Guidelines: CPIC, DPWG and drug labels

Reviewed by Dr. Daniel Wallerstorfer, molecular biologist and founder of Novogenia · Updated: 13 October 2026 · Reading time: 9 min

In short

Guidelines translate a genotype into a concrete recommendation. The international CPIC consortium lists 29 freely accessible guidelines for gene–drug pairs; they show how an available result can be used, not whom to test. The Dutch DPWG also assesses how important a test is before starting treatment. Both agree on most recommendations. PharmGKB, now part of ClinPGx, curates studies and drug labels from the FDA, EMA, Swissmedic and other agencies. Drug label and guideline belong together – the decision is made by the treating physician.

From genotype to prescription: guidelines as interpreters

For a long time, pharmacogenetics faced a strange problem: the knowledge was there, but it did not reach the prescription pad. A Dutch team described in 2008 that the first dose recommendations for antidepressants based on CYP2D6 results had already been developed in 2001. In everyday care they were of little help, because they were missing where decisions are made – at prescribing and at dispensing in the pharmacy (Swen et al., Clinical Pharmacology & Therapeutics, 2008). In the US, the initiators of CPIC came to the same conclusion: pharmacogenetic tests are slow to be adopted partly because there are no specific instructions on how to adjust medication based on the result (Relling & Klein, Clinical Pharmacology & Therapeutics, 2011).

Guidelines are therefore the interpreters of pharmacogenetics. They translate the language of the lab – *2/*4, activity score 1.0 – into the language of prescribing: standard dose, adjusted dose or a different drug. Today there are several such interpreters, and encouragingly often they speak with one voice.

CPIC: freely accessible guidelines for gene–drug pairs

The international Clinical Pharmacogenetics Implementation Consortium (CPIC), launched in 2009, grew out of the Pharmacogenomics Research Network of the US National Institutes of Health and the PharmGKB knowledge base. Its goal: peer-reviewed, updated, evidence-based and freely accessible guidelines for gene–drug pairs (Relling & Klein, 2011). As of October 2026, CPIC lists 29 guidelines; today their home is the ClinPGx platform, which brings together PharmGKB, CPIC and the PharmCAT analysis tool (CPIC guidelines).

One principle matters: CPIC guidelines show how an available test result can be used to optimize drug therapy – they do not recommend whom to test (Lee et al., Clinical Pharmacology & Therapeutics, 2022). This is how the guidelines are structured:

CPIC levels A to D: for every gene–drug pair, CPIC indicates whether genetic information should change prescribing (A), could change it (B), or whether no change is currently recommended (C, D). Prescribing recommendations exist only for pairs at levels A and B.
Strength of each recommendation: every individual recommendation carries its own classification – strong, moderate or optional.
Uniform terms: phenotypes have the same names in all guidelines, from poor to ultrarapid metabolizer. CPIC defined these terms in a consensus process with 58 experts (Caudle et al., Genetics in Medicine, 2017).
Three steps: the genotype becomes a phenotype, and the phenotype becomes a recommendation for the specific drug.

CPIC guidelines are updated regularly. In 2026, for example, revised versions were published on thiopurines and on the 5-HT3 receptor antagonists used against nausea (Moore et al., Clinical Pharmacology & Therapeutics, 2026).

DPWG: from the lab bench to the software

In the Netherlands, the Royal Dutch Association for the Advancement of Pharmacy (KNMP) founded the Dutch Pharmacogenetics Working Group (DPWG). After a systematic literature review, it published recommendations in 2011 for 53 drugs related to genes such as CYP2D6, CYP2C19, CYP2C9, TPMT, DPD, VKORC1, UGT1A1, HLA-B and CYP3A5 (Swen et al., Clinical Pharmacology & Therapeutics, 2011).

The title of its first publication says it all: “from bench to byte” – from the lab bench into the software. Recommendations only help if they are available at prescribing and dispensing (Swen et al., 2008).

Unlike CPIC, DPWG also assesses how important a test is before starting treatment. For DPYD and fluoropyrimidines, its verdict is “essential”: DPYD should be tested before starting treatment with 5-fluorouracil or capecitabine (Lunenburg et al., European Journal of Human Genetics, 2020). The DPWG recommendations also formed the basis of the European PREPARE study, presented in Pre-emptive or reactive testing?.

CPIC, DPWG and others: more agreement than you might think

Two groups, two methods – and yet mostly the same result. A joint comparison by both consortia found a high rate of concordance among published recommendations; differences arise mainly from different methods of guideline development (Bank et al., Clinical Pharmacology & Therapeutics, 2018). An overview from Utrecht set four groups side by side (Abdullah-Koolmees et al., Frontiers in Pharmacology, 2021):

Group	Origin	Focus
CPIC	International consortium from the US network PGRN and PharmGKB	Treatment recommendations for many drugs once a result is available
DPWG	Netherlands, pharmacists’ organization KNMP	Treatment recommendations for many drugs, plus recommendations on testing
CPNDS	Canadian network for drug safety	Also recommendations on testing
RNPGx	French network for pharmacogenetics	Recommendations on testing, partly for specific clinical situations

The groups are actively working toward harmonization. The former term “extensive metabolizer” became “normal metabolizer,” for example, and a CYP2D6 genotype with one non-functional copy (activity score 1.0) is now classified as an intermediate metabolizer. In this overview, the highest level of testing recommendation goes to abacavir (HLA), clopidogrel (CYP2C19), fluoropyrimidines (DPYD), thiopurines (TPMT), irinotecan (UGT1A1), codeine (CYP2D6) and cisplatin (TPMT). The authors’ conclusion: there are many evidence-based guidelines with clear recommendations – and looking at all of them together adds value.

PharmGKB and ClinPGx: the reference work

PharmGKB collects knowledge on genes and drugs and is now part of ClinPGx (PharmGKB). Two areas are especially useful in practice:

Clinical annotations: they summarize the evidence for a variant and a drug. Their level of evidence is calculated with a transparent scoring system that takes into account study size and p-value, among other factors (Whirl-Carrillo et al., Clinical Pharmacology & Therapeutics, 2021).
Drug label annotations: PharmGKB curates product information from the US FDA, the European Medicines Agency (EMA), Swissmedic, Japan’s PMDA and Health Canada. The Swiss texts are translated in collaboration with the University of Basel. Each label is assigned a category: testing required, testing recommended, actionable or informative.

Drug labels: what the marketing authorization says

In the EU, when a medicine is authorized, the authority approves its summary of product characteristics (SmPC) – the product information for health professionals (Art. 21(1) Directive 2001/83/EC). An EMA review of all 517 centrally authorized medicines showed that pharmacogenetic information appears in different sections depending on the product, and that it is steadily increasing (Ehmann et al., Pharmacogenomics Journal, 2015). Three examples from European public assessment reports (EPARs), as summarized by PharmGKB:

Abacavir: HLA-B*57:01 screening in all patients with HIV; carriers should not receive abacavir.
Capecitabine: phenotype and/or genotype testing recommended before starting treatment; complete DPD deficiency is a contraindication, and for partial deficiency a reduced starting dose should be considered.
Clopidogrel: CYP2C19 poor metabolizers form less of the active metabolite at recommended doses, and platelet function is inhibited less.

Switzerland: the EU directive does not apply there; Switzerland has its own drug labels, which PharmGKB curates separately. For codeine, according to this curation, the Swiss label lists CYP2D6 ultrarapid metabolizers as a contraindication. USA: the FDA maintains a table of pharmacogenomic biomarkers in drug labeling – the information appears there in sections such as warnings, dosage or clinical pharmacology. Alongside it, there is a table of pharmacogenetic associations in three sections: from associations that support therapeutic recommendations to those affecting pharmacokinetics only.

When guideline and drug label sound different

Drug labels and guidelines serve different purposes. The drug label describes the individual product within its authorized framework. Guidelines evaluate the entire body of evidence and translate it into recommendations for action. That is why they can go to different lengths: for clopidogrel, the European public assessment report describes the weaker effect in poor metabolizers, while the CPIC guideline specifically recommends an alternative in acute coronary syndrome and PCI.

The logical conclusion: both sources belong together. The drug label shows the authorized framework, the guideline the evaluated evidence. Weighing and deciding is the job of the treating physician – with the clinical picture in view.

How to use guidelines in everyday practice

Check the source: which guideline does the result rely on – CPIC, DPWG or both?
Check the version: guidelines are updated. An older result can be reread in light of the current version.
Match the terms: result and guideline should use the same phenotype terms.
Add the drug label: contraindications, testing recommendations and dosing information are found there.
Consider co-medication: inhibitors can slow an enzyme down further. The CPIC guideline on tamoxifen, for example, advises avoiding moderate and strong CYP2D6 inhibitors (Goetz et al., Clinical Pharmacology & Therapeutics, 2018).
Document the decision: decisions about medication are made by the treating physician.

How the guidelines fit into everyday practice is shown in Pharmacogenetics in medical practice; how to read a result step by step, in Understanding and communicating results.

Frequently asked questions

What is the difference between CPIC and DPWG?

Both translate genotypes into treatment recommendations, and they mostly agree. CPIC grew out of the US network PGRN and the PharmGKB database and explains how an available result can be used. DPWG was founded by the Dutch pharmacists’ organization KNMP and also assesses whether testing should take place before a treatment.

What do CPIC levels A to D mean?

They describe whether genetic information should change the prescribing of a drug (A), could change it (B), or whether no change is currently recommended (C and D). Prescribing recommendations exist only for pairs at levels A and B. Separately, each individual recommendation carries a strength: strong, moderate or optional.

Do the guidelines say who should be tested?

CPIC deliberately does not: its guidelines show how an available result can be used. DPWG, the Canadian CPNDS and the French RNPGx, by contrast, also make recommendations on testing; RNPGx even describes specific clinical situations.

Where does pharmacogenetic information appear in the drug label?

That depends on the product: under contraindications, dosage, warnings or pharmacokinetics. An EMA review of all 517 centrally authorized medicines classified the information by exactly these sections. The drug label annotations on PharmGKB are a faster route.

What should I do if the drug label and the guideline sound different?

Put them side by side: the drug label shows the authorized framework of the product, the guideline the evaluated body of evidence. The treating physician weighs both against the clinical picture and documents the decision.

Pharmacogenetics in your practice or pharmacy?

Tell us about your setting – we will show you how the offer can work for you.

Request a project →Offer for physicians

Note: This guide is general information for businesses and professionals and does not replace medical or legal advice. The genetic traits determined in a genetic analysis are well documented (each gene included is backed by at least three independent studies); recommendations derived from them are logical conclusions, not scientifically proven. Pharmacogenetic results do not replace a medical decision: any change to medication is made only by the treating physician. Regulations can change; errors and changes excepted.