The Pill That Could

Chapter 11: The Generalization Problem

Imatinib worked in CML for a reason. The reason is the subject of this chapter, and the reason is also why imatinib's success has not been quite as easy to repeat as the celebratory coverage of 2001 implied.

The reason is that CML is, by the standards of human cancer, almost embarrassingly simple. There is one driver. The driver is genetic. The driver is the same in essentially every patient. The driver is a kinase, and kinases are druggable. The cancer cells depend on the driver in a way they cannot escape, except by acquiring specific point mutations in the driver itself, which the field has, with some effort, been able to address through additional drugs. The cancer's natural history is slow enough — a chronic phase that lasts years, an accelerated phase that lasts months, a blast crisis that ends the disease — that drugging the chronic phase, where the cancer is most genetically homogeneous, can be done early enough to keep most patients out of the more lethal phases. CML is, in short, an unusually clean target for an unusually clean drug.

Most cancers are not like this.

Most cancers are driven by multiple genetic changes, accumulated over years, in shifting combinations across the population of cells in any individual tumor. Most cancers do not have a single dominant kinase that, if inhibited, brings the whole enterprise down. Most cancers, when you look at them at the molecular level, look less like an engine encased in concrete with one obvious wrench in the works, and more like a malfunctioning city in which several different systems have failed at different times in different parts.

What targeted therapy could do, after CML, depended on which cancers turned out to look more like CML and which turned out to look more like the general case.

Some did. Some didn't. The picture, twenty-five years after Gleevec's approval, is genuinely mixed in a way the 2001 coverage did not anticipate.

This chapter is about that mixture. It will look at where targeted therapy delivered something close to what was promised. It will look at where it did not. It will ask why — in a way the molecular logic the book has been laying out can begin to answer.

* * *

Start with the cancers where the targeted-therapy paradigm has worked best.

Several other hematologic cancers turned out, like CML, to be driven by single recurrent genetic events that could be drugged. Acute promyelocytic leukemia, a subtype of acute myeloid leukemia, is driven by the t(15;17) translocation Janet Rowley helped catalogue — a translocation that fuses two genes, PML and RAR-alpha, into a fusion that does to a different cell lineage what BCR-ABL does to myeloid stem cells. APL had been, until the late 1980s, one of the most lethal acute leukemias, with rapid bleeding complications often killing patients in the first weeks of presentation. The discovery that the disease could be treated, sometimes cured, with a combination of all-trans retinoic acid (a vitamin A derivative that could differentiate the leukemic cells) and arsenic trioxide (yes, the poison — in the right dose, against this specific kind of leukemia, an effective therapy) transformed APL into one of the most curable adult leukemias. The targeted approach in APL came first, chronologically, before imatinib; in some accounts of targeted therapy's history APL gets pride of place over CML. The cancer's molecular logic was similar: one fusion, one driver, one drug class that addressed it.

Gastrointestinal stromal tumors, or GISTs, are a rare type of soft-tissue sarcoma that, before 2001, had been one of the most stubbornly untreatable cancers. Conventional chemotherapy and radiation did essentially nothing. Surgery worked when the tumor was localized but most patients eventually had recurrence. In 1998, a group at Brigham and Women's Hospital in Boston — Jonathan Fletcher, Christopher Fletcher, and others — showed that GISTs were driven, in most cases, by activating mutations in a tyrosine kinase called c-KIT — the same c-KIT that, by happy coincidence, imatinib also inhibited as one of its handful of off-target activities. A clinician named George Demetri, also in Boston, took up the case for using imatinib in GIST. The first patient was treated in 2000. The response was dramatic. By 2002 — a year after Gleevec's approval in CML — the FDA had granted a second indication for the same drug in GIST, on the basis of a single small trial showing response rates that had been unimaginable in this cancer a few years before. Patients with metastatic GIST who would have been dead within a year went into durable remission on a daily pill. The drug had found a second cancer, almost by accident, that fit its mechanism.

Several lung cancers turned out, in the 2000s, to be driven by druggable kinase events in subsets of patients. The clearest case was EGFR-mutant non-small cell lung cancer, identified in 2004 by groups led by Bill Sellers, Pasi Jänne, and Bruce Johnson at Dana-Farber and Daniel Haber at Massachusetts General Hospital. About fifteen percent of lung adenocarcinomas in Western populations — and a higher fraction in East Asian populations — carried activating mutations in EGFR, the epidermal growth factor receptor. (EGFR, recall from chapter eight, is the first member of the HER family.) Inhibitors of EGFR — gefitinib, erlotinib, and later osimertinib — produced response rates in EGFR-mutant lung cancer of seventy or eighty percent, with substantial extensions of survival compared to chemotherapy. The first generation of these drugs had been developed before anyone knew which patients would benefit; what made them succeed was identifying the molecular subset, then giving the drug specifically to those patients.

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