Loading
Radical oncology is a horizon map of the wilder ideas in cancer, sorted by how close they are to mattering, with the reason each one might never arrive.
Most of what is celebrated as a breakthrough is an incremental gain on an existing modality. This roadmap collects the ideas that would change the shape of treatment rather than its slope: therapies that read DNA directly, living drugs, radiation delivered in milliseconds, and detection that runs continuously rather than annually.
The grouping is by horizon, from technologies already producing randomised data to ideas with no human evidence at all. Placement reflects the state of evidence in September 2026, not company timelines. Several steps will not happen: the failure record of hypoxia-activated prodrugs, metabolic therapy, and matrix-softening agents is a reminder that a clean mechanism is not a clinical benefit.
Microbiome modulation, spatially fractionated radiotherapy, sonodynamic therapy in glioblastoma, GDF-15 blockade for cachexia, and repurposed cheap drugs all have randomised or registrational trials running in 2026. These are the frontier ideas closest to a guideline: each has a defined population, a comparator, and a readout inside a few years. Cachexia therapy is the likeliest first approval in a domain with no approved drug at all.
Engineered bacteria, logic-gated cell therapies, molecular glue platforms, and de novo designed binders all have first-in-human programmes running. The question each faces is the same: does the elegant mechanism survive contact with a heterogeneous human tumour? Expect most to disappoint on response rate while establishing safety, which is how bispecific antibodies and ADCs also began.
Auger emitters, contained alpha nanogenerators, proton arc delivery, and very-high-energy electrons are all limited today by engineering rather than biology: isotope supply, daughter recoil, gantry speed, and dosimetry at ultra-high dose rate. Those are tractable problems with capital behind them. If FLASH sparing is real in humans, deep FLASH by electron or proton arc would be the largest change to radiotherapy since intensity modulation.
Fragmentomics, breath analysis, and near-continuous ctDNA sampling all push detection from an annual event towards a running signal, while spatial omics and organoid testing push treatment choice from genotype towards phenotype and architecture. The gating question for every one of them is not sensitivity but utility: acting earlier has to change outcomes, and no randomised trial has yet shown that for continuous monitoring.
In vivo base and prime editing, epigenetic silencing, antibody-oligonucleotide conjugates, and programmable DNA-targeting drugs share one blocker: getting a large, charged molecule into most cells of a solid tumour. Liver-directed editing is already in the clinic, so the chemistry works; the tumour delivery problem has resisted thirty years of effort. If it is solved, undruggable drivers such as MYC and TP53 loss become addressable and much of this roadmap is rewritten.
Stromal CAR-T, nerve blockade, senescence clearance, mechanical decompression, and targeting the tumour's own microbes all treat the environment a cancer needs rather than the cancer itself. The attraction is that the host does not mutate. The risk is visible in the record: PEGPH20 failed, FAP CAR-T caused cachexia in mice, and broad antibiotics blunt immunotherapy. Success here probably requires far better spatial measurement first.
DNA origami nanorobots, phage-based delivery, and quantum-dot imaging agents have striking preclinical demonstrations and no clinical footprint. They belong on the map because the failure modes are known and specific, nuclease degradation, rapid clearance, heavy-metal toxicity, rather than vague. Any of them could move a horizon if a single delivery or materials problem is solved.
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…
Shares Senolytics and senescence-directed therapy, Chronotherapy: timing treatment to the body clock, Auger-electron therapy, Breath and volatile-organic-compound detection.
Shares Photothermal (plasmonic) nanoparticle ablation, Histotripsy as an immune primer, Sonodynamic therapy, In situ vaccination.
Shares DNA origami nanorobots, Antibody-oligonucleotide conjugates, Engineered exosomes as drug carriers, Programmable DNA-targeting therapeutics.
Shares In vivo base and prime editing for cancer, Self-amplifying and circular RNA therapeutics, Engineered exosomes as drug carriers, Logic-gated therapeutics (AND, NOT gates).
Shares Lattice and GRID radiotherapy, Histotripsy as an immune primer, In situ vaccination.
Shares Magnetic nanoparticle hyperthermia, Radiodynamic therapy and radiosensitising nanoparticles, Devices and physical therapies roadmap: heat and light → electric fields and focused sound → drug-releasing implants, Radiotherapy roadmap: X-rays → shaped beams → fewer fractions, particles and FLASH.
Shares Alpha-emitter nanogenerators and daughter trapping, Auger-electron therapy, Total-body PET for screening and ultra-low-dose imaging.
Shares Radiodynamic therapy and radiosensitising nanoparticles, Sonodynamic therapy, Devices and physical therapies roadmap: heat and light → electric fields and focused sound → drug-releasing implants.