CYP2D6

What it metabolises, and the fact that it cannot be induced human data

CYP2D6 accounts for only a small fraction of total hepatic CYP protein yet handles a fifth to a quarter of prescribed drugs, and that mismatch is the source of its clinical character: there is not much enzyme, so it saturates easily and inhibits easily, and a strong inhibitor can remove the pathway almost entirely. Its substrates are overwhelmingly lipophilic bases — antidepressants, antipsychotics, beta blockers, antiarrhythmics, opioids and antiemetics. The second defining fact is a negative one that deserves stating loudly because it is the exception on this shelf: CYP2D6 is not meaningfully inducible. There is no rifampicin effect, no smoking effect, no St John’s wort effect of consequence here. Practically, that means CYP2D6 interactions run in one direction only — inhibition — and that the variability in this pathway is genetic and pharmacological rather than environmental. It also means that when a CYP2D6 substrate unexpectedly stops working, the explanation is more likely to be genotype or non-adherence than an inducer somewhere in the list.

Sources: Flockhart DA 2021 · U.S. Food 2023 · Zanger UM 2013 · Gasche Y 2004 · Koren G 2006 · Clinical Pharmacogenetics Implementation Consortium (Crews KR 2021* · Jin Y 2005

Drug inhibitors, with FDA potency classes human data

Because CYP2D6 cannot be induced, this table has only one column of interest. The strong inhibitors here are unusually common drugs, and two of them — paroxetine and fluoxetine — are antidepressants frequently co-prescribed with other CYP2D6 substrates, which makes this the most reliably reproducible interaction axis in psychiatry. Paroxetine is additionally a mechanism-based inhibitor of the enzyme that clears it, so its inhibition deepens with continued dosing; fluoxetine has an active metabolite with a one to two week half-life, so its inhibition persists for weeks after the last tablet. Neither fact is obvious from a potency class alone.

Sources: U.S. Food 2023 · U.S. Food 2020* · Flockhart DA 2021 · Zanger UM 2013

Botanical inhibitors: goldenseal is the one with human data contested human data

Goldenseal is the strongest botanical entry on this isoform and one of the very few botanicals with a controlled in-vivo human probe study behind it. In the Gurley design, volunteers took a standardised botanical supplement for 28 days and were phenotyped with probe substrates for CYP1A2, CYP2D6, CYP2E1 and CYP3A4/5 before and after. Goldenseal produced significant inhibition of both CYP2D6 and CYP3A4/5 — the largest botanical effect in that study — attributable to its isoquinoline alkaloids berberine and hydrastine. The same study is equally valuable for its negatives: kava, black cohosh and valerian did not significantly inhibit CYP2D6, and black cohosh in particular is frequently claimed to, so the null result is worth carrying. Cannabidiol is described in vitro as a potent atypical inhibitor of CYP2D6, meaning it does not fit the standard single-site competitive model, but that has not been translated into a demonstrated human CYP2D6 interaction. Curcuminoids inhibit CYP2D6 weakly in vitro.

Contested — caveat. Only the goldenseal entry rests on controlled in-vivo human data. The cannabidiol, THC, curcuminoid, piperine and nutmeg entries are in-vitro determinations whose magnitude at real human exposures is unestablished. The kava, black cohosh and valerian rows are recorded as null findings from one study of one standardised product each, which is not the same as proof of no effect for every preparation.

Sources: Gurley BJ 2005 · Yamaori S 2011* · Volak LP 2008 · Bhardwaj RK 2002 · Flockhart DA 2021 · Stout SM 2014 · Beyer J 2006 · Bailey DG 2013

Clinical consequence, concretely human data

Codeine is the case that has killed people and it runs in both directions. A CYP2D6 poor metaboliser given codeine after surgery gets no useful analgesia and is often labelled drug-seeking for saying so. An ultra-rapid metaboliser converts codeine to morphine faster than expected: there are published reports of life-threatening intoxication in an adult and of a fatal morphine poisoning in a breastfed neonate whose mother was an ultra-rapid metaboliser taking prescribed codeine, and those cases are why codeine is now contraindicated in breastfeeding and in young children in many jurisdictions. A patient on metoprolol who starts paroxetine develops bradycardia and fatigue a week later, with nothing in either prescription looking wrong. A patient on flecainide who starts fluoxetine can become proarrhythmic. A patient on tramadol who starts bupropion loses analgesia while keeping the seizure risk of both drugs. And the trap that recurs in the plant-medicine literature is the potentiation one: the same CYP2D6 inhibition that is deliberately used to intensify a tryptamine experience is the mechanism by which an intended dose becomes an unintended overdose, and the rule that predicts the stronger experience predicts the emergency identically.

Sources: Gasche Y 2004 · Koren G 2006 · Clinical Pharmacogenetics Implementation Consortium (Crews KR 2021* · Flockhart DA 2021 · U.S. Food 2023

Polymorphism: four phenotypes from one gene contested human data

CYP2D6 is the most variable drug-metabolising gene in the human genome, with more than a hundred named allelic variants spanning complete non-function, partial function, normal function, gene deletion and gene duplication or multiplication. The standard scheme converts a genotype into an activity score and then into one of four phenotypes: poor metaboliser (no functional enzyme, roughly 5 to 10 percent of European-ancestry populations), intermediate, normal, and ultra-rapid metaboliser (functional gene duplications, at frequencies reported up to around 10 percent in some Mediterranean, Middle Eastern and North African populations and near 25 percent in some studies of populations in the Horn of Africa). Two complications matter for anyone reading a genotype report. First, copy-number variation means the assay must count gene copies, not just detect variants, and older panels did not. Second, phenoconversion: a normal metaboliser on paroxetine or fluoxetine or bupropion or terbinafine is functionally a poor metaboliser, so the medication list can override the genotype entirely. Clinical guidelines exist for codeine, tramadol, the tricyclics, the SSRIs, atomoxetine and tamoxifen, and they are the best worked example on this shelf of pharmacogenomics actually changing a prescription.

• Four phenotypes: poor, intermediate, normal, ultra-rapid — derived from an activity score, not from a single variant.
• Gene duplications produce the ultra-rapid phenotype; allele and duplication frequencies vary widely between populations.
• Phenoconversion by a strong inhibitor can matter more than genotype.
• The enzyme is not inducible, so there is no environmental route to the opposite problem.
• Prodrug substrates (codeine, tramadol, tamoxifen) invert the direction of every genotype and interaction prediction.

Contested — caveat. Population frequency figures come from heterogeneous studies and are reported by broad ancestral grouping, which predicts nothing about an individual. The ultra-rapid frequency figures in particular vary widely between studies and assay generations, because older genotyping panels did not reliably count gene copies.

Sources: Zanger UM 2013 · Clinical Pharmacogenetics Implementation Consortium (Crews KR 2021* · Gasche Y 2004 · Koren G 2006 · Flockhart DA 2021

The tamoxifen controversy, because it is the honest example contested human data

Tamoxifen is a prodrug and CYP2D6 forms its most potent active metabolite, endoxifen. From that mechanism it follows that poor metabolisers, or patients taking a strong CYP2D6 inhibitor such as paroxetine or fluoxetine for hot flushes, should do worse on adjuvant tamoxifen — and the early pharmacokinetic and cohort work supported exactly that, making it one of the most cited examples of pharmacogenomics mattering. Then the large retrospective genotype analyses of the major adjuvant trials were published, and two of them found no significant association between CYP2D6 genotype and breast-cancer outcome. The field has not resolved this. Proposed explanations include genotyping from tumour rather than germline DNA in the trial analyses, endoxifen concentrations above a threshold in most patients regardless of genotype, adherence effects, and the sheer difficulty of detecting a modest effect on a long-horizon outcome. The reason it belongs on this page is that it is the clearest available demonstration that a mechanism can be completely correct and its clinical consequence still unproven — which is precisely the epistemic situation most of the botanical rows in this shelf are in, and it is better to show the reader a worked example of that than to pretend the distinction does not exist.

Contested — caveat. Genuinely unresolved. The metabolic pathway is not in doubt; the outcome association is, with the major trial genotype analyses and the earlier cohort studies disagreeing. Guidelines differ between bodies on whether to avoid strong CYP2D6 inhibitors in patients taking tamoxifen. Presented here as an open question, not as settled either way.

Sources: Jin Y 2005 · Regan MM 2012 · Zanger UM 2013 · Flockhart DA 2021