Some basic modeling results

From: "Willavize, Susan A" susan.a.willavize@pfizer.com Subject: [NMusers] Some basic modeling results Date: Thu, 5 Oct 2006 08:26:05 -0400 I will be meeting in about 3 hours time with a fellow student at Univ Buffalo. We will be discussing nonlinear mixed effects modeling of PK/PD data. I am relatively new to this area of study. There are two things that I recall being taught which I wish to discuss with him. 1.) that the FO method tends to provide more biased estimates of population mean parameters than FOCE. And 2.) that if we fit a between subject variance model (Omega) that is too simple (i.e. diagonal when it should be including covariances) then our estimates of the variability of the fixed effects parameters (and predictions of mean responses) are likely to be underestimated. Are these two things true? Can you show me literature references about this? Thank you in advance for helping me with my homework. Susan

From: Nick Holford n.holford@auckland.ac.nz Subject: Re: [NMusers] Some basic modeling results Date: Fri, 06 Oct 2006 10:09:37 +1300 Susan, Evidence for the bias in FO can be found in the earliest publications describing NONMEM's performance with simulated data. I cannot at the moment produce a published reference which directly compares the two estimation methods. I expect someone else can do that. The second issue is trickier. I have been working on a simulation project for some years now with Diane Mould and Joga Gobburu that has tried to answer your question. We have still not reached a fully satisfactory conclusion but my own impression is that simply using the diagonal for OMEGA does not consistently bias the fixed effect parameters when off-diagonal elements are included in OMEGA for simulating the data. Nick -- 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/ _______________________________________________________