The Pill That Could · Approval

Approval, part 2

All three of these conditions were unusually well met in the case of STI-571.

* * *

The phase 2 program for STI-571, which had launched in 1999 and continued through 2000, was designed to do, on a larger scale, what the phase 1 had done by accident: produce data showing that the drug worked at every clinically relevant stage of CML. There were three pivotal trials. One enrolled patients with chronic-phase CML who had failed interferon. One enrolled patients with accelerated-phase CML. One enrolled patients in blast crisis.

The trials were not randomized. This is, on paper, an unusual choice for a drug seeking approval; the modern standard for cancer drug evaluation is randomized controlled trials in which the new drug is compared head-to-head with the existing standard of care, and the comparison is what tells you whether the new drug is better. The phase 2 program for STI-571 used a different design — single-arm, with historical controls. Every patient on the trial got the drug. The response rates were compared not against a contemporary control group of patients on a different therapy but against historical data on what patients with the same disease, at the same stage, had been observed to do under standard care.

There is a serious objection to this kind of design. Historical controls are not ideal; the patients you compare against may differ from your trial patients in ways that are hard to control for, and the comparison may flatter or unflatter the new drug depending on which historical cohort you pick. Modern oncology, when it can, prefers randomization.

What made historical controls defensible in the case of STI-571, and what the FDA accepted as a basis for approval, was that the response rates the drug was producing were so far outside the historical range that no plausible adjustment for cohort differences could explain them away. CML in chronic phase, on standard therapy, produced complete cytogenetic responses in something like ten percent of patients, on a timescale of years. STI-571, in the phase 2 trial of patients who had failed interferon, was producing complete cytogenetic responses in roughly forty percent of patients, on a timescale of months. There was no historical cohort, no contemporary cohort, no patient subgroup that produced numbers like these on any other treatment. The drug's effect was so large that the question of whether the comparison was fair became, in a practical sense, moot.

This argument was made formally to the FDA, and the FDA accepted it. There is a regulatory path, called accelerated approval, that allows the agency to grant marketing authorization on the basis of a surrogate endpoint — in this case, cytogenetic response, which had not yet been formally validated as predictive of long-term survival but which everyone in the field believed would be — with the requirement that the sponsor complete confirmatory studies after approval. STI-571 was approved under accelerated approval. The confirmatory commitment was a phase 3 trial, not yet run, that would compare the drug head-to-head against interferon plus cytarabine, the standard of care, with overall survival as the endpoint. That trial would launch in 2000 under the name IRIS. We will come back to it.

* * *

The naming process for the molecule had been running in parallel with the regulatory process.

STI-571 was an internal Novartis code; it stood, as the previous chapter mentioned, for “signal transduction inhibitor.” It was a useful research label and an awkward consumer-facing name. Drugs that get to market need two names — a generic name, which is the official chemical name used in the medical literature and on the prescription label, and a brand name, which is the trademarked name the company uses to market the drug. The generic name is set by the United States Adopted Names Council, an independent body that follows international conventions for drug nomenclature. The brand name is chosen by the company, subject to FDA review for whether it might confuse patients or doctors with the names of other drugs.

The generic name assigned to STI-571 was imatinib mesylate. ‘Imatinib’ is the active molecule; ‘mesylate’ is the salt form, the methanesulfonate, that is actually formulated into the capsules. The ‘-tinib’ ending was new at the time and would, over the next two decades, become the standard suffix for tyrosine kinase inhibitors as a class — dasatinib, nilotinib, sorafenib, sunitinib, gefitinib, erlotinib, lapatinib, and dozens more would follow. In 2001, imatinib was the first.

The brand name was chosen, after the usual consumer-research process, as Gleevec. (In Europe and most of the rest of the world, the same drug was marketed under the slightly different brand name Glivec, an early lesson for the company about the difficulty of harmonizing brand names across markets with different linguistic preferences.) Both names are derived from the German verb ‘gleichen,’ meaning roughly to make even or to match — a play on the molecule's mechanism of fitting precisely into the active site of its target. Whether anyone outside the marketing department appreciated the etymology is unclear, but the name had the right syllabic shape — short, easy to say, hard to confuse with anything else.

From the point of approval forward, in this book, the molecule is Gleevec when it's a drug being sold and prescribed, and imatinib when it's being discussed as a chemical or as a research compound. CGP 57148 and STI-571 are retired. The molecule itself is the same molecule it has been since Jürg Zimmermann's chemists made it in Basel a decade earlier.

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