Diagnostics roadmap: stains → gene panels → blood tests that decide treatment

Overview

1. 1940s-1990shistoric

   Morphology, stains and the first companion test

   Haematoxylin and eosin, then immunohistochemistry and FISH, defined cancer by appearance and a handful of proteins. HER2 testing approved alongside trastuzumab in 1998 created the companion diagnostic: a test whose result is the gate to a drug. Every targeted therapy since has been launched with one.

   Hide era· 5 records

   Histopathology & immunohistochemistryCytogenetics and FISHHER2 IHC and ISH companion assays (HercepTest, PATHWAY 4B5, HER2 Dual ISH)Companion diagnosticsCore needle biopsy and fine-needle aspiration (FNA)
2. 2000s-2010shistoric

   Single-gene tests and gene-expression signatures

   PCR kits for EGFR, KRAS and BRAF matched the first kinase inhibitors to the right patients; the cobas EGFR test became the first blood-based companion diagnostic. Gene-expression signatures did the opposite job, identifying who could safely skip treatment: TAILORx (2018) and RxPONDER showed that most women with early hormone-positive breast cancer and a low Oncotype DX score gain nothing from chemotherapy.

   Show era· 7 records
3. 2017-2022current

   Comprehensive profiling and tumour-agnostic biomarkers

   Sequencing hundreds of genes at once (FoundationOne CDx, TruSight Oncology, Tempus xT) replaced serial single-gene tests, and the biomarker began to matter more than the organ: pembrolizumab for any mismatch-repair-deficient tumour and larotrectinib for any NTRK fusion made the test the indication. Guardant360 CDx did the same from blood. Variant knowledgebases and molecular tumour boards turned raw variants into decisions.

   Show era· 11 records
4. 2022-2026current

   Blood tests that change treatment in randomised trials

   Molecular residual disease testing crossed from prognosis to action. DYNAMIC (2022) halved adjuvant chemotherapy in stage II colon cancer with no loss of recurrence-free survival; CIRCULATE-Japan runs the same question at national scale; SERENA-6 switched endocrine therapy when an ESR1 mutation appeared in blood before a scan showed progression; and IMvigor011 delivered the first ctDNA-guided approval, in bladder cancer, in 2026. Tumour-informed (Signatera, RaDaR) and tumour-naive (Guardant Reveal) assays now compete on sensitivity and turnaround.

   Show era· 10 records
5. 2024-2028emerging

   Slides become data

   Whole-slide scanning made the microscope image computable, and foundation models trained on millions of slides now predict biomarkers, recurrence risk and treatment benefit from a routine stain. ArteraAI Prostate (2025) was the first AI test cleared to predict benefit from a therapy; ArteraAI Breast followed in 2026. MASAI showed in a randomised screening trial that AI reading finds more cancers with less radiologist workload. The open question is prospective proof that AI-derived biomarkers should change treatment, and a regulatory route for models that keep learning.

   Show era· 9 records
6. 2027-2032emerging

   Spatial, single-cell and protein layers guide the choice

   Genotype explains which drug could work; architecture and phenotype may explain which one will. Spatial and single-cell profiling map where immune cells sit relative to tumour cells, proteomics measures the drug's actual target, long-read sequencing resolves rearrangements and methylation together, and near-continuous ctDNA sampling turns monitoring into a running signal. Each is a research tool today; the work of this era is showing that any of them changes an outcome when used to choose treatment.

   Show era· 8 records
7. What sets the pacecurrent

   Validation, standardisation and payment

   Tests that decide who gets a drug are still often validated retrospectively, run differently in different laboratories, and paid for only when they add treatment rather than remove it. The fixes on the table: pre-registration of biomarker studies, a national platform every ctDNA-positive patient can join, coverage-with-evidence for residual disease tests, universal sequencing that feeds a shared learning system, and a clear regulatory status for laboratory-developed tests.

   Hide era· 10 records

   Biomarkers are not validated or standardisedDormant cells and minimal residual diseaseData silosTumour heterogeneity and clonal evolutionPre-register biomarker validation studies the way trials are registeredA national platform trial that every ctDNA-positive patient can joinPay for residual disease tests only inside a trial or registryUniversal tumour and germline sequencing at diagnosis feeding a shared learning systemFDA laboratory-developed test (LDT) ruleReference laboratories and companion-diagnostic testing

Every era's records, trial outcomes and papers, and every watch item, as JSON.