The Pill That Could · First in Human

First in Human, part 2

Charles Sawyers, running the trial at UCLA, was the first to mention to Druker, in a phone call sometime in the autumn of 1998, that his patients were doing strangely well. Druker, comparing notes, said his were too. Moshe Talpaz at MD Anderson, the third site, was seeing the same thing. Three sites, three independent groups of patients, all responding in ways that the design of the trial had not anticipated.

By late autumn, with the dose up around 300 milligrams, something had become unmistakable. Patients who had failed every other available treatment, patients who had been told they were running out of options, were having their CML go quiet. Their white counts were normalizing. Their bone marrows, on biopsy, were showing fewer leukemic cells. Some of them were starting to feel, for the first time in months or years, more or less well.

This was not what phase 1 trials were supposed to do. Phase 1 trials are supposed to demonstrate safety, not efficacy. Efficacy at phase 1 doses, in cancer drug development, is something you might catch a hint of, in a few patients, at the highest tolerable dose. It is not something you watch happen across multiple sites and multiple cohorts as you climb through the dose range.

Druker, in his lab notebooks and in his calls with Sawyers and Talpaz, kept noting the same thing: this is not what is supposed to be happening. He kept checking, the way he had checked the colony assays years earlier, that he was reading the data right. He kept being told, by Sawyers and Talpaz, the same numbers he was seeing himself.

* * *

The way doctors talk about response to leukemia therapy involves a small ladder of escalating goodness, and it is worth introducing the rungs of the ladder before going further, because the rungs are how the rest of this chapter, and most of the rest of this book, will measure what the drug does.

The lowest rung is hematologic response. This means the blood is starting to look more normal. The white count comes down toward the normal range. The differential — the breakdown of which kinds of white cell are present, in what proportions — starts to look more like a healthy person's blood than a CML patient's blood. The spleen shrinks. A complete hematologic response is when the blood looks essentially normal: white count under ten thousand, no immature cells circulating, platelets in range, spleen back to normal size. This is good but it is not the deepest kind of response, because the disease can still be present in the bone marrow without showing up in the peripheral blood. It is the first thing that responds, and it tells you the drug is working at the level of the cells that the body is actively producing.

The next rung is cytogenetic response. This is measured by karyotyping the bone marrow — the same Q-banding and G-banding techniques Janet Rowley had used in 1973, now standard in clinical hematology labs — and counting the proportion of cells that still carry the Philadelphia chromosome. In an untreated CML patient, essentially every dividing cell in the marrow has the Philadelphia chromosome. With effective therapy, you can watch that proportion fall. A partial cytogenetic response is when the proportion of Philadelphia-positive cells drops below about a third. A complete cytogenetic response is when no Philadelphia-positive cells are detectable on karyotyping at all. This is a much harder thing to achieve than a hematologic response, because it means the drug is not just normalizing the output of the marrow but actually clearing the leukemic cells from the marrow itself — reducing the population of malignant cells, not just the population of malignant cells that happen to be circulating today.

The third rung, which would only become measurable later in the story, is molecular response. The Philadelphia chromosome can be detected by karyotyping, which catches it in roughly a percent of cells — you have to count a hundred cells to find one carrier on average. But the BCR-ABL fusion gene can also be detected by a much more sensitive technique, called PCR, which can find the fusion in a tiny fraction of cells — one in a thousand, one in ten thousand, one in a hundred thousand, depending on how carefully you do it. Molecular response is what you measure when even cytogenetic response has gone all the way to zero and you want to know whether the disease is really gone or merely below the limit of the karyotype's resolution. We will come back to this in chapter ten.

What the trial in late 1998 was seeing was complete hematologic responses. Patients on STI-571, at doses in the 300-milligram range, were having their blood counts normalize. By early 1999, with patients holding dose for several months, partial cytogenetic responses were starting to show up too. By mid-1999, complete cytogenetic responses had appeared in some patients.

Patients whose CML had progressed past every other available treatment were having their leukemia, by every measure available to clinical hematology in 1999, disappear.

* * *

The trial's findings were not yet published, and the company's view of how widely to discuss them was, in the standard way of these things, conservative. Phase 1 data are preliminary; you do not want to make claims based on a few patients that might not hold up in larger trials. There was, internally at Novartis, a debate through the autumn of 1998 and the spring of 1999 about how to talk about what was happening.

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