modeling interoccasion variability

From: "Panetta, Carl" Carl.Panetta@stjude.org Subject: [NMusers] modeling interoccasion variability Date: 11/18/2003 11:24 AM I have a question on modeling interoccasion variability (IOV). I have read through and used the method of handling IOV discussed in Karlsson and Sheiner (JPP 1993) and it works fine when there are 2 or 3 occasions per individual. But, in situations where there are many more than 3 occasions per individual and each individual does not necessarily have the same set of occasions, the method becomes more difficult to implement and NONMEM does not always converge successfully and/or give acceptable results. For example, I am working with a set of data where there are 10 different occasions. Not every individual has PK samples for each of the 10 occasions, but each individual has about 3 to 5 occasions. Thus, while everybody has an occasion 1 and most have an occasion 3 only a few have occasion 2. Therefore, instead of having a measure of variability for each of the 10 occasions separately, I would like to have a single overall measure of variability for the IOV for clearance (and other PK parameters) in a similar manner as I can get a single measure of the interindividual variability measure. Does anybody have any suggestions, references, etc. on if and how this can be done? Thanks, J. Carl Panetta, Ph.D. Department of Pharmaceutical Sciences St. Jude Children's Research Hospital 332 N. Lauderdale St. Memphis, TN 38105 Phone: (901) 495-3172 Fax: (901) 525-6869 Carl.Panetta@stjude.org

From: "Panetta, Carl" Carl.Panetta@stjude.org Subject: RE: [NMusers] modeling interoccasion variability Date: 11/18/2003 1:12 PM After doing a bit more reading, I realized I was not using the SAME option in the OMEGA block! This is what I was looking for. While I think I have figured out my main problem, any other comments or suggestions on working with IOV would be appreciated. Thanks, J. Carl Panetta, Ph.D. Department of Pharmaceutical Sciences St. Jude Children's Research Hospital 332 N. Lauderdale St. Memphis, TN 38105 Phone: (901) 495-3172 Fax: (901) 525-6869 Carl.Panetta@stjude.org

From: Nick Holford n.holford@auckland.ac.nz Subject: RE: [NMusers] modeling interoccasion variability Date: 11/18/2003 1:17 PM Carl, It sounds like you should be using the SAME option: "instead of having a measure of variability for each of the 10 occasions separately, I would like to have a single overall measure of variability for the IOV for clearance" This means that you only estimate one OMEGA which is the same for all occasions. The ETA for each occasion will be different because they are sampled from a different distribution with the SAME variance. I don't know of any intrinsic reason why having different numbers of occasions for your subjects should make it hard to estimate BOV (between occasion variability aka IOV). Here is an example showing how to code this for NM-TRAN: $OMEGA .5 ; BSVCL $OMEGA BLOCK(1) .1 ; BOVCL1 $OMEGA BLOCK(1) SAME ;; BOVCL2 ... $OMEGA BLOCK(1) SAME ;; BOVCLn $PK IF (OCC.EQ.1) THEN BOVCL=ETA(2) ENDIF IF (OCC.EQ.2) THEN BOVCL=ETA(3) ENDIF ... IF (OCC.EQ.n) THEN BOVCL=ETA(n+1) ENDIF CL=POPCL*EXP(ETA(1) + BOVCL) -- Nick Holford, Dept Pharmacology & Clinical Pharmacology University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand email:n.holford@auckland.ac.nz tel:+64(9)373-7599x86730 fax:373-7556 http://www.health.auckland.ac.nz/pharmacology/staff/nholford/ _______________________________________________________