95% prediction interval and $OMEGA

Dear NONMEM users, I am trying to calculate and plot the 95% prediction interval (PI) for a single-subject multiple-dosing PO dataset by simulating 1000 DV values. It seems that bigger initial estimate of omega ($OMEGA) leads to wider 95% prediction band. I understand that OMEGA directs the variability in "ERR(1)" in single-subject data, but I am still confused. Could anyone help me pick up a $OMEGA to calculate 95% PI, or solve this problem in another way? Thanks! Below is the control stream (the best-fit THETA values were used as initials): -------------------------------------------------- $DATA po.csv IGNORE=C $INPUT ID TIME CONC=DV AMT MDV CMT $SUBROUTINE ADVAN2 TRANS2 $PK CL = THETA(1) V = THETA(2) KA = THETA(3) S2 = V F1 = 1 $ERROR IPRED=F Y=F+ERR(1) $THETA (0.398) $THETA (64.3) $THETA (0.425) $OMEGA 1.2 $SIMULATION (324422) SUBPROBLEMS=1000 $ESTIMATION METHOD=0 NOABORT MAXEVAL=9999 PRINT=0 $COVARIANCE $TABLE TIME DV IPRED NOPRINT NOHEADER FILE= --------------------------------------------------------- Any input would be greatly appreciated. Thanks! Sincerely -- Kelong Han PhD student

Kelong, If you are really and truly only interested in a single subject then the only source of random variability will be the residual error (ERR). I would therefore use the final estimate of OMEGA you obtained from fitting the single subject in order to compute the prediction interval. I should say that what you are doing is very unusual. In the real world people are usually interested in more than one subject and so there are random effects for the PK parameters (ETA) as well as for the residual error (ERR). Are you really and truly sure that you are only interested in one single subject all by himself? Note that even within one subject there is typically dose to dose variation in the parameters (especially KA and F1) which you can model by adding an ETA random effect for each different dose. See Karlsson MO, Sheiner LB. The importance of modeling interoccasion variability in population pharmacokinetic analyses. Journal of Pharmacokinetics & Biopharmaceutics. 1993; 21(6):735-50. Nick Han, Kelong wrote: > Dear NONMEM users, > > I am trying to calculate and plot the 95% prediction interval (PI) for a > single-subject multiple-dosing PO dataset by simulating 1000 DV values. > > It seems that bigger initial estimate of omega ($OMEGA) leads to wider 95% > prediction band. I understand that OMEGA directs the variability in > "ERR(1)" in single-subject data, but I am still confused. > > Could anyone help me pick up a $OMEGA to calculate 95% PI, or solve this > problem in another way? Thanks! > > Below is the control stream (the best-fit THETA values were used as > initials): > > -------------------------------------------------- > $DATA po.csv IGNORE=C > > $INPUT ID TIME CONC=DV AMT MDV CMT > > $SUBROUTINE ADVAN2 TRANS2 > > $PK > CL = THETA(1) > V = THETA(2) > KA = THETA(3) > S2 = V > F1 = 1 > > $ERROR > IPRED=F > Y=F+ERR(1) > > $THETA (0.398) > $THETA (64.3) > $THETA (0.425) > > $OMEGA 1.2 > > $SIMULATION (324422) SUBPROBLEMS=1000 > $ESTIMATION METHOD=0 NOABORT MAXEVAL=9999 PRINT=0 > $COVARIANCE > $TABLE TIME DV IPRED NOPRINT NOHEADER FILE= > --------------------------------------------------------- > > Any input would be greatly appreciated. > > Thanks! > > Sincerely > -- > Kelong Han > PhD student -- Nick Holford, Dept Pharmacology & Clinical Pharmacology University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand [EMAIL PROTECTED] tel:+64(9)373-7599x86730 fax:+64(9)373-7090 www.health.auckland.ac.nz/pharmacology/staff/nholford

Hello Kelong, Did you try to put TRUE=FINAL in the $simulation record. I suggest also to input you parameters using MSFI Bests, Samer

Just to elaborate slightly on this last suggestion. If you use a $MSFI to input your parameters, use TRUE=FINAL to use the final, as opposed to the initial, estimates from that preceding run. If $MSFI is not used, TRUE=FINAL has no meaning. Mats Mats Karlsson, PhD Professor of Pharmacometrics Div. of Pharmacokinetics and Drug Therapy Dept. of Pharmaceutical Biosciences Faculty of Pharmacy Uppsala University Box 591 SE-751 24 Uppsala Sweden phone +46 18 471 4105 fax +46 18 471 4003 [EMAIL PROTECTED]

Thank you very much for this suggestion. Unfortunately I am using NONMEM V which does not have TRUE=FINAL option in $SIMULATION, but I will keep exploring it. Thanks! Kelong Han > > Just to elaborate slightly on this last suggestion. If you use a $MSFI to > input your parameters, use TRUE=FINAL to use the final, as opposed to the > initial, estimates from that preceding run. If $MSFI is not used, > TRUE=FINAL > has no meaning. > > Mats > > > Mats Karlsson, PhD > Professor of Pharmacometrics > Div. of Pharmacokinetics and Drug Therapy > Dept. of Pharmaceutical Biosciences > Faculty of Pharmacy > Uppsala University > Box 591 > SE-751 24 Uppsala > Sweden > phone +46 18 471 4105 > fax +46 18 471 4003 > [EMAIL PROTECTED] > > > > > > > > >