Article

Top Article References

The article in this issue by White and Manitpisitkul (2001) is an excellent and balanced discussion of an important yet specialized technology, cassette dosing pharmacokinetic screening. This approach, developed in the early 1990s at Glaxo, is both widely used and vigorously debated. While this technology will have little application for the academic investigator, everyone can benefit from this work as an exercise in critical thinking. Too often new approaches or technologies are adopted without debate and dissection.

We must accept that, as defined, cassette dosing is a screening tool for the pharmaceutical industry. The utility of any tool is always directly related to the skill with which it is used and its appropriateness for the task at hand. The paper clearly delineates potential advantages, disadvantages, and assumptions important in evaluating the tool. The mathematics and enzyme kinetics are indisputable. The clear framework will allow both potential and current users to appreciate the assumptions inherent in using this approach, and therefore a more sophisticated application and interpretation of these studies should result. The real questions, however, are these: does this approach produce "good" pharmacokinetic data? and is it right for me? These must be answered in the context of alternative methods and the variability of the results produced, and of how well the approach meets its goal. One question is scientific, while the other is a decision based on individual tolerance for risk in new drug discovery. Answering these questions requires that we define our objectives precisely.

Scientifically, one central tenet of the manuscript is that that magnitude of interactions is potentially large, varying as much as 2- to 5-fold from the "true" values. But what are the true values, and what is the variability expected normally in their estimation? A cursory review of pharmacokinetic studies published in 2000 in Drug Metabolism and Disposition and other journals shows that the average number of animals studied after oral or i.v dosing is 3 to 6, and that the standard deviations of the reported pharmacokinetics range from ± 10% to 50%. Therefore, even discrete pharmacokinetics, or the plasma concentrations in individual animals, will range at best approximately 2-fold. The error potentially incurred with cassette dosing for any single compound may therefore be well within the variability encountered in discrete pharmacokinetic studies with relatively few animals. This is not an argument that all cassette dosing pharmacokinetics are as accurate as discrete pharmacokinetics, but rather points out that the variability observed in the published studies, cited in Table 1, is consistent with the inter- or even intra-animal or subject variability observed in discrete pharmacokinetic studies. Presenting or interpreting biological data, including discrete or cassette dosing pharmacokinetics, as physical constants is always dangerous.

The second question raised by the article, of whether this approach is right for me, is more difficult to answer. It is also indisputable that cassette dosing pharmacokinetics and the use of liquid chromatography/tandem mass spectrometry have greatly accelerated our ability to profile many more drug candidates with fewer resources. How good must the data be to justify the decisions? Many companies, including DuPont, have used this approach successfully, as defined by nominating candidates for clinical evaluation faster than could have been possible using conventional approaches. Providing medicinal chemistry and pharmacology with data and perspective rapidly is critical for reducing synthesis and testing cycle times. The suggestion by Obach and colleagues (1997), reinforced by Ward et al. (2001), to view potential drug candidates in terms of appropriate categories is a very practical and useful exercise, and it is supported further by White and Manitpisitkul. For those programs with many molecules to evaluate, winnowing the number for further evaluation may be more important than selecting "the" right one. Moreover, compounds are not chosen in a vacuum. The other physicochemical, biological, and pharmaceutical properties of the molecules are always considered in the decisions.

Should the limitations outlined by White and Manitpisitkul cause us to abandon cassette dosing? The answer must be no, because like all models, the final judgment lies in whether they are useful, not whether they represent absolute truth. This technology, if applied with the proper recognition of the pitfalls, can provide data that are meaningful, and in some cases, within the routine variability of all pharmacokinetics, discrete or otherwise. While other methods of accelerating candidate evaluation are promising, they are complementary and carry their own assumptions and limitations.