The Pill That Could · The Long Argument

The Long Argument, part 3

He also worked the academic publication channel relentlessly. The 1996 Nature Medicine paper was followed by additional papers — in Blood, in Clinical Cancer Research, in other venues — each one extending the preclinical case. The compound worked against primary cells from CML patients. It worked against cells from patients with the more aggressive accelerated and blast-phase forms of the disease. It worked against a related leukemia, Philadelphia-positive acute lymphoblastic leukemia, that shared the same fusion gene. It worked in mouse models. By 1998, the volume of preclinical data on CGP 57148 was greater than what most early-stage compounds carried into the clinic, and most of it had been generated in academic labs by people whose paychecks did not depend on Novartis's decision.

He also kept up direct contact with the people inside the company who were making the decisions. He flew to Basel. He hosted Novartis people in Oregon. He sent updates and reprints. He made it difficult for the program to quietly die, because every time someone inside the company tried to scale it back, there was an outside investigator with new data calling to ask why.

Lydon was doing the same thing from inside. He had, in 1997, left Novartis for a small biotechnology company in California, but he stayed in close contact with the program and its leadership. The departure removed a senior internal advocate, but Matter remained, and a small number of others.

* * *

There was, around 1996 or 1997, a problem with the dog tox studies.

The dose-finding work in dogs — the standard non-rodent species — turned up signs of liver toxicity at high doses. This was, depending on whom you talked to, either a deal-breaker or a nuisance. The deal-breaker reading went: a drug that hurts dog livers may hurt human livers, and a liver-toxic drug for a small disease is a non-starter. The nuisance reading went: the toxicity appeared at doses well above the therapeutic range, the safety margin was acceptable by ordinary pharmaceutical standards, and the relevant question was whether the drug would be safe at the doses where it actually worked. Both readings were consistent with the data. The two camps inside the company read the same numbers and disagreed about what they meant.

The argument took months. It was eventually resolved in favor of moving forward, on the grounds that the toxicity in dogs occurred at exposure levels well above what was projected to be needed in humans, that the molecular target was specific enough that off-target effects in human liver should be limited, and that the disease was severe enough to warrant proceeding into a phase 1 trial where careful monitoring could detect any problems early. The toxicology team signed off. The IND application could go forward.

But the dog issue had cost time. It had given the skeptical reading inside the company more ammunition; for several months it looked, from outside, as if the program might be terminated. Druker, hearing about the worries through Lydon and through his own contacts, was alarmed. He spent the period writing letters and making phone calls and giving talks.

From the perspective of someone with a patient sitting in their clinic, the question “is the dose at which dog livers show damage acceptable” and the question “is my patient going to live” are the same question. From the perspective of someone running a pharmaceutical company, they are not.

* * *

What seems to have shifted the situation, more than any single factor, was the accumulation. By 1998, the case for at least running a phase 1 trial of CGP 57148 had become very hard to argue against without sounding like the company was simply unwilling to develop the compound at all.

The preclinical data were exceptionally clean. The molecular target was thoroughly validated. The unmet medical need was evident. The toxicology, after months of argument, was acceptable. The chemistry could be made at the scale a phase 1 trial required. The investigators were in place; Druker at OHSU, Charles Sawyers at UCLA, Moshe Talpaz at MD Anderson Cancer Center in Houston — three sites, three established academic hematologists, all of whom understood the science and were prepared to recruit and treat patients carefully. The protocol had been worked out in detail.

Vasella, by some accounts, made the call to advance the program in late 1997 or early 1998. He has said in subsequent interviews that he saw the case as scientifically compelling and was willing to make the bet. The bet was, in his framing, about whether a careful and selective drug for a specific disease, if it worked, would justify the development cost — a question whose answer, he has said, he was willing to wait for the trial data to settle. The internal politics of how this decision actually got made, what conversations Vasella had with whom, who pushed back and who relented, are not fully captured in any public account, and probably never will be.

What is captured is that the program moved from “being kept alive” to “being actively developed” over the course of about a year. The IND was filed in early 1998. The FDA cleared it in due course. The phase 1 protocol — a dose-escalation study in CML patients who had failed interferon, the standard first-line therapy — was finalized. The first patient was scheduled to be enrolled in June.

The molecule, by this point, had been renamed. It was no longer CGP 57148. Internally at Novartis it had been given a new code: STI-571. The letters stood for “signal transduction inhibitor,” the broad class of drug Novartis was now describing it as. It was the same molecule. It had a slightly more marketable acronym.

* * *

There is a tendency, in the way this kind of story is told later, to compress everything that happened between 1993 and 1998 into a few sentences about delay and turn the company into the villain of the piece.

It is worth resisting that tendency, because it is wrong, and because it makes the actual story less interesting.

Novartis was not the villain. Novartis was a large, publicly-traded pharmaceutical company doing its best, by its own internal standards, to make rational decisions about which compounds to develop. CGP 57148, throughout the period in question, was a compound with real questions hanging over it: a small market, an unprecedented mechanism, real toxicology concerns at high doses, the fear of resistance. Reasonable people inside the company, looking at the same data Druker was looking at, came to a different conclusion about how to weight the evidence. They were not corrupt and they were not stupid. They were operating in a system that selects for, and that has historically been right to select for, a high bar of evidence before committing the resources required to develop a drug.

What Druker had, that allowed him to push past the bar, was three things. He had unusually clean data. He had unusual persistence — a willingness to spend years writing letters and giving talks and making phone calls and accepting that no single one of these activities would by itself change the outcome. And he had, inside the company, a small group of allies who were willing to argue his case in rooms he was not in. Without any of these three, the molecule probably would have stayed on the shelf.

Even with all three, it took five years from the first cell-line result to the first patient. The patient who walked into Druker's clinic in 1993, the one the prologue described, had — if she was lucky and her chronic phase was on the longer end — about that long before the disease progressed. Some of the patients Druker met in his early years at Oregon were alive when the trial started. Some were not. There is no way to recover the patients who weren't, and no point in pretending that the cost of the delay was zero. But there is also no clean alternative history in which a more decisive company would have produced a usable drug substantially faster, given the actual state of the science and the actual constraints on what the system can do quickly. The five years were too long. They were also, given everything, not crazy.

Druker himself, in interviews after the drug was approved, has been thoughtful about this. He has said that he understands why Novartis was cautious, that he believes the people he worked with there were trying to do the right thing as they saw it, and that the system that produces drugs is the system that produces drugs. He has also said that he does not regret pushing as hard as he did, and would do it the same way again. Both things can be true.

* * *

On a morning in late June of 1998, in Druker's clinic at Oregon Health Sciences University, the first patient enrolled in the phase 1 trial of STI-571 — a man whose CML had progressed past every other available treatment — swallowed the first capsule of a drug no human had ever taken before. The dose was very low. The starting dose for any phase 1 study is, by design, low; the goal of a phase 1 trial is to find the dose at which problems first appear, and you do that by starting well below where you expect problems and working upward in small steps.

At the dose he was given, on the basis of the animal studies, no one expected the drug to do much of anything. The point of the early doses was to demonstrate safety. Efficacy, if there was any, was not expected until the doses got considerably higher.

This is what nearly everyone running the trial believed in June 1998.

It is not what happened.

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