See what your DNA file can read before you pay. Free, on your own device. Check your file free →
Deutsch Sample report Support Log in
aimostı Check your file

Learn Explainer

Caffeine and CYP1A2: what the fast or slow label measures

One enzyme, CYP1A2, clears almost all the caffeine people drink, and the variant sold as the fast or slow caffeine gene sits inside it. What it changes is how far smoking or heavy coffee drinking can switch the enzyme up. The 2006 heart attack finding that made it famous did not hold in the largest cohort to test it.

Written by · 14 min read · Published 2026-10-10. Last reviewed .

Key takeaways

  • CYP1A2 carries out more than 95 percent of caffeine's first breakdown step, and caffeine's half-life is usually 4 to 5 hours[1].
  • rs762551, the CYP1A2*1F marker, mainly changes how far the enzyme can be induced. A pooled analysis of 3,570 people found a modest effect overall and a larger one in smokers[7].
  • In 786 Europeans, smoking more than 20 cigarettes a day multiplied caffeine clearance by 1.72 and oral contraceptives by 0.72[9].
  • The coffee GWAS found its CYP1A2 signal at rs2472297, between CYP1A1 and CYP1A2; rs762551 tracks it only weakly (r² 0.14 in Finns)[10, 13].
  • The 2006 Costa Rica finding, more heart attacks with coffee only in slow metabolisers, was followed by a UK Biobank analysis of 347,077 people that found no interaction between CYP1A2 genotype and coffee for cardiovascular disease[14, 16].

Caffeine is cleared almost entirely by the liver, and one enzyme, CYP1A2, carries out more than 95 percent of its first breakdown step[1]. Its activity differs widely between people. Few known CYP1A2 variants clearly explain that; tobacco smoke plainly turns it up, and oral contraceptives and some medicines turn it down[2]. The variant consumer DNA reports read, rs762551, affects how strongly the enzyme answers to that switching; in non-smokers, studies found no clear difference between its genotypes[3, 4].

One liver enzyme does most of the work

Caffeine leaves the body by being broken down in the liver. Its main route, 70 to 80 percent, is the removal of one methyl group to make paraxanthine, and CYP1A2 is responsible for more than 95 percent of caffeine's primary metabolism. The half-life is 4 to 5 hours, and can be much longer in newborns, in pregnancy and in liver disease[1]. CYP1A2 makes up about 13 percent of the cytochrome P450 protein in the liver[2]. The US Food and Drug Administration lists caffeine as a sensitive CYP1A2 substrate, the kind of drug whose levels rise steeply when the enzyme is blocked[5].

Because so much rides on one enzyme, caffeine is the standard probe for measuring CYP1A2. Researchers give a fixed dose and measure paraxanthine against caffeine in blood, saliva or urine some hours later; a higher ratio means more enzyme activity. The studies below compare that measured ratio between genotypes, not how anyone felt.

Terms this article relies on

CYP1A2
A cytochrome P450 enzyme made mostly in the liver. Besides caffeine it breaks down theophylline, melatonin, clozapine and tizanidine, among others[5].
Induction
An increase in how much of an enzyme the cells make, triggered by something they are exposed to. CYP1A2 is induced by smoking, by compounds in grilled meat, by cruciferous vegetables and by omeprazole[2]; its activity is regulated by the aryl hydrocarbon receptor, AHR[6].
rs762551 and *1F
A single-letter change, C to A, in the first intron of CYP1A2, also written -163C>A. The A allele defines the haplotype named CYP1A2*1F[2].

rs762551 changes how far the enzyme can be switched up

The variant was described in 1999 by a German group who gave 100 mg of caffeine to 185 non-smokers and 51 smokers. Among the non-smokers the urinary ratio did not differ significantly between the three genotypes. Among the smokers it did: the mean plasma paraxanthine to caffeine ratio was 1.37 with A/A, 0.88 with C/A and 0.82 with C/C. The authors called A/A a possible high-inducibility genotype and left open whether the variant itself or something linked to it was the cause[3].

1.37 vs 0.82

mean paraxanthine to caffeine ratio in smokers with A/A against C/C; no significant difference in non-smokers[3]

0.32 and 0.40

pooled effect of one and two A alleles on caffeine metabolism in 3,570 people (standardised mean difference)[7]

0.56 and 0.92

the same comparison restricted to smokers[7]

Later studies kept finding the same shape. In 194 Swedes and 150 Koreans, none of the CYP1A2 haplotypes tested explained enzyme activity, except *1F in Swedish smokers, where it went with higher inducibility; Koreans had lower activity than Swedes with the same genotype, smoking habit and contraceptive use[4]. In Serbian and Swedish non-smokers, heavy coffee drinkers with A/A had higher activity, and among people who drank less there was no difference between genotypes[8]. A 2018 meta-analysis of 3,570 people found higher activity with one or two A alleles, a standardised difference of 0.32 and 0.40, and roughly double that in smokers[7]. At 0.4 standard deviations, the genotype groups overlap heavily.

Whether rs762551 itself is the working change is still open. It sits in an intron, no mechanism has been shown, and the A allele also occurs on several other haplotypes (*1J, *1K, *21) that did not show the same effect; the PharmGKB curators suggest a nearby variant in linkage may be responsible, which would explain why results differ between populations[2].

Smoking, coffee itself and medicines move it further

The largest single measurement of what shifts caffeine clearance comes from a 1999 study that gave a test dose to 863 healthy Europeans in Germany, Bulgaria and Slovakia and analysed 786 of them. Clearance rose with every step of cigarette use and with coffee drinking, fell with oral contraceptives, and was 10 percent lower in women. Together the measured factors explained 37 percent of the variation, and an individual could still land anywhere from 0.46 to 2.2 times the predicted value. After adjusting for these factors the authors found no relevant genetic polymorphism for CYP1A2 activity[9].

Figure 1. Estimated change in caffeine clearance for each factor, in 786 healthy adults given a caffeine test dose. Bars to the right mean faster clearance, to the left slower[9].

Medicines move it too. The FDA's table of drug interactions lists fluvoxamine, an antidepressant, as a strong CYP1A2 inhibitor; ciprofloxacin, mexiletine and oral contraceptives as moderate inhibitors; and tobacco smoking, rifampin and phenytoin as moderate inducers[5]. The PharmGKB summary adds omeprazole and other proton pump inhibitors to the inducers[2]. A genotype read from a file says nothing about any of these, and they change from year to year in a way the genotype does not.

What the genome-wide studies found

The first genome-wide study of habitual caffeine intake pooled 47,341 people of European descent from five American studies. Two regions passed genome-wide significance: one near AHR on chromosome 7, at P = 2.4 × 10⁻¹⁹, and one between CYP1A1 and CYP1A2 on chromosome 15, at P = 5.2 × 10⁻¹⁴. The difference between the two homozygous genotypes was 44 mg of caffeine a day for the AHR variant and 38 mg for the CYP1A2 one; the authors counted a cup of coffee as 137 mg[6]. AHR regulates CYP1A2[6], so both hits lead back to the same enzyme.

A 2015 meta-analysis of up to 91,462 coffee drinkers, followed up in about 38,000 more, found eight regions with per-allele effects of 0.03 to 0.14 cups a day. The two strongest were again AHR and CYP1A1-CYP1A2, and the lead CYP1A2 variant was rs2472297, whose T allele added about 0.15 cups a day per copy in the discovery stage; the AHR variant rs4410790 changed intake by about 0.14 cups a day per copy[10]. A 2016 study of blood levels in up to 9,876 people closed the loop: the alleles at these two regions that go with more coffee go with lower plasma caffeine, that is, faster metabolism. The authors read this as blood caffeine levels steering how much people drink[11].

Figure 2. Where the variants sit on chromosome 15 (GRCh38). The genome-wide studies found their signal between the genes; the variant consumer DNA reports read is inside CYP1A2, and its genotype predicts the lead variant's only weakly[12, 6, 10, 13].

Neither the 2011 nor the 2015 study reported rs762551 among its lead variants[6, 10]. In 1000 Genomes data its correlation with rs2472297 is r² 0.14 in Finns, 0.15 in British and 0.10 in the CEU sample of Utah residents of northern and western European ancestry[13], so one genotype predicts the other poorly.

Coffee, genotype and the heart: which findings held

The study that made CYP1A2 a consumer trait compared 2,014 people in Costa Rica after a first non-fatal heart attack with 2,014 matched controls. Among carriers of the allele the authors called slow, 2 to 3 cups a day went with an odds ratio of 1.36 and 4 or more with 1.64, against less than one cup; with two copies of the fast allele, coffee showed no increase (interaction P = 0.04)[14]. The paper called the slow allele *1F, the standard name of the inducible A allele, and a later letter clarified that C carriers were the slow group[2].

An Italian study following 553 young adults with stage 1 hypertension for a median of 8.2 years pointed the same way: against abstainers with the same genotype, heavy coffee drinkers carrying the slow allele had a hazard ratio of 3.00 for hypertension, and heavy drinkers with the fast genotype 0.36[15].

The largest test came from UK Biobank: 347,077 people, of whom 8,368 developed cardiovascular disease. More than six cups a day went with 22 percent higher odds than one to two, but neither CYP1A2 genotype nor a genetic caffeine-metabolism score was associated with the disease or changed the coffee association (interaction P ≥ 0.53)[16]. In 498,134 participants, coffee drinking went with lower mortality, including at eight or more cups a day, across every level of a caffeine-metabolism score built from AHR, CYP1A2, CYP2A6 and POR variants[17].

Table 1. Studies that split coffee drinkers by caffeine-metabolism genotype
StudyPeopleOutcomeGenotype changed the coffee association?
Cornelis 2006, Costa Rica2,014 cases, 2,014 controlsFirst non-fatal heart attackYes: higher odds with coffee only in C carriers (interaction P = 0.04)
Palatini 2009, Italy553Hypertension needing treatmentYes: higher risk with coffee only in C carriers
Zhou 2019, UK Biobank347,077, of whom 8,368 casesCardiovascular diseaseNo (interaction P ≥ 0.53)
Loftfield 2018, UK Biobank498,134, of whom 14,225 diedDeath from any causeNo (heterogeneity P = 0.17)

Source: Each study's own abstract. The first two grouped people by the *1F variant alone; the UK Biobank analyses used CYP1A2 genotype and multi-gene caffeine-metabolism scores[14, 15, 16, 17].

The two early studies found an interaction; the far larger later ones, which recorded coffee habits years before any event, did not. Interactions need many more people than main effects to estimate reliably, and a case-control study asks about coffee after the heart attack.

Mendelian randomisation: using the variants as a natural experiment

Because the AHR and CYP1A1-CYP1A2 variants change how much coffee people drink, and genotypes are fixed at conception, they can be used to ask whether drinking more coffee causes a disease, in principle without the lifestyle differences that separate heavy from light drinkers. In 112,509 Danes, with international consortia bringing the heart disease analysis to up to 223,414 people, an allele score that raised coffee intake by 42 percent was not associated with ischaemic heart disease, stroke, cardiovascular death or death from any cause; the hazard ratios per allele ran from 1.01 to 1.02[18]. The observational data from the same Danes showed the U-shaped curve familiar from coffee research, lowest risk at moderate intake, which the genetic analysis did not reproduce[18].

The same two variants can be read the other way round: as raising the clearance of caffeine, and so lowering its level in the blood. A 2023 analysis used rs2472297 and rs4410790 as instruments for plasma caffeine. Higher genetically predicted plasma caffeine went with lower body mass index and a lower risk of type 2 diabetes, an odds ratio of 0.81 per standard deviation combining FinnGen and the DIAMANTE consortium, and with no strong association with any cardiovascular disease studied[19]. The authors estimated that about 43 percent of the diabetes effect ran through body mass index, and called for clinical studies to follow[19].

Neither design used rs762551. Both give population estimates, which hold only if the variants act on disease through caffeine alone.

How common each allele is in Finland and elsewhere

gnomAD, a public reference of sequenced genomes, splits its counts by genetic ancestry group, and Finland has a group of its own, of about 5,300 genomes for these positions. The A allele of rs762551 is the commoner one everywhere: 68.9 percent of Finnish alleles, and 47.5 percent of Finns carry two copies[20]. The intake-raising T allele of rs2472297 is about as common in Finns as in other Europeans, near 25 percent, and almost absent in East Asian genomes; the C allele of the AHR variant rs4410790 is commoner in Finns than in any other group[21, 22].

Table 2. Allele frequencies by gnomAD genetic ancestry group
Grouprs762551 A (*1F)A/A at rs762551rs2472297 Trs4410790 C
Finnish68.9%47.5%24.9%69.2%
Non-Finnish European71.2%50.7%24.3%62.6%
African / African American60.5%36.7%4.2%45.5%
Admixed American70.6%50.3%9.5%38.7%
East Asian65.1%43.1%0.1%42.1%
South Asian56.8%33.1%3.6%43.8%

Source: gnomAD v4.1 genomes (these non-coding positions have no exome data), read on 10 October 2026. The A/A share is the homozygote count divided by the people counted, not a Hardy-Weinberg estimate[20, 21, 22].

What the Aimosti report shows

The report reads rs762551 on two cards, in two vocabularies. The Traits card, Caffeine metabolism, labels A/A Faster metaboliser and both C/A and C/C Slower metaboliser; its evidence panel names PharmGKB as the source and says the variant is below genome-wide significance. The Frontier card, Caffeine metabolism: fast or slow (CYP1A2), grades the same genotype in three steps of inducibility and rates the evidence as mixed.

Figure 3. How the report's two caffeine cards would label Finns, by our arithmetic on gnomAD's counts. About half of Finns would read as slower metabolisers on the Traits card[20].
Table 3. What each file gives in the report, as of October 2026
Filers762551 (Traits and Frontier cards)CYP1A2 star allelesAHR coffee-intake card
Chip exportRead from the chip's row; a missing or no-call row is left unresolvedNot calledrs4410790, read the same way
Plain VCFA missing record is shown as C/C, with a note that this is an inferenceNot calledAs for rs762551
gVCFA missing record counts as C/C only where the file shows CYP1A2 was readA diplotype from 26 defining positions, with no phenotype or drug guidanceAs for rs762551
BAM or CRAM (Deep Read)From the VCF or gVCF the reads come withA diplotype called from the reads, with no phenotype or drug guidanceFrom the VCF or gVCF

Source: The report's Traits, Frontier, gVCF pharmacogene and Deep Read modules as they stand on 10 October 2026. CPIC lists no gene-drug pair for CYP1A2, which is why neither star-allele call carries a phenotype[23].

Three limits follow from the evidence above. Neither card knows about smoking, coffee habits or medicines, which moved clearance more than any genotype did. The report does not read rs2472297, the CYP1A2-region variant the genome-wide studies found, so the lead caffeine signal in that region is on neither card; the AHR card does read the other lead variant, rs4410790. And the free preview of a report shows the caffeine card too. From a chip it reads the chip's row the way the report does; from a VCF or gVCF it reads a missing record as C/C with the same inference note. Our sample report shows both caffeine cards for a fictional C/A chip file.

A further Frontier card reads ADORA2A, the adenosine receptor caffeine blocks; its TT genotype has been linked to more anxiety after a caffeine dose[1], which is about how caffeine is felt, not how fast it is cleared. It sits with the report's other sleep and body-clock cards, covered in a guide of their own. Why a missing VCF record is not a measured C/C is covered in a separate guide.

What Aimosti would (and wouldn't) show you

From a chip export, a plain VCF or a gVCF, the Traits card reads rs762551 and labels A/A a faster metaboliser and C/A or C/C a slower one; the Frontier card grades the same position on a three-step inducibility scale. A gVCF and Deep Read also call CYP1A2 star alleles, reported as a genotype alone because CPIC has no guideline for the gene.

What we won't claim

We won't tell anyone how much coffee or caffeine to have, read a heart, blood-pressure or pregnancy risk off rs762551, or present a fast or slow label as a measured clearance. The label describes a genotype; caffeine clearance itself was never measured from your file.

Bottom line. rs762551 is a real but conditional marker: its A allele lets CYP1A2 be induced further, which shows mainly in smokers and heavy coffee drinkers. The genome-wide studies found their caffeine signals at other variants, near CYP1A2 and near AHR, and the 2006 genotype-by-coffee heart finding did not replicate in UK Biobank.

Questions people ask

Is CYP1A2*1F the fast or the slow allele?

In standard nomenclature *1F is defined by the A allele of rs762551, the more inducible one, so A/A is the genotype usually called fast[2]. Some papers, including the 2006 heart attack study, used the name the other way round, which a later letter corrected[2].

Why do the Traits and Frontier cards label C/A differently?

The Traits card has two labels and puts C/A with C/C as Slower metaboliser. The Frontier card uses three and calls C/A intermediate inducibility. Both read the same position. The pooled studies found a difference from C/C with one A allele as well as with two, smaller with one[7], which fits the three-step reading better than the two-step one.

Does rs762551 matter for medicines?

CPIC, whose guidelines our medication cards restate, lists no gene-drug pair for CYP1A2[23]. Studies of rs762551 and antipsychotics cleared by CYP1A2 have disagreed, and a meta-analysis found no association[2]. Interacting medicines and smoking change CYP1A2 activity more reliably[5].

What if a chip export has no row for rs762551?

The report reads the row where an export has one. Where it has none, or the chip failed to call it, the card says the marker was not read instead of assuming a genotype. The CYP1A2 star-allele calls need a gVCF or aligned reads.

Does the AHR coffee variant measure caffeine metabolism?

Not directly. rs4410790 was found by its association with how much coffee people drink[6]. Its intake-raising allele also goes with lower blood caffeine and faster metabolism, which is how it connects back to CYP1A2[11].

References

  1. Thorn CF, Aklillu E, McDonagh EM, Klein TE, Altman RB. PharmGKB summary: caffeine pathway. Pharmacogenetics and Genomics, 2012. doi:10.1097/FPC.0b013e3283505d5e CYP1A2 is responsible for more than 95% of the primary metabolism of caffeine; 70 to 80% goes by N-3-demethylation to paraxanthine; half-life 4 to 5 h.
  2. Thorn CF, Aklillu E, Klein TE, Altman RB. PharmGKB summary: very important pharmacogene information for CYP1A2. Pharmacogenetics and Genomics, 2012. doi:10.1097/FPC.0b013e32834c6efd
  3. Sachse C, Brockmöller J, Bauer S, Roots I. Functional significance of a C→A polymorphism in intron 1 of the cytochrome P450 CYP1A2 gene tested with caffeine. British Journal of Clinical Pharmacology, 1999. doi:10.1046/j.1365-2125.1999.00898.x
  4. Ghotbi R, Christensen M, Roh HK, Ingelman-Sundberg M, Aklillu E, Bertilsson L. Comparisons of CYP1A2 genetic polymorphisms, enzyme activity and the genotype-phenotype relationship in Swedes and Koreans. European Journal of Clinical Pharmacology, 2007. doi:10.1007/s00228-007-0288-2
  5. Drug development and drug interactions: table of substrates, inhibitors and inducers. US Food and Drug Administration, 2026. Content current as of 29 May 2026, read 10 October 2026. CYP1A2 sensitive substrates include caffeine; strong inhibitor fluvoxamine; moderate inhibitors include ciprofloxacin, mexiletine and oral contraceptives; moderate inducers phenytoin, rifampin, teriflunomide and tobacco smoking.
  6. Cornelis MC, Monda KL, Yu K, et al. Genome-wide meta-analysis identifies regions on 7p21 (AHR) and 15q24 (CYP1A2) as determinants of habitual caffeine consumption. PLoS Genetics, 2011. doi:10.1371/journal.pgen.1002033 47,341 participants. Lead variants rs4410790 (AHR) and rs2470893 (CYP1A1-CYP1A2); 44 and 38 mg a day between homozygotes.
  7. Koonrungsesomboon N, Khatsri R, Wongchompoo P, Teekachunhatean S. The impact of genetic polymorphisms on CYP1A2 activity in humans: a systematic review and meta-analysis. The Pharmacogenomics Journal, 2018. doi:10.1038/s41397-017-0011-3
  8. Djordjevic N, Ghotbi R, Jankovic S, Aklillu E. Induction of CYP1A2 by heavy coffee consumption is associated with the CYP1A2 -163C>A polymorphism. European Journal of Clinical Pharmacology, 2010. doi:10.1007/s00228-010-0823-4
  9. Tantcheva-Poór I, Zaigler M, Rietbrock S, Fuhr U. Estimation of cytochrome P-450 CYP1A2 activity in 863 healthy Caucasians using a saliva-based caffeine test. Pharmacogenetics, 1999. PMID 10376760. Estimated multiples of basal clearance (a non-smoking man resident in Germany): coffee 1.45 per litre a day; smoking 1.22, 1.47, 1.66 and 1.72 for 1-5, 6-10, 11-20 and more than 20 cigarettes a day; oral contraceptives 0.72; female 0.90.
  10. Coffee and Caffeine Genetics Consortium, Cornelis MC, Byrne EM, et al. Genome-wide meta-analysis identifies six novel loci associated with habitual coffee consumption. Molecular Psychiatry, 2015. doi:10.1038/mp.2014.107 Table 1, stage 1 (Europeans): rs2472297 T +0.15 cups a day per allele; rs4410790 T -0.14.
  11. Cornelis MC, Kacprowski T, Menni C, et al. Genome-wide association study of caffeine metabolites provides new insights to caffeine metabolism and dietary caffeine-consumption behavior. Human Molecular Genetics, 2016. doi:10.1093/hmg/ddw334
  12. Gene: CYP1A2 (ENSG00000140505). Ensembl. Read 10 October 2026 through the Ensembl REST API. GRCh38: CYP1A2 canonical transcript 15:74,748,845-74,756,607, forward strand, exon 1 ending at 74,748,897; CYP1A1 15:74,719,542-74,725,528, reverse strand. rs762551 at 74,749,576 (intron variant, alleles C/A/G); rs2472297 at 74,735,539; rs2470893 at 74,727,108.
  13. Linkage disequilibrium between rs762551 and rs2472297, 1000 Genomes phase 3. Ensembl REST API. Read 10 October 2026. r² 0.142 in FIN, 0.150 in GBR, 0.096 in CEU.
  14. Cornelis MC, El-Sohemy A, Kabagambe EK, Campos H. Coffee, CYP1A2 genotype, and risk of myocardial infarction. JAMA, 2006. doi:10.1001/jama.295.10.1135
  15. Palatini P, Ceolotto G, Ragazzo F, et al. CYP1A2 genotype modifies the association between coffee intake and the risk of hypertension. Journal of Hypertension, 2009. doi:10.1097/HJH.0b013e32832ba850
  16. Zhou A, Hyppönen E. Long-term coffee consumption, caffeine metabolism genetics, and risk of cardiovascular disease: a prospective analysis of up to 347,077 individuals and 8368 cases. The American Journal of Clinical Nutrition, 2019. doi:10.1093/ajcn/nqy297
  17. Loftfield E, Cornelis MC, Caporaso N, Yu K, Sinha R, Freedman N. Association of coffee drinking with mortality by genetic variation in caffeine metabolism: findings from the UK Biobank. JAMA Internal Medicine, 2018. doi:10.1001/jamainternmed.2018.2425
  18. Nordestgaard AT, Nordestgaard BG. Coffee intake, cardiovascular disease and all-cause mortality: observational and Mendelian randomization analyses in 95 000-223 000 individuals. International Journal of Epidemiology, 2016. doi:10.1093/ije/dyw325
  19. Larsson SC, Woolf B, Gill D. Appraisal of the causal effect of plasma caffeine on adiposity, type 2 diabetes, and cardiovascular disease: two sample mendelian randomisation study. BMJ Medicine, 2023. doi:10.1136/bmjmed-2022-000335
  20. gnomAD v4.1, variant 15-74749576-C-A (rs762551, CYP1A2*1F). Genome Aggregation Database (gnomAD), 2024. Read 10 October 2026, genomes. Finnish 7,305 of 10,600 alleles, 2,519 homozygotes; non-Finnish European 48,375 of 67,968, 17,226.
  21. gnomAD v4.1, variant 15-74735539-C-T (rs2472297). Genome Aggregation Database (gnomAD), 2024. Read 10 October 2026, genomes. Finnish 2,625 of 10,540 alleles; East Asian 5 of 5,176.
  22. gnomAD v4.1, variant 7-17244953-T-C (rs4410790). Genome Aggregation Database (gnomAD), 2024. Read 10 October 2026, genomes. Finnish 7,304 of 10,554 alleles; non-Finnish European 42,536 of 67,940.
  23. CPIC gene-drug pairs: query for CYP1A2. Clinical Pharmacogenetics Implementation Consortium, 2026. Read 10 October 2026: the query returns no pairs. The CPIC data snapshot of 13 June 2026 that our report quotes also holds no CYP1A2 pair.

Last reviewed . Every number on this page links to the source it comes from; if one of them has moved, tell us.

This is the kind of answer we give. See what your file says.

See a sample report Get your report · $39