Almost Everyone Carries a Gene Variant That Affects Medication Response

In the 2016 Ji et al. analysis of 1,013 people, 99% carried an actionable pharmacogenetic variant in at least one of five common drug-response genes. Individual genes ranged from 30% (SLCO1B1) to 79% (CYP2D6). That does not mean 99% of that cohort was at immediate risk. It means nearly everyone in that cohort carried a result that only becomes a warning when a matching medication is ordered.

A pharmacogenetic variant is not a diagnosis. It is a conditional: if this gene meets that drug, then change the dose or the drug.

What percentage of people have pharmacogenetic variants?

In the Ji et al. cohort, the percentage was 99% when testing SLCO1B1, CYP2C19, CYP2C9, VKORC1 and CYP2D6 together. For a single gene, the chance of carrying an actionable variant was lower: 30% for SLCO1B1 at the low end and 79% for CYP2D6 at the high end. The other three genes fell between those points. The 99% figure comes from combining five genes, so a person could have a result in one, three, or all of them.

The word actionable is doing the work here. It does not mean the person had a drug response, an allergy, or a disease. It means the test found a gene version that could change a clinical decision for a specific medication.

What 'actionable' actually means

In pharmacogenetics, actionable means there is enough evidence that a particular genotype should change a drug choice, a starting dose, or a monitoring plan. It is not a list of current problems. A CYP2C19 loss-of-function result may be actionable if you are prescribed clopidogrel, because clopidogrel is a prodrug that CYP2C19 activates. If no clopidogrel is ordered, the same result can sit in your file silently for years.

The common mistake is to read a pharmacogenetic report as a list of things wrong with you. It is not that. It is a list of future conditional events.

Why the mechanism matters more than the label

These five genes code for proteins that handle drugs. CYP2D6, CYP2C19 and CYP2C9 are cytochrome P450 enzymes in the liver. They oxidize or activate many drugs. SLCO1B1 is a transporter that moves some statins into liver cells. VKORC1 is the enzyme that warfarin blocks, so variants near it change warfarin sensitivity.

A variant can reduce the amount of working protein or knock it out completely. Some variants create extra copies of the gene instead. Codeine is the clearest example. It is a prodrug. CYP2D6 converts codeine into morphine, which is the molecule that actually relieves pain. A poor metabolizer with two low-activity CYP2D6 copies makes little morphine and may get weak pain relief. An ultrarapid metabolizer with multiple active copies makes morphine too fast, which is why some guidelines tell prescribers to avoid codeine in that group. Clopidogrel is also a prodrug. CYP2C19 converts it into its active metabolite, so a loss-of-function variant means less active drug, not more.

What you would see in a real report

A pharmacogenomic report usually shows: - Gene and genotype, such as CYP2D6 4/4 or SLCO1B1 c.521T>C - Predicted phenotype, such as poor metabolizer, intermediate metabolizer, normal metabolizer, or ultrarapid metabolizer - A drug-specific note only when a drug is in the report's scope, such as 'simvastatin: increased myopathy risk, consider a lower dose'

It will not list every possible future drug, and it will not say you are sick. Many results will be normal or intermediate. Even a high-risk genotype only becomes clinically relevant when the matching prescription is written.

Does a result change what you can take?

Usually no. Most of your prescriptions will not touch the gene in question. A CYP2C19 variant matters for clopidogrel activation, and some guidelines also use it for certain proton pump inhibitors and antidepressants. A SLCO1B1 variant matters mainly for simvastatin myopathy risk. A CYP2C9 or VKORC1 result matters mostly for warfarin dosing. If you do not take those, the result is dormant.

When a matching drug is ordered, the result may change the starting dose or the drug choice, and sometimes the monitoring. That conversation is worth having with a pharmacist or prescriber, not with a search engine at 2am.

Where the evidence thins out

The 99% number is solid for this study, but it comes from one clinical pilot of 1,013 people. It is not a global population estimate. Variant frequencies differ by genetic ancestry, so a low-activity allele common in one population may be rare in another. Second, actionable depends on evidence and guidelines at the time. Some gene-drug pairs have strong backing; others are still debated. Third, genotype is only one input. Age, weight, liver and kidney function, and other drugs all change how a person responds.

Preemptive pharmacogenetic testing is not yet standard of care everywhere. That is part of why a result can sit in a chart until a prescription triggers an alert.

See it in your own file

If you already have raw DNA from 23andMe, AncestryDNA, MyHeritage or a whole-genome VCF, one option is to check the gene-drug conditionals that matter to your actual prescriptions. DTC raw files cover only a subset of PGx variants, not clinical-grade testing. The free HealthOS DNA tool supports raw DNA files from 23andMe, AncestryDNA, MyHeritage or whole-genome VCF and reports medication response and carrier-relevant findings. This is product information from HealthOS, not from the cited study. Do not treat every result as a current problem. Treat it as a set of if-this-drug-then-that-dose rules.