CANCER R&D IS BOOMING RIGHT NOW. At a time of poor ratings on both Wall Street and Main Street, pharma can at least point proudly to its oncology pipeline as proof that it still takes big risks to make big advances against big killers—and win. According to a recent IMS report, the cancer pipeline contains 380 compounds, with nearly 100 in Phase III. The long-established standard of care—surgery, radiotherapy, and chemo—is fast giving way to a high-tech array of targeted therapies. These molecules and antibodies are designed to block specific disease pathways, and they are proving both far more effective and far more tolerable than the sledgehammer status quo. Since 1996, the overall survival rate for patients has jumped by 30 percent, from one-half to two-thirds.

Given this gold rush, it's a bit of a shock to hear distinguished cancer researcher Joseph Bolen bemoan the current state of his art. The chief scientific officer of Millennium—the Cambridge, MA–based biotech that shot out Velcade, for multiple myeloma, in a mere four and a half years—points out that oncology R&D has the lowest success rate of any therapeutic area. But Bolen is not one to sit back and complain. As chairman of last August's 12th Annual World Congress on Drug Discovery and Development of Innovative Therapeutics, in Boston, Bolen held a panel called "Why Is Cancer Drug Discovery So Difficult?" We caught up with him afterward to get the answer—and what to do about it.

So tell us—why is cancer drug discovery so difficult?

So that's quite a challenge from a number of points. Once you've made the substantial investment to do all the early work that goes into filing the IND [Investigational New Drug application], the compound doesn't change. Either you've gotten it right or it's going to fail—there's nothing different about oncology there. What is unique about oncology is that Phase I clinical trials, for the most part, are in patient populations rather than in healthy volunteers. And, of course, the patients that you're seeing in Phase I, which is supposed to be just a safety study, are basically patients who have failed virtually every other therapy. So there is a real hurdle in order to get any glimpse of efficacy at all. That has put a tremendous burden on trying to sort out relevant biomarkers.

What we're faced with in Phase I is the accumulated evolution of tumors that have survived previous therapies and metastasized. Trying to assay what's going on inside the tumor is also quite difficult because, again, these patients are usually not feeling well. The idea of a biopsy to see if your drug is doing something is quite a hurdle to get over.

Generally, the majority of our current animal models are poorly predictive of what we are looking at in the clinic. You have a great variety of tumor phenotypes in the same patient—rarely are they clonal [genetically identical]. And unlike many diseases, cancer is a collection of proliferative diseases that represent hundreds, if not thousands, of different specific diseases. There are also enormous differences among patients.

So it's almost miraculous that you see anything in Phase I to figure out early on what your compound is actually doing. What you're really trying to do is move to a Phase II trial where you are picking specific cancer types, and you're also trying to determine a dose and a schedule for patients.

And if you manage to get to Phase III, you're rarely using a single agent, because cancer patients are normally treated with multiple drugs at the same time. You take the standard of care and add your drug on top of it. So it's little wonder that, with the complexity here so much greater than for many other indications, cancer drugs generally fail in Phase III—the most expensive part of drug development.