Pharmacogenetic testing and medication matching: an association is not a prescription
A pharmacogenetic result can identify selected variants associated with drug metabolism, exposure, response, or adverse-event risk. It does not test every relevant variant, rank every medication, or replace diagnosis, current medicines, organ function, monitoring, and clinical response.
Pharmacogenetic testing can identify selected inherited variants associated with how certain drugs are metabolized, transported, tolerated, or—in narrower cases—managed. It does not test every relevant variant, determine whether a medication is indicated, rank all drugs for one person, or replace dose history, other medicines, liver and kidney function, age, diagnosis, monitoring, and clinical response. Never start, stop, or change a medication from a color-coded report without the prescriber who owns the treatment plan.12
The useful output is not a green, yellow, or red bin. It is a traceable chain from an accurately measured genotype to a supported gene–drug association, the current drug label or other clinical evidence, and a patient-specific action considered by a qualified prescriber.
Audit four links, not one dashboard
| Evidence link | Question the report must answer |
|---|---|
| Analytical test | Which genes and variants were measured, by what method, with what call rate and confirmation policy? |
| Genotype to phenotype | How does the laboratory translate variants into metabolizer, transporter, immune-risk, or other functional status? |
| Phenotype to drug | What evidence connects that status to exposure, adverse events, response, or a management recommendation for this exact medication? |
| Drug to patient | How do diagnosis, dose, formulation, other medicines, organ function, ancestry, prior response, and monitoring change the decision? |
If any link is missing, the report may accurately measure DNA while overstating what the result can decide.
FDA authorization follows the exact claim
FDA has authorized direct-to-consumer pharmacogenetic systems for defined variant reports and limitations. In its first such De Novo authorization, FDA said the test provided information about variants associated with medication metabolism to inform a discussion; it did not determine whether a medication was appropriate, predict response to a specific drug, or provide medical advice, and results required independent confirmation before medical decisions.3
FDA’s current direct-to-consumer page says it has not authorized a DTC pharmacogenetic test that predicts whether a person will respond to or have adverse reactions from a specific therapeutic drug.5 A seller should not stretch an authorization for variant reporting into a promise to “find the right antidepressant,” “eliminate trial and error,” or choose an anesthesia plan.
In 2019, FDA warned a laboratory over tests marketed to predict responses to named medications without reviewed support and emphasized the potential harm of changing treatment on those claims.4 That enforcement record is not a verdict on every PGx test. It illustrates why the exact assay and exact claim matter.
The CLIA-versus-FDA guide shows how to verify the laboratory separately from the test’s premarket status.
FDA’s association table has three evidence levels
FDA organizes pharmacogenetic associations into categories that support different statements: some support therapeutic-management recommendations, some indicate a potential impact on safety or response, and some show a potential pharmacokinetic effect only.1 Those columns should not be collapsed into one “actionable” label.
Even inclusion in the table does not mean FDA recommends pre-prescription testing unless the test is a companion diagnostic. FDA notes that most listed associations have not been evaluated for whether testing improves clinical outcomes and directs prescribers to current approved labeling.1
For each medication on a report, ask:
- Is the gene–drug pair in FDA labeling or the FDA association table?
- Does the evidence concern drug concentration, efficacy, toxicity, or a specific management action?
- Does the label require, recommend, mention, or merely describe the biomarker?
- Is the tested variant the one in that evidence?
- Is the proposed change supported for this diagnosis, dose, route, and population?
FDA’s pharmacogenomic-labeling table includes many biomarkers, but the agency states that some labels contain a specific action and others do not.2 Presence in a label is not itself a dosing instruction.
Test coverage is not gene-name coverage
A report can list CYP2D6, CYP2C19, or another familiar gene while testing only selected variants. Different assays may cover different alleles, copy-number changes, structural variants, or rare variants. “Normal” means no reportable finding under that assay’s design; it does not mean the gene or pathway is globally normal.
Ask for the full variant and allele list, version of the translation rules, whether copy number is assessed, which populations informed validation, and how an uncertain or no-call result appears. Some allele frequencies and phenotype-prediction performance differ across ancestral backgrounds. A test should describe those limitations rather than using race as a genetic shortcut.
Keeping the underlying calls matters because interpretation can be updated while germline DNA remains mostly constant. A new dashboard may reclassify a result without a new sample.
Drug interactions can change the functional phenotype
A genetically predicted metabolizer phenotype is one input. Another medication can inhibit or induce an enzyme, changing drug exposure in practice. Dose, formulation, smoking, pregnancy, age, liver or kidney function, adherence, and interacting foods or supplements can also matter. The genotype does not absorb those variables.
This is sometimes called phenoconversion: observed metabolic behavior can differ from the inherited category because of medications or other conditions. A prescriber should review the complete current list, including over-the-counter drugs, supplements, cannabis, hormones, and intermittent medicines—not only the medication the panel highlights.
A historical medication “failure” also needs context: indication, dose, duration, adherence, side effects, concurrent treatment, and reason for stopping. A genotype should not retroactively rewrite every clinical experience.
Panels should not rank unrelated evidence as one score
Commercial panels may combine strong label-directed associations, pharmacokinetic-only signals, limited or conflicting clinical evidence, proprietary combinatorial algorithms, and non-genetic factors. A single traffic-light color can hide those differences.
Ask whether the algorithm is FDA reviewed, peer-reviewed, independently replicated, and validated prospectively for the proposed outcome. If several genes are weighted together, request the model version, output definition, comparison group, and evidence that acting on that composite improves care. “Clinically validated” should identify a study and endpoint.
No test can list every possible medicine as equally studied. A blank drug, gray result, or green category can mean no relevant variant was detected, no association was assessed, ordinary labeling applies, or evidence is insufficient. The legend must say which.
Timing depends on a real medication decision
Testing may be most useful when a specific medication with a supported association is being considered, a relevant adverse event occurred, or a result could change monitoring or dose under an established pathway. Broad preemptive testing can also create a durable record, but only if results are accessible, interpretable, and revisited by future prescribers.
Buying a panel “for wellness” without a named decision can produce information that is hard to act on and easy to overread. Before ordering, name the question and both branches: what will change if the relevant variant is present, and what will remain unchanged if it is absent?
Privacy also belongs in the decision. Germline results can reveal information about relatives and persist for life. Review sample storage, secondary research, recontact, law-enforcement request policy, account deletion, raw-data download, and whether data enter the clinical record.
Use a prescriber-owned action sheet
- Name the medication decision. Identify the drug, diagnosis, current or proposed dose, prior response, and what testing could plausibly change.
- Verify the assay. Record FDA status, laboratory, genes, exact variants, copy-number capability, specimen, no-call rate, and confirmation.
- Grade the association. Separate management recommendations, safety or response associations, and pharmacokinetic-only evidence.
- Open the current drug label. Find whether the biomarker is required, recommended, actionable, descriptive, or absent for this use.
- Add non-genetic inputs. Review all medicines, organ function, age, diagnosis, adherence, prior outcomes, and required clinical monitoring.
- Document—not automate—the decision. The prescriber should record whether and why the result changes selection, dose, monitoring, or nothing.
The best pretest question is: “For which exact medication decision could this exact variant result change the current label-based plan—and who will interpret it with my full medication and clinical record?”
Sources
- U.S. Food and Drug Administration. Table of pharmacogenetic associations. Current gene–drug evidence categories, limitations, companion-diagnostic distinction, label priority, and warning against medication changes without a prescriber. Accessed .
- U.S. Food and Drug Administration. Table of pharmacogenomic biomarkers in drug labeling. Current drug-label inventory and distinction among biomarker information, exposure, response, adverse-event, dosing, and action language. Accessed .
- U.S. Food and Drug Administration. FDA authorizes first direct-to-consumer test for variants associated with medication metabolism. Exact De Novo authorization boundaries, confirmation requirement, and limits on using a DTC result for treatment decisions. Accessed .
- U.S. Food and Drug Administration. FDA warning letter on genetic tests claiming to predict responses to specific medications. Enforcement example distinguishing variant detection from unsupported claims to predict response or select named drugs. Accessed .
- U.S. Food and Drug Administration. Direct-to-consumer tests. Current FDA boundaries for direct-to-consumer pharmacogenetic reports, including the absence of authorization to predict response or adverse reactions to a specific therapeutic drug. Accessed .