Mixture models in NONMEM

From: Subject: [NMusers] Mixture models in NONMEM Date:Thu, 17 Apr 2003 16:09:18 -0400 Dear all: I am trying out some mixture models in NONMEM, however, the models appear to be somewhat too sensitive to initial estimates. Specifically, this is a two compartment oral model with mixture on CL and V2, using the MIX=MIX function with the Advan4, Trans4 subroutine. Can someone pls. shed some light on the utility of the mixture models and the apparent sensitivity to initial estimates? Any feedback is much appreciated. Best regards, Rajesh Rajesh Krishna, PhD, FCP US Clinical Pharmacokinetics Aventis Pharmaceuticals, Inc. Mailstop: BWM-303A 1041 Route 202-206 P. O. Box 6800 Bridgewater, NJ 08807-0800

From: "Bhattaram, Atul" Subject:RE: [NMusers] Mixture models in NONMEM Date:Thu, 17 Apr 2003 16:32:13 -0400 Hello Rajesh It would be interesting to know why you selected MIXTURE model. Was it based on some prior information (phenotype) or some assumption testing. I remember a reference in which they used MIXTURE model because they had polymorphism in metabolism. Use of mixture model might help in identifying why you have two proportions which must be reflected as bimodality etc in your plots. Venkatesh Atul Bhattaram CDER, FDA.

From:"Sam Liao" Subject: RE: [NMusers] Mixture models in NONMEM Date: Thu, 17 Apr 2003 19:14:20 -0400 Hi Rajesh: The problem may not be the mixture model. The two-compartment oral model itself is very sensitive to the initial estimates. You need to find ways to get good initial estimates. Also try to model F1 with the mixture model because the first-pass metabolism may be affected by genotype. We may be able to help you further if you can share your control file. Best regards, Sam Liao, Ph.D. PharMax Research 20 Second Street, PO Box 1809, Jersey City, NJ 07302 phone: 201-7983202 efax: 1-720-2946783 _______________________________________________________