Industries / Health / Healthcare / Medical labs / Genetic testing and diagnostics
Building and approving tests
Designing a genetic or other marker test, showing that it measures what it claims and that its result means something for patients, and getting it approved to sell or use.
7
Milestones achieved, from 1970s to 2023.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1970s
Couples tested before pregnancy
Tay-Sachs disease is an inherited nerve disorder that kills in early childhood, and a child is at risk only when both parents carry a faulty copy of the gene. From 1970, voluntary programs offered young adults a blood enzyme test for carrier status, with counseling and, for couples who both carried it, testing during pregnancy. By 2000 more than 1.4 million people had been screened and more than 1,400 at-risk couples found, and M. M. Kaback reported that the disease's incidence among Jewish births in the United States and Canada had fallen by more than 90%. The enzyme test gave false positives in carriers of two harmless "pseudo-deficiency" variants, so DNA tests were added.
LessonA screening program for healthy adults needs education and counseling before and after the test, and a second method for the results the first one gets wrong.
Credit
M. M. Kaback (University of California, San Diego) and the International TSD Data Collection Network
1997
Fetal DNA in mother's blood
Y. M. Lo and colleagues at the John Radcliffe Hospital, University of Oxford, looked in the blood of 43 pregnant women for a Y-chromosome DNA sequence, which only a male fetus could supply. They found it in the plasma of 24 of the 30 women carrying boys (80%), using only 10 microliters of plasma, and in none of the 13 carrying girls or the 10 women who were not pregnant. The authors noted that the finding could allow prenatal testing from a blood draw instead of a needle into the womb, the basis of today's cell-free DNA screening.
Credit
Y. M. Lo, J. S. Wainscoat and colleagues, University of Oxford
2013
Gene-guided warfarin falls short
Warfarin is a blood thinner whose right dose varies widely, and setting the first doses by a patient's genes was meant to reach it sooner. In the COAG trial, 1,015 patients had their first five days of doses set by an algorithm that used clinical factors plus genotype, or clinical factors alone. At four weeks the share of time spent in the target clotting range was 45.2% with genotype and 45.4% without, and among Black patients it was lower in the genotype group.
LessonA gene-based rule has to beat a good rule without genes, in every group of patients who will use it.
Credit
S. E. Kimmel and the COAG investigators; funded by the US National Heart, Lung, and Blood Institute
2016
Blood picks the lung drug
A companion diagnostic is a test that decides who may receive a particular drug. On 1 June 2016 the FDA approved Roche's cobas EGFR Mutation Test v2 to find EGFR mutations in tumor DNA circulating in plasma and pick lung cancer patients for erlotinib, and on 28 September 2016 added the T790M mutation for osimertinib. The label tells labs to send patients whose blood result is negative for a tissue biopsy. By 2026 the FDA's list of companion tests included many plasma tests, among them Guardant360 CDx and FoundationOne Liquid CDx.
LessonA blood test that finds a tumor mutation can spare a biopsy, but one that misses it cannot rule the mutation out.
Credit
Roche Molecular Systems; US FDA
2022
Europe phases in test rules
The EU's In Vitro Diagnostic Regulation (IVDR), Regulation (EU) 2017/746, became applicable on 26 May 2022, replacing a 1998 directive. Under the directive about 8% of tests on the market, the highest-risk ones, needed review by a notified body, an independent assessor; under the regulation about 80% do, most for the first time. In October 2021 the European Commission found that member states, health institutions, notified bodies and makers would not be ready, and proposed deadlines staggered by risk class, from 2025 for the highest-risk tests to 2027 for lower-risk ones; a further extension followed over the shortage of notified bodies.
LessonA rule that sends ten times as many tests to outside reviewers needs those reviewers in place before it starts.
oeil.secure.europarl.europa.euhealth.ec.europa.eupmc.ncbi.nlm.nih.gov
Credit
European Commission, European Parliament and Council of the EU
2023
A 12-gene panel cuts reactions
In the PREPARE study, 6,944 patients starting a common drug at 18 hospitals, 9 community health centers and 28 community pharmacies in seven European countries were genotyped for 50 variants in 12 genes. In the study group doctors adjusted drugs and doses by the Dutch Pharmacogenetics Working Group's guidelines; the control group received standard care. Among patients with an actionable result for their drug, 21.0% had a clinically relevant adverse drug reaction within 12 weeks, against 27.7% with standard care (odds ratio 0.70). The study was open-label, so patients and doctors knew which group they were in.
Credit
J. J. Swen, H.-J. Guchelaar and the Ubiquitous Pharmacogenomics Consortium, Leiden University Medical Centre; funded by the EU's Horizon 2020
2023
Rules for inherited cancer panels
On 29 September 2023 the FDA granted De Novo authorization, its route for new kinds of low- to moderate-risk devices, to the Invitae Common Hereditary Cancers Panel, which reads 47 genes from a blood sample for inherited cancer risk. The FDA said it was the first test of its kind to receive marketing authorization; Invitae had tested more than 9,000 clinical samples. With it the FDA set special controls for all such tests, among them at least 99.0% positive agreement and 99.9% negative agreement with a validated second method for single-letter changes, small insertions and deletions, and copy number changes.
Credit
Invitae; US FDA
Watch for
Gene-guided prescribing by family doctors
NHS England's PROGRESS pilot tests patients who start a new medicine in primary care, mostly statins or SSRI antidepressants, for variants in four genes (CYP2C19, CYP2D6, SLCO1B1 and CYP2C9). In a talk on 2 October 2026, NHS England said the pilot had enrolled more than 1,325 patients at 21 sites; 91% carried a clinically relevant variant in one of the four genes, 9.8% received a "red flag" advising a switch of medicine and 30.5% an "orange flag" on dose, clinicians acted on 90% of the advice, and one patient in four had a medicine changed.
Achieved whenthe full results, including whether side effects or failed treatments fell, appear in a peer-reviewed journal or an outside evaluation.
Claimed by
NHS England, October 2026
Open
01
Clinical validity, as registered
A public table, rebuilt each month from the full data file of the NIH Genetic Testing Registry (GTR), of what every registered clinical test says about its clinical validity, meaning how well a result predicts the condition. For each test it records whether the section is filled in, whether it gives a number (detection rate, sensitivity, positive predictive value), whether it cites a study, and whether the number matches the cited study. Results are grouped by lab and by test type: inherited cancer, carrier, prenatal, pharmacogenomic and others.
The briefCredit · unclaimed02
A census of in-house tests
A public register of the declarations that EU hospitals and hospital labs must publish for the tests they make and use themselves (in-house tests) under Article 5(5) of the IVDR, starting with genetic and molecular tests. Each row gives the institution, the test, its intended use, the date and whether the declaration lists requirements not fully met. It is rebuilt each year.
The briefCredit · unclaimed
Open 01 · the brief
Clinical validity, as registered
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table, rebuilt each month from the full data file of the NIH Genetic Testing Registry (GTR), of what every registered clinical test says about its clinical validity, meaning how well a result predicts the condition. For each test it records whether the section is filled in, whether it gives a number (detection rate, sensitivity, positive predictive value), whether it cites a study, and whether the number matches the cited study. Results are grouped by lab and by test type: inherited cancer, carrier, prenatal, pharmacogenomic and others.
Why it is still open
The US lab law, CLIA, checks a lab's analytical validity, meaning whether the test measures the DNA correctly, and "does not address the clinical validity of any test", in the words of CMS, which runs it, and most lab-developed tests reach patients without FDA review (the Medical labs file covers the 2025 court ruling that struck down the FDA's rule). GTR makes analytical validity a required field and clinical validity only a recommended one, and a 2023 study that counted 129,624 US test versions in GTR did not measure how often the fields are filled. A count for this file of GTR's full XML of 18 November 2026 found 63,241 clinical test records: every one has an analytical validity section, and 3,960 (6.3%) have a clinical validity section. Searches found no study of these fields. The Medical labs file's GTR audit checks the genes on each test against expert panels, not what labs claim a result means.
Built with what exists
GTR's full XML file (gtr_ftp.xml.gz, 176 MB) and its monthly archive; Python with lxml; a current AI model, such as Claude or Mistral Large 4 with open weights (see the Note "A trillion open parameters"), to sort each free-text statement into a fixed schema and pull out its number and citation; the Europe PMC API to fetch the cited abstracts for checking.
First step
Sort GTR's 3,960 clinical validity sections by test type and hand-code 50 of them against the studies they cite.
Done looks like
every clinical test in GTR classified, with the table and code published under an open license. Two genetic counselors or lab directors hand-code 200 random records and agree with the table on at least 95% of fields. Each lab sees its rows before publication, and the table reruns monthly for six months with a list of changes.
Open 02 · the brief
A census of in-house tests
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public register of the declarations that EU hospitals and hospital labs must publish for the tests they make and use themselves (in-house tests) under Article 5(5) of the IVDR, starting with genetic and molecular tests. Each row gives the institution, the test, its intended use, the date and whether the declaration lists requirements not fully met. It is rebuilt each year.
Why it is still open
The EU's Medical Device Coordination Group says each health institution must draw up a declaration naming each in-house device and its intended use and make it public, for example on its website, but leaves the format to national rules, so no central list exists. European lab societies (EFLM, 2025) asked the Commission to drop most of Article 5(5)'s conditions to protect specialty and rare-disease tests, and the Commission's December 2025 proposal would loosen them, yet searches found no count of how many in-house tests EU hospitals declare. Single hospitals have counted their own: UZ Leuven, a Belgian university hospital, says 25% of the 2,200 tests it runs are in-house devices, and found in 2020 that 71.9% of them had no CE-marked alternative.
health.ec.europa.eueuropepmc.orgcovingtonblogs.combiomedeurope.org
Built with what exists
web search APIs and an open crawler such as Scrapy; a current AI model to read declarations in the EU's 24 languages and fill a fixed schema; the MDCG 2023-1 model declaration as the schema; the public EUDAMED database to flag in-house tests that have a CE-marked equivalent.
First step
Pick one country, list its university hospital laboratories, and collect by hand the Article 5(5) declarations of ten of them.
Done looks like
declarations from at least 300 health institutions in at least five member states, each row linked to its source page with the date retrieved. A second person checks 100 random rows against the source and finds at least 95% of fields right. Counts of declared genetic tests by country are published with the data.