Re: Design Evaluation for PK Substudies in Phase III

From mats@eddie.biof.uu.se Thu Aug 1 23:06:06 1996 Subject: Re: Design Evaluation for PK Substudies in Phase III Dear Ken, If your interest is not only to estimate the parameters of your preliminary model, but also to investigate the presence/absence of covariate relationships and/or the relationship between some exposure parameter and effects/side-effects in the trial, then I would try to include that in the simulations too. I wouldn't be surprised if you find that if you are most interested in relationships with average exposure (CL or CL/F) then a design is favoured that collects more information about CL in every subject, than a design for just the basic parameters would do. One simple way of doing it would be to create hypothetical covariates (e.g. 50/50 or 90/10 distribution) with similar PK characteristics, then estimate the relative CL values and look for deviations from one. More appropriate would of course be to use the empirical distributions of your covariates of interest in your study population, but that may not be easy to get. With respect to another of your questions, I would tend to look at SE's if I've had a small N, if it is large looking at SE's shouldn't be necessary. Best regards, Mats Karlsson