Δ⁹-Tetrahydrocannabinol (THC)

THC is a partial agonist at the CB1 and CB2 cannabinoid receptors, which the body's own cannabinoids act on. CB1 receptors are concentrated in the brain, which is where the psychoactive effects come from.

Oral THC is almost completely absorbed, but the liver removes most of it before it reaches the bloodstream, so only 10 to 20% of an oral dose gets there. Concentrations peak between 30 minutes and 4 hours. The first phase of elimination has a half-life of about 4 hours, but THC is fat soluble and leaves fat slowly, so the final phase takes 25 to 36 hours [1]. Inhaled THC reaches the blood without passing through the liver first, so onset is quicker and less of the dose is lost.

Which enzymes do this has been measured in liver tissue rather than assumed. In human liver microsomes and with individual recombinant enzymes, THC is converted to at least four metabolites: CYP2C9 and CYP2C19 both form 11-hydroxy-THC, while CYP3A4 and CYP3A5 make the other three [2]. A modeling study of pooled human liver microsomes put the CYP2C9 share of THC clearance at 91%, with CYP2D6 accounting for the rest [3]. Estimates of the CYP2C9 share range from about 70% upward, partly because a liver protein that binds fatty acids, FABP1, shifts how much each enzyme contributes [2].

CYP2C9 is a liver enzyme that also clears warfarin, phenytoin and several anti-inflammatories. Two variants account for most reduced function: CYP2C9*2 (rs1799853, p.Arg144Cys) and CYP2C9*3 (rs1057910, p.Ile359Leu). CPIC gives *2 an activity value of 0.5 for decreased function and *3 a value of 0 for no function [4]. The values add up to the activity score Gene2Rx reports: 2 for *1/*1, 1.5 for *1/*2, 1 for *1/*3 or *2/*2, 0.5 for *2/*3 and 0 for *3/*3. In 1000 Genomes data, *3 is on 7.3% of gene copies in Europeans, 10.9% in South Asians and 0.2% in Africans [5], and *2 on 12.4% in Europeans and under 4% outside Europe and the Americas [6]. Reduced-function results are therefore not rare in Europeans, while in people of African ancestry other, less commonly tested variants matter more.

What the laboratory and human data show

In expressed enzymes, the intrinsic clearance of THC by the CYP2C9.2 and CYP2C9.3 proteins was about 30% of the normal form [7]. That is a measurement of the enzyme, not an association.

The human pharmacokinetic evidence is one study of 43 healthy volunteers given oral THC. CYP2C9*3/*3 homozygotes had a median area under the curve three times that of *1/*1, and 70% less of the inactive carboxy metabolite, which is what a slowed first step would produce. There was a trend toward more sedation in *3 carriers, and no difference by *2 status [8]. The dronabinol label puts it more cautiously: published data indicate potentially 2 to 3 times higher exposure with diminished CYP2C9 function, and section 8.6 advises monitoring those patients for increased adverse reactions [1]. Dronabinol with CYP2C9 appears in the FDA's table of pharmacogenomic biomarkers in drug labeling [9].

This is the strongest link on the page: a measured enzyme effect, a mechanism, a human study pointing the same way, and a drug label that acts on it. Its weakest point is size, since a handful of *3/*3 volunteers in one trial is thin ground for a precise number. The *2 row also deserves care: that study found no difference for *2, while the activity score counts it as decreased function on the basis of CYP2C9 work with other drugs [4].

Fatty acid amide hydrolase breaks down anandamide, one of the body's own cannabinoids. It does not metabolize THC. The reason for reporting it is that someone whose own cannabinoid signaling runs higher may respond differently to an outside cannabinoid. Gene2Rx reads rs324420, a C to A change at position 385 that swaps a conserved proline for threonine (p.Pro129Thr, often written C385A). The A allele is on 36.8% of gene copies in Africans, 35.2% in admixed Americans, 21.1% in Europeans, 19.5% in South Asians and 17.6% in East Asians [10].

The functional work is solid. The variant protein is catalytically normal but is broken down faster inside the cell [11], and lymphocytes from homozygotes had less than half the usual FAAH protein and activity [12]. Brain imaging with a tracer that binds FAAH found 23% lower binding in A-allele carriers than in C/C homozygotes, in 24 participants [13]. A mouse carrying the human variant showed the same reduced expression, with parallel brain findings in human carriers [14].

The weak link is everything after that. The original report associated the variant, in homozygous form, with problem drug use [11], and a case-control study of 249 people with multiple drug addictions supported it [15]. For cannabis specifically, the findings run the other way. Among 749 adults of European ancestry, A/A carriers were less likely to be THC dependent than C/C or C/A carriers (11% against 26%; odds ratio 0.25 after controlling for nicotine and sedative use) [16]. In 105 daily cannabis smokers, the C/C group reported the larger increase in craving after five days of abstinence [17]. And in 720 people, rs324420 was not associated with psychotic-like or euphoric experiences from cannabis [18]. So the variant lowers FAAH activity, but the claim that it intensifies reward or dependence risk is not supported by the cannabis studies published so far.

AKT1 is a signaling kinase that works downstream of dopamine D2 receptors. The variant Gene2Rx reads, rs2494732, sits in an intron, so it does not change the protein, and what it does to AKT1 activity in people has not been established. The C allele is on 73.4% of gene copies in East Asians, 62.5% in South Asians, 61.0% in admixed Americans, 49.9% in Africans and 43.8% in Europeans [19], so roughly a fifth of Europeans and over half of East Asians are C/C, the genotype the report labels Poor Function.

The human evidence is a set of gene-by-environment studies in psychosis. In 740 unaffected siblings of 801 patients with psychosis, rs2494732 was one of three variants out of 152 tested that survived correction for multiple testing in an interaction with recent cannabis use, and C/C individuals who had used cannabis had about twice the odds of a psychotic disorder diagnosis [20]. A case-control study of 489 first-episode psychosis patients and 278 controls found the variant was not associated with psychosis or cannabis use on its own, but modified the effect of use: among people who had ever used, C/C carriers had 2.18 times the odds of a psychotic disorder compared with T/T, and among daily users the odds ratio was 7.23 (95% confidence interval 1.37 to 38.12) [21]. In 442 healthy young people tested while intoxicated with their own cannabis and again when drug free, psychotomimetic symptoms rose with the number of C alleles [22].

Three designs pointing the same way is a real signal, and the acute-intoxication study is the most useful, because its participants were not psychiatric patients. The weaknesses are serious all the same: the variant has no known biological function, the odds ratio for daily users rests on small subgroups, and a later study of 720 people found no association with cannabis-induced psychotic-like experiences [18].

The exposure matters more than the genotype. Daily cannabis use was associated with 3.2 times the odds of a psychotic disorder in an 11-site European study, rising to 4.8 times for daily use of cannabis with 10% or more THC [23]. That risk tracks frequency and potency in everyone.

Catechol-O-methyltransferase clears dopamine, and it matters most in the prefrontal cortex, where other routes are sparse. The rs4680 variant replaces valine at position 158 with methionine. In 108 postmortem prefrontal cortex samples, Val158Met affected protein abundance and enzyme activity, making Val the main determinant of higher COMT activity and so of lower synaptic dopamine [24]. The Met allele is on 50.0% of gene copies in Europeans, 44.1% in South Asians, 28.1% in Africans and 28.0% in East Asians [25]. Gene2Rx labels Val/Val Normal Function, Val/Met Intermediate Function and Met/Met Decreased Function.

Two early studies put Val carriers at higher risk. In a birth cohort followed to adulthood, Val-allele carriers who used cannabis in adolescence were more likely to show psychotic symptoms and schizophreniform disorder, while use had no such effect in people with two Met copies [26]. In a placebo-controlled crossover experiment in 74 people, Val carriers were more sensitive to THC-induced psychotic experiences, though only among those who already showed psychosis liability, and to memory and attention impairment [27].

The follow-up literature does not hold together. A case-only study of 493 people with schizophrenia found no interaction between cannabis use and COMT genotype [28]. A meta-analysis of 13 studies found a significant interaction only in case-only designs, which the authors noted have lower validity and tend to overstate effects; there the risk genotype was Met/Met rather than Val/Val, with an odds ratio of 1.45, and the authors concluded the evidence remains unconvincing [29]. A pooled analysis of three placebo-controlled imaging studies found the strongest dysphoric reactions to THC in Met/Met participants [30], again the opposite direction, and the study of 720 cannabis users found no COMT association [18].

This is the weakest chain on the page. The enzyme effect of rs4680 is well established; its direction of effect on reactions to THC is not. The report places the adverse-reaction flag on Val/Val, following the two early studies, and the later literature does not support that choice confidently.

Gene2Rx wrote this guideline from the research above, and the recommendations are exploratory. They are written for a prescriber considering a cannabinoid medicine, with the strength the report assigns each row. Nothing here is advice about using cannabis, and no genotype makes cannabis use safe. Do not start, stop or change any medicine on the basis of a genetic result without talking to your prescriber.

The activity score 0 row carries the report's strongest wording. Its direction is well founded, but the exact 25% figure is a judgment rather than a measured threshold.

One A allele is enough for a Decreased Function result here. The functional part of this row is well supported; the dependence and craving part is contradicted by the two cannabis studies cited above [16][17].

These labels describe an association with reactions, not a measured loss of enzyme function, because rs2494732 is intronic. A strength of strong also sits uncomfortably with a finding that one later study did not reproduce [18].

These rows follow the two early studies, so the Decreased Function row should not be read as protective [29][30].

The CYP2C9 part of this guideline would be unremarkable for any other drug: one enzyme does most of the clearance, two common variants reduce its activity, and the label for the prescription form of the same molecule tells prescribers to monitor people with reduced CYP2C9 function [1][9]. THC looks unusual only because most people exposed to it are not holding a prescription. The pharmacology does not change with the setting: a poor metabolizer eating an edible is taking an oral dose of a CYP2C9 substrate, and the same first-pass step is the slow one.

The other three genes are there for a different reason. People vary a great deal in how cannabis affects them, and these candidate genes have been studied for that variation for twenty years. Reporting them is defensible only alongside what the studies support: a documented change in FAAH activity whose consequences for cannabis use are unresolved, an AKT1 interaction found in three studies and absent in a fourth, and a COMT interaction whose direction flips between studies. The report's own text for AKT1 and COMT calls the findings inconclusive, though the strength values on those rows are higher than the evidence warrants.

The practical case for including them is narrow. These three results do not tell anyone whether to use cannabis. They can add weight to a conversation, particularly for someone young, with a family history of psychosis, or considering a cannabinoid medicine alongside other prescriptions. The larger and better measured risk factors are frequency of use, potency and starting in adolescence [23].

If a cannabinoid medicine is being considered, the CYP2C9 result is the part worth taking to your prescriber, along with a list of everything else you take [1].

Other drugs can produce the same effect as a reduced-function genotype, which is called phenoconversion. The dronabinol label names amiodarone and fluconazole as CYP2C9 inhibitors and ketoconazole, itraconazole, clarithromycin, ritonavir, erythromycin and grapefruit juice as CYP3A4 inhibitors, and advises monitoring when they are taken together; inducers of the same enzymes may reduce exposure instead [1]. A normal metabolizer taking fluconazole may clear THC more like an intermediate metabolizer.

Two interaction risks have nothing to do with genotype and come up more often. Taking THC with alcohol, benzodiazepines, barbiturates, opioids, antihistamines or other sedating drugs adds to the dizziness, confusion and drowsiness, and taking it with drugs that affect the circulation, including tricyclic antidepressants, can produce additive effects on blood pressure and heart rate, with reports of fainting and fast heart rate [1]. The label also advises weighing the risk before prescribing dronabinol to someone with a history of seizures, and avoiding it in people with a psychiatric history, because it has been reported to worsen mania, depression and schizophrenia [1].

Interaction can run the other way too. In human liver microsomes and expressed CYP2C9, THC and both of its main metabolites increased the rate of phenytoin hydroxylation, which suggests the combination could lower phenytoin concentrations [7]. That is an in vitro finding, not a clinical study, but phenytoin has a narrow margin for error, so it is worth raising with a prescriber.

Non-genetic factors shape the outcome as well: the route and amount, the potency of the product, its cannabidiol content, tolerance, age, liver function and other medicines.

Gene2Rx reads the variant calls in an uploaded file. For CYP2C9 it assesses a defined set of star alleles, so a rare reduced-function variant outside that set is not seen, and a Normal Metabolizer result means only that none of the assessed variants was found among the positions the data covers. Consumer arrays (23andMe, AncestryDNA, MyHeritage) read fixed positions and are not clinical-grade tests; their raw data is labeled not for medical use. Array processing infers some variants by imputation, but a variant at a position the data does not cover cannot be seen.

For the other three genes one position is read in each: rs324420, rs2494732 and rs4680. Each result describes that position and nothing else about the gene. None of the three is a test for psychosis risk, for cannabis use disorder or for how cannabis will affect you. Psychotic illness, panic reactions, cyclic vomiting, dependence and cognitive effects occur in people with every one of these genotypes. A normal result across all four genes is not a reason to use cannabis or to use more of it.

If a cannabinoid medicine is part of a treatment plan and the CYP2C9 result would change the decision, a clinical-grade pharmacogenetic test ordered through your clinician can confirm it.

Does a CYP2C9 poor metabolizer result mean edibles hit me harder?

It means more THC is likely to reach your bloodstream from a swallowed dose and to stay there longer, because the liver step that removes most of it first is slower [8][1]. The research measured concentrations rather than experiences.

My AKT1 result says Poor Function. Am I going to develop psychosis?

No, and the result should not be read that way. The variant was not associated with psychotic disorder by itself, only with the effect of cannabis use, and a later study found no association at all [21][18]. If psychosis risk concerns you, frequency of use, potency and family history are what to discuss with a clinician [19].

Does this report cover CBD?

No. The guideline is about THC. Cannabidiol is handled by a partly different set of enzymes and is not assessed.

Why does my report give a dose percentage for a substance I am not prescribed?

The guideline is written for prescribers considering a cannabinoid medicine such as dronabinol, where there is a defined dose to adjust. Gene2Rx does not recommend using cannabis, and the percentages are not guidance for non-medical use.

1. Dronabinol capsules: prescribing information. DailyMed. Pharmacokinetics, drug interactions, and sections 8.6 and 12.5 on CYP2C9.
2. Yabut KCB et al. CYP2C9, CYP3A and CYP2C19 metabolize Δ9-tetrahydrocannabinol to multiple metabolites but metabolism is affected by human liver fatty acid binding protein (FABP1). Biochem Pharmacol (2024). Which enzymes form which metabolites.
3. Patilea-Vrana GI, Unadkat JD. Quantifying hepatic enzyme kinetics of THC and its psychoactive metabolite 11-OH-THC through in vitro modeling. Drug Metab Dispos (2019). Fraction of THC clearance by CYP2C9.
4. Theken KN et al. CPIC guideline for CYP2C9 and nonsteroidal anti-inflammatory drugs. Clin Pharmacol Ther (2020). The activity-score system and the function assignments for *2 and *3.
5. dbSNP: rs1057910 (CYP2C9*3). 1000 Genomes allele frequencies.
6. dbSNP: rs1799853 (CYP2C9*2). 1000 Genomes allele frequencies.
7. Bland TM et al. CYP2C-catalyzed delta9-tetrahydrocannabinol metabolism: kinetics, pharmacogenetics and interaction with phenytoin. Biochem Pharmacol (2005). Intrinsic clearance of the variant proteins, and the phenytoin finding.
8. Sachse-Seeboth C et al. Interindividual variation in the pharmacokinetics of delta9-tetrahydrocannabinol as related to genetic polymorphisms in CYP2C9. Clin Pharmacol Ther (2009). Oral THC in 43 volunteers.
9. FDA table of pharmacogenomic biomarkers in drug labeling. Lists dronabinol with CYP2C9.
10. dbSNP: rs324420 (FAAH). 1000 Genomes allele frequencies.
11. Sipe JC et al. A missense mutation in human fatty acid amide hydrolase associated with problem drug use. Proc Natl Acad Sci U S A (2002). The P129T variant and its instability.
12. Chiang KP et al. Reduced cellular expression and activity of the P129T mutant of human fatty acid amide hydrolase. Hum Mol Genet (2004). Protein and activity in cells from homozygotes.
13. Boileau I et al. The fatty acid amide hydrolase C385A variant affects brain binding of the PET tracer [11C]CURB. J Cereb Blood Flow Metab (2015). Brain FAAH binding by genotype.
14. Dincheva I et al. FAAH genetic variation enhances fronto-amygdala function in mouse and human. Nat Commun (2015). Knock-in mouse and parallel human findings.
15. Flanagan JM et al. The fatty acid amide hydrolase 385 A/A (P129T) variant: haplotype analysis and validation of risk for drug addiction. Hum Genet (2006). Case-control support for the addiction association.
16. Tyndale RF et al. The fatty acid amide hydrolase C385A (P129T) missense variant in cannabis users. Am J Med Genet B (2007). THC dependence by genotype.
17. Haughey HM et al. Marijuana withdrawal and craving: influence of the CNR1 and FAAH genes. Addiction (2008). Craving after abstinence by genotype.
18. Hindocha C et al. Do AKT1, COMT and FAAH influence reports of acute cannabis intoxication experiences in patients with first episode psychosis, controls and young adult cannabis users? Transl Psychiatry (2020). Null findings for all three genes in 720 people.
19. dbSNP: rs2494732 (AKT1). Intron variant; 1000 Genomes allele frequencies.
20. van Winkel R. Family-based analysis of genetic variation underlying psychosis-inducing effects of cannabis. Arch Gen Psychiatry (2011). Sibling and proband analysis of the interaction.
21. Di Forti M et al. Confirmation that the AKT1 (rs2494732) genotype influences the risk of psychosis in cannabis users. Biol Psychiatry (2012). Odds ratios in ever-users and daily users.
22. Morgan CJA et al. AKT1 genotype moderates the acute psychotomimetic effects of naturalistically smoked cannabis. Transl Psychiatry (2016). Acute symptoms by C-allele count.
23. Di Forti M et al. The contribution of cannabis use to variation in the incidence of psychotic disorder across Europe (EU-GEI). Lancet Psychiatry (2019). Odds ratios for daily and high-potency use.
24. Chen J et al. Functional analysis of genetic variation in catechol-O-methyltransferase (COMT). Am J Hum Genet (2004). Val158Met effects on protein and activity in human brain.
25. dbSNP: rs4680 (COMT). 1000 Genomes allele frequencies.
26. Caspi A et al. Moderation of the effect of adolescent-onset cannabis use on adult psychosis by a functional polymorphism in the COMT gene. Biol Psychiatry (2005). The birth-cohort interaction finding.
27. Henquet C et al. An experimental study of COMT Val158Met moderation of delta-9-THC-induced effects on psychosis and cognition. Neuropsychopharmacology (2006). Placebo-controlled crossover study.
28. Zammit S et al. Genotype effects of CHRNA7, CNR1 and COMT in schizophrenia: interactions with tobacco and cannabis use. Br J Psychiatry (2007). No cannabis by COMT interaction.
29. Vaessen TSJ et al. The interaction between cannabis use and the Val158Met polymorphism of the COMT gene in psychosis: a transdiagnostic meta-analysis. PLoS One (2018). Interaction significant only in case-only designs.
30. Pelgrim TAD et al. Acute effects of Δ9-THC on resting state connectivity networks and impact of COMT genotype. Drug Alcohol Depend (2023). Subjective and connectivity effects by genotype.