Loading
Targeted drugs switch off the specific broken protein a cancer depends on. The first ones turned a leukaemia into a chronic condition; the field then learned that resistance is the rule, designed drugs around it, and has now reached the drivers that were called impossible to target.
Tamoxifen (1977) and trastuzumab (1998) were targeted therapies before the term existed, but imatinib (2001) defined the category: a pill against the one enzyme a cancer cannot live without, matched to the patients whose tumours carry it. EGFR, ALK and BRAF followed within a decade, and with them the lesson that shaped everything since: nearly every targeted drug stops working within months to years, and the mechanism of escape can be read from the tumour and drugged in turn.
The second generation was designed for resistance and the brain: osimertinib, alectinib and lorlatinib, then their use after surgery (ADAURA, ALINA) and after chemoradiation (LAURA). PARP inhibitors made an inherited DNA-repair defect a treatable target; tissue-agnostic approvals made the mutation, not the organ, the indication. The third generation is reaching targets that were called undruggable: KRAS G12C (sotorasib 2021), then G12D and pan-RAS(ON) inhibitors, HIF-2 in kidney cancer, menin in leukaemia, and protein degraders (vepdegestrant, the first approved PROTAC, 2026) that remove a protein rather than block it.
The pace is set by resistance biology, by the combinatorial space of pairings that trials cannot search, by prices that compound over years of therapy, and by biomarkers that must be validated before a drug can be matched to a patient.
Tamoxifen blocked the oestrogen receptor and trastuzumab (1998) the HER2 protein, each with a test to find the patients whose tumours depended on them. Imatinib (2001) proved the concept in its purest form: a pill against BCR-ABL, the one enzyme chronic myeloid leukaemia cannot live without, that turned a fatal disease into a chronic one and made Brian Druker's Oregon clinic the birthplace of precision oncology.
Numbers are from the trial as recorded here; see the source links in the table below. This is orientation, not medical advice: ask your team how closely the trial population matches you.
Numbers are from the trial as recorded here; see the source links in the table below. This is orientation, not medical advice: ask your team how closely the trial population matches you.
EGFR mutations (2004) explained why gefitinib and erlotinib worked spectacularly in a minority of lung cancers; crizotinib for ALK (2011) and vemurafenib for BRAF (2011) followed within months of the target being found. Then nearly every responder relapsed: T790M in EGFR, gatekeeper mutations in ALK, MAPK reactivation in BRAF. Resistance became the central problem of the field, and reading it from a biopsy became the route to the next drug.
Osimertinib was built to hit T790M and cross into the brain; alectinib and lorlatinib did the same for ALK. Given after surgery, the pills halved the risk of death in EGFR-mutant lung cancer (ADAURA) and sharply cut recurrence in ALK-positive disease (ALINA); LAURA extended them to after chemoradiation. Olaparib (2014) made an inherited BRCA defect a target in ovarian, breast and prostate cancer (SOLO-1, OlympiA, PROfound). Larotrectinib's approval for any NTRK-fusion tumour (2018) made the mutation, not the organ, the indication; RET (LIBRETTO-431) and ROS1 followed. COMBI-AD proved a year of BRAF-MEK pills after melanoma surgery halves relapse; DREAMseq settled that immunotherapy should come first.
KRAS, mutated in a quarter of cancers and called undruggable for forty years, yielded to sotorasib (2021) and adagrasib, which lock the G12C mutant in its off state; CodeBreaK 200 and KRYSTAL-12 showed they beat chemotherapy, and divarasib beat both head to head (Krascendo 1). The G12D mutation, the commonest in pancreatic cancer, has its first drug in zoldonrasib, and daraxonrasib inhibits every RAS in its active state (RASolute 302). Elsewhere belzutifan drugged HIF-2, menin inhibitors reached leukaemia, and vepdegestrant (2026) became the first approved PROTAC, removing its target rather than blocking it. MARIPOSA and FLAURA2 showed that combinations beat osimertinib alone.
Neladalkib is designed to work after lorlatinib (ALKOVE-1); pirtobrutinib works after other BTK inhibitors and a BTK degrader is in phase 3 (CaDAnCe-304); the first MET-directed ADC is in its confirmatory trial. SERENA-6 changed endocrine therapy on a blood test before the scan changed, and INSIGHT is matching GIST drugs to the resistance mutation found in blood. The next trial designs assign treatment by how the tumour escaped rather than where it started.
Molecular glues and degraders reach proteins without a druggable pocket; degrader-antibody conjugates deliver them into the tumour; antisense and siRNA silence the gene rather than the protein; designed proteins grip surfaces no small molecule can. Synthetic lethality finds a second gene a cancer needs only because its first is broken: PRMT5 inhibitors for MTAP-deleted tumours and WRN inhibitors for mismatch-repair-deficient ones are the leading examples. Chemoproteomics is finding covalent handles on KRAS and beyond.
If the next resistance mutation can be predicted, as flu strains are, it could be blocked or vaccinated against before it takes over; adaptive dosing that keeps the sensitive clone dominant is in randomised phase 2. MYC and mutant p53, the drivers still without a drug, are the open frontier, and eprenetapopt's failure to refold p53 shows how hard it is. A ready-made KRAS vaccine after pancreatic surgery missed its primary goal, but the idea of vaccinating against a driver mutation remains live.
Every targeted drug meets resistance; a tumour is many tumours, so the resistant clone is usually already there. There are thousands of possible combinations and sequences and trials can test a few dozen a year. A drug cannot be matched without a validated test, and tests lag drugs. And targeted therapy is priced per month for years, so the cost of a cure-like outcome compounds; the biosimilar and generic wave for the first generation is only now arriving.
Every era's records, trial outcomes and papers, and every watch item, as JSON.
Probability ranges are named estimates that the claim is borne out on roughly a five-year horizon. They are meant to be argued with: propose a revision with your name and reasoning via a pull request to src/data/confidence.ts.
Loading this step…
Loading this step…
Loading this step…
Loading this step…
Loading this step…
Loading this step…
Loading this step…
Loading this step…
The second driver in lung cancer, and the one that proved the first was not a special case. It also established mutual exclusivity as a working assumption: a tumour usually has one driver, so finding it tells you what to give.
The first time a targeted biological agent extended survival in lung cancer, and the first time median survival in the advanced setting crossed twelve months. It also set the pattern that a drug's exclusion criteria can matter as much as its mechanism.
The template for every driver mutation since: find the responders, sequence them, and give the drug only to people whose tumour carries the lesion it was built for. Gefitinib had been close to abandonment on the strength of unselected trials.
Why a drug can look useless in one trial and transformative in another: the trials had different proportions of the patients the drug was for. It is the argument for genotyping before drawing conclusions from a response rate.
Shares Divarasib, Olomorasib, Identification of the transforming EML4-ALK fusion gene in non-small-cell lung cancer, Lazertinib.
Shares CodeBreaK 200, CodeBreaK 300, KRAS roadmap: undruggable → G12C → pan-RAS, Adagrasib.
Shares PRMT5 (MTAP-deleted cancers), Synthetic lethality approaches, Sotorasib, ALK.
Shares Divarasib, Elironrasib, Olomorasib, RASolute 302.
Shares MRTX1133, Elironrasib, Zoldonrasib, RASolute 302.
Shares LIBRETTO-431, MARIPOSA, ROS1, Selpercatinib.
Shares MARIPOSA, FLAURA2, ADAURA, Amivantamab.
Shares Identification of the transforming EML4-ALK fusion gene in non-small-cell lung cancer, FLT3, ROS1, Selpercatinib.