← All educational resources

Caffeine, Alcohol, and Cannabis by Genotype: Why the Same Dose Does Different Things to Different People

2026-11-22

Same post, two depths. Switch any time; the research version has the enzyme names, lab markers, and doses. Written from a research and education perspective: a model of what may be going on and what the published literature suggests would address it if the model is right, not medical advice or a treatment plan.

One person has an espresso after dinner and sleeps fine. Another has a cup at noon and is awake at 2 am. One person flushes red after half a beer; another drinks a bottle of wine and feels nothing until morning. One person relaxes with cannabis; another gets paranoid and panicky from the same joint. A lot of this is genetics, and the specific genes are known, well studied, and present on the raw data file you may already have.

This page covers the three most-used psychoactive substances and the handful of genes that set how each one hits you: how fast it clears, how strongly the receptors respond, and what the long-term risks look like in each group. It is written from a research perspective; none of it is a recommendation to use any of them, and in a few places (alcohol in ALDH2 carriers, cannabis in some COMT and AKT1 carriers) the genetics argue for not using them at all.

Caffeine

The clearance gene: CYP1A2

Caffeine is cleared almost entirely by one liver enzyme, CYP1A2. A common variant (rs762551) splits people into fast metabolizers (AA) and slow metabolizers (AC or CC). In fast metabolizers caffeine's half-life is around four hours; in slow metabolizers it can be eight to twelve. Smoking, cruciferous vegetables, and some medications speed the enzyme up; oral contraceptives, pregnancy, fluvoxamine, and ciprofloxacin slow it down dramatically.

  • Slow metabolizer signature: afternoon coffee ruins sleep, a second cup causes jitters and palpitations, caffeine anxiety, headache if you skip it, and in the Cornelis study, higher heart attack risk with over two cups a day, while fast metabolizers showed a protective association at the same intake.
  • What the research says: slow metabolizers do best under 100 to 200 mg (one modest cup) before noon, or none; fast metabolizers tolerate more and clear it in time for sleep. The anxiety post here and migraine post here both come back to caffeine timing.

The sensitivity gene: ADORA2A

Caffeine works by blocking adenosine receptors, and a variant in the A2A receptor gene (rs5751876) makes some people far more anxious and sleep-disrupted from the same dose, independent of how fast they clear it. People with the TT genotype are over-represented in caffeine-induced anxiety and panic studies. If caffeine makes you anxious even in small doses cleared quickly, this is the likely reason.

Everyone

Caffeine blocks adenosine, which is the sleep-pressure signal, so it does not give energy, it hides tiredness; it raises adrenaline and cortisol; it pulls magnesium and calcium into the urine; and daily use produces withdrawal headache and fatigue within a day of stopping. The cleanest test of what it is doing to you is two weeks off and a single dose on, with a log.

Alcohol

The two enzymes: ADH1B and ALDH2

Alcohol is converted to acetaldehyde (toxic, the thing that makes you flush and feel sick) by alcohol dehydrogenase, then acetaldehyde is cleared by aldehyde dehydrogenase. Two variants set the whole experience:

  • ALDH2*2 (rs671), carried by around 30 to 40 percent of people of East Asian descent and rare elsewhere, cripples acetaldehyde clearance. Carriers flush, get a racing heart, nausea, and headache from small amounts, and, because acetaldehyde is a carcinogen, have a sharply higher risk of esophageal and head-and-neck cancer if they drink through the flush. The research position is unambiguous: this genotype is a reason not to drink, and the flush is a warning, not a quirk.
  • ADH1B*2 (rs1229984) makes the first enzyme fast, so acetaldehyde builds quickly; carriers feel unwell from alcohol sooner and are protected against alcohol use disorder as a result.

The dependence genes

Variants in the GABA receptor (GABRA2), the opioid receptor (OPRM1 A118G, which increases the reward from alcohol and predicts response to naltrexone), dopamine receptors, and others shift the risk of alcohol becoming a problem. None is deterministic; family history remains the strongest predictor.

Everyone

Alcohol is a GABA drug with a glutamate and histamine rebound a few hours later, which is why it helps you fall asleep and wakes you at 3 am (the 3am post here); it is a histamine load (wine especially); it depletes thiamine, magnesium, folate, zinc, and B6; it blocks fat burning while it is being cleared; it raises GGT; and for the mast cell, migraine, POTS, and perimenopause populations on this site it is one of the most consistent symptom triggers. There is no genotype on which alcohol is a health food.

Cannabis

The receptor and breakdown genes: CNR1 and FAAH

THC acts on the CB1 receptor (CNR1), the same receptor used by the body's own cannabinoid anandamide. Anandamide is broken down by the enzyme FAAH, and a common variant (rs324420, C385A) lowers FAAH activity, raising baseline anandamide. People with the A variant tend to have lower anxiety at baseline, extinguish fear faster, and, in several studies, are less likely to enjoy or become dependent on cannabis, possibly because their system is already partly "on."

The psychosis-risk genes: COMT and AKT1

This is the important one. THC raises dopamine in the mesolimbic pathway (the neurotransmitter map here), and the genes that handle that dopamine shape the risk of the worst outcome.

  • COMT Val158Met: in the Dunedin cohort, adolescents with the Val/Val genotype who used cannabis had a substantially higher risk of later psychotic disorder than Met carriers who used the same amount. Later studies have been mixed, but the direction has held in several.
  • AKT1 (rs2494732): the C/C genotype roughly doubled psychosis risk with cannabis use in the GAP study and predicted acute psychotic-like symptoms and cognitive impairment with THC in experimental studies.

Together with adolescent use (the developing brain is the biggest risk factor of all), high-THC products, and a family history of psychosis, these genotypes define who should not use cannabis.

The metabolism genes: CYP2C9 and CYP3A4

THC is cleared mostly by CYP2C9; the 2 and 3 variants (common in people of European descent) slow it, so the same edible lasts longer and hits harder. CBD is cleared by CYP3A4 and CYP2C19, and CBD inhibits both, which is why it interacts with a long list of medications (anti-seizure drugs, blood thinners, some antidepressants and statins) and why "it's just CBD" is not a safety statement.

Everyone

THC impairs memory formation, driving, and adolescent brain development; daily use raises the risk of cannabis use disorder (around 10 percent of users, higher with early and heavy use), cannabis hyperemesis syndrome (cyclic vomiting relieved by hot showers, increasingly common with high-potency products), and a lasting amotivation in some. CBD has trial evidence for specific seizure disorders and some for anxiety at high doses; the rest is mostly unstudied. Cannabis also affects mast cells and the gut in ways that are only beginning to be mapped.

The practical summary

Substance Gene Variant that changes the picture What it means
Caffeine CYP1A2 rs762551 C allele (slow) Half-life doubles; cap at one morning cup or none; cardiovascular risk with higher intake
Caffeine ADORA2A rs5751876 TT Caffeine anxiety and sleep disruption at low doses
Alcohol ALDH2 rs671 A allele Flush, acetaldehyde buildup, esophageal cancer risk; the genotype argues for abstinence
Alcohol ADH1B rs1229984 A allele Faster acetaldehyde; unpleasant early, protective against dependence
Alcohol OPRM1 rs1799971 G allele More reward from alcohol; better naltrexone response
Cannabis FAAH rs324420 A allele Higher baseline anandamide; less cannabis reward and dependence
Cannabis COMT rs4680 Val/Val Higher psychosis risk with adolescent cannabis use
Cannabis AKT1 rs2494732 CC Higher psychosis risk and acute cognitive impairment with cannabis
Cannabis CYP2C9 2, 3 Slower THC clearance; edibles last longer

The pharmacogenomics here report covers these, and the raw DNA tutorial here shows how to look them up yourself.

How to use this

Not as permission. As a reason to stop treating your reaction to these things as a character flaw or a mystery, and as a way to identify the people for whom the research says the answer is simply no: ALDH2 carriers and alcohol; Val/Val COMT or CC AKT1 adolescents and cannabis; slow CYP1A2 metabolizers and afternoon caffeine if sleep and anxiety matter to you.

Educational content written from a research perspective, not medical advice. I am a PhD scientist, not a physician, and I do not diagnose, treat, cure, or prevent disease. Nothing here is a recommendation to use caffeine, alcohol, or cannabis; cannabis legality varies by location, and CBD interacts with many prescription drugs.

Caffeine

Clearance. Over 95 percent of caffeine is N-demethylated by hepatic CYP1A2 to paraxanthine (the principal active metabolite), theobromine, and theophylline. CYP1A2 is highly inducible; activity varies more than 40-fold between individuals. The intronic variant rs762551 (CYP1A2*1F, A allele) confers higher inducibility and is commonly reported as "fast"; C carriers show lower inducible activity and longer half-life (roughly 4 to 6 hours in AA, extending to 8 to 12 in CC, with wide variance). Inducers: tobacco smoke (polycyclic aromatic hydrocarbons; CYP1A2 activity falls within days of quitting, which is a common unrecognized cause of caffeine toxicity in new ex-smokers), cruciferous vegetables, omeprazole, carbamazepine. Inhibitors: fluvoxamine (profound), ciprofloxacin, oral contraceptives and pregnancy (half-life up to 15 hours in the third trimester), cimetidine, grapefruit (modest). Caffeine is also a CYP1A2 phenotyping probe.

Outcomes by genotype. Cornelis and colleagues (JAMA 2006): among slow metabolizers (C carriers), caffeine intake over 2 to 3 cups daily was associated with increased myocardial infarction risk (odds ratio about 1.4 to 1.6 at over 4 cups), whereas fast metabolizers showed no increase or a protective association, consistent with caffeine's adenosine antagonism and pressor effects persisting longer in slow metabolizers. Associations with hypertension, impaired fasting glucose, and, in athletes, blunted ergogenic effect in slow metabolizers have been reported with variable replication.

Receptor sensitivity. Caffeine is a non-selective adenosine A1 and A2A antagonist; A2A antagonism in the striatum disinhibits D2 signaling. ADORA2A rs5751876 (1976C>T) TT genotype is associated with caffeine-induced anxiety at 150 mg in placebo-controlled challenge (Alsene 2003, Childs 2008), with greater sleep disruption, and with lower habitual intake; ADORA2A variants also associate with panic disorder. ADORA1 and the dopamine receptor DRD2 variants modify reinforcing effects.

Physiology relevant to this site. Adenosine accumulation is the homeostatic sleep signal; antagonism masks sleep pressure without reducing it. Caffeine raises catecholamines and cortisol, is a diuretic with magnesium and calcium loss, raises intraocular and blood pressure acutely, and in daily use produces withdrawal headache, fatigue, and dysphoria within 12 to 24 hours of cessation (DSM-5 caffeine withdrawal). Recommendations by genotype: C carriers and ADORA2A TT, under 100 to 200 mg before noon or abstinence where anxiety, insomnia, migraine, POTS, or cardiovascular risk is relevant; AA, moderate intake with a 6-hour pre-sleep cutoff. The anxiety post here and migraine post here apply.

Ethanol

Metabolism. Ethanol is oxidized to acetaldehyde by class I alcohol dehydrogenases (ADH1A, ADH1B, ADH1C) and, at higher concentrations and with chronic use, by CYP2E1 (which generates reactive oxygen species); acetaldehyde is oxidized to acetate by mitochondrial ALDH2. Acetaldehyde is a Group 1 carcinogen (IARC), forms DNA adducts, and drives the flushing response through histamine and prostaglandin release.

ALDH2*2 (rs671, Glu504Lys). The Lys allele produces a dominant-negative enzyme; heterozygotes retain about 10 to 40 percent of activity, homozygotes essentially none. Allele frequency is roughly 30 to 40 percent in East Asian populations (highest in southern China, Japan, Korea) and under 1 percent in most others. Carriers experience facial flushing, tachycardia, nausea, and hypotension at low doses. Heterozygotes who drink habitually have substantially increased risk of esophageal squamous cell carcinoma (meta-analytic odds ratios of 3 to 7 at moderate intake, rising steeply with heavier use), head and neck cancer, and gastric cancer; the flush is also associated with higher blood pressure and, in some studies, osteoporosis and Alzheimer's risk. Alcohol-interactive drugs (metronidazole, disulfiram) act by inhibiting ALDH2 and reproduce the phenotype. Research position: ALDH2*2 carriage is a reason to abstain; the flush should not be overridden with antihistamines, which mask the symptom without reducing acetaldehyde exposure.

ADH1B*2 (rs1229984, Arg48His) increases ADH activity roughly 40- to 100-fold, raising acetaldehyde formation rate; carriers (common in East Asian, and about 5 to 10 percent in European, populations) experience earlier aversive effects and have markedly lower rates of alcohol use disorder (one of the strongest protective genetic associations in psychiatry). ADH1C*1 has a similar smaller effect. CYP2E1 variants modify oxidative injury.

Dependence and treatment-response loci. GABRA2 (rs279858 and haplotypes) associates with alcohol dependence and subjective response; OPRM1 A118G (rs1799971, G allele) increases ethanol-induced dopamine release and euphoria and predicts better response to naltrexone in several (not all) trials; DRD2 Taq1A (ANKK1 rs1800497), SLC6A4, and KCNJ6 modify risk; polygenic scores explain a modest fraction of heritability (roughly 50 percent for alcohol use disorder in twin studies).

Physiology relevant to this site. Acute GABA-A potentiation and NMDA antagonism with rebound glutamate excess, sympathetic activation, and histamine release during clearance (the 3am post here); histamine and tyramine content of wine and beer; thiamine, magnesium, folate, zinc, and pyridoxine depletion; preferential oxidation that suppresses fat oxidation and raises de novo lipogenesis; GGT induction; mast cell degranulation; disruption of slow-wave and REM sleep; dose-dependent increase in breast cancer risk through estrogen and acetaldehyde pathways. No genotype confers net health benefit from ethanol; the J-shaped curve has largely been attributed to confounding and sick-quitter effects in Mendelian randomization studies using ADH1B and ALDH2 as instruments.

Cannabinoids

Receptors and endocannabinoid tone. THC is a partial agonist at CB1 (CNR1; dense in hippocampus, prefrontal cortex, basal ganglia, cerebellum, and on GABAergic and glutamatergic terminals) and CB2 (immune cells, microglia, mast cells). The endocannabinoids anandamide and 2-AG are retrograde messengers degraded by FAAH and MAGL respectively. FAAH rs324420 (C385A, Pro129Thr) reduces enzyme stability and activity, raising anandamide; A carriers show lower trait anxiety, enhanced fear extinction, altered amygdala reactivity, and in several studies lower cannabis liking and lower rates of cannabis dependence, with inconsistent replication; the variant also modifies response to the FAAH inhibitor class and to CBD. CNR1 polymorphisms (rs1049353, rs806380, AAT repeats) associate with cannabis dependence, obesity, and hedonic response in some cohorts.

Psychosis risk modification. THC increases striatal dopamine release and disrupts prefrontal GABAergic and glutamatergic signaling. COMT Val158Met: Caspi and colleagues (Dunedin, 2005) reported that adolescent cannabis use raised risk of schizophreniform disorder in Val/Val carriers (odds ratio about 10 versus about 2.5 in Met/Met); subsequent studies were mixed, with meta-analytic support for a Val-allele interaction in some but not all. AKT1 rs2494732: the C/C genotype roughly doubled psychotic disorder risk in cannabis users (Di Forti 2012, GAP study) and predicted THC-induced psychotic-like symptoms and cognitive impairment in experimental and observational studies (Bhattacharyya 2012, Morgan 2016). Adolescent onset, daily use, high-THC products (over 10 to 15 percent), and family history of psychosis are the dominant risk factors, with genotype as a modifier; the research position is that Val/Val COMT or CC AKT1 carriage with adolescent or heavy use defines a group in which the risk argues against use.

Metabolism and interactions. THC is hydroxylated to 11-OH-THC (active, more potent, the reason edibles differ) by CYP2C9 and to a lesser extent CYP3A4; CYP2C92 and 3 (combined allele frequency roughly 20 to 30 percent in European populations) reduce clearance, with 3/3 showing about threefold higher THC exposure. CBD is metabolized by CYP3A4 and CYP2C19 and is a potent inhibitor of CYP3A4, CYP2C19, CYP2C9, CYP2D6, and UGT enzymes, raising concentrations of clobazam (documented), valproate (hepatotoxicity signal), warfarin, tacrolimus, and many others; the pharmacogenomics here report covers CYP2C9, CYP2C19, CYP3A4, and COMT.

Clinical pharmacology relevant to this site. THC impairs encoding and working memory acutely, impairs driving, and with adolescent onset is associated with persistent cognitive effects in longitudinal cohorts; cannabis use disorder develops in about 10 percent of users (up to 17 percent with adolescent onset, higher with daily use); cannabis hyperemesis syndrome (cyclic vomiting, compulsive hot bathing) is increasingly reported with high-potency products and resolves with cessation; cannabis withdrawal (irritability, insomnia, anorexia, anxiety) is recognized in DSM-5. CBD has Grade A evidence (Epidiolex) in Dravet, Lennox-Gastaut, and tuberous sclerosis seizures, Grade B at 300 to 600 mg for social anxiety in small trials, and little else; most consumer products are under-dosed relative to trials and variably labeled. Cannabinoids modulate mast cell activity (CB2 and TRPV1 effects) and gut motility; the clinical implications for the MCAS and POTS populations are not established.

Summary table

Substance Gene (variant) Effect Practical read
Caffeine CYP1A2 rs762551 C Lower inducibility, longer half-life Cap and morning-only; MI risk with high intake
Caffeine ADORA2A rs5751876 TT Receptor sensitivity Caffeine anxiety at low doses
Ethanol ALDH2 rs671 A Acetaldehyde accumulation Abstinence; cancer risk with drinking through the flush
Ethanol ADH1B rs1229984 A Rapid acetaldehyde formation Aversive early; protective against AUD
Ethanol OPRM1 rs1799971 G Greater reward Naltrexone response
Ethanol GABRA2 haplotypes Subjective response, AUD risk Family history still dominates
Cannabis FAAH rs324420 A Higher anandamide Lower anxiety; lower cannabis reward
Cannabis COMT rs4680 Val/Val Slower dopamine clearance Psychosis risk with adolescent use
Cannabis AKT1 rs2494732 CC Downstream dopamine signaling Psychosis risk; acute cognitive impairment
Cannabis CYP2C9 2, 3 Reduced THC clearance Prolonged and intensified effect, especially oral
CBD CYP3A4, CYP2C19 inhibition Drug interactions Review all co-medications

The raw DNA tutorial here shows how to find each rsID in a consumer file, with the strand and orientation cautions that apply to several of these loci.

Educational and research content, not medical advice. The author holds a PhD in cellular and molecular nutrition, is not a licensed physician, and does not diagnose, treat, cure, or prevent disease. Nothing here constitutes a recommendation to use caffeine, ethanol, or cannabinoids; legality of cannabis varies by jurisdiction; CBD interacts with many prescription drugs and is reviewed with the prescriber.