NIH All of Us study finds pharmacogene variants in every participant
A large NIH analysis of nearly 100,000 people in the All of Us Research Program found a pharmacogenetic signal in every participant. Each person carried at least one variant in a set of 16 pharmacogenes, and about 1 in 4 had already taken a medication that could be affected by those variants. The result does not mean every prescription needs a genetic test. It does mean the inherited biology shaping drug handling is far more common, and far more clinically relevant, than many routine prescribing workflows currently reflect.
The study expanded the view beyond older genetic reference sets by drawing on a more diverse ancestry mix than databases such as the U.K. Biobank. That matters because drug-response genetics is not evenly distributed across populations, and a narrow dataset can miss variants that are clinically important in real-world practice. In this cohort, investigators identified thousands of new variants in genes that influence response to SSRIs, statins, chemotherapy and clopidogrel. The scope of the finding is striking: pharmacogene variation was not an exception tucked into a small subgroup. It was universal.
Universal variation, uneven clinical use
The central message is not that genetics overrides clinical judgment. It is that pharmacogenes are present in nearly everyone, while pharmacogenomic decision support remains unevenly available. The study makes that gap hard to ignore. When a patient carries a variant in a drug-metabolizing or drug-target gene, the consequence can range from reduced efficacy to toxicity, depending on the medication and the direction of the effect. A normal-looking prescription can behave very differently in two patients because the liver, receptor, or transporter pathways underneath it are not the same.
For clinicians, the most practical implication sits at the point of prescribing. SSRIs may require closer attention when metabolism differs. Statins can be affected when transporter or metabolism variants alter exposure. Clopidogrel remains a classic example of a prodrug whose activation can be blunted by certain variants, potentially leaving patients underprotected against thrombosis. In oncology, inherited variation can change how some patients handle chemotherapy, raising the stakes for dose selection and toxicity monitoring. The biology is familiar, but the breadth of the All of Us data shows how often these pathways may be relevant across a general population rather than an academic specialty clinic.
One participant in four had already been prescribed a medication touched by the variants identified in the study.
Why ancestry breadth changed the variant count
Older pharmacogenetic datasets have often leaned heavily on less diverse populations, which creates blind spots. By contrast, the All of Us program includes a broader mix of ancestries, and that diversity surfaced variants earlier databases missed. The result reinforces a simple principle of human genetics: if the sample is narrow, the map is incomplete. Drug-response variants do not disappear in underrepresented groups; they become harder to see.
This has direct implications for how precision medicine is operationalized. A pharmacogene panel built from a limited reference population may label a variant as rare or miss it altogether, even when it is common in another ancestry group. The consequence is not just a database problem. It can become a bedside problem when a dose is selected without recognizing altered metabolism, altered activation, or altered target response. In those settings, treatment failure may be misread as disease resistance, adherence problems, or unexplained idiosyncrasy.
The study also pushes back against the idea that genetics is the sole explanation for a poor drug response. Senior author Philip Empey cautioned against that oversimplification. Gene variants are one part of the response equation, alongside age, organ function, comorbid disease, polypharmacy, adherence, and environmental factors. A patient can carry a clinically relevant variant and still respond normally; another can have no known variant and still experience toxicity or lack of effect. The point is not determinism. It is risk stratification.
What this means for routine prescribing
The current barrier is not just sequencing. It is implementation. Many patients have never had their genomes sequenced, and many clinicians do not have an easy way to translate a variant into an actionable prescribing choice. That creates a familiar tension in modern medicine: biologic knowledge is advancing faster than workflow. Even when the evidence is strong enough to influence care, it can remain outside the electronic health record, outside the pharmacy interface, and outside the clinician’s usual decision pathway.
In practice, pharmacogenomic use is most helpful when a drug has a narrow therapeutic window, a well-established gene-drug relationship, or a serious downside to getting the first dose wrong. The study’s list of implicated medication classes includes several examples where early precision can matter. Clopidogrel is one, because underactivation can translate into inadequate platelet inhibition. Certain chemotherapy regimens are another, because toxicity can be severe and dose-limiting. SSRIs and statins are common enough that even modest improvements in matching drug to patient could have broad effects across primary care, psychiatry and cardiometabolic medicine.
The study does not solve the clinical workflow problem. It sharpens it. Wider variant detection only becomes useful when results can be interpreted quickly, integrated into prescribing systems and explained in a way that changes what happens in the room, not just in the lab report.
Precision medicine still depends on interpretation
A pharmacogene variant is not a diagnosis, and it is not a prescription by itself. It is a piece of evidence that can explain why one patient needs a different dose, a different agent or closer monitoring. The All of Us findings strengthen the case for broader pharmacogenomic literacy, particularly in areas where adverse effects or nonresponse have outsized consequences. They also suggest that the future of medication safety may depend less on finding rare genetic outliers and more on recognizing that variation is the rule.
The unresolved challenge is operational: who gets tested, when results are returned, and how often they are available before the first prescription is written. Until those questions are answered in routine care, the clinical value of universal variant carriage will remain partly theoretical, even as the biology becomes harder to dispute.