Not every variant is a risk. What matters is what follows from it.

A finding matters when it triggers a consequence. By that criterion three groups remain. First, monogenic variants with high penetrance, meaning ones that actually cause the condition in most carriers: BRCA1 and BRCA2 in breast and ovarian cancer, the Lynch genes in bowel cancer, LDLR, APOB and PCSK9 in familial hypercholesterolaemia, HFE in haemochromatosis. These findings are covered by guidelines and change screening intervals or treatment. Second, pharmacogenetics: CYP2C19, DPYD, TPMT and CYP2D6 demonstrably change the dosing of certain drugs. Third, carrier status for recessive conditions, relevant for family planning. Everything else provides context without justifying an action.

A monogenic risk traces back to a variant in a single gene and often has a large effect. A pathogenic BRCA1 variant markedly raises lifetime breast cancer risk, which is why concrete screening recommendations follow from it. A polygenic risk arises from hundreds or thousands of variants each with a tiny effect, combined into a score. The difference is not one of degree but of kind: a monogenic finding can justify a diagnosis, a polygenic score shifts a probability. Polygenic scores are also calibrated to the cohorts in which they were developed, which is why their predictive performance outside European populations is far less well documented.

This is one of the few genetic questions where the honest answer is: it depends what you intend to do with it. ApoE4 raises the risk of Alzheimer’s disease, particularly in homozygous carriers. At the same time there is no approved therapy initiated on the basis of ApoE status, and many carriers never develop the disease. What the finding can change: the priority you give to cardiovascular health, sleep, hearing loss and exercise, meaning precisely the factors that count without genetics too. What it does not deliver: a prognosis. Which is why ApoE belongs in a conversation before the test, not in a report afterwards.

It is the part of genetics with the clearest immediate benefit, because it changes an action rather than a probability. Four examples. CYP2C19 determines whether clopidogrel works for you; a poor metaboliser gains little protection from the standard drug. DPYD variants substantially raise the risk of severe side effects on fluoropyrimidines in chemotherapy, which is why testing is recommended before treatment starts. TPMT concerns thiopurines in autoimmune conditions. CYP2D6 affects numerous antidepressants and tamoxifen. The finding holds for life, making it one of the few genetic values you only need to establish once.

Considerably less robust than the marketing suggests. These variants are common, their effect sizes are small, and the leap from a gene variant to a supplement recommendation is not supported by the studies. The MTHFR C677T variant, for instance, has a measurable influence on homocysteine levels in some carriers, but the clinical consequence is disputed and a raised homocysteine value can be measured directly anyway. That is precisely the point: where a laboratory value answers the question directly, the detour via genetics is the worse route. We therefore classify these findings as research-adjacent and derive no recommendations from them.

The finding is discussed in a personal medical consultation, never by portal or PDF alone. Where a pathogenic variant in a high-risk gene is found, three things follow. First, an adjusted screening schedule oriented to the guidelines of the relevant specialist society, such as shorter MRI intervals for BRCA carriers. Second, coordination of onward care: we refer to human genetics centres or specialist clinics where specialised management is needed. Third, the question of family, because a hereditary finding concerns siblings and children and their testing requires counselling of its own. A right not to know remains in place throughout.