I am writing this because of what showed up on my phone this past week.
On Sunday morning, in a windowless plenary room at the McCormick Place convention centre in Chicago, the data from a trial called RASolute 302 was presented to the annual meeting of the American Society of Clinical Oncology. The trial tested a new oral drug, daraxonrasib, against standard chemotherapy in patients with previously treated metastatic pancreatic cancer — the hardest setting in oncology, the setting in which median overall survival has not moved in fifteen years. The drug doubled it.
Within hours, my feed had the story in twenty different framings. There was the press-release version, in which the word "breakthrough" arrived at speed, attached to stock photographs of pipettes, and the algorithm was suggesting to readers that pancreatic cancer had been cured between lunch and tea. There was the cynical version — a lot of likes, fast — which went something like, Six months to thirteen. Oh wow, real revolution. So I'll die in twelve months instead of six. Big medicine, big party. There was the small, careful subset of posts from people who had actually been in the room and who were trying to explain what it meant, usually buried beneath the other two. By Wednesday, the discourse had picked up sponsors, a podcast, and the predictable parade of accounts explaining to each other a study they had not read.
Both extremes are wrong, in opposite directions, and they are wrong about the same thing — the structure of how oncology actually progresses, and what a result like RASolute 302 means inside that structure. The breathless reading does not understand that "from six months to thirteen" is, in fact, not a small thing in a disease where the curve has not moved in fifteen years. The cynical reading does not understand that an effect of that size, in a disease this difficult, is the rare exception, made possible by forty years of slow, almost-invisible work that did not look like progress while it was happening, and that very few inflections of this kind ever land in a working oncologist's career.
I want to lay out the forty years. Not because daraxonrasib is a cure — it is not — and not because it will be available in India tomorrow — it will not — but because the structure of the work that produced it is the structure of almost all real progress in medicine, and almost nobody outside the field gets to see it. We see the inflection points. We do not see the work.
In the Room
For those who haven't sat through one, the ASCO Annual Meeting holds a single curated session each year called the Plenary. It is reserved for a handful of trials — three or four, occasionally a few more — judged most likely to change cancer practice. They are selected months in advance, presented in the largest auditorium the meeting can fill, and almost always paired with a simultaneous publication in the New England Journal of Medicine or the Lancet, embargoed until the speaker takes the stage.
In the weeks before the meeting, medtwitter — the cancer-medicine corner of social media — argues about which of the plenary abstracts will turn out to be the one. The audience's verdict, in the room, is the standing ovation. It is rare. Most plenary presentations are received with respectful applause and a question or two. The handful that are received with the room on its feet are the ones every oncologist there knows, instantly, will be the talked-about result of the year.
Dr. Brian Wolpin, the principal investigator of RASolute 302, walked to the podium. He is an oncologist at Dana-Farber Cancer Institute in Boston, and he has been part of the small community of physicians and scientists working on RAS inhibition in pancreatic cancer since before there was a community to speak of. He worked through the trial design quickly: patients with metastatic pancreatic cancer who had received one prior line of treatment were randomised, in a roughly two-to-one ratio, to receive either daraxonrasib — a new oral inhibitor of mutant RAS, taken once a day — or the chemotherapy that the treating oncologist would otherwise have chosen.
The Kaplan–Meier curve for overall survival came next. It was the kind of slide one looks at and then looks at again, to be sure one is seeing it correctly. The daraxonrasib curve sat well above the chemotherapy curve, with the gap widening as the months went on. Median overall survival on daraxonrasib was about thirteen months. On chemotherapy, about six. The hazard ratio was well into the territory of a clinical effect that is not subtle.
The room was quiet for a beat longer than I expected. Then a row stood, and then several rows, and then the entire auditorium. The applause did not stop. It went past a minute, then past a minute and a half, then close to two. Dr. Wolpin stood at the podium and waited, plainly a little overcome. When the room eventually sat down and the applause subsided enough for him to be heard, he leaned into the microphone and said, with the dryness of a man who has waited a long time to be allowed to say something like this:
"I will not be held responsible for the two minutes we have just lost to this unruly behaviour."
The auditorium laughed. The chair laughed. And the New England Journal of Medicine paper, embargoed until that moment, became all at once what every oncologist in the room would be reading in their hotel that evening.
When applause for a clinical-trial result lasts long enough to interrupt the schedule, it is almost never about the drug itself. It is about what the drug means. What daraxonrasib means is the conclusion of an argument that began a long time before any of us in that room were oncologists.
How Thin the Cupboard Has Been
To understand why a room of oncologists stood up, you need to know how thin the cupboard has been.
In 1996, the FDA approved gemcitabine — a chemotherapy drug that had been investigated for years in other cancers — as a treatment for advanced pancreatic cancer. It was the first time, ever, that any chemotherapy had been approved specifically for this disease. The pivotal trial, run by Dr. Howard Burris and his colleagues and published in the Journal of Clinical Oncology in 1997, compared gemcitabine with 5-fluorouracil, the only other agent that had any track record. The median overall survival on gemcitabine was 5.6 months. On 5-fluorouracil, 4.4 months. The difference was a little over a month.
What carried the approval over the line was an unusual co-primary endpoint the trial had been designed around — something called "clinical benefit response," which was a composite measure built from improvements in the patient's pain, in their performance status, and in their weight. Twenty-three per cent of patients on gemcitabine had a clinical benefit response. Fewer than five per cent of patients on 5-fluorouracil did. The field, in other words, was so short of options for this cancer that the FDA accepted a regulatory standard built partly around how much less the patient hurt. In almost every other modern oncology approval, the central question is whether the patient lives longer. In pancreatic cancer in 1996, the question being asked was whether the patient suffered less while dying — and a drug got through because the answer, modestly, was yes.
For the next twenty years, almost every step forward in metastatic pancreatic cancer was an attempt to add another chemotherapy on top of gemcitabine. FOLFIRINOX — a four-drug regimen — was shown in 2011 to extend survival further, in the patients well enough to tolerate it. Gemcitabine with nab-paclitaxel was approved in 2013 on a similar logic. Both regimens, real as their gains were, measured improvement in weeks. Median survival from a diagnosis of metastatic pancreatic cancer drifted from about six months in the late 1990s to about eleven months by the mid-2010s. And then it stopped. For the last fifteen years, the curve has not moved.
The mathematics of pancreatic-cancer practice — what every oncologist tells every family — has been built on top of that plateau. Cytotoxic chemotherapy is the mainstay. We know it will not be enough. We talk to patients about months, not years. We have wanted, for thirty years, for there to be something else.
The Argument Behind the Drug
The argument behind daraxonrasib is about a single gene called RAS.
What RAS does inside a cell is simple enough to describe. It is a switch. When a growth signal arrives at the cell, RAS turns "on" and tells the cell's growth machinery to start. When the signal ends, RAS turns "off" again, and the machinery stops. The switch is fast and tightly controlled. It has to be — because if it gets stuck on, the cell keeps growing when it should not.
In 1982 and 1983, three independent teams of scientists — at the National Cancer Institute in Maryland, at Cold Spring Harbor on Long Island, and in Dr. Mariano Barbacid's group in Spain — demonstrated that human cancers carried mutated versions of RAS in which the switch had been jammed permanently in the "on" position. The discovery was, at the time, electrifying. Here, at last, was a molecular target that an entire class of human cancers seemed to share. Six years later, in 1988, Dr. Carmen Almoguera and colleagues published the seminal paper showing that this same mutated RAS — specifically the variant called KRAS — was present in more than ninety per cent of pancreatic cancers. The most lethal common cancer in the world had a single most-common mutation, and the field knew what it was.
The field also failed to do anything about it for the next thirty-one years.
The Undruggable Decades
The reason for the failure was, in retrospect, simple. Most drugs work by lodging into a pocket on the surface of the target protein, the way a key fits into a lock. RAS, when laboratories around the world examined its structure, did not appear to have one. By the early 2000s, the consensus in pharmaceutical chemistry was that RAS was not druggable in any practical sense. The view was so firmly held that it had a textbook name. The undruggable target. It appeared in lectures, in editorials, in research-grant rejections. A generation of oncologists trained partly around the assumption that RAS would always be unreachable.
What kept the work alive, against this consensus, was a stubborn refusal by a small number of scientists to accept that the consensus was a fact. They worked in places — Texas, California, Cambridge in England — where heretical projects were sometimes tolerated. They tried fragment screens, in which thousands of tiny chemical pieces are washed across the protein in the hope that one of them will stick somewhere unexpected. They tried covalent strategies, in which a drug binds permanently rather than reversibly. They built more refined crystallographic methods so that they could see what RAS looked like under conditions other than the textbook ones. Most of what they tried did not work. Their careers were quieter than the careers of their colleagues. Their grants were difficult.
In May, 2013, a group led by Dr. Kevan Shokat, a chemist at the University of California, San Francisco, published a paper in Nature. They had been studying a particular mutant version of KRAS called G12C, and they reported something nobody had reported before: when KRAS-G12C was in a certain brief intermediate state — between its "on" and "off" positions — a small, transient pocket appeared on its surface. The earlier images had simply not been taken at the right moment. Dr. Shokat's group designed a small molecule that fit into this pocket and locked the protein in its inactive state. The molecule was crude, by drug-development standards. But it was the first published demonstration, anywhere, that a small drug could bind RAS in a way that mattered.
The next decade was a sprint. Other groups confirmed the cryptic pocket. Pharmaceutical chemistry adapted its methods to look for transient features rather than static ones. Computational protein-structure prediction — which had been improving steadily through the 1990s and 2000s — was transformed in 2021 by AlphaFold and its successors, deep-learning systems that could predict the three-dimensional shape of a protein at near-experimental accuracy. Suddenly, the cryptic pockets that had been invisible for forty years could be predicted in silico, mapped in detail, and exploited. Drug design, as a discipline, was moved a step closer to engineering.
In 2021, the FDA approved sotorasib, a small-molecule inhibitor of KRAS-G12C developed by Amgen, for use in patients with lung adenocarcinoma carrying the G12C mutation. A year later, adagrasib, developed by Mirati, was approved. The patient population was small — roughly thirteen per cent of lung adenocarcinomas, more or less — and the clinical benefit was modest, with median survival improvements measured in months. The reaction in the oncology world, however, was disproportionate to the size of the benefit. Sotorasib was not a great drug. It was a real drug, against a target everyone had given up on, in patients who had previously had nothing left to try. The principle that mattered — that RAS could be drugged — had been proved in clinic.
RASolute 302
The problem for pancreatic cancer was that the version of RAS jammed on in pancreatic tumours was a different one from the version sotorasib hit. The dominant pancreatic mutations are G12D and G12V; the mutation that sotorasib was designed against, G12C, is mostly a lung-cancer mutation. Drug chemists would need to invent different drugs for the pancreatic versions. The pancreatic version of the story would have to wait.
It didn't have to wait long. Through the early 2020s, two strategies emerged. The first was to design narrowly targeted drugs that bind only one specific mutant form. The second, more ambitious, was to design a single drug that could block several mutant forms of RAS at once — a so-called pan-RAS inhibitor. Daraxonrasib, developed by a small biotechnology company called Revolution Medicines in California, belonged to the second category. It was built to bind the active "on" form of RAS regardless of the exact mutation. The phase one and two trials, presented at ASCO over the past three years, had shown encouraging response rates and tolerability. The phase three trial — RASolute 302 — was designed by Dr. Wolpin and his colleagues to be small, fast, and definitive: a head-to-head against the chemotherapy the treating oncologist would otherwise have chosen, in the second-line setting where the unmet need was greatest.
It had been persuasive. More than persuasive — definitive. The Kaplan–Meier curve I had watched on the plenary screen showed the kind of separation that pancreatic oncology had not seen in fifteen years. The benefit held in patients whose tumours carried any of the major RAS mutations, and it held, somewhat unexpectedly, in patients whose tumours did not carry a detectable RAS mutation at all — a finding that hinted at a more general role for RAS-state signalling than the textbook biology had suggested. Fewer patients discontinued the drug for toxicity than discontinued chemotherapy. The time to a clinically meaningful worsening of pain — a deceptively important endpoint in pancreatic cancer, where pain is the dominant late-stage burden — was significantly delayed. The authors recommended daraxonrasib as a new standard of care for previously treated metastatic pancreatic adenocarcinoma.
And behind RASolute 302, in active or planned phase three development, were eight more trials — RASolute 303, 304, 305, and 309; HRS-4642 from Hengrui; Astellas's 3082 programme; Incyte's DAWN-303 trial; and Genfleet's GFH375 — each testing a slightly different version of the same idea, in a slightly different clinical setting. A drug class had been born, and the number of arrows in the quiver was no longer one.
The Lesson
What is the lesson, if there is one?
The lesson, I think, is about the structure of how this kind of progress actually happens, and how poorly that structure is understood by anyone not embedded in the field. Almost all of medical progress — almost all of useful medical progress, the kind that ends up changing what we tell a patient and their family in clinic on a working morning — is the result of a long, mostly invisible, often unrewarded body of work, performed by a small number of people, for years and sometimes decades, before the work suddenly translates into a result the rest of us notice. There are no breakthroughs in oncology, in any meaningful sense. There are only inflections, and the inflections are produced by an enormous amount of work that has been happening below the surface for a long time. We notice the inflections because they make headlines. We don't notice the work because it makes graphs in obscure journals.
This matters for a particular reason. When patients and families read headlines about a "breakthrough" cancer drug, they often suspect, correctly, that the headline is overselling. They are then left with one of two unsatisfying conclusions. Either the headline is true and modern medicine has become a series of miracles, which their own experience tells them is not so. Or the headline is false and modern medicine is mostly hype, which is closer to the texture of their own experience but underestimates the work. Both readings miss the structure underneath. There are, in any given decade, fifty or a hundred quiet inflections in oncology, almost all of them the result of long perseverance, each one moving the curve a little or a lot, in one disease or several. Daraxonrasib is one of these. So is sotorasib. So is the slow improvement in adjuvant chemotherapy regimens for early-stage breast cancer over the past twenty years. So is the development of CAR-T cells. So is the careful refinement of dose-reduction strategies in elderly leukaemia patients. So is the patient, grinding, decade-long work on supportive care that has cut hospitalisations and quietly extended lives.
If there is a single thing I would like patients and their families to take from the story of pancreatic cancer this June, it is this: incremental progress is what real medical progress looks like, almost all of the time, and it is much, much better than the alternative. The reason to keep paying attention to a difficult cancer is not because a cure is around the corner. The reason is that the next ten years are usually built on the previous forty. The undruggable years of RAS were not wasted years. They were the years in which the methods, the chemistry, the structural biology, and the human community capable of making the next thing were quietly assembled. A scientific community that stops doing the unglamorous middle work — because the headlines are not big enough, because the grants are not coming in, because somebody important has declared the problem unsolvable — is a community that will not have a next inflection. We get the future we keep showing up for.