tools and science

From davide@ariel.ucsf.edu Tue Jul 9 08:38:11 1996 Subject: tools and science Dear Rik, in reference to your "summary" "-some examples of the use of NONMEM in drug development are: "-analyses which allow simulations of proposed dosage regimens before the drug is marketed. "-meta-analysis. "-to allow analysis of experimental design for phase I PK "studies, with reduced sampling times because of ethical and/or "economical reason. "-where very little pharmacokinetics are available from formal "Phase I pharmacokinetic studies and data from Phase II/III "clinical trials are used for population pharmacokinetic "analysis. I think that what might make people suspicious about the claim of "essentiality" emerges from this "summary" where many items are pooled together without really separate what is specific to nonlinear mixed effect modeling and what it is not. Going through the items: simulations: they can be done with many software packages meta-analysis: this is a really complicated topic which is not even well accepted outside PK/PD modeling. The reliability of meta-analysis to reach scientific conclusions has little to do with the tools used in the meta-analysis itself. The problem is related to mixing different studies (with, e.g. different quality control) not to how to analyze the mixture! reduced sampling: I think that a discussion is needed about what one can realy achieve doing a spare sampling phase I (or II/III) study. Of course drug companies love it, but for the wrong reason: it saves their money. That is why they are all buying it! To think otherwise is really naive. Some questions here are: Can "the mechanic" of a drug (as Jane Wade put it) be determined using sparse sampling? Experience with modeling suggests that this is not in general the case. (It is easy to come up with examples, and it is strange to think that people in regulatory agencies think otherwise.) Can sparse sampling be used to obtain, necessarily approximate, rough estimates of population mean PK parameter (Volume, Clearance, ...) when some mechanic of drug is already known. Experience suggests: yes, and this might be all that is necessary to find out from a regulatory point of view. (note that none of the items are related, per se, to nonlinear mixed effect) overall: I agree with Holford/Steimer. NONMEM is a tool and a real good one! What is somewhat annoying is to see some people talk or use it as if it was something magic which can make by itself bad (or good) science turn into something better. Bests, Davide