CPIC Guidelines and CPIC-Concordant PGx Reporting: A Lab Director's Guide

CPIC guidelines are the closest thing pharmacogenomics has to a rulebook for turning a genotype into an actionable prescribing recommendation. For a lab, "CPIC-concordant" is not a marketing phrase, it is a testable property of every report you sign out, and the labs that treat it that way build interpretation that clinicians and payers can trust.

What CPIC guidelines actually are

CPIC is an international consortium that publishes peer-reviewed, freely available guidelines for specific gene-drug pairs. Each guideline does one job: it assumes the genotype is already known and tells the prescriber what to do with it. That is a deliberate scope. CPIC does not decide who should be tested or argue the cost-effectiveness of a panel; it answers the question a clinician faces once a result is in hand, which is "given this phenotype, how should I prescribe this drug?"

As of 2025 the evidence base CPIC anchors is substantial: roughly 34 genes, 164 drugs, and 28 guidelines, with each recommendation graded by strength of evidence. That structure is what makes CPIC usable in a clinical report. A guideline gives you the gene, the drug, the phenotype-specific action (for example, select an alternative agent, adjust dose, or use standard dosing with monitoring), and a clear statement of how strong the underlying evidence is.

Diplotype to phenotype: the core interpretive step

The heart of a pharmacogenomic report is the translation from a diplotype, the pair of star alleles a patient carries, into a phenotype expressed in CPIC's standardized terminology. For metabolizing enzymes that means terms like poor, intermediate, normal, rapid, and ultrarapid metabolizer; other gene families use their own standardized categories. This standardization matters because it is the shared language that lets a guideline written by one group be applied consistently by another.

This is where reporting quality is won or lost. A defensible report does not jump from a raw genotype straight to "avoid this drug." It shows the diplotype, assigns the phenotype using CPIC's terminology, and then attaches the phenotype-driven recommendation. Each link in that chain should be inspectable. Our PGx reporting approach is built around exactly this traceability, and you can see how a finished interpretation reads in a sample report.

What "CPIC-concordant" really means

CPIC-concordant reporting means that for a given gene-drug pair and assigned phenotype, your report's recommendation matches what CPIC publishes, in CPIC's own phenotype language. The important word is verifiable. Concordance is not a claim you make about your software, it is a property a reviewer can check by following the trail from diplotype to phenotype to the specific guideline that produced the recommendation.

This matters more than it might sound, because concordance is not automatic. Peer-reviewed work has documented commercial PGx tools diverging from CPIC on a meaningful share of gene-drug recommendations, often because proprietary interpretive logic is applied that the reader cannot see. When the logic is opaque, a clinician cannot tell whether a recommendation reflects CPIC, the vendor's house rules, or an error. Concordance-by-design, where the report is structured so that CPIC's recommendation is the default and any deviation is explicit and justified, is therefore a genuine differentiator rather than table stakes.

Why CPIC and FDA labeling sometimes disagree

Labs often assume CPIC and FDA pharmacogenomic labeling should say the same thing. They frequently do not, and the reason is structural. CPIC and the FDA write for different purposes and from different inputs. FDA labeling reflects what a drug's sponsor submitted and what the agency approved; it tends to be more conservative, is sometimes silent on a specific phenotype, and is updated on a regulatory cadence. CPIC synthesizes the broader published evidence into an actionable recommendation for a known genotype.

The gap is not theoretical. Published comparisons of overlapping drugs have found CPIC and FDA labeling concordant on only a minority of pairs. That means a report built solely on the FDA pharmacogenomic table will, for some drugs, give different guidance than one built solely on CPIC, and neither alone tells the whole story. A note on the current regulatory backdrop: the FDA's laboratory-developed test rule was vacated in March 2025 and rescinded in September 2025, so CLIA is again the operative framework for labs. FDA labeling remains an essential evidence source for interpretation, but labs are not building their reporting program around FDA device clearance.

Reconciling CPIC, FDA, DPWG, and PharmGKB

Because no single source is sufficient, a high-quality interpretation reconciles four established bodies of evidence rather than simply listing them. CPIC provides graded, actionable gene-drug recommendations. FDA pharmacogenomic labeling, which annotates well over 250 germline-relevant drugs, carries regulatory weight. DPWG, the Dutch Pharmacogenetics Working Group, contributes 100-plus gene-drug recommendations that sometimes differ from CPIC in useful ways. PharmGKB, now part of ClinPGx, aggregates more than 23,000 curated annotations that underpin the whole field.

Reconciliation means the report makes a defensible call when these sources align and is transparent when they diverge. The reader should be able to see which source drove the recommendation and why. This is the difference between an interpretation that draws on the evidence base and one that quietly picks a winner behind the curtain. Our interpretation layer is designed to surface these sources rather than collapse them, and the custom lab report software renders that reconciliation in language a clinician can act on.

• CPIC — graded, phenotype-specific prescribing recommendations; the interpretive backbone.
• FDA pharmacogenomic labeling — regulatory context and approved-use language for germline-relevant drugs.
• DPWG — an independent set of gene-drug recommendations useful for cross-checking and coverage.
• PharmGKB / ClinPGx — the curated annotation layer that connects variants to drug response evidence.

Why source-cited interpretation beats a black box

A pharmacogenomic report is judged against the ACMG 2022 clinical pharmacogenomics technical standard, which spans nomenclature, testing, interpretation, and reporting, and which recommends sign-out by a board-certified professional in a CLIA-certified setting. Every one of those pillars rewards transparency. A report whose reasoning is visible can be checked against the standard; a black box cannot.

Transparency also addresses the single biggest reason clinicians hesitate to adopt these tools, which is trust in the interpretation. When a recommendation cites the specific guideline and shows the diplotype-to-phenotype logic, a prescriber can evaluate it on the merits. When it does not, the report asks for blind faith, and clinicians, reasonably, decline. The same traceability that earns clinical trust also supports the billing side: molecular and PGx claims increasingly require MolDX/DEX Z-codes from Medicare and major commercial payers, and clean, well-documented reporting is what keeps those claims defensible. You can see how reporting connects to reimbursement in our billing workflow.

Where AI fits, and where it does not

AI earns its place in PGx reporting by doing the mechanical, error-prone assembly well: calling the diplotype, mapping it to the correct CPIC-standardized phenotype, pulling the matching guideline and FDA, DPWG, and PharmGKB context, and drafting source-cited interpretive language at a speed no manual process matches. Our own lab-trained AI layer, built on leading models by the team behind a CLIA-certified genetic lab, is tuned specifically for this concordance-and-citation work.

What AI does not do is make the clinical decision. Qualified lab staff validate the call, approve the interpretation, and sign out the report; the AI never signs out and never overrides a reviewer. That human-in-the-loop design is not a compliance afterthought, it is the architecture, and it is what makes "CPIC-concordant, source-cited, lab-signed" a claim a director can stand behind. If you want to walk through how this works end to end, book a demo and we will run a real interpretation against your panel.

Frequently asked questions

What does CPIC-concordant PGx reporting actually mean?

It means the report's prescribing guidance matches the recommendation CPIC publishes for a given gene-drug pair and phenotype, using CPIC's standardized phenotype terminology. Concordance is verifiable: a reader can trace the diplotype to the assigned phenotype and the phenotype to the specific CPIC guideline that drove the recommendation. Studies have documented commercial PGx tools diverging from CPIC on a meaningful share of gene-drug recommendations, so concordance-by-design is a real differentiator rather than an assumed default.

Why do CPIC and FDA pharmacogenomic labeling sometimes disagree?

CPIC and the FDA write for different purposes. CPIC assumes a genotype is already known and tells the clinician what to do with it, while FDA labeling reflects what a drug's sponsor submitted and what the agency approved, which is often more conservative or simply silent on a specific phenotype. Published comparisons have found the two concordant on only a minority of overlapping drugs, which is exactly why a defensible report reconciles CPIC, FDA labeling, DPWG, and PharmGKB rather than relying on any single source.

Does AI write the PGx interpretation, or does the lab?

AI assists with assembling the diplotype call, mapping it to a CPIC-standardized phenotype, and drafting source-cited interpretive language, but qualified lab staff validate, approve, and sign out every report. The AI never makes the clinical decision. This human-in-the-loop model, combined with a CLIA-certified lab and traceable citations, is what keeps the report defensible and aligned with the ACMG clinical pharmacogenomics standard.