The Poison That Cures

A history of chemotherapy, from the magic bullet to the antibody-drug conjugate

Explore this history

Chemotherapy was born from a weapon of war, became the first treatment to cure cancer that had spread, was pronounced obsolete by the targeted-therapy revolution — and has returned, strapped to antibodies, as the most versatile killer in oncology.

At the start of the twentieth century, Paul Ehrlich described an ambition that had occupied him for two decades: a chemical that would seek out a pathogen and kill it without touching the patient's own cells. He called it a Zauberkugel — a magic bullet. Ehrlich was not thinking about cancer. He was thinking about the spirochete of syphilis, and about the dyes he had used to stain cells in the 1880s, which had taught him that different tissues had different chemical affinities. If a dye could colour one kind of cell and not another, then perhaps a drug could poison one kind of cell and not another.

The idea was revolutionary, and it was also, for cancer, premature. Ehrlich's own magic bullet, Salvarsan, was an arsenic compound that worked against a foreign organism. Cancer is not foreign. It is the body's own tissue, gone wrong, and it offers few chemical handles that normal tissue does not also offer. The magic bullet would have to wait half a century for a target, and it would find its first target not in a laboratory but in the trenches of the First World War.

The history of chemotherapy is the history of that delay and its consequences. It is a story about a therapy that was discovered by accident, developed by trial and error, rationalised by cell kinetics, nearly displaced by molecular biology, and then reinvented — twice. Along the way it produced the first cures of disseminated cancer, the first medical oncology departments, and the first generation of cancer drugs that are still, in 2026, among the most widely used medicines in the world.

It is also a story about a single number: the therapeutic index, the ratio between the dose that kills a tumour and the dose that kills the patient. Radiotherapy widened that ratio with physics and imaging. Chemotherapy had to widen it with pharmacology — with schedules, combinations, supportive drugs, targeting devices, and, most recently, a blood test that tells the doctor whether to treat at all.

A primer: what chemotherapy actually does

The word "chemotherapy" now covers a large and confusing territory. Strictly, it means the treatment of disease with chemical compounds, and it includes antibiotics and antihistamines. In oncology it has come to mean cytotoxic chemotherapy — drugs that kill dividing cells — as distinct from targeted agents, hormones, immunotherapies, and radiotherapy. The distinction is historically contingent and biologically blurry, but it is the one most patients and doctors still use.

The classical cytotoxic drugs fall into a few families, each with a different way of damaging a cell:

Alkylating agents — nitrogen mustard, cyclophosphamide, chlorambucil, melphalan, busulfan, the nitrosoureas, and the platinum compounds. They attach alkyl groups to DNA, cross-linking strands and blocking replication.
Antimetabolites — methotrexate, 6-mercaptopurine, 5-fluorouracil, cytarabine, gemcitabine, pemetrexed. They masquerade as the building blocks of DNA and RNA, and starve or poison the synthesis of both.
Topoisomerase inhibitors — the anthracyclines (doxorubicin, daunorubicin), etoposide, irinotecan, topotecan. They trap the enzymes that unwind DNA, producing double-strand breaks.
Tubulin-binding agents — the vinca alkaloids (vincristine, vinblastine) and the taxanes (paclitaxel, docetaxel). They interfere with the microtubules that pull chromosomes apart at mitosis.

These drugs are not selective in any elegant sense. They kill any cell that is dividing, which is why they work on tumours and also on bone marrow, gut lining, hair follicles, and gonads. Their therapeutic index is narrow, and much of their history is the history of learning to manage that narrowness.

Two ideas organise the pharmacology.

The first is the log-kill hypothesis, formulated by Howard Skipper and colleagues in the 1960s from experiments on mouse leukaemia. A drug kills a constant fraction of cancer cells, not a constant number. If a tumour has a billion cells and a drug kills 99%, ten million remain; another cycle kills 99% of those, and so on. Cure therefore requires repeated cycles, and it requires starting when the tumour burden is small — which is why adjuvant chemotherapy, given after surgery when the tumour is microscopically small, works better than the same drugs given for bulky disease.

The second is combination chemotherapy. If a tumour contains a billion cells, and random mutation gives some of them resistance to any single drug, then a combination of drugs with different mechanisms and non-overlapping toxicities reduces the probability that any cell survives all of them. The Goldie-Coldman hypothesis (1979) formalised the logic: the larger the tumour, the more likely it harbours resistant clones, so combinations should be given early and, if possible, alternated. The clinical corollary was blunt: do not use one drug at a time if you can use four.

Everything else in the history — dose intensity, schedule, supportive care, targeted delivery — is a refinement of those two ideas.

1. The magic bullet, 1899–1935

Ehrlich's programme was the first systematic attempt at rational drug design. He synthesised hundreds of arsenic compounds, tested them against trypanosomes and spirochetes, and in 1909 found compound 606 — arsphenamine, marketed as Salvarsan — which was effective against syphilis. It was not a true magic bullet: it had to be given intravenously, it caused rashes and liver damage, and it was eventually replaced by penicillin. But it demonstrated the principle that a chemical could be designed to attack a pathogen selectively, and Ehrlich's term chemotherapy entered the language.

Cancer was a different problem. Ehrlich himself had noticed that tumours could be stained with dyes, and others had tried injecting dyes, acids, and metals into tumours in the hope of poisoning them. The results were uniformly disappointing. The reason is fundamental: cancer cells are the patient's own cells, and the differences between a cancer cell and a normal cell are differences of degree and of molecular wiring, not of kind. There is no equivalent of a bacterial cell wall to attack.

The first effective cancer drug therefore did not come from a rational search for a cancer-specific target. It came from a search for a weapon, and from a pathologist's curiosity about why soldiers exposed to gas died with shrunken spleens.

2. Mustard gas, 1917–1946

On 12 July 1917, the German army released sulfur mustard gas near Ypres. It was a horrific weapon: it blistered skin, blinded eyes, and destroyed the lining of the airways. Thousands of soldiers were injured, and many died — but not immediately, and not only of burns.

The American pathologist Edward Krumbhaar, studying the victims, noticed something that would change oncology. The gas had a profound effect on the blood and the lymphoid organs: the bone marrow was depleted, the lymph nodes were shrunken, the white cell count collapsed. Mustard gas was not just a blistering agent; it was a bone marrow poison and a lymphoid poison. It killed the cells that divide fastest.

The observation lay dormant for two decades, then returned through a second catastrophe. On 2 December 1943, German bombers attacked the harbour at Bari, Italy. Among the ships hit was the American Liberty ship SS John Harvey, which was carrying a secret cargo of mustard gas bombs. The gas escaped, and hundreds of sailors and civilians were burned and blistered. Lieutenant Colonel Stewart Alexander, a pathologist, performed autopsies and found the same thing Krumbhaar had seen: the victims' lymphoid tissue and bone marrow had been almost wiped out.

The American Chemical Warfare Service had already been studying nitrogen mustards — synthetic analogues of sulfur mustard — as potential weapons. The Bari findings reinforced the interest, and, through a wartime contract, the Chemical Warfare Service was also funding pharmacology at Yale. There, Louis Goodman, Alfred Gilman, and F. S. Philips had been testing nitrogen mustard in animals and had shown that it produced the same lymphoid depletion. They reasoned that a drug that destroyed lymphoid tissue might shrink a lymphoma.

On 27 August 1942, a patient at Yale-New Haven Hospital — a man with advanced lymphosarcoma, known in the records as "J.D." — received the first intravenous dose of nitrogen mustard. The tumour masses softened within days. The response was temporary, and the patient died months later, but the principle was established: a synthetic chemical could shrink a human cancer.

The results were classified. The Yale team published them in 1946, after the war, in the Journal of the American Medical Association, alongside related reports from other centres. The modern era of cancer chemotherapy is usually dated from that publication — or, more accurately, from the realisation it forced: that cancer was not only a surgical disease, and that a systemic poison could, for a time, control it.

3. Antifolates and the first cure, 1947–1960

Nitrogen mustard shrank tumours but did not cure them. The first cure came from a different direction: vitamins.

Sidney Farber was a pathologist at Boston Children's Hospital, and he had noticed that leukaemia cells seemed to depend on folic acid. In the 1940s, folic acid was known to stimulate the proliferation of bone marrow cells, and Farber reasoned that an anti-folate — a molecule that blocked folic acid metabolism — might starve leukaemia cells. In 1947, his team treated children with acute lymphoblastic leukaemia with aminopterin, a folic acid antagonist. Several children went into remission: their blasts disappeared from the blood, their bone marrow recovered, and they returned home, briefly.

The 1948 paper in the New England Journal of Medicine reported temporary remissions in a majority of the children treated. The remissions were short — weeks to months — and the drug was toxic. But the result was electrifying: a childhood leukaemia, universally fatal within months, had been pushed back.

Aminopterin was soon replaced by methotrexate, a less toxic analogue that became one of the most widely used drugs in oncology. And in the 1950s, methotrexate produced something even more remarkable. Min Chiu Li and Roy Hertz, at the National Cancer Institute, treated women with choriocarcinoma — a cancer of the placenta that kills quickly and spreads widely — with methotrexate. The tumours disappeared. Li kept treating until the pregnancy hormone hCG was undetectable in the urine, a decision that his superiors considered excessive and that led to his dismissal from the NCI in 1957. The patients stayed in remission. Li was vindicated, and the principle he had insisted on — treat until the marker of disease is gone, not until the scan looks better — became a foundation of curative chemotherapy.

Choriocarcinoma was the first disseminated solid tumour cured by drugs. The lesson was that chemotherapy could be curative, not merely palliative, if the tumour was sensitive and the treatment was complete.

4. The golden age of drug discovery, 1950–1970

The 1950s and 1960s produced most of the drugs that are still in use today. It was a period of systematic screening, medicinal chemistry, and, in a few cases, genuine rational design.

The most important partnership was at Burroughs Wellcome, where Gertrude Elion and George Hitchings built drugs by understanding the biochemistry of nucleic acids. They reasoned that if you knew how a cell made DNA, you could design a molecule that looked like a building block but could not be used. In 1951 they synthesised 6-mercaptopurine, which blocked purine synthesis and produced remissions in childhood leukaemia; it was approved in 1953. Their work also produced thioguanine, allopurinol, and, eventually, azathioprine for transplant rejection. Elion and Hitchings shared the Nobel Prize in Physiology or Medicine in 1988.

Charles Heidelberger, at the University of Wisconsin, took a similar approach to pyrimidines. In 1957 he produced 5-fluorouracil, a uracil analogue that blocked thymidylate synthesis. It became — and remains — the backbone of chemotherapy for colorectal, gastric, breast, and head and neck cancers.

The vinca alkaloids came from a plant. Vincristine and vinblastine were isolated in the late 1950s from Catharanthus roseus, the Madagascar periwinkle, whose leaves had been used in folk medicine for diabetes. They bound tubulin and blocked mitosis, and vincristine became a cornerstone of childhood leukaemia and lymphoma regimens. The antibiotics followed: actinomycin D from Streptomyces, mitomycin C, daunorubicin from Streptomyces peucetius in the 1960s, and doxorubicin — the most important of them — in 1969. Doxorubicin became the backbone of breast cancer, lymphoma, and sarcoma regimens, and its cardiotoxicity became one of the defining toxicities of the era.

The alkylating agents expanded too: cyclophosphamide (1958), melphalan, busulfan, chlorambucil, and the nitrosoureas. And in 1965, Barnett Rosenberg, a biophysicist at Michigan State, noticed that bacteria exposed to an electric field stopped dividing and grew into long filaments. The effect was not caused by the field but by a platinum compound leaching from the electrodes. Cisplatin was born, and by 1978 it was approved for testicular and ovarian cancer.

By the end of the 1960s, oncology had a cabinet of active drugs — and almost no idea how to use them together.

5. The kinetic model and the cure of Hodgkin lymphoma, 1960–1980

The idea that transformed chemotherapy was quantitative. Howard Skipper and his colleagues at the Southern Research Institute, working with mouse leukaemia, showed that a drug killed a constant fraction of cells, and that cure required repeated cycles to reduce the burden from billions to zero. The log-kill hypothesis made two predictions: dose and schedule mattered, and early treatment — when the tumour was small — was more effective than late treatment.

At the National Cancer Institute, Emil Frei, Emil Freireich, and colleagues tested the logic in childhood leukaemia. They combined drugs with different mechanisms — methotrexate, 6-mercaptopurine, vincristine, prednisone — and gave them in cycles, with supportive care for the infections and bleeding that followed. At St. Jude Children's Research Hospital, Donald Pinkel went further with "total therapy": combination chemotherapy plus cranial irradiation to prevent leukaemia from returning in the brain. The results, reported through the 1960s and early 1970s, turned a universally fatal disease into one with a substantial cure rate.

The most dramatic proof came in Hodgkin lymphoma. In 1964, Vincent DeVita and colleagues at the NCI began treating patients with advanced Hodgkin's disease with a four-drug combination — mechlorethamine, vincristine (Oncovin), procarbazine, and prednisone — which became known as MOPP. Advanced Hodgkin's had been uniformly fatal. MOPP produced complete remissions, and a substantial fraction of those patients never relapsed. When the long-term results were published in 1970, they showed that a disseminated cancer could be cured by drugs alone.

Other regimens followed. CHOP — cyclophosphamide, doxorubicin, vincristine, prednisone — became the standard for diffuse large B-cell lymphoma. In testicular cancer, Lawrence Einhorn and John Donohue at Indiana University combined cisplatin, vinblastine, and bleomycin (PVB), and reported in 1977 that a majority of patients with metastatic non-seminomatous germ cell tumours could be cured. Testicular cancer became the model of a curable solid tumour, and cisplatin became the most important new drug of the decade.

The theoretical scaffolding arrived alongside. The Goldie-Coldman hypothesis (1979) explained why combinations should be given early and alternated. The Norton-Simon model (1976) combined log-kill with the observation that tumours grow in a Gompertzian curve — fast when small, slow when large — and predicted that dose-dense schedules, with shorter intervals between cycles, would be more effective than the same total dose given less often. The concept of dose intensity, formalised by W. Hryniuk in the 1980s, made the prediction testable.

The political context mattered too. In 1971, President Nixon signed the National Cancer Act, which created the National Cancer Institute's comprehensive cancer centre programme and massively expanded federal funding for cancer research. The "War on Cancer" was declared, and chemotherapy — with its dramatic cures and its terrible side effects — became its most visible weapon.

The war was oversold, and the backlash was fierce. Critics pointed out that age-adjusted cancer mortality had barely moved since the 1950s, that the new drugs worked in a handful of rare tumours and hardly at all in the common ones, and that the rhetoric of "cure" had raised expectations the science could not meet. The statistician John Bailar published a famously bleak assessment in 1997, arguing that the war on cancer had been a qualified failure. The critique was partly right and partly wrong: mortality from several cancers did fall, but slowly, and often because of earlier detection and better surgery rather than because of new drugs. What the era did produce, unmistakably, was an infrastructure — cooperative groups, cancer centres, trial networks, and a generation of trained oncologists — that made everything afterwards possible.

6. Adjuvant therapy and the breast cancer wars, 1970–1990

The next conceptual leap was to treat cancer that could not be seen.

By the 1970s, it was clear that many patients who had "curative" surgery for breast cancer later died of metastatic disease. The explanation was micrometastases: cancer cells that had escaped the breast before surgery and were too small to detect. If chemotherapy worked best against small tumours, then giving it after surgery, before the micrometastases grew, should improve survival.

Gianni Bonadonna and colleagues in Milan tested this in the early 1970s. Their trial of cyclophosphamide, methotrexate, and fluorouracil (CMF) after surgery for node-positive breast cancer, published in 1976, showed a reduction in recurrence. It was the first convincing evidence that adjuvant chemotherapy worked, and it changed the practice of breast cancer irrevocably.

The National Surgical Adjuvant Breast and Bowel Project (NSABP), led by Bernard Fisher, extended the logic. Fisher's work established that breast-conserving surgery plus radiation was equivalent to mastectomy in appropriate patients, and that adjuvant chemotherapy and tamoxifen improved survival. The randomised trial became the engine of oncology, and medical oncology became a specialty with its own clinics, its own training programmes, and its own culture.

Adjuvant chemotherapy spread to other cancers: osteosarcoma in the 1970s, colorectal cancer in the late 1980s (5-fluorouracil plus levamisole, later oxaliplatin), and gastric, lung, and bladder cancer in the 1990s and 2000s. The principle was the same everywhere: treat the invisible disease while it is still small.

The same trial machinery became a moral instrument as well as a statistical one. The 1970s and 1980s also brought a reckoning with research ethics: the Tuskegee syphilis study was exposed in 1972, the Belmont Report codified the principles of respect for persons, beneficence, and justice in 1979, and institutional review boards became mandatory for federally funded research. Oncology trials, which often enrolled dying patients and tested toxic drugs, were at the centre of that reckoning. The cooperative groups — the NSABP, the Eastern Cooperative Oncology Group, the Southwest Oncology Group, the European Organisation for Research and Treatment of Cancer — became the machinery through which a single institution's experience was turned into generalisable evidence. The adjuvant era was built as much on statistics as on pharmacology.

7. The price, 1970–2000

Chemotherapy's successes were purchased at a cost that the early literature often understated.

The acute toxicities were brutal and familiar: nausea and vomiting, mouth ulcers, hair loss, fatigue, and the destruction of the bone marrow that left patients defenceless against infection and dependent on transfusions. In the 1970s and 1980s, a course of chemotherapy could mean weeks in hospital with fever and neutropenia, and a substantial number of patients died of treatment-related infections.

Supportive care changed that, and it is one of the great unsung achievements of oncology. The 5-HT3 antagonists, introduced with ondansetron in 1991, transformed the control of chemotherapy-induced nausea. The NK1 antagonists, beginning with aprepitant in 2003, improved it further. Granulocyte colony-stimulating factor (G-CSF), approved in 1991, reduced the duration and severity of neutropenia and allowed dose-dense schedules. Erythropoietin reduced transfusion dependence. Dexrazoxane, approved in 1995, reduced anthracycline cardiotoxicity. Antibiotic prophylaxis, antifungals, and better supportive care turned chemotherapy from a hospital-bound ordeal into, for many patients, an outpatient regimen.

The late toxicities were slower to be recognised. Anthracyclines damaged the heart, sometimes years later. Alkylating agents and topoisomerase inhibitors caused secondary leukaemias. Cisplatin damaged the kidneys, the nerves, and the hearing. Taxanes caused peripheral neuropathy that could persist for years. Many patients became infertile. Some developed cognitive changes — "chemobrain" — that were long dismissed and are still poorly understood.

The emotional and cultural cost was equally real. Chemotherapy became a symbol of cancer itself: the bald head, the vomiting, the venous access port, the phrase "going through chemo". The therapy that cured some patients terrified all of them, and the fear shaped how patients consented to treatment, how oncologists communicated, and how the public imagined the disease.

8. Resistance and the limits of "more is better", 1980–2000

By the 1980s, the easy victories had been won. The curable cancers — testicular, Hodgkin and non-Hodgkin lymphoma, childhood leukaemia, choriocarcinoma, osteosarcoma — were being cured in most patients. The common solid tumours — breast, lung, colon, prostate — responded to chemotherapy but rarely disappeared. The field confronted resistance.

The first explanation was pharmacological. In 1976, Juliano and Ling identified P-glycoprotein, an efflux pump that cancer cells overexpressed and that pumped drugs out of the cell. The discovery launched a decade of research into multidrug resistance modulators — verapamil, cyclosporine, and others — and a decade of clinical trials that largely failed. The pumps were real, but blocking them did not reverse resistance in patients.

The deeper explanation was evolutionary. Tumours are genetically unstable, and a billion cells contain a vast reservoir of variants. Peter Nowell's 1976 model of clonal evolution described cancer as a Darwinian process: treatment selects for resistant clones, which then repopulate. Resistance was not a single mechanism but a moving target.

The clinical response was to give more: higher doses, more drugs, shorter intervals, and, in the 1990s, high-dose chemotherapy with autologous stem cell rescue. The logic was appealing — if log-kill is proportional to dose, then a massive dose should eliminate the last cell. Breast cancer became the testing ground, and thousands of women underwent high-dose chemotherapy with stem cell transplantation. Randomised trials in the late 1990s and early 2000s showed that it did not improve survival and increased toxicity. The episode is now remembered as a cautionary tale about the seduction of dose and the importance of randomised evidence.

New drugs did arrive — paclitaxel in 1992, docetaxel in 1996, irinotecan and topotecan in 1996, gemcitabine in 1996 — and they improved outcomes at the margins. But the plateau was real. By the end of the 1990s, the great era of cytotoxic discovery was over, and the field was looking elsewhere.

9. The targeted revolution and chemotherapy's identity crisis, 1990–2020

The "elsewhere" turned out to be molecular biology.

In 1997, rituximab became the first monoclonal antibody approved for cancer. In 1998, trastuzumab was approved for HER2-positive breast cancer. In 2001, imatinib transformed chronic myeloid leukaemia from a fatal disease into a chronic condition managed with a daily pill. The magic bullet had finally found its targets: the fusion protein BCR-ABL, the receptor HER2, the B-cell marker CD20.

The rhetoric of the era was that chemotherapy was obsolete — "targeted therapy" would replace the shotgun with a sniper rifle. The reality was more complicated. Targeted drugs worked brilliantly in tumours that depended on a single oncogene, and much less well in the common solid tumours, which depended on many. And even where they worked, they were usually combined with chemotherapy: trastuzumab with a taxane, rituximab with CHOP, cetuximab with radiotherapy or platinum.

Chemotherapy's role changed rather than disappeared. It became the backbone on which targeted drugs were added, the debulking agent that made immunotherapy more effective, and the fallback when resistance emerged.

The immunotherapy partnership was the most consequential. In 2018, the KEYNOTE-189 trial showed that adding pembrolizumab to platinum-based chemotherapy improved overall survival in metastatic non-small-cell lung cancer, and similar results followed in triple-negative breast cancer, gastric cancer, and other diseases. Chemotherapy, once thought to be immunosuppressive, turned out to be a useful partner: it releases tumour antigens, depletes immunosuppressive cells, and can make immunologically "cold" tumours more visible to the immune system. The combination is now standard in several cancers, and the chemotherapy component is often what makes the immunotherapy work.

At the same time, the field began to question whether everyone needed chemotherapy at all. Genomic assays — Oncotype DX (2004) and MammaPrint (2007) — measured the expression of genes associated with recurrence in early breast cancer and predicted who would benefit from adjuvant chemotherapy. The TAILORx trial (2018) showed that most women with intermediate-risk, hormone-receptor-positive, HER2-negative breast cancer did not benefit from chemotherapy added to endocrine therapy. MINDACT (2016) and RxPONDER (2021) produced similar messages. De-escalation became a theme: not more treatment, but the right treatment.

10. The magic bullet, armed: antibody-drug conjugates, 2000–2026

The most important development in chemotherapy in the last two decades is, paradoxically, a way of making chemotherapy targeted.

An antibody-drug conjugate (ADC) is a monoclonal antibody that binds a protein on the surface of cancer cells, linked to a cytotoxic payload. The antibody provides the targeting; the payload provides the killing. The payload is usually a classical chemotherapy drug — monomethyl auristatin E, emtansine, deruxtecan, SN-38, exatecan — attached by a linker that is stable in the bloodstream and cleaved inside the tumour cell.

The concept dates from the 1980s, but the first success came in 2011, when brentuximab vedotin was approved for Hodgkin lymphoma and anaplastic large-cell lymphoma. Trastuzumab emtansine (T-DM1) followed in 2013 for HER2-positive breast cancer. Then came trastuzumab deruxtecan (T-DXd) in 2019, which produced unprecedented responses in HER2-positive and, later, HER2-low breast cancer and in gastric cancer, and sacituzumab govitecan in 2020 for triple-negative breast cancer.

An ADC has three parts, and each is an engineering problem. The antibody must bind an antigen that is abundant on tumour cells and scarce on vital normal tissue. The linker must be stable in the bloodstream — otherwise the payload leaks and poisons the patient — but cleavable inside the tumour cell. The payload must be potent enough to kill a cell at very low concentrations, because only a small fraction of the drug reaches the target. The drug-to-antibody ratio, the choice of payload, and the chemistry of the linker determine whether the drug is a precise weapon or a slow-release poison. Some ADCs also have a "bystander effect": the payload diffuses out of the target cell and kills neighbouring tumour cells that do not express the antigen, which helps against heterogeneous tumours but can widen toxicity.

By 2026, ADCs had become the fastest-growing class in oncology, and the results had begun to challenge chemotherapy directly.

In May 2026, the FDA approved datopotamab deruxtecan, a TROP2-directed ADC, for previously untreated metastatic triple-negative breast cancer in patients who are not candidates for immunotherapy. The TROPION-Breast02 trial randomised 644 patients to the ADC or to investigator's choice of chemotherapy. Median progression-free survival was 10.8 months versus 5.6 months; median overall survival was 23.7 months versus 18.7 months; the response rate was 64% versus 30%. An ADC had beaten chemotherapy in a head-to-head first-line trial.

Five months later, at the European Society for Medical Oncology Congress in October 2026, the OptiTROP-Lung06 trial reported that sacituzumab tirumotecan, a TROP2 ADC, combined with pembrolizumab, more than doubled progression-free survival compared with chemotherapy plus pembrolizumab in PD-L1-negative non-small-cell lung cancer: 15.0 months versus 5.7 months. Grade 3 or higher adverse events were less frequent with the ADC combination (47.6% versus 60.6%), although stomatitis was a common side effect. The result was published simultaneously in the New England Journal of Medicine.

The irony is exquisite. The "chemo-free" future that targeted therapy promised in 2001 has arrived, but it is carrying chemotherapy inside an antibody. The cytotoxic payload is still doing the killing; the antibody is merely the delivery van. The magic bullet, it turns out, needed a poison after all.

ADCs are not a panacea. They require a target that is expressed on tumour cells and not on vital normal tissue; they cause interstitial lung disease, neuropathy, and myelosuppression; they are expensive; and resistance emerges. But they have changed the arithmetic of chemotherapy: the therapeutic index can now be widened by targeting, not just by scheduling.

11. The molecular residual disease era, 2015–2026

The other revolution of the past decade is diagnostic. Circulating tumour DNA (ctDNA) — fragments of tumour DNA shed into the blood — can be detected after surgery to identify patients with microscopic residual disease. If chemotherapy is most effective against small tumours, then a blood test that finds the small tumour should tell us who needs treatment.

The landmark trial was DYNAMIC, published in 2022. In stage II colon cancer, a ctDNA-guided approach reduced the use of adjuvant chemotherapy from 28% to 15% without compromising recurrence-free survival. The result suggested that many patients were being overtreated, and that a blood test could safely spare them.

The PEGASUS trial, published in Nature Cancer in 2026, extended the idea to high-risk stage II and stage III colon cancer. In 135 patients, ctDNA was measured after surgery; 26% were positive. Patients who were negative on two consecutive tests received a less intensive, oxaliplatin-free regimen. At three years, 83% of the ctDNA-negative patients were disease-free and 94% were alive — despite receiving less treatment. Among ctDNA-positive patients, disease-free survival was about 62%, even with intensified therapy.

The results are not conclusive, and the field is still arguing about sensitivity, false negatives, and whether escalation actually helps the patients who test positive. A negative ctDNA test is not a guarantee of cure — the assay can miss tumours that shed little DNA — and a positive test does not prove that more treatment will work. The assays are not standardised, the costs are high, and the trials that would show a survival benefit are still maturing; at least one 2026 strategy trial missed its primary endpoint. But the direction is clear: chemotherapy is moving from "give it to everyone with stage III disease" to "give it to the people whose blood says they need it". The same logic is being tested in breast, lung, and bladder cancer, and the first ctDNA-guided trials have begun to report both successes and setbacks.

12. What chemotherapy does today

The cumulative result is a therapy that is narrower in ambition but broader in use than its pioneers imagined.

Cure. Chemotherapy still cures a remarkable list of cancers: testicular germ cell tumours (cure rates of roughly 80–90% even in metastatic disease), Hodgkin and some non-Hodgkin lymphomas, acute leukaemias, choriocarcinoma, osteosarcoma, and many childhood cancers. In these diseases, the old drugs — cisplatin, doxorubicin, vincristine, methotrexate, cyclophosphamide — remain the backbone, and no targeted drug has replaced them.

Adjuvant and neoadjuvant treatment. Chemotherapy given before or after surgery improves survival in breast, colorectal, gastric, oesophageal, lung, and bladder cancer. In rectal cancer, neoadjuvant chemoradiation can shrink tumours enough to allow sphincter-preserving surgery, or, in selected patients, a watch-and-wait strategy.

Palliation. In advanced disease, chemotherapy can shrink tumours, relieve pain, and prolong life by months to years. The trade-off between benefit and toxicity is the central ethical question of medical oncology, and it is decided differently for a 40-year-old with a responsive tumour and an 85-year-old with a chemoresistant one.

Combination with immunotherapy. Chemotherapy is not immunosuppressive in the simplistic sense once believed. It can release tumour antigens, deplete regulatory T cells, and sensitise tumours to checkpoint inhibitors. Chemo-immunotherapy combinations are now standard in non-small-cell lung cancer, triple-negative breast cancer, gastric cancer, and several other diseases, and the chemo component is often the part that makes the immunotherapy work.

Oral and metronomic regimens. Capecitabine, temozolomide, and oral targeted agents have made some chemotherapy home-based. Low-dose metronomic chemotherapy, given continuously, may work partly by anti-angiogenic and immune mechanisms rather than by direct killing.

De-escalation. Genomic assays, ctDNA, and clinical trials have made it possible to give less. The field is learning that the therapeutic index can be widened by withholding treatment as well as by intensifying it.

Less chemotherapy. In a few diseases, chemotherapy has begun to recede. Imatinib replaced interferon and cytotoxic drugs in chronic myeloid leukaemia. In chronic lymphocytic leukaemia, targeted agents — ibrutinib, venetoclax — have displaced chemoimmunotherapy for many patients. In diffuse large B-cell lymphoma, the POLARIX trial (2021) showed that replacing vincristine with the antibody-drug conjugate polatuzumab vedotin improved outcomes, and pola-R-CHP is now a standard option. In HER2-positive breast cancer, trastuzumab deruxtecan has produced responses in patients who have failed several lines of therapy. The direction of travel is toward less cytotoxic chemotherapy and more targeted payloads — but not, yet, to none.

13. Access, shortages, and the economics of old drugs

Chemotherapy's oldest drugs are also its most essential, and that is precisely why they are vulnerable.

In 2023, the United States experienced a severe shortage of cisplatin and carboplatin, two of the most widely used cytotoxic drugs. A single manufacturer's plant shutdown, combined with the fragility of the generic sterile-injectable market, left oncology practices rationing doses and delaying treatment. The shortage persisted into 2024 and 2025, and surveys of oncology practices reported that a substantial proportion had been forced to modify regimens. Similar shortages have affected 5-fluorouracil, methotrexate, and other generics.

The economics explain the problem. Generic cytotoxic drugs are cheap, off-patent, and difficult to manufacture safely; manufacturers have little incentive to maintain capacity, and a single quality failure can remove a large share of supply. The newest drugs — ADCs, checkpoint inhibitors, CAR-T cells — cost tens or hundreds of thousands of dollars per patient per year, and the health systems that cannot afford them are often the same ones that cannot reliably obtain cisplatin.

The geography matters too. A large share of the world's generic chemotherapy is manufactured in India and China, and a disruption in one plant can ripple across continents. The COVID-19 pandemic exposed how little slack the supply chain had. International bodies have called for stockpiles, diversified manufacturing, and price guarantees, but the incentives remain misaligned: the drugs that matter most are the ones that earn the least.

The World Health Organization's Essential Medicines List includes many cytotoxic drugs, and global access remains deeply unequal. In high-income countries, the debate is about the price of innovation; in low- and middle-income countries, it is about whether a child with Burkitt lymphoma can get cyclophosphamide at all.

14. The open questions

Can resistance be reversed? The dream of the 1980s — a drug that blocks multidrug resistance — has not been realised. The modern approach is to characterise the resistance mechanism in each patient and attack it specifically: a second-generation inhibitor after a first-generation target fails, an ADC with a different payload, a combination that exploits a synthetic lethal vulnerability. Resistance is now understood as an evolutionary process, and the goal is to stay ahead of it rather than to abolish it.

Can toxicity be reduced without losing efficacy? Supportive care has improved enormously, but neuropathy, cardiotoxicity, infertility, and cognitive effects remain. The ADCs reduce some toxicities and introduce others. The measurement of toxicity — patient-reported outcomes, quality-adjusted survival — is becoming as important as the measurement of response.

Are ADCs the future of chemotherapy, or a transitional technology? The head-to-head wins in 2026 are impressive, but ADCs are not effective in every tumour, and their payloads are still cytotoxic. The next generation may use different payloads — immunomodulators, radionuclides, degraders — or may eliminate the payload entirely.

Who should not be treated? The hardest question in oncology is when to stop. Genomic assays and ctDNA can identify low-risk patients who can safely avoid chemotherapy, but they can also create anxiety and overtreatment. The integration of biology, patient preference, and cost is unresolved.

Can AI discover better drugs? Machine learning is being used to design molecules, predict toxicity, and match patients to trials. The first AI-derived drugs are in early trials. Whether the technology will produce a genuinely new class of cytotoxic agents, or merely accelerate the existing pipeline, is unknown.

Will the cytotoxic itself be replaced? The most interesting alternative is not a better poison but a different mechanism. Targeted protein degraders — PROTACs and molecular glues — do not damage DNA; they mark a cancer-driving protein for destruction by the cell's own recycling machinery. Several are in clinical trials, and they represent a genuinely new pharmacology rather than a refinement of the old one. If they succeed, the next history of "chemotherapy" may be a history of drugs that do not kill dividing cells at all, but simply remove the proteins that make a cell cancerous.

Can the world afford the drugs it has? The shortage of generic chemotherapies is a market failure, not a scientific one, and it is solvable with policy. The affordability of ADCs and immunotherapy is harder.

Timeline

Year	Event	Why it mattered
1909	Ehrlich develops Salvarsan	The first rationally designed chemotherapeutic
1917	Mustard gas used at Ypres	The observation that alkylating agents destroy bone marrow and lymphoid tissue
1919	Krumbhaar describes mustard-induced marrow aplasia	The link between chemical warfare and cancer therapy
1942	First patient treated with nitrogen mustard (Yale)	The birth of cancer chemotherapy
1943	Bari harbour disaster	Confirms the lymphoid toxicity of nitrogen mustard
1946	Yale nitrogen mustard papers published	The field becomes public
1947–48	Farber's aminopterin trials in childhood leukaemia	The first drug-induced remissions in leukaemia
1949	Methotrexate introduced	A less toxic antifolate; the workhorse of the field
1951–53	6-Mercaptopurine (Elion and Hitchings)	Rational antimetabolite design
1956–58	Min Chiu Li cures choriocarcinoma with methotrexate	The first cure of a disseminated solid tumour
1957	5-Fluorouracil (Heidelberger)	The backbone of gastrointestinal chemotherapy
1958	Vincristine and cyclophosphamide	New classes: tubulin binders and alkylators
1964	Log-kill hypothesis (Skipper)	The quantitative basis of dose and schedule
1964–70	MOPP for Hodgkin lymphoma (DeVita)	The first cure of advanced Hodgkin lymphoma
1965–69	Cisplatin discovered (Rosenberg)	The most important new cytotoxic of the era
1969–74	Doxorubicin introduced	The anthracycline backbone
1971	National Cancer Act	The War on Cancer; expansion of clinical research
1976	Bonadonna's CMF adjuvant trial	Adjuvant chemotherapy proven in breast cancer
1976–79	Norton-Simon and Goldie-Coldman models	Dose density and combination logic
1977	PVB for testicular cancer (Einhorn)	A curable solid tumour
1978	Cisplatin approved	Testicular and ovarian cancer treatment
1980s	CHOP becomes standard for lymphoma	Combination chemotherapy for common lymphomas
1991	Ondansetron and G-CSF approved	Supportive care transforms tolerability
1992–96	Paclitaxel, docetaxel, irinotecan, gemcitabine	The last great wave of cytotoxics
1997–2001	Rituximab, trastuzumab, imatinib	The targeted-therapy era begins
2004–07	Oncotype DX and MammaPrint	Genomic selection for adjuvant chemotherapy
2011	Brentuximab vedotin approved	The first modern ADC
2013	Trastuzumab emtansine approved	ADCs enter breast cancer
2016–18	MINDACT and TAILORx	De-escalation of adjuvant chemotherapy
2019	Trastuzumab deruxtecan approved	ADCs become a major class
2021	RxPONDER	Many postmenopausal patients can safely avoid chemotherapy
2022	DYNAMIC trial	ctDNA-guided adjuvant therapy
2023–26	Global shortages of cisplatin, carboplatin, 5-FU	The fragility of generic supply
2025	Datopotamab deruxtecan approved for breast cancer	TROP2 ADCs enter practice
2026	TROPION-Breast02; OptiTROP-Lung06; PEGASUS	ADCs beat chemotherapy head-to-head; ctDNA guides treatment

Coda

Chemotherapy was born from a poison, rationalised by a mouse leukaemia model, and vindicated by a handful of cures that seemed, at the time, almost miraculous. It was then declared obsolete by a generation of scientists who had found the molecular targets it lacked — and it survived, because the targets were rare and the poison was reliable.

The deepest lesson of its history is about the relationship between selectivity and toxicity. Radiotherapy could aim; surgery could cut; chemotherapy could only poison, and it had to make poisoning tolerable. It did so by scheduling, by combining, by supporting, by targeting, and finally by attaching the poison to an antibody that knows where to go.

That is why the ADCs are the fitting end of the story — or rather, its latest chapter. They are not a repudiation of chemotherapy. They are chemotherapy's oldest idea, selective toxicity, finally equipped with a guidance system. Ehrlich would have recognised the ambition. He might have been surprised that it took a hundred years, a world war, and a periwinkle to get there.

The poison still cures. It is simply getting better at knowing where to go.

Research Blog - Terms of Use

Synergies for Cancer Treatment

Pages