Phase-1 Metabolism and the CYP Interaction Axis

What phase-1 oxidative metabolism is human data

Almost everything foreign that enters the body by mouth is fat-soluble enough to cross a membrane, which is precisely why the body cannot excrete it: the kidney filters water-soluble molecules and reabsorbs lipophilic ones. Metabolism solves that by making the molecule more polar in two stages. Phase 1 is functionalisation — the insertion of an oxygen, the unmasking of a hydroxyl, an N- or O-dealkylation, an epoxidation — carried out overwhelmingly by the cytochrome P450 superfamily, haem-containing monooxygenases embedded in the endoplasmic reticulum of hepatocytes and enterocytes. The catalytic cycle binds the substrate, accepts two electrons from NADPH via cytochrome P450 reductase, splits molecular oxygen, and delivers one oxygen atom into the substrate while the other leaves as water. Phase 2 then attaches a large water-soluble group — glucuronic acid, sulfate, glutathione, an acetyl or a methyl — to the handle phase 1 created, and the conjugate is excreted in bile or urine.

• Phase 1 makes a handle. Phase 2 attaches the exit tag. Most interaction tables model only the first half.
• The same reaction is detoxification for one molecule and activation for another: phase 1 turns codeine into morphine and paracetamol into the liver-killing quinone imine NAPQI.
• CYP nomenclature: family number, subfamily letter, gene number. CYP2D6 is family 2, subfamily D, gene 6. Above about 40 percent amino-acid identity shares a family; above about 55 percent shares a subfamily.

Sources: Guengerich FP 2008 · Zanger UM 2013 · Lynch T 2007*

Why the CYP family is the dominant interaction axis human data

Interactions can happen anywhere along the path a substance takes — absorption, protein binding, transport, metabolism, renal elimination, or at the receptor itself. They concentrate on CYP for three structural reasons. First, convergence: a small number of isoforms handle the clearance of most marketed drugs, so unrelated substances share an enzyme and therefore compete. CYP3A4 alone is implicated in the metabolism of roughly half of drugs in clinical use. Second, saturability and inhibitability: these are enzymes with finite capacity and an accessible active site, so a second molecule with affinity for that site changes the first molecule fate immediately. Third, inducibility: CYP expression is under transcriptional control by ligand-activated receptors — the pregnane X receptor (PXR), the constitutive androstane receptor (CAR), and the aryl hydrocarbon receptor (AhR) — so a substance can raise the amount of enzyme present over days, not just occupy what is there. Nothing else in pharmacokinetics combines convergence, competition and inducibility in one place.

• Convergence is why a grapefruit and an antifungal and a protease inhibitor and a piperine capsule all produce the same class of consequence.
• PXR and CAR are xenobiotic sensors: they exist to upregulate clearance in response to a chemical load. Induction is a defence mechanism that becomes an interaction.
• AhR is the receptor that polycyclic aromatic hydrocarbons in smoke activate, which is why smoking is a CYP1A2 interaction and nicotine is not.

Sources: Zanger UM 2013 · Moore LB 2000 · Flockhart DA 2021 · U.S. Food 2023

Inhibition and induction: opposite mechanisms, opposite clinical consequences human data

These two are not degrees of the same thing. Inhibition reduces the activity of enzyme that is already present; onset is as fast as the inhibitor reaches the enzyme, which for an intestinal effect can be the first dose, and the consequence is that the substrate is cleared more slowly and accumulates. Induction increases the amount of enzyme by increasing its transcription; onset takes days to a couple of weeks as new protein accumulates, offset takes a similar period after the inducer stops, and the consequence is that the substrate is destroyed faster and its concentration falls. The clinical failures are mirror images and the second one is more dangerous precisely because it is quiet. An accumulation produces symptoms — sedation, bleeding, rhabdomyolysis, arrhythmia — which prompt investigation. A loss of exposure produces nothing at all until the transplanted organ is rejected, the viral load rebounds with resistance, the seizure happens, the INR collapses, or the contraception fails. Nobody phones a pharmacist about the absence of a side effect.

Sources: U.S. Food 2020* · Zanger UM 2013 · Moore LB 2000 · Ruschitzka F 2000 · Piscitelli SC 2000 · Bailey DG 2013

Reversible, quasi-irreversible, and mechanism-based inhibition human data

Reversible inhibition — competitive, non-competitive or mixed — tracks the concentration of the inhibitor. When the inhibitor is cleared, the enzyme is free again, so the interaction has roughly the duration of the inhibitor half-life plus a margin. Mechanism-based inhibition is different in kind: the enzyme metabolises the inhibitor into a reactive species that then covalently modifies the haem or the apoprotein, or coordinates the haem iron so tightly that the complex does not dissociate on any useful timescale. The enzyme molecule is destroyed. Activity returns only as the cell synthesises new protein, which means recovery is governed by the enzyme turnover rate and not by the inhibitor pharmacokinetics at all. This is the single most misunderstood point in practical interaction avoidance: a mechanism-based inhibitor outlasts its own presence in the blood. Grapefruit furanocoumarins are the canonical dietary example — intestinal CYP3A4 protein falls measurably after one glass, the inhibitor itself is long gone within hours, and CYP3A activity takes on the order of one to three days to return. Separating the juice from the tablet by a few hours therefore does not avoid the interaction, which is the part almost every patient gets wrong.

• Reversible: fluconazole, cimetidine, quinidine. Duration follows the inhibitor.
• Mechanism-based or quasi-irreversible: grapefruit furanocoumarins, clarithromycin and erythromycin (nitrosoalkane-iron complex), ritonavir, paroxetine on CYP2D6, disulfiram metabolites on CYP2E1. Duration follows enzyme resynthesis.
• Practical corollary: for a mechanism-based inhibitor there is no safe dosing gap within the day. For a reversible one, timing sometimes helps and sometimes does not.
• Bergamottin and 6,7-dihydroxybergamottin are the identified furanocoumarin mediators; a furanocoumarin-free grapefruit juice does not reproduce the interaction, which is the experiment that settled the question.

Sources: Orr STM 2012 · Greenblatt DJ 2003 · Bailey DG 2013 · Paine MF 2006 · Edwards DJ 1996 · Lown KS 1997 · Lundahl J 1995

The prodrug inversion: when an inhibitor gives you LESS drug human data

The reflex rule — inhibitor means higher level means more effect — is wrong for every substance whose activity lives in a metabolite rather than in the molecule administered. If CYP2D6 converts codeine to morphine, then inhibiting CYP2D6 does not produce a codeine overdose; it produces an absence of analgesia, because the opioid effect of codeine is almost entirely the morphine it becomes. The same inversion applies to tramadol and its O-desmethyl metabolite, to clopidogrel, which requires CYP2C19 for its two-step bioactivation to the active thiol, and to tamoxifen, whose potent antioestrogen endoxifen is formed by CYP2D6. Add an inhibitor and the patient on codeine is in pain, the patient on clopidogrel has less platelet inhibition than the prescription implies and therefore more stent thrombosis risk, and the drug has quietly stopped being the drug. Induction inverts the same way in the other direction: an inducer raises the active metabolite of a prodrug, which is why a CYP2D6 ultra-rapid metaboliser given codeine can reach morphine concentrations that have proved fatal.

• Prodrugs on CYP2D6: codeine to morphine, tramadol to O-desmethyltramadol, tamoxifen to endoxifen.
• Prodrug on CYP2C19: clopidogrel to its active thiol metabolite.
• Prodrug on CYP2C9: losartan to its more potent carboxylic acid metabolite E-3174.
• Prodrug on CYP3A4: a great many, including several statin lactones and some antivirals.
• Reading rule: before predicting the direction of an interaction, ask whether the administered molecule or its metabolite is the active one. The direction flips.

Sources: Flockhart DA 2021 · U.S. Food 2023 · Gasche Y 2004 · Koren G 2006 · Mega JL 2009 · Clinical Pharmacogenetics Implementation Consortium (Crews KR 2021*

Intestinal versus hepatic CYP3A4: why route of administration changes everything human data

CYP3A4 is expressed in two places that matter, and they are not interchangeable. Enterocytes lining the small intestine carry a substantial CYP3A complement — CYP3A4 is the dominant isoform of the intestinal CYP pie — and an orally administered substrate must traverse that layer before it reaches the portal vein, then pass the liver before it reaches the systemic circulation. Some drugs lose most of their dose to this double first pass; oral midazolam bioavailability is limited by both gut and hepatic CYP3A, and the gut contribution is large. The consequences are specific. An inhibitor that acts only in the gut lumen — grapefruit juice being the example, since its furanocoumarins are destroyed or diluted before reaching the liver in meaningful concentration — raises the oral bioavailability of a high-first-pass substrate dramatically while barely changing the clearance of the same drug given intravenously. That is why the grapefruit effect is enormous for oral felodipine and simvastatin, why the same juice does essentially nothing to an intravenous dose, and why the size of the effect depends on how much of the dose was being destroyed in the gut in the first place. Route also decides whether the interaction exists at all: inhaled or sublingual cannabinoid bypasses intestinal CYP3A4 and hepatic first pass, so an oral cannabinoid and an inhaled one are not the same pharmacokinetic object even at the same delivered dose.

• Big oral first pass plus a gut inhibitor equals a large AUC increase. Small first pass means little room to move.
• Intravenous dosing removes the intestinal component and shrinks or abolishes gut-limited interactions.
• Grapefruit juice raised simvastatin AUC roughly sixteen-fold in one high-exposure protocol; pravastatin, rosuvastatin and fluvastatin are not primarily CYP3A4 substrates and are the usual way round it.
• The same logic explains why oral THC and inhaled THC have different metabolite ratios: oral dosing generates much more 11-hydroxy-THC on first pass.

Sources: Paine MF 2006 · Thummel KE 1996 · Bailey DG 2013 · Lilja JJ 1998 · Huestis MA 2007

Pharmacogenomic polymorphism: the same pair, opposite outcomes in different people contested human data

Three of the six isoforms on this shelf are strongly polymorphic in a way that changes clinical outcome: CYP2D6, CYP2C19 and CYP2C9. CYP2D6 is the most dramatic — it has null alleles, reduced-function alleles, and gene duplications, so the population spans poor metabolisers with no functional enzyme through to ultra-rapid metabolisers with extra gene copies, and the span of exposure for a CYP2D6 substrate across that range is more than an order of magnitude. CYP2C19 has common loss-of-function alleles (notably the star-2 and star-3 variants, with poor-metaboliser frequencies far higher in East and South-East Asian populations than in European ones) and a common increased-function promoter variant. CYP2C9 reduced-function alleles lower warfarin dose requirement and raise bleeding risk. The consequence for an interaction table is uncomfortable but has to be said plainly: the direction of an interaction is a property of the mechanism, but its magnitude is a property of the individual. Adding a strong CYP2D6 inhibitor to a poor metaboliser changes very little, because there was no functional enzyme to inhibit; adding it to a normal metaboliser converts them phenotypically into a poor metaboliser, a phenomenon called phenoconversion; adding it to an ultra-rapid metaboliser may merely normalise them. The same pair of substances, with three different outcomes, in three people with the same prescription.

Contested — caveat. Allele frequencies vary substantially between populations and the older literature reports them by crude continental groupings that do not map cleanly onto individuals. Genotype predicts phenotype imperfectly: concurrent inhibitors, inflammation, liver disease and age all shift measured activity away from what the genotype predicts. Treat the table as the direction of an effect, not as a prediction for a person.

Sources: Zanger UM 2013 · Desta Z 2002 · Mega JL 2009 · Aithal GP 1999 · International Warfarin Pharmacogenetics Consortium (Klein TE 2009* · Kuehl P 2001 · Wang D 2011 · Gunes A 2008 · Sachse-Seeboth C 2009

Narrow therapeutic index: where a small shift becomes an event human data

An interaction matters in proportion to how little room the substrate has. For a drug with a wide margin between the concentration that works and the concentration that harms, a two-fold change in exposure is usually absorbed without consequence. For a narrow-therapeutic-index drug, the therapeutic and toxic ranges nearly touch, and a modest change in clearance is a clinical event. This is why the NTI column on the substrate tables in this shelf is the most important column on the page. Warfarin, phenytoin, digoxin, ciclosporin, tacrolimus, lithium, theophylline, carbamazepine, levothyroxine, clozapine and the antiarrhythmics are the recurring names. Two of them add a second trap: phenytoin has saturable, non-linear kinetics, so near the top of its range a small decrease in clearance produces a disproportionately large rise in level, and digoxin becomes more toxic at an unchanged blood level when serum potassium falls, because potassium and digoxin compete for the same site on the sodium-potassium ATPase. The second case is worth holding onto because it breaks the habit of treating the blood level as the whole story.

• NTI plus an inhibitor is the highest-consequence combination in this shelf.
• Phenytoin: saturable kinetics mean the dose-level relationship is a curve, not a line.
• Digoxin: hypokalaemia raises toxicity without raising the level — anything that lowers potassium raises digoxin risk.
• Ciclosporin and tacrolimus: both directions are catastrophic. Inhibition gives nephrotoxicity, induction gives graft rejection.

Sources: Holbrook AM 2005 · Flockhart DA 2021 · U.S. Food 2023 · Ruschitzka F 2000 · Roden DM 2004

Fractional contribution (fm): why an inhibitor only matters in proportion human data

The most useful single quantitative idea in interaction prediction is the fraction of a substrate total clearance that runs through the affected enzyme, conventionally written fm. Complete inhibition of a pathway that carries 90 percent of clearance leaves 10 percent, a ten-fold rise in exposure. Complete inhibition of a pathway carrying 30 percent of clearance leaves 70 percent, a rise of about 1.4-fold — the same strong inhibitor, the same enzyme, a clinically trivial result. The general relation, for a competitive inhibitor, is that the ratio of exposure with and without the inhibitor equals one divided by the quantity fm divided by one plus the inhibitor concentration over its inhibition constant, plus the remainder one minus fm. Two consequences follow, and they are the two facts that most interaction lists obscure. First, a strong inhibitor is only strong against a high-fm substrate: strong and moderate potency classes are defined by what an inhibitor did to a sensitive index substrate, so calling an inhibitor strong describes the inhibitor, not the outcome for your particular drug. Second, redundancy protects: a drug cleared by three CYP isoforms and glucuronidation in roughly equal share is nearly interaction-proof on any one of them, whereas a drug cleared almost entirely by CYP3A4 is hostage to it. When a substrate table in this shelf calls something a sensitive index substrate, that is the fm statement in disguise — it means fm for that pathway is high.

• High fm plus strong inhibitor equals a large, clinically meaningful change.
• Low fm plus strong inhibitor equals a measurable but usually unimportant change.
• Potency class describes the inhibitor tested against a sensitive substrate; it does not transfer unchanged to every substrate of that enzyme.
• Multiple parallel routes of clearance are the pharmacokinetic equivalent of redundancy in engineering.
• Renally cleared drugs have low fm on every CYP: lithium is not a CYP interaction at all, it is a sodium, water, NSAID and diuretic interaction.

Sources: Ito K 1998 · U.S. Food 2020* · Zanger UM 2013

Why a mechanism table generalises where a pairs list cannot human data

A pairs list answers the question it was typed to answer and nothing else. CYP3A4 inhibition, stated once as a mechanism, explains grapefruit, ketoconazole, itraconazole, ritonavir, clarithromycin, verapamil, and the piperine in black pepper, against every CYP3A4 substrate at once. The equivalent pairs list needs one row for every inhibitor multiplied by every substrate, and the rows nobody typed are invisible — a clean result from a pairs list is indistinguishable from a gap in it. That asymmetry is the whole argument for the architecture used here and in the live checker. It matters most exactly where the commercial databases are thinnest: there is no product label for a kava beverage, no monograph for a novel cannabinoid, no FDA-approved comparator for most of the botanicals in this library. A mechanism rule reaches them anyway, because the question it asks is not what has this pair been reported to do but what enzyme does this substance act on and what else uses that enzyme. The cost of the mechanism approach is honesty about magnitude: it tells you an interaction exists and in which direction, and it is much weaker at telling you how large it will be in a specific person. That is why the potency class, the evidence grade and the contested markers are carried on every row rather than averaged away.

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

What this reference is, and what it is not human data

This is a research reference. It is not a clearance, not an approval, not a prescription, and not a substitute for a pharmacist or a physician who can see the whole medication list, the liver and kidney function, the genotype if it is known, and the person. Nothing on these pages tells anyone to take or to stop taking anything. Read the absence of a substance from a table as exactly what it is: this dataset does not contain a documented mechanism for it. That is not a finding of safety. It can mean the interaction is not documented, that it is documented in a source not yet incorporated, that the substance has never been studied — which is the normal condition for most plant preparations and for every novel cannabinoid — or that the product in hand is not reliably what its label says, which is a separate and common failure mode that no interaction table can detect. The corresponding caution runs the other way too: a documented mechanism is not a prediction of harm in a given person, and marking a row contested is not a hedge, it is the actual state of the evidence.

• Absence from a table is absence of documentation, never a clearance.
• A mechanism gives direction reliably and magnitude poorly.
• Product identity and standardisation are prerequisites this dataset cannot verify. An unstandardised botanical extract has an unknown dose of an active constituent.
• Education and harm reduction. No diagnosis, no prescribing, no individualised advice.

Sources: U.S. Food 2023 · Flockhart DA 2021 · Teschke R 2011