Of the drugs in development, or even on the market, which best represent the direction you think that discovery in oncology is moving?

The poster child, of course, is Gleevec [approved for chronic myeloid leukemia, or CML]. There, you can make a diagnosis and if the patient has the Philadelphia chromosome, then you know exactly what subtype of disease you're looking at. You have a specific drug that targets the specific molecular alteration, and that means its side effects are few and its efficacy is extraordinarily high. So we're trying to turn more common cancers into the situation that you see with CML.

Some other new targeted therapies include Tarceva [approved for non-small cell lung cancer and pancreatic cancer] and the anti-VEGFs [vascular endothelial growth factors], like Avastin. There are additional interesting compounds hitting other pathways, such as the IGF [insulin-like growth factor] pathway, the PI-3 kinase pathway, the WNT pathway, the ubiquitin and ubiquitin-like pathways, to name just a few.

We'll wrestle this to the ground. We'll have an up-front therapy and a whole range of strategies: If the cancer comes back, we're going to use this combination of drugs, and when it comes back after that, we're going to use that combination. The idea is to turn what could be acutely life-threatening into, basically, a chronic disease that we manage.

What do companies need to be doing now to make that happen? Or is it more than companies? Is it academic researchers? Is it FDA?

With respect to the application of biomarkers and patent stratification to early clinical trials, it absolutely must be done at the company level—but in close partnership with the clinical centers where the trials are being conducted. Over the three days that we talked about projects at this drug discovery and development meeting, I can't remember a single company that does not have a clear biomarker strategy to figure out whether their drug hits the target intended. So it is now embedded in every project that I'm aware of. The old standard of imaging analysis to see if the tumor shrinks does not work anymore, because a lot of these new compounds are not cytotoxic but cytostatic—they're intended to stop the tumor from growing but do not necessarily result in single-agent tumor shrinkage.

All across the board, biomarker strategies and new ways to design clinical trials are ongoing. And, again, at the heart of targeted therapy is the idea of understanding the biology of cancer and then translating that into clinical medicine.

When biomarkers are available, does that offer a way out of the trap of having to start every drug with the sickest patients?

No, it doesn't. You still have to deal with the same patient populations, but it allows you to ask specific questions about your compound to make sure you've gotten enough compound into the patient at a safe enough level to see biological implications. And then—whether by taking blood, skin, or tumor samples—you can ask, "Has the compound actually hit the target, and if so, has it shut off the pathway that leads to either the growth or the survival of those specific types of cancers?"

And that's one way to think about setting the doses of these therapeutics as opposed to just giving increasing doses over the course of time until something toxic happens and you have to stop. It's a more informed way of doing clinical research, but you're still starting with the same patient population.

Do you also work with them in Phase III?

In principal, once you define your dose and schedule and can relate the plasma levels of a drug to an efficacious exposure, then perhaps the Phase III design might be less dependent upon biomarkers. However, this assumes that the drug combination chosen for the Phase III trial will not somehow modify the behavior of the experimental compound. This is one of the reasons that companies are conducting larger, randomized Phase II trials with designs similar to the potential Phase III design but on a smaller scale.