and after end of infusion were too sparse. > My dataset also includes concentration measurements after daily oral intake > and 2-hour infusion of the drug. An active metabolite of the drug is also > captured in my model. Both compounds could be best described with a three > compartment model. Visual predictive checks demonstrate that the parent drug > measured after 2-hour infusion is well described by the model (after oral > administration, no parent drug above lloq was observed in plasma), but after > 48-hour long-term infusion, variability is highly inflated (please see > attached PNG file). > This is why I was thinking about to implement variability on infusion > duration of the long-term infusion, but I am also thankful for any other > suggestion to improve the model fit. RE is modelled as additive error in the > log space. > > Thanks and best regards, > Patricia > > $SUBROUTINES ADVAN6 TOL=5 > > $MODEL > NCOMP=7 > COMP=(DEPOT,DEFDOSE) > COMP(CENTPRNT) > COMP (PERPRNT1) > COMP (PERPRNT2) > COMP (CENTMETB) > COMP (PERMETB1) > COMP (PERMETB2) > > $PK > ;; PK Parameters > TVKA=THETA(1) > KA=TVKA*EXP(ETA(6)) > > TVV2=THETA(2) > V2=TVV2*EXP(ETA(3)) > > TVCL1=THETA(3) > CL1=TVCL1*EXP(ETA(1)) > > TVQ3=THETA(4) > Q3=TVQ3 > > TVV3=THETA(5) > V3=TVV3 > > TVQ4=THETA(6) > Q4=TVQ4 > > TVV4=THETA(7) > V4=TVV4 > > FMET=0.6 > > F1=THETA(20) > IF(STDY.EQ.2) F1=(0.8*FMET) > > TVV5=THETA(8) > V5=TVV5*EXP(ETA(4)) > > TVQ6=THETA(9) > Q6=TVQ6 > > TVV6=THETA(10) > V6=TVV6*EXP(ETA(5)) > > TVQ7=THETA(11) > Q7=TVQ7 > > TVV7=THETA(12) > V7=TVV7*EXP(ETA(7)) > > TVCL2=THETA(13) > CL2=TVCL2*EXP(ETA(2)) > > TVALAG1=THETA(14) > ALAG1=TVALAG1 > > ;;scaling parameter > S2=V2/1000 > S5=V5/1000 > > ;;microconstants > K23=Q3/V2 > K32=Q3/V3 > K24=Q4/V2 > K42=Q4/V4 > K56=Q6/V5 > K65=Q6/V6 > K57=Q7/V5 > K75=Q7/V7 > K50=CL2/V5 > > $DES > C2=A(2)/S2 > C5=A(5)/S5 > > DADT(1) = - KA*A(1) > DADT(2) = - K23*A(2) + K32*A(3) - K24*A(2) + K42*A(4) > -((1-FMET)*((CL1/V2)*A(2))) - (FMET*((CL1/V2)*A(2))) > DADT(3) = K23*A(2) - K32*A(3) > DADT(4) = K24*A(2) - K42*A(4) > DADT(5) = KA*A(1) + (FMET*((CL1/V2)*A(2))) - K50*A(5) - K56*A(5) + K65*A(6) > - K57*A(5) + K75*A(7) > DADT(6) = K56*A(5) - K65*A(6) > DADT(7) = K57*A(5) - K75*A(7) > > $ERROR > IPRED=-5 > IF(F.GT.0) THEN > IPRED=LOG(F) > ENDIF > > IF(STRAT1.EQ.1) THEN ; PRNT after 2 hour infusion > W=SQRT(THETA(15)**2) > Y = (IPRED + W*EPS(1)) > ENDIF > IF(STRAT1.EQ.2) THEN ; METB after 2 hour infusion > W=SQRT(THETA(16)**2) > Y = (IPRED + W*EPS(2)) > ENDIF > IF(STRAT1.EQ.3) THEN; PRNT after 48 hour infusion > W=SQRT(THETA(17)**2) > Y = (IPRED + W*EPS(3)) > ENDIF > IF(STRAT1.EQ.4) THEN; METB after 48 hour infusion > W=SQRT(THETA(18)**2) > Y = (IPRED + W*EPS(4)) > ENDIF > IF(STRAT1.EQ.5) THEN; METB after oral administration > W=SQRT(THETA(19)**2) > Y = (IPRED + W*EPS(5)) > ENDIF > > IRES = DV-IPRED > DEL=0 > IF(W.EQ.0) DEL=0.0001 > IWRES = (IRES/(W+DEL)) > >> On Wed, 5 Aug 2020 14:55:02 -0500 >> Ayyappa Chaturvedula <ayyapp...@gmail.com> wrote: >> Hi Patricia, >> What is the purpose of your modeling exercise? I am not sure your scenario >> could be assigned to any particular distribution. If you intend to simulate >> population from the model, then your assumptions would not be reasonable. If >> you have rich data, you may try individual modeling approach to estimate >> duration and fix in population model. Regards, >> Ayyappa >>>> On Aug 5, 2020, at 1:04 PM, Bill Denney <wden...@humanpredictions.com> >>>> wrote: >>> Similar to Leonid's solution, you can try using an exponential distribution: >>> D1 = DUR*(1-EXP(-EXP(ETA(1)))) >>> The exponential within an exponential gives left skew and ensures that D1 ≤ >>> DUR. >>> For subjects who you know had an incomplete infusion duration, I would add >>> an indicator variable (1 if incomplete, 0 if full duration) so that the >>> subjects with complete duration have the known complete duration. >>> D1 = DUR*(1 - Incomplete*EXP(-EXP(ETA(1)))) >>> Thanks, >>> Bill >>> -----Original Message----- >>> From: owner-nmus...@globomaxnm.com <owner-nmus...@globomaxnm.com> On Behalf >>> Of Leonid Gibiansky >>> Sent: Wednesday, August 5, 2020 12:51 PM >>> To: Patricia Kleiner <pkle...@uni-bonn.de>; nmusers@globomaxnm.com >>> Subject: Re: [NMusers] Variability on infusion duration >>> may be >>> D1=DUR*EXP(ETA(1)) >>> IF(D1.GT.DocumentedInfusionDuration) D1=DocumentedInfusionDuration >>>>> On 8/5/2020 12:18 PM, Patricia Kleiner wrote: >>>> Dear all, >>>> I am developing a PK model for a drug administered as a long-term >>>> infusion of 48 hours using an elastomeric pump. End of infusion was >>>> documented, but sometimes the elastomeric pump was already empty at >>>> this time. Therefore variability of the concentration measurements >>>> observed at this time is quite high. >>>> To address this issue, I try to include variability on infusion >>>> duration assigning the RATE data item in my dataset to -2 and model >>>> duration in the PK routine. Since the "true" infusion duration can >>>> only be shorter than the documented one, implementing IIV with a >>>> log-normal distribution >>>> (D1=DUR*EXP(ETA(1)) cannot describe the situation. >>>> I tried the following expression, where DUR ist the documented >>>> infusion >>>> duration: >>>> D1=DUR-THETA(1)*EXP(ETA(1)) >>>> It works but does not really describe the situation either, since I >>>> expect the deviations from my infusion duration to be left skewed. I >>>> was wondering if there are any other possibilities to incorporate >>>> variability in a more suitable way? All suggestions will be highly >>>> appreciated! >>>> Thank you very much in advance! >>>> Patricia > > > > <VPC_48h_infusion.PNG>
