Genomic medicine and screening

Picture a family who spend most of a decade moving between specialists. Their daughter has seizures, developmental delay and a feeding problem nobody can join up; three working diagnoses each prove wrong. Then one blood test sequences her genome and names the condition in a matter of weeks. Clinicians call those years the diagnostic odyssey.

The Human Genome Project took thirteen years and roughly three billion dollars, and finished in 2003. Sequencing a genome now costs a few hundred pounds and takes days. Once a test is cheap enough to run on a population, the hard questions stop being technical: a genetic result is about a patient and simultaneously about relatives who never entered the clinic, it can predict without being treatable, and it can be handed to someone who would rather not have it.

What changed, and where the UK sits

The 100,000 Genomes Project, run by Genomics England from 2013 and completed in December 2018, sequenced whole genomes from NHS patients with rare disease or cancer alongside their families. In the published pilot, around a quarter of rare disease participants received a diagnosis they had not had before, and for a minority of those it changed management.

In 2018 NHS England established the NHS Genomic Medicine Service, turning that research into infrastructure: Genomic Laboratory Hubs across England, and a National Genomic Test Directory setting out which test is commissioned for which indication. England became one of the first national health systems to offer whole genome sequencing as ordinary care. Scotland, Wales and Northern Ireland run their own genomic services.

Two caveats keep this honest. Interpretation, not sequencing, is now the bottleneck — clinical scientists, genetic counsellors and reference data, increasingly assisted by algorithms, which imports every question raised in AI in medicine. Those datasets also skew heavily towards European ancestry, so variants in patients of African or South Asian ancestry more often return as uncertain: the pattern described in health inequalities.

The four things genomics is actually used for

Ending the diagnostic odyssey

Rare diseases are individually uncommon and collectively are not — UK policy documents commonly put the lifetime figure at around one person in seventeen — and most have a genetic cause. The point candidates miss is that a diagnosis is valuable even when untreatable: it ends further investigation, replaces a void with a prognosis, and gives the family accurate information for future pregnancies.

Cancer genomics and targeted treatment

Sequencing a tumour identifies the mutation driving it, and where a drug exists against that target, treatment is chosen by the mutation rather than the organ: HER2 in breast cancer, EGFR in lung cancer, BRAF in melanoma, BRCA1 and BRCA2 predicting response to PARP inhibitors. Germline testing answers a different question — whether relatives should be offered cascade testing. Many tumours have no actionable target, resistance emerges, and the drugs must clear NICE appraisal inside a fixed budget.

Pharmacogenomics

The least glamorous application, already routine in the NHS, and the one to lead with because almost nobody does. Some adverse drug reactions are predictable from a patient’s genotype before the first dose.

• DPYD before fluoropyrimidine chemotherapy. Reduced enzyme activity means capecitabine and 5-fluorouracil are not cleared normally, risking severe or fatal toxicity. NHS England introduced testing before treatment in 2020.
• TPMT before azathioprine. Low enzyme activity causes profound bone marrow suppression at standard doses, and testing has been routine for years — which shows pharmacogenomics is not new.
• HLA-B*57:01 before abacavir. Carriers risk a serious hypersensitivity reaction, so the test decides whether the drug is used at all.
• CYP2C19 and clopidogrel. Poor metabolisers activate less of the prodrug, and NICE diagnostics guidance current at the time of writing supports genotype testing to guide antiplatelet choice after ischaemic stroke or transient ischaemic attack.

The direction of travel is away from one test per drug and towards a pharmacogenomic profile in the record, consulted at every prescription — the strongest answer when an interviewer asks what genomics changes about your working life.

Prenatal screening

Non-invasive prenatal testing analyses placental DNA circulating in the mother’s blood from around ten weeks, offered in England as a second-line test after a higher-chance result for trisomy 21, 18 or 13. It is screening, not diagnosis: confirmation still needs chorionic villus sampling or amniocentesis, which carry a small but real risk of miscarriage.

The counter-argument comes largely from disability rights organisations and is not about accuracy. A safe, routinely offered screen becomes something a woman is expected to accept rather than choose, and a programme aimed at detecting a condition carries a message about which lives are anticipated. The Nuffield Council on Bioethics has examined this, and both halves are strong.

Newborn screening, and the debate that is genuinely live

The UK already screens every newborn. The blood spot test, from a heel prick at around day five, covers nine conditions in England at the time of writing, including sickle cell disease and cystic fibrosis. They were chosen not for being genetic but because each is serious, the test is accurate, and treatment before symptoms works better than treatment after.

The Generation Study, run by Genomics England with NHS England and recruiting from 2024, is exploring whether to go further: whole genomes from 100,000 newborns, with parental consent, looking for around 200 rare conditions that present in early childhood and have an effective intervention. Be precise about its status — it is research into whether this should be done, not an adopted programme, and screening policy follows the advice of the UK National Screening Committee.

The ethics, which is where the marks are

Incidental findings and the right not to know

Sequence a genome for epilepsy and you may also find a cancer predisposition or a neurodegenerative disease that will present in middle age. The usual threshold for reporting is actionability: would knowing change management. That is defensible and not clean — actionability moves as treatments appear, a finding useless to the patient may matter greatly to a sibling, and a threshold set by clinicians is still a decision taken on someone else’s behalf.

Against it sits the right not to know. Autonomy protects refusing information as much as receiving it, and a patient who does not want to spend thirty years waiting for a disease has made a coherent choice. The usual resolution is procedural: decide during consent what will be reported, and make the opt-out real. That is a partial fix rather than a clean one, because nobody consenting to a test can picture every finding it might return.

A result about one person is a result about a family

This is the idea to carry out of the article. Everything in confidentiality, and when it breaks assumes information belongs to the person it is about. A pathogenic variant is simultaneously a partial result for siblings, parents and children, none of whom consented and some of whom may not want to know.

The leading UK case is ABC v St George’s Healthcare NHS Trust. A man diagnosed with Huntington’s disease refused permission to tell his pregnant adult daughter; she learned later, tested positive, and sued. The Court of Appeal allowed the claim to proceed in 2017, and at trial in 2020 the High Court accepted that clinicians owed her a duty to weigh her interest in knowing against her father’s confidentiality, but found on the facts that it had not been breached. The rule is not that you must tell relatives, but that you must genuinely balance and be able to show it.

UK practice handles most of this at the front end, by explaining before testing that results may matter to relatives, so sharing is the expectation rather than a later confrontation. Where a patient refuses, you explore why, offer to help them tell the family or to contact relatives without naming them, and treat disclosure without consent as a last resort.

Insurance, employment and genetic discrimination

In the UK this runs through the Code on Genetic Testing and Insurance, agreed between the government and the Association of British Insurers and in force since 2018. Insurers must not require or pressure an applicant to take a predictive genetic test, and will not use the result of one, with a single exception at the time of writing: Huntington’s disease, for life cover above £500,000.

Know the limits. The Code is voluntary rather than statutory, it does not stop insurers asking about family history or diagnosed conditions, and it says nothing about employment, where the UK has no dedicated statute and an asymptomatic person with a predisposition sits awkwardly inside the Equality Act 2010 definition of disability. Fear of these effects deters people from tests that would benefit them: a harm from policy rather than biology.

Predicting a disease nobody can prevent

Huntington’s disease is the canonical case: autosomal dominant, caused by an expanded CAG repeat in the HTT gene, so each child of an affected parent has a 50 per cent chance of inheriting it. Onset is usually in adult life, the test is close to definitive, and at the time of writing no licensed treatment alters the course, though disease-modifying therapies are in trials — so a predictive test still largely delivers certainty and nothing else. Some people want exactly that; the consistent international finding is that only a minority of at-risk adults, commonly reported at well under a quarter, choose to be tested. That is autonomy exercised in the other direction, and testing runs through counselling across several appointments.

Testing children is the nearest thing here to settled practice. UK professional guidance works from a strong presumption that asymptomatic minors are not tested for adult-onset conditions when nothing would change in childhood: there is no benefit during the years the parents are deciding, and it removes a choice the child would otherwise make as an adult. It is a presumption rather than an absolute bar — requests are considered case by case after counselling, and there is a real argument that a blanket rule underweights parental responsibility and the strain of prolonged uncertainty in a family. Where a result does change childhood care — familial adenomatous polyposis, with bowel surveillance from adolescence — testing a child is expected.

Consumer tests, and data at population scale

Consumer tests create two problems. Technically, many genotype a selected panel rather than sequencing a gene, so a reassuring cancer-risk result from a chip reading a handful of variants is not a negative test though it reads like one, and findings flagged in raw data often fail to confirm in an accredited laboratory. Clinically, nobody is holding the patient: ancestry results can reveal misattributed parentage with no counselling attached, and the NHS absorbs the follow-up. Argue the other side too, because it is not weak — people are entitled to information about their own bodies without a gatekeeper, some tests do surface a variant the NHS would never have looked for, and the paternalism of saying otherwise is exactly what medicine has spent decades unlearning. The position most people land on is not prohibition but accredited confirmation before anything is acted on.

A genome is also the strongest identifier a person has: it cannot be anonymised the way a name is redacted, it does not expire, and it partially identifies relatives who were never asked. Genomics England keeps researchers working inside a controlled environment rather than sending copies out. Trust is slow to rebuild — the care.data programme was abandoned in 2016 after opposition to how records would be shared.

Running the four pillars through it

Genomics rewards practice because the pillars from the four pillars of medical ethics do real work here rather than being recited over the top of an answer.

• Autonomy cuts several ways. It includes the right not to know, so consent must specify what will be reported; parents exercise it for a newborn who cannot object; and relatives who share the result never consented at all.
• Beneficence is concrete: a diagnosis after nine years without one, a drug chosen because the tumour carries its target, a chemotherapy dose corrected before it causes harm.
• Non-maleficence covers harms that are informational rather than physical: anxiety from an uncertain variant, investigation of a well person, a prediction nobody can act on, insurance consequences, family relationships altered by a result nobody asked for.
• Justice covers uneven access, reference datasets skewed towards European ancestry, the opportunity cost of genomic spending inside a fixed budget, and who captures the value from a population’s data.

Notice the pattern: almost every problem here comes from information rather than intervention. Genomics reads the genome and tells you things, and that alone strains confidentiality, unsettles screening policy and reaches into insurance law. The next stop on the path, CRISPR and gene editing, is about what happens when the technology stops reading and starts rewriting — at which point the consent problem you have just met, about a person who cannot yet object, becomes considerably sharper.