The Pill That Could · A Different Cancer

A Different Cancer, part 5

The phase 1 trial of the humanized antibody began in 1992. The phase 2 trials, including a study testing the antibody in combination with chemotherapy, ran through the mid-1990s. The pivotal phase 3 trial, completed in 1998, compared chemotherapy alone to chemotherapy plus the antibody in women with HER2-amplified metastatic breast cancer. The trial showed that adding the antibody extended progression-free survival and, in subsequent follow-up, overall survival. The combination was approved by the FDA in September 1998.

The drug's generic name was trastuzumab. Its brand name was Herceptin.

It was, by some accounts, the first humanized monoclonal antibody approved for any cancer indication. (Rituxan, a different humanized antibody for non-Hodgkin lymphoma, had been approved in November 1997, about ten months before Herceptin; the priority depends on how you count.) It was also, importantly for our story, the first targeted therapy for breast cancer in the modern molecular sense — a drug aimed at a specific genetic alteration in a specific subset of patients, given on the basis of a molecular test.

* * *

The molecular test was a piece of clinical infrastructure that had to be built alongside the drug.

Herceptin's approval was, from the beginning, restricted to patients whose tumors were HER2-amplified. The FDA approval was contingent on the use of a companion diagnostic — a laboratory test that classified each tumor as HER2-positive or HER2-negative, and only HER2-positive patients were supposed to receive the drug. The original test was an immunohistochemistry assay called HercepTest, which scored tumor sections on a 0 to 3+ scale based on how much HER2 protein the tumor cells displayed; tumors scoring 3+ were considered candidates for Herceptin. A more accurate but more expensive test, fluorescence in situ hybridization (FISH), counted the actual number of HER2 gene copies. The two tests gave somewhat different answers in borderline cases; the field gradually settled into a workflow in which immunohistochemistry was done first, with FISH used to clarify the 2+ borderline cases.

The point, for our purposes, is that Herceptin established the model that targeted therapy would follow for the next two decades: every targeted drug came paired with a test that identified which patients had the alteration the drug was aimed at. You did not give the drug to all breast cancer patients; you gave it to the quarter or so whose tumors were HER2-positive, identified by the test. The patients without the amplification didn't benefit and weren't supposed to receive the drug. The patients with the amplification benefited substantially. The pairing of a specific drug with a specific test for a specific molecular alteration in a specific cancer became the template.

Imatinib, which would be approved three years later, fit the same template, although the test in CML's case was simpler — essentially every CML patient had the Philadelphia chromosome by definition, so the test for the disease was the test for the target. In other cancers that would follow — EGFR-mutant lung cancer, BRAF-mutant melanoma, ALK-rearranged lung cancer, and many more — the same pairing held. A targeted drug came with a companion test. The model was, by the late 2000s, the dominant paradigm in oncology drug development.

* * *

It is worth, before closing this chapter, noticing what is similar and what is different between the Gleevec story and the Herceptin story.

What is similar is the structural argument. Both diseases were driven, in a substantial subset of cases, by a single molecular event — the BCR-ABL fusion in CML, the HER2 amplification in HER2-positive breast cancer. Both molecular events produced a specific protein doing too much of a specific kind of signaling — in both cases, signaling driven by tyrosine kinase activity. Both could be addressed by drugs that targeted the responsible protein selectively. Both drugs, when developed, produced clinical benefit so much larger than what the previous standard of care delivered that the comparison was not really close.

What is different is the modality. Imatinib is a small molecule. It is a single chemical compound, made by chemical synthesis, that binds inside the BCR-ABL kinase and turns it off. It is taken as a pill once a day. Trastuzumab is a protein — a humanized antibody — made in cultured cells in large bioreactors, purified through a complicated downstream process, given by intravenous infusion every three weeks. It binds the outside of HER2, on the cell surface, and stops the receptor from doing what it does. The two drugs work on the same general kind of target — a tyrosine kinase — but from opposite sides of the cell membrane and through quite different mechanisms.

The two modalities have different practical implications. Small molecules can typically be given orally and reach essentially all tissues, including the brain, although they often need to be taken every day to maintain effective levels. Antibodies are large; they can't be given orally (they would be digested), they don't cross the blood-brain barrier (which is why antibody therapies often fail to control brain metastases that small-molecule therapies might reach), but they have very long half-lives in the bloodstream, which is why they can be given every few weeks rather than every day. Small molecules are cheaper to manufacture; antibodies are much more expensive. Small molecules tend to have more off-target effects; antibodies tend to be more selective. Each kind of drug has its strengths.

What targeted therapy turned out to mean, in practice, was both kinds of drug, used together where useful, deployed in different cancers based on what the cancer's molecular drivers were and what kinds of drugs would reach those drivers most effectively. Some targets were druggable only with antibodies (most cell-surface receptors). Some were druggable only with small molecules (most kinases inside the cell). Some could be approached with either or both. The toolkit, by the early 2000s, was bigger than it had been in 1990, and the question for any new molecular target was no longer ‘is there a way to drug this’ but ‘what kind of drug, and which one will get there first.’

* * *

Herceptin was approved in September 1998. Gleevec was approved in May 2001. The phase 1 trial of Gleevec started in June 1998, three months before Herceptin's approval. The pivotal phase 3 trial of Herceptin had completed enrollment by the time Druker dosed his first patient.

If you had been an oncology drug developer in 1998, looking at these two programs side by side, you would have seen a coherent picture taking shape. Two cancers. Two different driver alterations. Two completely different kinds of drug — small molecule and antibody — aimed at the same general class of target. Both programs had been developed on the strength of strong molecular evidence and had produced clinical results so striking that the regulatory path had compressed in front of them. Both established the principle that a drug could be developed for a defined molecular subset of a disease, paired with a diagnostic that identified the right patients, and produce benefit substantially better than the cytotoxic standard of care that had been the backbone of oncology for the previous thirty years.

By 2001, with Gleevec approved, the picture had become even clearer. The targeted-therapy era had not just arrived; it had begun to scale. Other compounds were already in development, aimed at other molecular targets in other cancers. Other companies were retooling their oncology pipelines around the idea that finding the molecular driver was the first step in finding the drug. The 1990s, in oncology drug development, had been the decade in which a specific kind of bet — the bet on selective targeting — went from being a contrarian hypothesis to being the default frame.

What this looked like, in the years after 2001, is the subject of chapter eleven — what targeted therapy did and did not deliver across the broader landscape of cancer.

But before that, two things have to happen in the CML story itself, and they are the subjects of chapters nine and ten.

Some patients on imatinib started to relapse.

And, for the patients who did not relapse, a question began to take shape that the trials of 2001 had not been long enough to answer: what does it actually mean, in this disease, in the long run, to keep taking the drug?

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