LOINC pharmacogenomics codes give your lab a shared vocabulary for what a PGx result is: the gene studied, the genotype, the predicted phenotype and the interpretation. The core of the set is the Pharmacogenomics result panel, LOINC 82118-1, together with gene-specific interpretation codes such as 79714-2 for CYP2C19 and CPIC-aligned answer codes for phenotypes like poor metabolizer. This guide shows which codes matter, how they relate to SNOMED CT and FHIR, and where labs go wrong when they map them.
TL;DR
- LOINC identifies what was observed (the question); the phenotype values (the answers) come from a CPIC-aligned answer list, which is how a receiving system can read a result without parsing free text.
- Build around the Pharmacogenomics result panel (82118-1), then use gene-specific codes such as 79714-2 (CYP2C19) and 79715-9 (CYP2D6).
- LOINC is not a transport protocol. Pair it with HL7 FHIR for exchange and RxNorm for medications, and use SNOMED CT where your interface specification calls for it.
- Keep genotype and therapeutic implication as separate, linked observations, not one merged field.
- Coding the result is only half the work: the interpretation behind it has to stay current as CPIC guidance changes.
Table of Contents
- What LOINC does for a PGx result
- The codes most labs reach for first
- LOINC versus SNOMED CT for PGx
- LOINC is a language, not a protocol
- How the codes sit inside FHIR
- A worked example: CYP2C19 and clopidogrel
- Where implementations go wrong
- Keeping coded results current
- A practical mapping checklist
- Our perspective
- Building this with SignalPGx
- FAQ
- Sources
What LOINC does for a PGx result
LOINC, maintained by the Regenstrief Institute, is a terminology for identifying laboratory tests and observations. In pharmacogenomics its value is practical. A result that arrives as a PDF can be read by a clinician but not by a clinical decision support (CDS) rule. A result that arrives as a coded observation can fire an alert when a prescriber orders a drug that the patient's genotype affects.
The effort to make that possible was formal. Regenstrief and the Clinical Pharmacogenetics Implementation Consortium (CPIC) worked together to create LOINC codes for reporting pharmacogenomic results in a standard format, covering both phenotypes such as poor metabolizer and gene interpretations such as CYP2D6 gene product metabolic activity. The CPIC term standardization project that fed this work is described in a CPIC presentation on LOINC codes.
If you are new to the genotype-to-phenotype step that sits behind these codes, our genotype to phenotype reporting guide covers it from the lab side.
The codes most labs reach for first
Below are the codes that appear in the LOINC and CPIC material we reviewed. Always confirm the current long common name and status on loinc.org before you build, because LOINC terms can change status.
The panel
LOINC 82118-1, Pharmacogenomics result panel, is the structure that bundles genotype, phenotype and therapeutic-implication observations. At the time of the page we reviewed, its status was listed as TRIAL. That does not stop you from using it, but it means you should track the term and plan for revisions rather than hard-code assumptions.
Gene-level and general codes
| Purpose | LOINC code | Notes |
|---|---|---|
| Pharmacogenomics result panel | 82118-1 | Panel that groups PGx observations |
| CYP2C19 phenotype interpretation | 79714-2 | Determined by the reporting lab and returned with structured results |
| CYP2D6 phenotype | 79715-9 | Listed in LOINC's PGx panel material |
| UGT1A1 gene product metabolic activity interpretation | 79718-3 | Same pattern, different gene |
| Gene studied ID | 48018-6 | Identifies the gene tested |
| Overall gene product metabolic activity interpretation | 51971-0 | Overall predicted phenotype for drug metabolism across markers and alleles |
Codes 79715-9, 79718-3 and 48018-6 are drawn from the LOINC panel page for 82118-1.
Phenotype answers
The CPIC-standardized phenotype terms have their own LOINC answer codes, as listed in the LOINC announcement:
- Ultrarapid metabolizer: LA10315-2
- Rapid metabolizer: LA25390-8
- Normal metabolizer: LA25391-6
- Intermediate metabolizer: LA10317-8
- Poor metabolizer: LA9657-3
Note the difference in prefix. Codes beginning LA are answer codes, the values that populate a result. The numeric codes with a check digit, like 79714-2, name the observation itself. Mixing the two up is a common first mapping error.
LOINC versus SNOMED CT for PGx
Labs searching for "LOINC vs SNOMED" usually want to know which one to pick. The usual answer is that they do different jobs. As one health IT explainer on lab reporting puts it, LOINC identifies the test or observation, while SNOMED CT standardizes many non-numeric result values, organisms and specimen descriptions.
Applied to pharmacogenomics:
| Question | LOINC | SNOMED CT |
|---|---|---|
| What is being reported? | Primary role: test and observation identity (for example, the CYP2C19 interpretation) | Not the usual choice for the observation name |
| What is the result value? | Answer lists such as the CPIC phenotype set | Often used for coded qualitative values where an interface calls for it |
| Typical home in a message | Observation code | Value or supporting concept |
The scoping review on health data standards in pharmacogenomics systems argues that standardizing PGx data with these terminologies means a patient's results stay accessible and usable when they change providers, which avoids repeat testing, since a germline PGx result does not change over a lifetime. That argument holds whichever terminology carries the value. What matters is that your lab chooses deliberately, documents the choice, and applies it consistently.
A note of caution: which value set a particular EHR expects is a local decision. Ask the receiving organization's interface team before you finalize your value bindings.
LOINC is a language, not a protocol
LOINC's own guidance on how it works with other standards makes the point that it names the observation but does not move it. For a complete, actionable result you combine:
- LOINC for the observation identity.
- HL7 FHIR for the transport and structure.
- RxNorm for the medication a recommendation refers to.
- SNOMED CT where your profile or customer requires coded values.
The Council on Pharmacy Standards data standards chapter adds a practical caveat: LOINC does not dictate which PGx panel you offer, and codes often need to be mapped into your internal LIS or LIMS, especially when downstream EHRs do not natively support FHIR genomics reporting.
How the codes sit inside FHIR
The HL7 Genomics Reporting Implementation Guide shows how a PGx report links the genotype finding to the clinical interpretation. The model uses Task resources for medication usage impacts and ties them to observations through LOINC-coded interpretations. The result is a traceable chain from the patient's test, through the gene-level interpretation, to a drug-specific recommendation.
Two design rules follow from that guide.
Keep genotype and implication apart. The same HL7 page describes keeping the biological finding and the therapeutic recommendation as separate, linked observations. Labs often merge them into one text field because it is easier to print. It is also much harder to query, and it breaks the moment guidance changes and only the implication needs updating.
Start small. You do not need to code every gene on day one. Our post on mapping one gene-drug pair in FHIR first walks through proving the pattern on a single pair, and the FHIR genomics reporting implementation guide goes deeper on the resource model. For the document side of delivery, see our notes on FHIR DocumentReference for genomics.
A worked example: CYP2C19 and clopidogrel
Take a patient whose CYP2C19 result is a poor metabolizer. A well-structured report would carry:
- A gene-studied observation (48018-6) identifying CYP2C19.
- The CYP2C19 interpretation observation (79714-2) with the answer Poor metabolizer (LA9657-3).
- A linked medication usage impact for clopidogrel, with the drug identified through RxNorm and the recommendation text coming from the current guideline.
- All of it grouped under the Pharmacogenomics result panel (82118-1).
Only items 1, 2 and 4 are LOINC-coded; item 3 is where the interpretation layer lives, and that is the part that must be reviewed by your medical director. For the clinical reasoning, see CYP2C19 and clopidogrel: a clinical lab guide and the clinician actions for poor metabolizers. A parallel build for a different drug class is in building a standards-compliant warfarin PGx report.
Where implementations go wrong
Research on PGx implementation suggests the obstacle is rarely a shortage of codes. A paper in The Pharmacogenomics Journal notes that PGx data standards exist in SNOMED CT and LOINC, yet many labs and EHRs stay with manual report delivery instead of integrating structured results into CDS. In our experience, the failure points cluster like this:
- Answer codes used as observation codes, or the reverse.
- Codes selected by guesswork. A survey of LOINC code selection practices is a reminder that incorrect code choices are common in real lab submissions; check each code against loinc.org rather than copying from another lab's file.
- Ignoring status. A TRIAL term can change; keep a mapping table with version and review date.
- Static interpretation text. The code stays valid while the guideline behind it moves.
- One field for everything, as discussed above.
Keeping coded results current
Coding fixes the shape of a result. It does not keep the content right. CPIC guidelines are living documents, updated as evidence emerges, and a lab maintaining interpretation text by hand risks version drift: reports in the field that quietly disagree with current guidance. The CPIC term standardization material reflects that guidelines and terms evolve together.
The broader discussion of why reanalysis matters is covered in this piece on delivering the benefits of genomic data reanalysis, which addresses implementation challenges and integrating results into healthcare systems. For PGx specifically, a reasonable internal standard is to log which guideline version each report used, so that when guidance changes you can identify exactly which reports are affected.
A practical mapping checklist
- Inventory every PGx analyte and result field in your LIS.
- Match each to a LOINC observation code and confirm its current status on loinc.org.
- Bind phenotype values to the CPIC-aligned answer codes.
- Decide, with your receiving EHR teams, where SNOMED CT values are required.
- Separate genotype, phenotype and therapeutic implication into linked observations.
- Identify medications with RxNorm.
- Record the guideline version and medical director sign-off with each report.
- Test with one gene-drug pair end to end before scaling.
Our perspective
We see the same pattern in lab after lab: the team treats LOINC mapping as a one-time interface project, finishes it, and then discovers the harder problem is that the clinical content underneath keeps moving. A correctly coded result that carries an out-of-date recommendation is worse than an uncoded one, because the code lends it authority.
So we design in the other direction. The coded structure should be generated from the same evidence layer that produces the narrative, so the two cannot drift apart. Evidence is fused from multiple authoritative sources, reanalysis alerts you when guidance changes, and your medical director stays in control of final approval, with an audit trail. Codes are the plumbing. The judgment is still your lab's, and the platform should make that judgment faster and easier to defend, not replace it.
— Tarek
Building this with SignalPGx
If your lab wants coded, FHIR-ready PGx reports without building the evidence and reporting layer yourself, White-Label PGx Reporting is built for it: branded reports, HL7/FHIR integration, living reanalysis and medical-director review in one deployment. Plans are Pilot, Standard, Scale and Enterprise, with pricing available on request. For labs that need scoping first, our project services include Discovery & Strategy from $7,000, System Design from $9,000 and Deployment & Integration from $15,000, each one-off per project.
FAQ
What is the main LOINC code for a pharmacogenomics report?
The Pharmacogenomics result panel, LOINC 82118-1, is the panel used to group genotype, phenotype and therapeutic-implication observations. Its status on loinc.org was TRIAL when we reviewed it, so check the current status before you build.
Are LOINC and SNOMED CT interchangeable for PGx?
No. LOINC mainly names the test or observation, while SNOMED CT is commonly used to standardize coded result values. Many implementations use both, and the right split depends on the receiving system's specification.
Do I need FHIR to use LOINC PGx codes?
No, LOINC codes can be used in HL7 v2 messages and other formats. FHIR adds a structured model, including the Genomics Reporting guide, that links genotype findings to medication recommendations, which makes CDS integration easier.
Which LOINC codes cover CPIC phenotypes?
CPIC-aligned answer codes include LA10315-2 (ultrarapid), LA25390-8 (rapid), LA25391-6 (normal), LA10317-8 (intermediate) and LA9657-3 (poor metabolizer). These populate the result for gene interpretation observations such as 79714-2.
Sources
- New LOINC codes for reporting pharmacogenomics results
- LOINC 82118-1 Pharmacogenomics result panel
- LOINC 79714-2
- LOINC 51971-0
- CPIC Term Standardization: LOINC Codes
- Understanding LOINC and SNOMED CT in laboratory reporting
- Health data standards in pharmacogenomics systems: a scoping review
- LOINC and other standards
- Council on Pharmacy Standards: Data Standards
- HL7 Genomics Reporting IG: Pharmacogenomic Reporting
- Pharmacogenomics cascade testing, The Pharmacogenomics Journal
- A Survey of LOINC Code Selection Practices