Two Molecules, Two Routes: How the Body Handles THC and CBD Differently

THC is oxidized principally by CYP2C9, with a contribution from CYP3A4, to 11-hydroxy-THC, which is itself psychoactive, then further oxidized to the inactive acid 11-nor-9-carboxy-THC. CBD takes a different route: human liver microsome studies identify CYP3A4 and CYP2C19 as the dominant enzymes, producing a family of monohydroxylated products rather than one principal active metabolite.

The receptor picture diverges even more sharply. THC is a partial agonist at CB1 and CB2 receptors. CBD has low direct affinity for either, and a comprehensive systematic review of its molecular pharmacology concluded it is very unlikely to exert its effects in neurological disease through the endocannabinoid system at all.

Product marketing and a good deal of clinical shorthand treat THC and CBD as interchangeable ingredients on a label. The metabolism says otherwise: they load different enzymes, carry different interaction risks, and generate different downstream chemistry.

The mechanism literature for CBD in particular is enormous and mostly in vitro. Separating the handful of findings with plausible human relevance from the many that occur only at concentrations no oral dose produces is the difference between informed practice and citation theater.

The THC pathway was mapped first and is the simpler of the two. Work by Wall and colleagues, published in Clinical Pharmacology and Therapeutics in 1983, described microsomal hydroxylation converting THC to 11-hydroxy-THC, which the authors called both a key intermediate and a potent psychoactive metabolite in its own right. Liver alcohol dehydrogenase enzymes then oxidize it further to 11-nor-9-carboxy-THC, and from there the final products are that acid and related, more polar acids.

The CBD pathway is broader and messier. Jiang and colleagues, in Life Sciences in 2011, incubated CBD with pooled human liver microsomes and identified eight monohydroxylated metabolites. Four of them predominated: 6-alpha-hydroxy, 6-beta-hydroxy, 7-hydroxy, and 4-double-prime-hydroxy CBD. Of 14 recombinant human cytochrome P450 enzymes tested, seven metabolized CBD: CYP1A1, CYP1A2, CYP2C9, CYP2C19, CYP2D6, CYP3A4, and CYP3A5.

Correlation and inhibition experiments narrowed that to two principal contributors. CYP3A4 was chiefly responsible for forming 6-beta-hydroxy and 4-double-prime-hydroxy CBD, and CYP2C19 predominantly catalyzed 7-hydroxylation, the step that produces 7-hydroxy-CBD. THC’s pathway runs mainly through CYP2C9; CBD’s runs mainly through CYP3A4 and CYP2C19. They share a liver, not a route.

For THC there is a clean human demonstration that the enzyme identification has consequences. Sachse-Seeboth and colleagues, in Clinical Pharmacology and Therapeutics in 2009, gave oral THC to 43 healthy volunteers genotyped for CYP2C9. People carrying two copies of the reduced-function CYP2C9*3 allele had a median THC area under the curve three times higher than people with two normal-function copies, and a 70 percent lower area under the curve for 11-nor-9-carboxy-THC. The CYP2C9*2 allele made no measurable difference.

Those numbers behave exactly as the enzymology predicts and are the best available evidence that individual variation in cannabinoid response has a partly genetic basis.

No comparable controlled study exists for CBD. A 2025 retrospective chart review in the Journal of Cannabis Research examined 71 patients who used oral cannabis and had undergone pharmacogenomic testing, finding atypical CYP2C19 variants in 37 of them and CYP3A4 variants in 6. The authors stated plainly that the absence of published clinical outcomes in this field made it difficult to estimate the clinical significance of what they found. That is the correct level of confidence to carry.

Pertwee’s review in the British Journal of Pharmacology in 2008 laid out the comparison directly. THC behaves as a partial agonist at both CB1 and CB2 receptors, and its effects are strongly shaped by how many receptors are expressed, how efficiently they signal, and how much endogenous cannabinoid is already being released. That is classical receptor pharmacology and it explains a good deal of the variability patients report.

CBD behaves nothing like that. In the same review, CBD displayed unexpectedly high potency as an antagonist of CB1 and CB2 receptor agonists in cells and tissues expressing those receptors, which is a different relationship than simply binding weakly. Its low direct affinity combined with meaningful functional antagonism has driven a great deal of subsequent work on allosteric and indirect mechanisms.

The most rigorous attempt to sort out CBD’s targets is the systematic review by Ibeas Bih and colleagues in Neurotherapeutics in 2015. They searched the literature for original reports of CBD’s molecular pharmacology and appraised each proposed target for validity. More than 65 discrete molecular targets had been reported. After excluding those demonstrated only at concentrations difficult to reach in vivo, and those asserted through experimental designs showing correlation rather than causation, a relatively limited number remained plausible, and the survivors clustered around regulation of and response to intracellular calcium, including voltage-dependent anion channel 1, GPR55, and CaV3 channels.

The 65-target problem is worth stating plainly, because it is the single most common source of confident but unsupportable claims about CBD. A compound tested at high micromolar concentrations against a panel of receptors will hit many of them. That does not mean a patient taking an oral dose reaches those concentrations at those targets.

The Ibeas Bih review made this its central methodological point. Some CBD effects at proposed targets manifest only at high concentrations that may be difficult to achieve in vivo, particularly given CBD’s relatively poor bioavailability. Several targets were asserted through designs demonstrating only correlation with a given target rather than causal proof. And when the plausible targets were then evaluated for whether they were relevant to the neurological diseases in question, some had little or no established link to those conditions.

The bioavailability point is not a technicality. A systematic review by Millar and colleagues in Frontiers in Pharmacology in 2018 retrieved 792 articles and found only 24 reporting human pharmacokinetic parameters for CBD. Reported half-life ranged from 1.4 to 10.9 hours after oromucosal spray, 2 to 5 days after chronic oral dosing, 24 hours after intravenous administration, and 31 hours after smoking. Bioavailability following smoking was 31 percent, and no study had attempted to establish absolute oral bioavailability in humans despite intravenous formulations being available. That is a thin foundation under a very large mechanistic literature.

Amid a great deal of in vitro speculation, one CBD interaction has been documented in patients with clear numbers. Geffrey and colleagues, in Epilepsia in 2015, reported on 13 children with refractory epilepsy taking clobazam who were given CBD under an expanded access protocol. Clobazam levels rose by a mean of 60 percent with a standard deviation of 80, and levels of the active metabolite norclobazam rose by a mean of 500 percent with a standard deviation of 300, at four weeks.

Side effects occurred in 10 of the 13 children and were relieved by reducing the clobazam dose. Nine of 13 had a greater than 50 percent reduction in seizures, and clobazam doses were reduced in 10 of the 13 over the course of treatment. The authors’ conclusion was operational: monitoring of clobazam and norclobazam levels is necessary when the two drugs are used together.

Phase II metabolism adds a second interaction layer. Nasrin and colleagues, in Drug Metabolism and Disposition in 2021, found that CBD inhibited several UDP-glucuronosyltransferase enzymes with binding-corrected IC50 values as low as 0.12 micromolar against UGT1A9, and that THC and cannabinol also strongly inhibited UGT1A9 and UGT2B7. Those are laboratory measurements rather than patient outcomes, and they point in the same direction as the clobazam data: CBD is a more consequential perpetrator of interactions than THC.

Popular summaries of cannabinoid biochemistry tend to include a tidy table of neurotransmitter and hormone effects: THC raises dopamine, CBD modulates serotonin, THC suppresses testosterone, CBD acts on estrogen receptors. The underlying evidence is far less tidy.

Take dopamine, the best studied of these. Bloomfield and colleagues reviewed the evidence in Nature in 2016 and described conflicting evidence from human and animal studies, complicated by the complexity of cannabinoid and dopamine interactions. Their summary was that acute THC administration causes increased dopamine release and neuron activity while long-term use is associated with blunting of the dopamine system. That is a time-dependent, bidirectional picture, not a simple arrow.

For hormonal effects the situation is worse. The claims that circulate generally trace to small human studies from the 1970s or to animal and cell work, and have not been replicated in modern controlled human designs. This article does not reproduce them. A table that assigns a direction and a dose threshold to each hormone receptor implies a level of knowledge the literature does not contain.

The enzyme identification is strong on both sides. Recombinant enzyme panels combined with isoform-selective inhibitors, blocking antibodies, and correlation analysis across individual human liver microsome preparations is the accepted method, and independent laboratories agree on the principal assignments.

The human pharmacokinetic base is much thinner for CBD than for THC. A systematic review found only 24 human pharmacokinetic reports and noted discrepancies among them, and absolute oral bioavailability has never been measured in people.

The mechanistic target literature for CBD is the weakest tier despite being the largest in volume. When it was systematically appraised, most of the more than 65 reported targets did not survive scrutiny for concentration plausibility or causal design.

Human liver microsomes are a preparation, not a liver. They lack transporters, competing endogenous substrates, and the concentration gradients of an intact organ, and they tend to overstate how cleanly one isoform dominates a reaction.

The clobazam interaction data came from 13 children in an expanded access program, not a controlled trial, and the standard deviations were larger than the means in one case. The direction of the effect is convincing; the magnitude in any individual patient is not predictable from these numbers.

Comparisons of CBD potency across studies are hazardous because cannabinoids bind nonspecifically to proteins and plasticware. Work that does not correct for that binding reports weaker inhibition than the same chemistry shows when it is corrected, and much of the older literature did not correct.

This literature does not establish a clinical mechanism of action for CBD. It identifies candidate targets and eliminates many of them, which is a narrowing exercise, not an answer.

It does not show that THC and CBD have opposing hormonal effects, or that either has a defined dose threshold for effects on estrogen, testosterone, adrenaline, or acetylcholine signaling. Those claims appear in popular summaries without a supporting body of controlled human evidence, and they are omitted here for that reason.

It does not tell you what a given ratio of THC to CBD in a product will do in a given patient. Metabolic pathway maps describe chemistry, not clinical response.

It does not establish that CYP genotyping improves cannabinoid dosing. The THC association with CYP2C9*3 is real and no trial has tested whether acting on it changes outcomes.

The practical consequence of two distinct routes is that THC and CBD carry different interaction profiles. CBD’s reliance on CYP3A4 and CYP2C19, combined with its inhibitory potency at several phase I and phase II enzymes, makes it the more consequential of the two in a complex medication list, which is the opposite of the intuition many patients bring in.

There is also a lesson about how mechanism literature should be read. CBD accumulated more than 65 proposed molecular targets in roughly two decades, most of which did not survive systematic appraisal. A compound with a plausible story at every target has, in practice, no established story at any of them. The narrowing work matters more than the accumulation.

The most common misconception I correct in clinic is that CBD is the harmless one. On the metabolism, it is the opposite. THC has one main enzyme and a reasonably predictable arc. CBD loads CYP3A4 and CYP2C19, inhibits several enzymes at concentrations that matter, and has a documented interaction with clobazam large enough that dose reduction is routinely required.

That does not make CBD dangerous. It makes it a drug, which is how it should have been treated from the beginning. When a patient on a substantial medication list tells me they started high-dose CBD, that is a medication reconciliation conversation, not a wellness one.

On the receptor side I try to keep patients away from mechanism stories entirely. Nobody knows how CBD produces the effects it produces, and a confident explanation involving serotonin or the endocannabinoid system is not something I can support. What I can do is watch the response and the interactions, which is what practice actually consists of.

THC and CBD share a liver but not a pathway. THC is oxidized principally by CYP2C9 to the active metabolite 11-hydroxy-THC and then to an inactive acid, with a threefold exposure difference between CYP2C9 genotypes. CBD is oxidized principally by CYP3A4 and CYP2C19 into a family of hydroxylated products, and it inhibits several enzymes potently enough to have produced a documented clinical interaction with clobazam. On targets, THC is a CB1 and CB2 partial agonist while CBD is not meaningfully acting through those receptors, and most of the many alternative targets proposed for it do not survive systematic appraisal.

Carry forward the enzyme assignments and the interaction implications, because those are well supported and directly actionable. Carry forward the finding that CBD’s mechanism remains genuinely unresolved. Leave behind any tidy table matching cannabinoids to hormones and neurotransmitters with assigned directions and dose thresholds, because that table describes a level of knowledge that the human literature does not support.