The Pill That Could · The Generalization Problem

The Generalization Problem, part 4

Triple-negative breast cancer — breast cancer that is negative for estrogen receptor, progesterone receptor, and HER2 amplification — is the breast cancer subset for which targeted therapy has had the least to offer. About fifteen percent of breast cancers are triple-negative; they tend to be more aggressive than ER-positive or HER2-positive disease, and they affect younger women disproportionately. The molecular drivers of triple-negative breast cancer are heterogeneous and largely not druggable with current approaches. Treatment has remained, in many patients, conventional chemotherapy. Recent developments — PARP inhibitors in BRCA-mutant subsets, immunotherapy combinations, antibody-drug conjugates targeting other surface markers — have begun to make a dent. But triple-negative disease has, in 2026, no equivalent of Herceptin or imatinib; no clean molecular driver, no transformative single drug.

These are not isolated cases. Many of the most common solid tumors — most colorectal cancers, most stomach cancers, most ovarian cancers (outside the BRCA-mutant subset), the majority of lung cancers (those without an actionable kinase mutation) — have not been transformed by the targeted-therapy paradigm in the way CML and HER2-positive breast cancer were. They have benefited, sometimes substantially, from improvements at the margins. They have not been remade.

* * *

Why some cancers have yielded and others have not is a question with a partial answer the field has, by 2026, become reasonably comfortable with.

The cancers that yield to targeted therapy tend to share several features. They are typically driven by a single dominant genetic event, present in essentially every cell of the tumor. The driver is typically a kinase or a comparable signaling protein with a druggable structural feature — an ATP pocket, an allosteric site, an extracellular domain accessible to an antibody. The cancer cells are typically dependent on the driver in a way that they cannot easily replace by activating some other pathway. The tumor's mutational burden — the total number of additional mutations beyond the driver — is moderate enough that the cells have not already accumulated alternative ways to grow.

The cancers that resist targeted therapy tend to share the opposite features. They are typically driven by combinations of changes, with no single dominant event. Their drivers are often not kinases but transcription factors, structural proteins, or proteins with shapes that don't accept conventional small-molecule inhibitors well. The cells have multiple redundant pathways for growth and survival, so blocking any one of them gets compensated for by another. The mutational burden is high, the tumor's genetic heterogeneity is high, and within any one patient there are usually many subpopulations of cells with different vulnerabilities, so a single drug can't catch them all.

There is also, layered on top of this, the question of stage. Targeted therapy generally works better in earlier-stage disease, where the cell population is smaller and more genetically homogeneous, than in advanced disease, where the cells have had more time and more opportunity to evolve away from any single intervention. CML in chronic phase responded beautifully to imatinib; CML in blast crisis responded less well. EGFR-mutant lung cancer in early-stage disease, on adjuvant osimertinib, has dramatic recurrence-prevention benefits; EGFR-mutant lung cancer in advanced disease eventually evolves resistance, in most patients, to even the best current drugs. The cancers that have generalized targeted-therapy success best are the ones where the molecular insight could be deployed early, against a smaller and more genetically uniform target population, before the cancer had a chance to become as heterogeneous as advanced cancers tend to be.

There are, finally, the cancers where targeted therapy has been tried, has worked partially, and has hit a ceiling. HER2-positive breast cancer is, in some ways, the prototype of this. The drug works, the responses are durable in many patients, but a substantial fraction of metastatic patients eventually progress, and the resistance mechanisms are heterogeneous and harder to address with second-generation drugs than the analogous mutations in CML were. The newer antibody-drug conjugates have extended the franchise, but they have not closed it. HER2-positive breast cancer in 2026 remains a treatable but, in advanced settings, often ultimately fatal disease.

* * *

It is worth being clear about the larger frame this chapter has been moving toward.

The 2001 Time cover suggested that targeted therapy was a new paradigm that would transform cancer treatment. Twenty-five years later, this is true and not true. Targeted therapy has transformed treatment for a real and meaningful subset of cancers — the chronic myeloid leukemia patients on imatinib, the GIST patients on imatinib, the EGFR-mutant lung cancer patients on osimertinib, the BRAF-mutant melanoma patients on combination BRAF/MEK therapy, the HER2-positive breast cancer patients on trastuzumab and its successors, the BRCA-mutant ovarian cancer patients on PARP inhibitors, and a growing list of others. For these patients, the change has been real and large. Some of them have effectively had their cancer converted from a fatal disease to a chronic one, or from a chronic one to a curable one. Some of them are alive, in 2026, who would not have been alive in 1995, and the difference is the molecular logic and the drugs that came out of it.

Targeted therapy has not transformed treatment for most cancers, in the sense that it has not delivered, for most patients with most cancers, the kind of clean before-and-after benefit it delivered for CML. The pancreatic cancer patient diagnosed in 2026 has, as a function of pancreatic cancer therapy specifically, a slightly better median survival than the patient diagnosed in 2001, but only slightly. The glioblastoma patient has not benefited materially. The triple-negative breast cancer patient has benefited modestly. The ovarian cancer patient outside the BRCA subset has benefited modestly. For the bulk of solid-tumor cancer, targeted therapy has been an additive, incremental improvement, not a paradigm shift.

What the field has done, in response, is broaden its attention. Targeted therapy as the dominant frame of cancer drug development — the frame that the imatinib story established and the Time cover propagated — has, since the mid-2010s, shared the stage with several other approaches. Immunotherapy, particularly the class of drugs called immune checkpoint inhibitors, has produced transformative results in some cancers (melanoma, certain lung cancers, certain kidney cancers, Hodgkin lymphoma) and modest results in others. Antibody-drug conjugates — antibodies that deliver cytotoxic payloads selectively to cells expressing a particular surface marker — have expanded the targetable space substantially, particularly in HER2-positive disease, and are now in clinical development for many other markers. CAR-T cell therapies — patient T cells engineered to recognize and kill tumor cells — have produced striking results in certain blood cancers and are being adapted for solid tumors with mixed success. Each of these approaches has its own logic, its own successes, and its own limitations. None of them, individually, has been the universal solution that targeted therapy was sometimes thought to be in 2001.

The cancer therapy landscape in 2026, taken as a whole, is dramatically better than the cancer therapy landscape in 1995. The improvements are not uniform across cancers and they are not, in any single drug or approach, the kind of complete transformation that the early imatinib results produced in CML. The improvements are, more accurately, a thousand smaller transformations, each in a particular subset of a particular cancer, each driven by a particular molecular insight, each producing benefit measured in months or years for a particular group of patients. Cumulatively, this has added years of life and quality of life to enormous numbers of people. It has not, on the whole, made cancer go away.

* * *

The deepest thing the targeted-therapy era has done, perhaps, is not a particular drug but a particular habit of mind.

Cancer drug development before imatinib was largely empirical. You took compounds with broadly toxic activity, gave them to patients, watched what happened. The drugs that worked, you used. The drugs that didn't, you didn't. The mechanisms by which the drugs worked were often only partly understood, and the question of which patients would benefit from which drug was, in most cases, not well answered before the patient got the drug. Cancer drug development was a slower, more gradual, more accidental enterprise than what came after.

Cancer drug development after imatinib has been, increasingly, a different kind of enterprise. The standard sequence — identify the molecular driver, validate it experimentally, design a drug to inhibit it, identify which patients have the driver, give the drug to those patients — has become the default frame. The imatinib story established that this sequence was possible. The subsequent success stories established that it was reproducible, in the right cancers. The failure stories established that it was not universal, that some cancers were going to require different approaches. The combination of successes and failures has shaped what cancer drug development looks like in 2026 in a way it would not have been shaped without imatinib.

This is a real legacy. It is also a different legacy than the one the 2001 coverage promised. The promise was that targeted therapy would solve cancer. The actual delivery is that targeted therapy solved CML, transformed several other cancers, modestly improved many more, and gave the field a habit of asking molecular questions about cancer drugs that has, slowly and unevenly, raised the floor of what cancer treatment can achieve. The patients who have benefited have benefited substantially. The patients who have not benefited are still waiting.

The patients who have benefited — the ones who entered Druker's clinic in 1993 with three to five years to live, and who are now, in 2026, living more or less normal lives on a daily pill — are the subject of the next chapter, and the last chapter of this book. They are the chapter the book has been moving toward from the start. They are the people for whom the long argument from 1960 to 1998 to 2001 was, in the end, the argument that mattered.

Some of them are still in Druker's clinic.

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