Dear Peter,
Thank you for your interesting question which is indeed a design problem.
Using one of the tool available for optimal design in nonlinear mixed effect models (PFIM, WinPOT, ....) , you can also evaluate a given design (with a priori values of the pop parameters and the model) and then get what is the expected RSE for the estimation of the mean clearance (I rather RSE than CV to separate from variability). If nothing is changed in the strcuture of the elementary design (for instance all patients with the same elementary designs), this RSE varies with the square root of the number of subjects (you need to multiply by 4 the number of patient to decrease by 2 the RSE)
We have discussed a rather similar approach which was based on the a priori power for a test of one covariate in population analysis in the following paper.
Retout S, Comets E, Samson A, Mentre F. Design in nonlinear mixed effects models: Optimization using the Fedorov-Wynn algorithm and power of the Wald test for binary covariates. Stat Med. 2007 May 8; [Epub ahead of print]
The power and number of subjects needed were computed using the RSE of the fixed effect for the covariate. We used PFIM but the tool for evaluation of the matrix when there is a covariate is not avilable on the website yet.
You might want to send the question also to the PopDesign list (You can subscribe to the list at the website: http://lists.otago.ac.nz/listinfo/popdesign ), as many of the statistician of this list are not on the NONMEM user list
Best wishes
France Mentré and Emmanuelle Comets
PFIM: www.pfim.biostat.fr
---------------------------
France Mentré
Professor of Biostatistics
INSERM U738 - Université Paris 7
UF de Biostatistiques - CHU Bichat Claude Bernard - AP-HP
UFR de Médecine - Site Bichat
16 rue Henri Huchard
75018 Paris, France
tel: 33 (0) 1 44 85 62 71
sec: 33 (0) 1 44 85 62 80
fax: 33 (0) 1 44 85 62 83
email : [EMAIL PROTECTED]
PFIM: www.pfim.biostat.fr
MONOLIX: www.monolix.org
npde: www.npde.biostat.fr