The Number That Should Keep Healthcare Up at Night
Three hundred and sixty-two. That is the number of FDA drug labels that currently require or recommend pharmacogenomic testing before a medication is prescribed. These are not experimental guidelines or academic recommendations. They are federal regulatory mandates printed on the labels of drugs prescribed millions of times per day across the United States.
Clopidogrel, the blood thinner prescribed to nearly every cardiac stent patient, carries an FDA label requiring CYP2C19 genotyping. Codeine, one of the most commonly prescribed opioids, requires CYP2D6 testing. Fluorouracil, a cornerstone chemotherapy agent, requires DPYD testing because patients with certain variants face fatal toxicity. Simvastatin, prescribed to millions for cholesterol management, carries SLCO1B1 guidance because certain genotypes increase myopathy risk up to 17-fold.
The FDA has done its job. The science is settled. The labels are printed. And in the vast majority of cases, nobody checks.
The Scale of the Disconnect
Over 40 million Americans have already undergone consumer genetic testing through companies like 23andMe and Ancestry. That number grows by millions each year. These consumers have pharmacogenomic data sitting in their accounts right now — CYP2C19 status, CYP2D6 metabolizer phenotypes, SLCO1B1 variants, DPYD deficiency markers — data that directly maps to FDA drug label requirements.
And yet, when these same individuals walk into a clinic, an emergency room, or a pharmacy, that data does not follow them. The prescriber has no visibility into their pharmacogenomic profile. The pharmacist fills the prescription without genomic context. The patient receives a medication that may be ineffective, may cause a preventable adverse reaction, or may be explicitly contraindicated by their genome.
This is not a future problem. This is happening today, in every hospital, on every floor, with every prescription written without pharmacogenomic context. The data exists. The guidelines exist. The infrastructure to connect them at the point of care does not.
What the Gap Actually Costs
Adverse drug reactions are the fourth leading cause of death in the United States, responsible for an estimated 100,000 deaths and over 2 million hospitalizations annually. A significant and growing body of evidence demonstrates that a substantial portion of these events are pharmacogenomically predictable. They are not random. They are not idiosyncratic. They are the direct consequence of prescribing a drug to a patient whose genome metabolizes it differently than assumed.
Consider the clinical reality. A patient with a cardiac stent is prescribed clopidogrel. They are a CYP2C19 poor metabolizer. The drug fails to achieve therapeutic effect. The stent thromboses. The patient returns to the emergency department with a myocardial infarction. The total cost of that re-admission — catheterization, ICU stay, extended hospitalization, rehabilitation — exceeds $150,000. The pharmacogenomic test that would have identified the risk costs less than $200.
Or consider a patient prescribed fluorouracil for colorectal cancer. They carry a DPYD variant that impairs dihydropyrimidine dehydrogenase activity. Without testing, they receive a standard dose. The result is severe, potentially fatal mucositis and myelosuppression. The FDA label says to test. The oncologist may not have a system that surfaces the requirement at the point of prescribing. The test is never ordered. The patient is harmed.
These are not edge cases. DPYD deficiency affects roughly 3-5% of the population. CYP2C19 poor metabolizer status affects 2-15% depending on ancestry. CYP2D6 poor or ultrarapid metabolizer phenotypes affect up to 10% of certain populations. These are common variants with well-characterized clinical consequences and FDA-mandated testing requirements that are routinely ignored.
Why the Test Never Gets Ordered
The failure is not one of awareness. Most physicians understand that pharmacogenomics is clinically relevant. The failure is infrastructural. It is a systems problem with three distinct layers.
The Data Layer
Genomic data exists in silos. Consumer DNA data sits in 23andMe and Ancestry accounts. Clinical whole-genome sequencing results sit in laboratory information systems. Neither is integrated into the electronic health record in a format that a prescribing physician can act on in real time. The EHR knows the patient's blood pressure, their medication list, their allergy history. It does not know their CYP2C19 status.
The Translation Layer
Even when genomic data is available, it arrives in formats that clinical systems cannot interpret. Variant data is expressed in rsIDs, HGVS notation, star alleles, and chromosomal coordinates. Clinical systems speak LOINC, RxNorm, and ICD-10. Without a nomenclature translation layer that maps between these systems in real time, the data is present but unintelligible at the point of care.
The Decision Layer
The final gap is clinical decision support. Even if the data were integrated and translated, the prescriber needs a system that surfaces the relevant pharmacogenomic finding at the moment of prescribing — not buried in a tab, not available on request, but presented as part of the prescribing workflow. The finding must be deterministic, traceable to peer-reviewed evidence, and actionable within the clinical context. Today, that system does not exist at scale in any health system in the United States.
The Ancestry Dimension
The pharmacogenomic gap is not distributed equally. Variant frequencies differ dramatically across ancestries. CYP2C19 poor metabolizer prevalence ranges from approximately 2% in European populations to over 15% in certain East Asian and Oceanian populations. CYP2D6 ultrarapid metabolizer prevalence is significantly higher in East African and Middle Eastern populations. NUDT15 variants — which cause life-threatening myelosuppression on thiopurine drugs — affect approximately 20-25% of East Asian populations compared to less than 1% of European populations.
A prescribing system without ancestry-calibrated pharmacogenomic context is not only incomplete. It is inequitable. It systematically underserves the populations most likely to carry high-consequence variants because the one-size-fits-all approach to prescribing was built on clinical trials that disproportionately enrolled European-descent participants.
Pharmacogenomic decision support that incorporates population-specific variant frequencies is not a feature. It is a health equity requirement.
What Closing the Gap Requires
The solution is not another test. The tests exist. The solution is not another guideline. The guidelines exist. The FDA labels are printed. The Clinical Pharmacogenetics Implementation Consortium has published dosing guidelines for over 20 gene-drug pairs with Level 1A evidence.
What is missing is infrastructure. Specifically, three capabilities that do not exist at scale today:
First: a consumer-to-clinical bridge that takes existing DNA data from the 40 million Americans who already have it and makes it available, with patient consent and cryptographic identity protection, in a format that clinical systems can ingest.
Second: a cross-nomenclature translation layer that maps between the hundreds of thousands of variant identifiers used across genomic databases, clinical coding systems, and drug label specifications — in real time, at the point of care.
Third: a clinical decision support engine that resolves a patient's pharmacogenomic profile against the full landscape of FDA drug label requirements, CPIC guidelines, and clinical rules in seconds — not days — and surfaces actionable findings within the prescribing workflow.
These are engineering problems. They are solvable. They require infrastructure, not invention. The science is settled. The regulatory framework is in place. The data exists. The gap is the bridge between them.
The Human Cost of Waiting
Every day that passes without pharmacogenomic infrastructure at the point of care, patients are harmed by medications that their own genome could have flagged as dangerous. Not hypothetically. Not theoretically. Measurably, documentably, and preventably.
A newborn with spinal muscular atrophy has a therapeutic window measured in weeks. Zolgensma, at $2.1 million per dose, is most effective before three months of age. Without newborn genomic screening, the diagnosis comes after symptom onset — often too late for maximum therapeutic benefit. The drug exists. The screening infrastructure to identify the infant who needs it in time does not.
A patient with iron deficiency and compensatory polycythemia receives two years of therapeutic phlebotomy — a treatment that depletes the very resource their body is struggling to maintain. A pharmacogenomic analysis of their variant landscape would have identified the contraindication before the first procedure. Instead, the absence of genomic context turned a treatable condition into an iatrogenic spiral.
These are not anecdotes. These are the predictable consequences of a healthcare system that sequences genomes, publishes guidelines, approves targeted therapies, prints FDA labels requiring genetic testing — and then prescribes without checking.
362 Labels. One Gap. Zero Excuses.
The question has never been whether pharmacogenomic data matters. The FDA settled that question when it began requiring genetic testing on drug labels. The question has never been whether the data exists. Forty million Americans have already been genotyped. The question has never been whether the guidelines are ready. CPIC has published Level 1A evidence for the highest-consequence gene-drug interactions.
The only remaining question is infrastructure. Who builds the bridge between the genome and the point of care? Who connects the data that exists to the decisions that matter? Who closes the gap?
The science is decades ahead of the infrastructure. That gap is where patients fall through.
NomosLogic Inc. — Sovereign Clinical Genomics Infrastructure — Salt Lake City, Utah
© 2026 NomosLogic Inc. All rights reserved.
