Pharmacogenomics — why the same drug works differently in different people

The problem pharmacogenomics solves

Standard drug dosing assumes a mythical "average patient." A doctor prescribes 10mg of Drug X because clinical trials found that dose effective in the average trial participant. But drug response varies enormously between individuals — not randomly, but in ways that are substantially genetically determined.

The consequences of ignoring this variation:

Adverse drug reactions (ADRs) are the fourth leading cause of death in the United States, responsible for approximately 100,000 deaths and 2 million serious injuries per year. A significant fraction of serious ADRs are pharmacogenomically predictable — they occur because a patient metabolizes a drug far more slowly or quickly than expected, or because they carry a variant that makes them unusually sensitive to a drug's mechanism.

Treatment failure is the other side: patients who don't respond to standard doses because they metabolize a drug so rapidly that it never reaches therapeutic concentrations, or because their disease biology is driven by a mechanism the drug doesn't reach in their genetic context.

Pharmacogenomics promises to replace "average patient" dosing with genotype-informed prescribing — the right drug, at the right dose, for the right patient, from the first prescription rather than after trial and error.

---

Drug metabolism — the CYP450 system

When a drug enters the body, it must be processed before it can act and before it can be eliminated. This processing — drug metabolism — primarily happens in the liver and involves a family of enzymes called cytochrome P450 (CYP450) enzymes.

CYP450 enzymes modify drug molecules through oxidation, reduction, and hydrolysis reactions. These reactions typically convert an active drug into an inactive form (inactivation) or convert an inactive prodrug into an active drug (activation). The metabolized products are then excreted in urine or bile.

The human genome encodes 57 CYP450 enzymes, but a handful account for the metabolism of roughly 75% of clinically used drugs:

These enzymes are highly polymorphic — they vary substantially between individuals in ways that dramatically affect metabolic capacity. This is where pharmacogenomics lives.

---

The four metabolizer phenotypes

Pharmacogenomics classifies individuals into metabolizer categories based on their genotype at drug-metabolizing genes. The standard system uses four categories:

Poor Metabolizers (PM) Carry two non-functional alleles — no functional enzyme is produced. Drugs that require this enzyme for inactivation accumulate to toxic levels. Drugs that require this enzyme for activation (prodrugs) are never converted to the active form and produce no effect.

Intermediate Metabolizers (IM) Carry one functional and one reduced-function (or non-functional) allele. Metabolism is slower than normal but not absent. Approximately halfway between normal and poor metabolizer phenotype in many cases.

Normal Metabolizers (NM) (also called Extensive Metabolizers, EM) Carry two functional alleles. Standard dosing is designed for this group.

Ultrarapid Metabolizers (UM) Carry extra copies of a functional gene (gene duplication) or carry alleles with enhanced function. Metabolize drugs extremely rapidly — active drugs are cleared before reaching therapeutic concentrations; prodrugs are converted so rapidly that toxic concentrations of the active metabolite form.

The critical insight: the same variant causes opposite problems for drugs vs. prodrugs.

---

CYP2D6 — the paradigm case

CYP2D6 is the best-studied pharmacogenomic gene and illustrates the consequences of metabolizer variation most clearly.

CYP2D6 genetics:

• More than 100 known variant alleles, designated CYP2D61 (reference/normal), 2, 3, 4, etc.
• 3, 4, *5 are the most common loss-of-function alleles in European populations
• Gene deletion (*5) and gene duplication (up to 13 copies reported) are possible
• Allele frequencies vary dramatically by ancestry — 4 is common in Europeans (~20%) but rare in East Asians; 17 and *29 are common in Africans but rare elsewhere

CYP2D6 population frequencies (approximate):

• Poor Metabolizers: 5–10% European, 1–2% East Asian, 3–8% African
• Ultrarapid Metabolizers: 1–2% European, 1% East Asian, 5–30% East African (some Ethiopian populations approach 30%)

Codeine — the canonical example:

Codeine is a prodrug. CYP2D6 converts codeine to morphine, the active analgesic. This means:

• Poor Metabolizers: Codeine is never converted to morphine → no pain relief. The patient may be labeled "drug-seeking" or non-compliant when in fact they simply cannot metabolize the drug.
• Normal Metabolizers: Standard analgesic effect.
• Ultrarapid Metabolizers: Extremely rapid conversion produces dangerously high morphine concentrations. Several deaths have been reported, including a nursing infant who died after receiving breast milk from a UM mother prescribed codeine postpartum.

In 2013, the FDA issued a black box warning for codeine in breastfeeding mothers and children after multiple deaths. In 2017, the FDA banned codeine in children under 12 for pain and cough.

Tamoxifen:

Tamoxifen is the most widely used endocrine therapy for hormone receptor-positive breast cancer. It is also a prodrug — CYP2D6 converts tamoxifen to endoxifen, the active metabolite responsible for most of the drug's anti-cancer effect.

Poor metabolizers of CYP2D6 produce minimal endoxifen → reduced tamoxifen efficacy → higher breast cancer recurrence rates. Multiple studies have found that CYP2D6 PM status is associated with significantly worse disease-free survival in patients on tamoxifen.

The clinical implication: a CYP2D6 PM patient with ER-positive breast cancer may be better served by switching to an aromatase inhibitor (which does not require CYP2D6 metabolism) rather than tamoxifen. CPIC (Clinical Pharmacogenomics Implementation Consortium) provides exactly this recommendation.

---

Other critical gene-drug pairs

CYP2C19 and clopidogrel:

Clopidogrel (Plavix) is an antiplatelet drug widely used to prevent blood clots after cardiac stents or strokes. It is a prodrug requiring CYP2C19 activation.

Poor metabolizers of CYP2C19 cannot adequately activate clopidogrel → insufficient antiplatelet effect → higher rates of stent thrombosis and recurrent stroke. The FDA added a black box warning in 2010 noting that poor metabolizers have higher cardiovascular event rates.

CYP2C192 and 3 are the primary loss-of-function alleles. Their frequencies vary substantially by ancestry: *2 is found in ~15% of Europeans, ~30% of Asians, and ~15% of Africans. This means East Asian patients are at particularly elevated risk of clopidogrel treatment failure.

Alternative antiplatelet drugs (prasugrel, ticagrelor) do not require CYP2C19 activation and are recommended for CYP2C19 poor metabolizers by CPIC guidelines.

CYP2C9/VKORC1 and warfarin:

Warfarin is the most widely prescribed anticoagulant in the world and the drug with the most well-established pharmacogenomic dosing guidance. It has an extremely narrow therapeutic window — too little and it fails to prevent clots; too much and it causes life-threatening bleeding.

Two genes are central:

• CYP2C9: Metabolizes warfarin; poor metabolizers have slower clearance → require lower doses
• VKORC1: Encodes the warfarin target (vitamin K epoxide reductase); variants affect how sensitive the target is to warfarin

The FDA-approved warfarin label includes a dosing table based on CYP2C9 and VKORC1 genotype. FDA-cleared clinical decision support tools calculate warfarin starting doses using these genotypes plus clinical factors (age, weight, INR target). VKORC1 variant frequencies differ markedly between Europeans and Asians — largely explaining why Asian patients on average require lower warfarin doses.

HLA-B and severe hypersensitivity reactions:

Not all pharmacogenomics involves drug metabolism. Some involves immune recognition. The HLA (human leukocyte antigen) genes encode proteins that present peptides to the immune system. Certain HLA alleles cause the immune system to treat a drug or drug-protein complex as a foreign antigen — triggering severe, life-threatening hypersensitivity reactions.

• HLA-B\57:01 and abacavir: Abacavir is an antiretroviral drug for HIV. Approximately 5–8% of individuals of European ancestry carry HLA-B\57:01. In these individuals, abacavir triggers a severe hypersensitivity reaction (fever, rash, respiratory symptoms) that can be fatal on rechallenge. Prospective HLA-B\57:01 screening before abacavir prescription is now standard of care globally — it has essentially eliminated abacavir hypersensitivity. CPIC: do not use abacavir in HLA-B\57:01 carriers.

• HLA-B\15:02 and carbamazepine: Carbamazepine is an anticonvulsant. In carriers of HLA-B\15:02 — found almost exclusively in Southeast and East Asian populations (prevalence ~6–8% in Han Chinese, Thai, Malaysian populations) — carbamazepine can cause Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN), severe blistering conditions with 10–30% mortality. The FDA label requires HLA-B\*15:02 testing before prescribing carbamazepine to patients of Asian ancestry.

---

Clinical implementation — CPIC and PharmGKB

The science of pharmacogenomics has outpaced clinical implementation for decades. The two most important resources for translating PGx into practice:

CPIC (Clinical Pharmacogenomics Implementation Consortium)

CPIC publishes peer-reviewed, freely available guidelines for gene-drug pairs with sufficient evidence for clinical action. Guidelines are tiered by evidence strength and provide specific prescribing recommendations based on genotype. Current CPIC guidelines cover >25 gene-drug pairs including CYP2D6/codeine, CYP2C19/clopidogrel, CYP2C9/VKORC1/warfarin, HLA-B/abacavir, and many others.

CPIC's goal: when a pharmacogenomic result is known, the clinician should not have to re-derive the clinical implication — CPIC provides the interpretation and recommendation directly.

PharmGKB (Pharmacogenomics Knowledgebase)

PharmGKB is a curated database of the relationships between genetic variation, drugs, and clinical outcomes. It aggregates evidence from thousands of studies, curates gene-drug relationships by evidence level (1A = highest, 4 = lowest), and hosts CPIC guidelines.

PharmGKB is also where variant-drug annotations live: you can look up a specific variant and find all documented drug interactions with that variant.

Pre-emptive genotyping

The emerging clinical model is pre-emptive PGx genotyping: testing patients for a panel of pharmacogenomic variants before they need any specific drug, storing the results in the electronic health record, and automatically surfacing relevant alerts when a relevant drug is prescribed.

Large health systems (Vanderbilt, Stanford, Mayo Clinic, St. Jude Children's Research Hospital) have implemented pre-emptive PGx programs. Vanderbilt's PREDICT program has genotyped over 100,000 patients and demonstrated clinical utility across multiple gene-drug pairs. The cost of a 12-gene PGx panel is now under $200 — often less than a single adverse drug reaction hospitalization.

---

Ancestry and pharmacogenomics — a compounding inequity

Pharmacogenomics has a serious ancestry problem, and it intersects with everything you learned in Modules 4 and 7.

The research base is European-skewed: Most early pharmacogenomic studies were conducted in European-ancestry populations. Allele frequencies, clinical associations, and dosing recommendations were derived from this base.

Allele frequencies vary dramatically by ancestry: This isn't a minor correction — in some cases the clinically important allele is essentially absent in one population and common in another.

Clinical consequences of ancestry mismatch:

1. Standard dosing recommendations calibrated to European allele frequencies systematically mis-dose patients of other ancestries — not because of any biological inferiority, but because the calibration dataset didn't include them.

1. Clinical decision support tools that use self-reported race/ethnicity as a proxy for pharmacogenomic prediction (rather than actual genotyping) amplify these errors — they assume individuals within a racial group are genetically homogeneous, which is false.

1. Testing programs that are not universally offered but are recommended specifically for "patients of Asian ancestry" (as with carbamazepine/HLA-B\*15:02) create equity issues: who determines ancestry? What about admixed individuals?

The solution is universal genotyping — test everyone, store results, apply regardless of ancestry. But pre-emptive programs have been implemented primarily at well-resourced academic medical centers, leaving community hospitals and under-resourced health systems behind.

---

Beyond metabolism — pharmacogenomics of targets and transporters

Drug metabolism (CYP450) is only one layer of pharmacogenomics. Two additional layers:

Drug targets: Genetic variants in the gene encoding a drug's molecular target can alter how effectively the drug binds or acts. The most important example in oncology is precision oncology: tumors are sequenced to identify somatic mutations in drug targets, then drugs are matched to the mutation.

• EGFR mutations in non-small cell lung cancer → EGFR inhibitors (erlotinib, osimertinib) work dramatically better in EGFR-mutant than wild-type tumors
• BRAF V600E in melanoma → BRAF inhibitors (vemurafenib, dabrafenib)
• BCR-ABL in CML → imatinib (Gleevec) — the founding example of targeted oncology

This is pharmacogenomics of the tumor genome, not the germline — and it's the dominant model in oncology today.

Drug transporters: Proteins that move drugs into and out of cells affect drug concentrations at the site of action. Variants in transporters like SLCO1B1 (encoding a hepatic uptake transporter) dramatically affect statin concentrations in the liver.

SLCO1B15 — a common variant — reduces hepatic uptake of simvastatin, causing higher blood concentrations → increased risk of statin-induced myopathy (muscle damage). CPIC recommends lower simvastatin doses or switching to pravastatin/rosuvastatin (less affected by SLCO1B1 variation) in SLCO1B15 carriers.

---

Check yourself

1. A 35-year-old woman of East African descent is prescribed codeine for pain after minor surgery. She experiences severe respiratory depression requiring emergency treatment. Her CYP2D6 genotype later reveals she is an ultrarapid metabolizer. Explain the molecular mechanism of her adverse reaction step by step.

2. A Han Chinese patient is prescribed carbamazepine for epilepsy without HLA-B\*15:02 testing. She develops Stevens-Johnson syndrome. Her physician says the FDA label only recommends testing for "patients of Asian ancestry" and he didn't know her ancestry. List three systemic failures this case represents, beyond individual physician error.

3. A pharmaceutical company designs a clinical trial for a new antiplatelet drug that activates via CYP2C19. They recruit 10,000 participants — 95% European ancestry. The drug passes Phase 3 with excellent efficacy. When the drug is later prescribed to East Asian patients, it underperforms. What happened, and what trial design decision caused it?

4. A hospital implements a pre-emptive PGx program and genotypes 50,000 patients for 12 genes. They find a patient who is a CYP2D6 ultrarapid metabolizer. The patient is not currently on any medications. What should happen with this result, and what are the risks of doing nothing vs. acting on it proactively?

---

Key facts to remember

• CYP450 enzymes metabolize ~75% of clinically used drugs; CYP2D6, CYP2C19, CYP2C9, CYP3A4/5 are the key ones
• Four metabolizer phenotypes: Poor (PM), Intermediate (IM), Normal (NM), Ultrarapid (UM)
• PM + active drug = accumulation/toxicity; PM + prodrug = no effect; UM + active drug = no effect; UM + prodrug = toxicity
• CYP2D6/codeine: codeine is a prodrug; UM → fatal morphine toxicity (FDA black box warning); PM → no analgesia
• CYP2D6/tamoxifen: PM → reduced endoxifen → worse breast cancer outcomes; switch to aromatase inhibitor
• CYP2C19/clopidogrel: prodrug; PM → no antiplatelet effect; FDA black box warning 2010; CYP2C19*2 common in Asians (~30%)
• HLA-B\*57:01/abacavir: pre-screening eliminates hypersensitivity — universal standard of care
• HLA-B\*15:02/carbamazepine: SJS/TEN risk; common in Southeast/East Asian populations; FDA mandates testing for Asian patients
• CPIC: clinical guidelines for gene-drug pairs; PharmGKB: curated evidence database
• Allele frequencies vary dramatically by ancestry → standard dosing recommendations can systematically mis-dose non-European patients

---

Primary sources & references

• Relling, M. V. & Evans, W. E. (2015). "Pharmacogenomics in the clinic." Nature, 526, 343–350.
• Crews, K. R. et al. (2014). "Clinical Pharmacogenetics Implementation Consortium guidelines for cytochrome P450 2D6 genotype and codeine therapy." Clinical Pharmacology & Therapeutics, 91, 321–326.
• Scott, S. A. et al. (2013). "CPIC guidelines for CYP2C19 and clopidogrel therapy." Clinical Pharmacology & Therapeutics, 94, 317–323.
• Mallal, S. et al. (2008). "HLA-B\5701 screening for hypersensitivity to abacavir." NEJM, 358*, 568–579.
• Relling, M. V. et al. (2022). "CPIC guideline for CYP2D6 and tamoxifen therapy." Clinical Pharmacology & Therapeutics, 111, 1296–1302.
• Dunnenberger, H. M. et al. (2015). "Pre-emptive clinical pharmacogenomics testing: PREDICT." Mayo Clinic Proceedings, 90, 36–45.