The Pill That Could · First in Human

First in Human, part 4

Hematologic response rate of one hundred percent. There is no comparable number for any other cancer drug at any other phase 1 trial. There are very few cancer drugs at any phase of any trial that produce response rates in the high nineties, and most of those drugs are tested in much more selected patient populations than the STI-571 phase 1 was. Druker, Sawyers, and Talpaz had a result that was, by any historical standard, off the chart.

* * *

The phase 1 results were presented at the American Society of Hematology annual meeting in December 1999, in New Orleans. The session in which Druker presented was scheduled in a room of the type ASH typically uses for sessions on uncommon diseases or specialized methodology — a room of moderate size, several hundred seats, in a part of the convention center where you would expect a moderate audience of specialists. By the time the session began, the room was overflowing. Hematologists who had heard rumors about what STI-571 was doing had come from other sessions to see the data. The hallway outside the room was crowded. The session, by the standards of ASH, was a sensation.

Druker presented carefully. The data were presented as preliminary, with the appropriate caveats: the follow-up was short, the patient numbers were modest, the durability of responses was unknown, the safety profile would need to be confirmed in larger studies. None of this kept the audience from understanding what they were seeing. The cytogenetic response rates, in particular — the proportion of patients who were not just responding clinically but actually clearing the Philadelphia chromosome from their bone marrow — were the numbers that the audience kept fixing on. Cytogenetic response had been, in CML therapy, the holy grail of interferon: the durable cytogenetic responses produced by interferon were the ones that translated into long-term survival, and those responses occurred in only a small fraction of patients. STI-571 was producing them, in the early data, in most of the patients at therapeutic doses, in a matter of months.

By the end of the session, several hundred hematologists had taken in the same set of data. Most of them came away with the same conclusion: this was, in the technical sense of a phrase that gets overused but that occasionally turns out to be appropriate, a paradigm shift. Cancer therapy did not normally work this way. It was working this way now.

The 2000 New England Journal of Medicine paper, when it appeared in April of that year with Druker, Talpaz, Sawyers, and others as authors, made the same case in print. It reported the chronic-phase results of the phase 1 trial. The results were what the audience at ASH had heard about: complete hematologic response in essentially every patient at doses of 300 milligrams or more, complete cytogenetic response in a substantial fraction. The drug was well tolerated. Toxicity was manageable. The maximum tolerated dose had not, in fact, been reached; the trial had escalated past 600 milligrams a day without the kind of dose-limiting toxicity that would normally cap a phase 1 study. The drug was simply working, at every dose at which it was given, with no real ceiling on how much you could give.

A companion paper in the same issue reported on the blast-crisis arm of the trial. Blast crisis is the late, transformed phase of CML, the phase in which the disease has acquired additional genetic changes and become refractory to most therapy. The hope had been, going into the trial, that STI-571 would do something useful in blast crisis even if other therapies could not. In the event, blast-crisis responses were less durable than chronic-phase responses; many patients responded initially and then relapsed within months, with the leukemic cells acquiring resistance to the drug. (Why this happens, and what to do about it, is the subject of chapter nine.) But even in blast crisis, where the prognosis had been measured in weeks, STI-571 was producing meaningful responses in a substantial fraction of patients. The drug was useful at every stage of the disease where it had been tested.

Together, the two New England Journal papers established — in the formal venue where this kind of thing gets established — that something unprecedented had happened in a phase 1 trial.

* * *

It is worth pausing here, because the story we are now in is going to keep accelerating, and there is a thing about the phase 1 results that is easy to lose track of in the rush.

What had happened, from June 1998 to April 2000, was that a hypothesis articulated in the late 1980s had been confirmed in human beings. The hypothesis was this: a small molecule that selectively inhibits BCR-ABL will turn off CML in patients without seriously harming them. There were many ways the hypothesis could have failed. The drug could have been too toxic. It could have failed to reach its target in vivo. The cells could have responded by upregulating other pathways. The disease could have evolved out from under the inhibitor. Patients could have died of side effects unrelated to BCR-ABL that no one had anticipated. The hypothesis had to survive, in turn, every one of these possible failures.

It survived all of them. The drug was well tolerated, reached its target, shut down BCR-ABL signaling in the cancer cells, and produced clinical and cytogenetic responses on a scale that nobody, including the people running the trial, had expected. The forty-year arc from Nowell and Hungerford's small chromosome to Druker's clinic had ended in a drug that did, in patients, exactly what the science said it should do.

This kind of thing is rare in cancer medicine. The history of cancer drug development is largely a history of drugs that work less well than the preclinical data suggested they would, of compounds that failed in the move from animal to human, of mechanisms that turned out to be more complicated in patients than in dishes. The instances in which the molecular logic translated cleanly into clinical benefit, with no major surprises along the way, are few enough to be enumerable. STI-571 in CML was one of the cleanest.

It was also, although nobody quite knew this yet in April 2000, a stress test of the regulatory system. A phase 1 result this striking creates a problem for the FDA. Phase 1 trials are not, by design, the trials on which approval decisions are made; the standard regulatory path requires phase 2 and then phase 3 randomized trials, with proper comparators, before a drug is approved for general use. But STI-571 had effectively skipped past the kind of question phase 2 is meant to answer. There was no doubt that the drug worked in CML. The question was how to get it into patients' hands as quickly as possible without breaking the framework that protects patients from drugs that don't work.

The next chapter is about what the FDA did about that question.

It is a story about a regulator confronting a drug whose evidence of efficacy was so unambiguous that the usual procedural questions about how much evidence is enough became, briefly, almost beside the point.

It is also a story about how, on May 10, 2001, in a press release that would become a small landmark in oncology, a drug called Gleevec was approved for the treatment of chronic myeloid leukemia, two and a half months after Novartis filed its application.

Two and a half months.

Patients in Druker's clinic, by then, had been on the drug for almost three years.

All 13 chapters of The Pill That Could · Library · All courses