Learn Explainer
Clopidogrel and CYP2C19: what the gene changes, what the trials found
Clopidogrel (Plavix) is switched on by CYP2C19, and two no-function copies of that gene have carried an FDA boxed warning since 2010. The guideline is strict about carriers, the big trials are less clear, and a 23andMe or AncestryDNA file may or may not contain the gene.
Key takeaways
- Clopidogrel is a prodrug. The label says its active metabolite is formed mostly by CYP2C19, and that the inactive route through esterases accounts for 85 percent of circulating metabolites[1].
- The FDA boxed warning is about poor metabolizers, people with two no-function alleles such as
*2/*2. The label gives their prevalence as about 2 percent of White patients and 14 percent of Chinese patients[1]. - CPIC's 2022 guideline also covers intermediate metabolizers: for ACS and PCI it recommends avoiding standard-dose clopidogrel if possible, at strong strength[7, 4].
- TAILOR-PCI found 4.0 percent events against 5.9 percent in carriers given ticagrelor instead of clopidogrel, a hazard ratio of 0.66 with a confidence interval of 0.43 to 1.02[8]. POPular Genetics found less bleeding, 9.8 against 12.5 percent, when the genotype was used to step down[9].
- In gnomAD's Finnish genomes
*2has a frequency of 18.65 percent,*1718.49 percent, and*3appeared in none of 10,550 alleles[11, 12, 13]. - Four of the ten chip exports we measured read all three alleles: the three from 23andMe and a 2024 AncestryDNA export[14].
Clopidogrel is a platelet-blocking tablet given after a heart attack or a stent. It does nothing until the liver converts it, and one enzyme, CYP2C19, does most of that work[1]. The label's recent-changes list dates its boxed warning about poor metabolizers to March 2010[2]. What follows is what that warning, the main guideline and the large trials say, with the gaps left in.
One tablet, two routes
Most of what circulates after a dose of clopidogrel is a dead end. The label describes two metabolic routes. Esterases hydrolyze the drug into an inactive carboxylic acid derivative, 85 percent of circulating metabolites. Cytochrome P450 enzymes oxidize it to an intermediate, 2-oxo-clopidogrel, and then to an active thiol metabolite[1]. CYP2C19 is involved in both oxidation steps, and the label says the active metabolite is formed mostly by CYP2C19, with smaller contributions from CYP1A2, CYP2B6 and CYP3A[1].
The active metabolite binds the P2Y12 receptor on platelets, and the label says that inhibition is irreversible and lasts for the life of the platelet[1]. A prodrug that depends on one polymorphic enzyme has a built-in source of variation: less working CYP2C19 should mean less active metabolite from the same tablet. The rest of this guide checks that prediction.
The three alleles that carry the signal
CPIC grades CYP2C19 star alleles by function. Three account for most of what a consumer file can show. *2 is rs4244285, a G-to-A change that causes a splicing defect, and *3 is rs4986893, which creates a premature stop codon (p.W212X). CPIC grades both as no function. *17 is rs12248560, a promoter change written -806C>T that CPIC grades as increased function[3]. A copy carrying none of the tested variants is called *1, which means "none of the listed variants found"; our star alleles guide explains why that is weaker than it sounds.
Each person has two copies, and CPIC translates the pair into a group. Two no-function copies (*2/*2, *2/*3, *3/*3) make a poor metabolizer. One no-function copy makes an intermediate metabolizer, including when the other copy is *17: both *2/*17 and *3/*17 are intermediate, so *17 does not cancel a no-function copy. One *17 with a *1 copy is rapid, and two *17 copies make an ultrarapid metabolizer[3].
Terms this guide relies on
- No-function allele
- A version of a gene that makes an enzyme with no activity. CPIC grades CYP2C19
*2and*3this way[3]. - Poor metabolizer
- CPIC's group for two no-function copies. The label defines it as being homozygous for nonfunctional alleles of the CYP2C19 gene[1].
- ACS and PCI
- Acute coronary syndrome, which includes heart attack, and percutaneous coronary intervention, the procedure that places a stent. CPIC's recommendation table for this group is its "CVI ACS PCI" population[4].
The boxed warning, then and now
The August 2010 revision of the label lists its boxed warning under "Recent major changes" with the date 03/2010. It was headed "Diminished effectiveness in poor metabolizers"[2].
Poor metabolizers with acute coronary syndrome or undergoing percutaneous coronary intervention treated with Plavix at recommended doses exhibit higher cardiovascular event rates than do patients with normal CYP2C19 function.
The same box said genotype tests "can be used as an aid in determining therapeutic strategy" and asked prescribers to "consider alternative treatment or treatment strategies" for poor metabolizers[2]. The current label, revised May 2025, is headed "Diminished antiplatelet effect in patients with two loss-of-function alleles of the CYP2C19 gene". It no longer repeats the sentence about event rates. It says tests are available to identify poor metabolizers and to consider another P2Y12 inhibitor in them[1].
Two details limit how far the warning reaches. It concerns two loss-of-function alleles, and one copy is outside its scope. And the label counts only part of the genetic picture: the 2010 text said *2 and *3 account for 85 percent of the reduced-function alleles in White poor metabolizers and 99 percent in Asian ones, and named *4 to *8 among the less frequent others[2]. The current label gives the prevalence of poor metabolizers as about 2 percent of White patients, 4 percent of Black patients, and 14 percent of Chinese patients[1].
The first studies
The warning came from comparing carriers with noncarriers among people already taking clopidogrel. In 162 healthy volunteers, carriers of at least one reduced-function allele, about 30 percent of the group, had a 32.4 percent lower exposure to the active metabolite and 9 percentage points less reduction in maximal platelet aggregation than noncarriers. In 1,477 people with acute coronary syndromes in the TRITON-TIMI 38 trial who took clopidogrel, carriers had death, heart attack or stroke at 12.1 percent against 8.0 percent (hazard ratio 1.53, 95% confidence interval 1.07 to 2.19), and stent thrombosis at 2.6 against 0.8 percent (hazard ratio 3.09, 1.19 to 8.00)[5].
A meta-analysis of nine studies then pooled 9,685 patients, 91.3 percent of whom underwent PCI. Of these, 71.5 percent carried no reduced-function allele, 26.3 percent carried one and 2.2 percent carried two[6]. Both groups of carriers did worse than noncarriers, and the point estimates were higher with two alleles than with one.
What the CPIC guideline says
The Clinical Pharmacogenetics Implementation Consortium (CPIC) is the expert group whose guidelines turn genotypes into prescribing recommendations. Its 2022 update of the clopidogrel guideline extended the indications covered, raised the strength of its recommendation for intermediate metabolizers and updated the genotype-to-phenotype translation[7]. CPIC adds that the same recommendations may also be considered for elective PCI[7]. The table below is read from the CPIC database for the ACS and PCI indication[4]. It states what the guideline tells prescribers.
| Group | Examples from *2, *3, *17 | CPIC recommendation, quoted | Strength |
|---|---|---|---|
| Ultrarapid | *17/*17 | If considering clopidogrel, use at standard dose (75 mg/day) | Strong |
| Rapid | *1/*17 | If considering clopidogrel, use at standard dose (75 mg/day) | Strong |
| Normal | *1/*1 | If considering clopidogrel, use at standard dose (75 mg/day) | Strong |
| Intermediate | *1/*2, *1/*3, *2/*17, *3/*17 | Avoid standard dose (75 mg) clopidogrel if possible. Use prasugrel or ticagrelor at standard dose if no contraindication. | Strong |
| Poor | *2/*2, *2/*3, *3/*3 | Avoid clopidogrel if possible. Use prasugrel or ticagrelor at standard dose if no contraindication. | Strong |
| Indeterminate | no group | No recommendation | No recommendation |
Source: CPIC database, guideline 100411, recommendation text and strength for the "CVI ACS PCI" population, read on 10 October 2026; diplotype translation from the same database[4, 3]. Examples are those the three tested alleles can produce.
For the intermediate and poor groups CPIC's stated implication is reduced or significantly reduced formation of the active metabolite, higher platelet reactivity while on treatment, and an increased risk of adverse cardiac and cerebrovascular events. For rapid and ultrarapid metabolizers it states increased formation, lower platelet reactivity and no association with higher bleeding risk[4]. CPIC also lists "likely" intermediate and poor groups and gives them the same recommendations and strength as the confirmed ones[4].
What the randomised trials found
Carriers who happened to take clopidogrel cannot show whether a different drug would change their outcomes. Two large randomised trials tested that, from opposite directions. In TAILOR-PCI clopidogrel was the default, and carriers of *2 or *3 were switched to ticagrelor. In POPular Genetics ticagrelor or prasugrel was the default, and noncarriers were stepped down to clopidogrel.
| Trial | Who and how | What the genotype-guided arm found |
|---|---|---|
| TAILOR-PCI | 5,302 PCI patients, 82% with ACS. Carriers of *2 or *3 got ticagrelor, noncarriers clopidogrel. Control: clopidogrel for all. | Among 1,849 carriers, the primary composite occurred in 35 of 903 (4.0%) against 54 of 946 (5.9%): hazard ratio 0.66, 95% CI 0.43 to 1.02, P = .06. Carrier bleeding 1.9% against 1.6%. |
| POPular Genetics | 2,488 primary PCI patients. Carriers of *2 or *3 got ticagrelor or prasugrel, noncarriers clopidogrel. Control: ticagrelor or prasugrel for all. | Net adverse clinical events 5.1% against 5.9%, noninferior (P < 0.001). Bleeding 9.8% against 12.5%: hazard ratio 0.78, 95% CI 0.61 to 0.98. |
Source: TAILOR-PCI, JAMA 2020[8]; POPular Genetics, NEJM 2019[9]. Percentages and intervals are as the trials' abstracts report them.
TAILOR-PCI missed its primary target: the interval for carriers includes 1, and the trial had 85 percent power to detect a hazard ratio of 0.50, a much larger effect than the one it saw[8]. Among all 5,302 patients the composite was 4.4 percent against 5.3 percent, with a hazard ratio of 0.84 (0.65 to 1.07)[8]. POPular Genetics succeeded at what it set out to test, that the genotype-guided strategy was no worse than giving everyone the stronger drug, and it did so with less bleeding[9].
6,734
participants in a 2026 pooled analysis of the two trials, which found no difference in the primary safety or efficacy endpoints overall[10]
0.77
hazard ratio for bleeding under genotype-guided de-escalation, as in POPular Genetics, in that analysis (95% CI 0.62 to 0.97)[10]
In that pooled analysis, escalation as in TAILOR-PCI performed similarly to conventional therapy, while de-escalation reduced bleeding and net adverse events with no significant difference in major cardiovascular events[10]. Both trials defined carriers by *2 and *3 only. A person with *17 counted as a noncarrier in both[8, 9].
How common the alleles are, in Finland and elsewhere
We read the three alleles' counts in gnomAD, a public collection of sequenced genomes. In its Finnish genomes *2 was present in 1,934 of 10,370 alleles (18.65 percent) and *17 in 1,943 of 10,506 (18.49 percent). *3, a no-function allele that is common in East Asia, appeared in none of 10,550 Finnish alleles[11, 12, 13].
The table pairs them at random to estimate each metabolizer group.
| Group in gnomAD | Poor | Intermediate | Normal | Rapid | Ultrarapid |
|---|---|---|---|---|---|
| Finnish | 3.5 | 30.3 | 39.5 | 23.2 | 3.4 |
| Non-Finnish European | 2.2 | 25.2 | 39.9 | 27.8 | 4.8 |
| East Asian | 13.9 | 46.8 | 38.2 | 1.2 | 0.0 |
Source: Our arithmetic, an estimate: no-function frequency is *2 plus *3, increased function is *17, every other allele is counted as *1, and the two copies are assumed to pair at random. Rarer alleles are ignored. Inputs from gnomAD v4 genomes[11, 12, 13].
The poor-metabolizer estimates land near the label's prevalences of about 2 percent for White patients and 14 percent for Chinese patients[1], a sanity check on the method and nothing more. In the Finnish genomes the estimate is 3.5 percent, about 35 people in 1,000, and 30 percent in the intermediate group. Our antidepressant gene-test guide compares this estimate with a measured Finnish cohort.
What an Aimosti report shows, by file type
The report reads the same three alleles from a chip export, a plain VCF and a gVCF, and differs in how it treats a position it cannot see. The CYP2C19 gene page and the clopidogrel drug page give the standing description; this is the file-by-file detail.
| File | What is read | When the file is silent | Clopidogrel line |
|---|---|---|---|
| Chip export | *2, *3, *17, where the array carries the position | A group is given only if all three positions can be read, or if both copies already carry a variant it did read. Otherwise the card names what it read and gives no group. | CPIC's clopidogrel row for the group, paraphrased |
| Plain VCF | The same three alleles | A reference result is labelled an unconfirmed inference, because a variant-only file cannot separate the reference allele from an uncovered position. | As above |
| gVCF | The same three alleles | Reference blocks show whether a position was covered, so a reference result can be supported by the file. | As above |
| BAM or CRAM (Deep Read) | CYP2C19 star alleles called from the reads with PyPGx | Per-gene callability is measured from the reads. | CPIC's row, quoted |
Source: aimosti/analysis/pgx.py and aimosti/content/pgx/cyp2c19.toml (chip, VCF, gVCF), aimosti/genomics/pypgx.py and aimosti/content/deep_read/cyp2c19.toml (Deep Read), as of 10 October 2026.
The chip column is where the answer to "does 23andMe test CYP2C19" sits. Of the ten public chip exports we measured, the three 23andMe exports and a 2024 AncestryDNA export contained all three positions. The other six all missed *17, and two of those read only *3, the allele no Finnish genome in gnomAD carried; one export read none of the three[14]. A chip that cannot see *17 cannot tell a normal metabolizer from a rapid one, which is why the card leaves the group open. Whether a vendor's own report interprets the positions is a separate question that a raw-data file does not answer. Our data page lists each export.
The report's chip, VCF and gVCF reading covers three alleles. The 2010 label named *4 to *8 among the less frequent alleles with absent or reduced function[2], so a *1/*1 result there means that none of the three tested variants was found, not that the gene has no variants. Deep Read's star-allele calling from aligned reads goes beyond the three. Our guide to genome file types explains why the files differ, and CYP2D6 and your medicines covers the neighbouring gene a chip cannot settle.
Where a genotype stops
A CYP2C19 group predicts how much active metabolite a standard dose is likely to make. It does not measure platelet reactivity in a given person. The label's warnings section says drugs that inhibit CYP2C19, such as omeprazole or esomeprazole, can also impair the conversion, and says to avoid concomitant use of Plavix with them[1]. In a post hoc analysis of TAILOR-PCI, a score combining age, body mass index, chronic kidney disease, diabetes and genotype helped identify patients on clopidogrel at higher risk, which the authors say needs confirmation in prospective studies[15].
The trials also limit what can be said about an individual. In TAILOR-PCI, 35 of 903 carriers on genotype-guided therapy and 54 of 946 on clopidogrel had the primary event[8]. Most carriers did not, and the trial could not rule out chance as the reason for the gap.
What Aimosti would (and wouldn't) show you
From a chip export, a VCF or a gVCF the report reads *2, *3 and *17, gives a metabolizer group when the file settles all three, and restates CPIC's clopidogrel row for that group. When it cannot, it names the position it could not read and gives no group. From a BAM or CRAM, Deep Read calls the gene's star alleles from the reads and quotes CPIC's row.
What we won't claim
We won't say whether anyone should take, change or stop clopidogrel, and we won't treat a consumer-file result as a clinical genotype test. CPIC and the FDA write for prescribers who are choosing a drug for a patient with a specific heart or brain indication; we restate their text and attribute it.
Bottom line. CYP2C19 makes the active form of clopidogrel, and carriers of no-function alleles make less of it. Whether acting on the genotype improves outcomes depends on the question asked. Switching carriers to ticagrelor was not statistically better in TAILOR-PCI, while POPular Genetics' genotype-guided arm, stepping noncarriers down to clopidogrel, recorded less bleeding. A consumer file can read the gene only when the array carries the right positions.
Questions people ask
Does 23andMe test CYP2C19?
The 23andMe raw-data exports we measured, versions 3, 4 and 5, contain the three positions that define *2, *3 and *17[14]. Whether 23andMe's own report interprets them is a separate matter. A 2018 AncestryDNA export read only *3; a 2024 one read all three[14].
What is a poor metabolizer for Plavix?
Someone with two no-function CYP2C19 alleles, such as *2/*2. The label defines it as being homozygous for nonfunctional alleles and says clopidogrel forms less of its active metabolite in such patients[1]. The label gives the prevalence as about 2 percent of White patients, 4 percent of Black patients and 14 percent of Chinese patients[1].
Does the FDA warning cover everyone with a CYP2C19 variant?
No. The boxed warning is headed "two loss-of-function alleles" and concerns poor metabolizers[1]. CPIC's 2022 guideline also gives a strong recommendation for intermediate metabolizers, who have one no-function allele, for ACS and PCI[7, 4].
Did genotype-guided treatment improve outcomes in trials?
It depends on the trial. TAILOR-PCI found 4.0 percent events in carriers given ticagrelor against 5.9 percent on clopidogrel, which did not reach statistical significance (hazard ratio 0.66, 95% CI 0.43 to 1.02)[8]. POPular Genetics found noninferior thrombotic outcomes and less bleeding when the genotype was used to step down from ticagrelor or prasugrel[9].
Does *17 matter for clopidogrel?
CPIC's ACS and PCI rows for rapid and ultrarapid metabolizers read "If considering clopidogrel, use at standard dose (75 mg/day)", the same as for normal metabolizers, and describe increased active metabolite formation with no association with higher bleeding risk[4]. *17 also does not offset a no-function copy: *2/*17 is an intermediate metabolizer[3]. The two big trials counted only *2 and *3 as carrier alleles[8, 9].
References
- Plavix (clopidogrel bisulfate) tablets: prescribing information. DailyMed, Sanofi-Aventis U.S., 2025. Label revised May 2025, read on 10 October 2026 from the DailyMed SPL: boxed warning, section 5.1 and sections 12.3 and 12.5.
- Plavix (clopidogrel bisulfate) tablets: prescribing information, revision of August 2010. U.S. Food and Drug Administration, Drugs@FDA, 2010. Highlights list the boxed warning under recent major changes as 03/2010; the full text of the boxed warning and section 12.5 were read on 10 October 2026.
- CPIC database: CYP2C19 allele function, diplotype-to-phenotype translation and variant locations. Clinical Pharmacogenetics Implementation Consortium, 2026. Allele function for *1, *2, *3 and *17; diplotype results for *1/*1, *1/*2, *1/*3, *1/*17, *2/*17, *3/*17, *17/*17, *2/*2, *2/*3; variant effects for rs4244285, rs4986893 and rs12248560. Read from the CPIC API on 10 October 2026.
- CPIC database: recommendations for CYP2C19 and clopidogrel (guideline 100411). Clinical Pharmacogenetics Implementation Consortium, 2026. Implication, recommendation text and strength for each phenotype and population (ACS or PCI, other cardiovascular, neurovascular). Read from the CPIC API on 10 October 2026.
- Mega JL, Close SL, Wiviott SD, et al. Cytochrome P-450 polymorphisms and response to clopidogrel. The New England Journal of Medicine, 2009. doi:10.1056/NEJMoa0809171
- Mega JL, Simon T, Collet JP, et al. Reduced-function CYP2C19 genotype and risk of adverse clinical outcomes among patients treated with clopidogrel predominantly for PCI: a meta-analysis. JAMA, 2010. doi:10.1001/jama.2010.1543
- Lee CR, Luzum JA, Sangkuhl K, et al. Clinical Pharmacogenetics Implementation Consortium Guideline for CYP2C19 Genotype and Clopidogrel Therapy: 2022 Update. Clinical Pharmacology and Therapeutics, 2022. doi:10.1002/cpt.2526
- Pereira NL, Farkouh ME, So D, et al. Effect of genotype-guided oral P2Y12 inhibitor selection vs conventional clopidogrel therapy on ischemic outcomes after percutaneous coronary intervention: the TAILOR-PCI randomized clinical trial. JAMA, 2020. doi:10.1001/jama.2020.12443
- Claassens DMF, Vos GJA, Bergmeijer TO, et al. A genotype-guided strategy for oral P2Y12 inhibitors in primary PCI. The New England Journal of Medicine, 2019. doi:10.1056/NEJMoa1907096
- Galli M, Pereira NL, Lennon RJ, et al. Genotype-guided vs conventional oral P2Y12 inhibitors in acute coronary syndrome: a combined analysis of TAILOR-PCI and POPular Genetics. JACC: Cardiovascular Interventions, 2026. doi:10.1016/j.jcin.2025.11.029
- gnomAD v4 variant 10-94781859-G-A (rs4244285, CYP2C19*2). Genome Aggregation Database. Genomes: Finnish 1,934 of 10,370 alleles (18.65%); non-Finnish European 10,042 of 67,874 (14.80%); East Asian 1,608 of 5,156 (31.19%). Read 10 October 2026.
- gnomAD v4 variant 10-94761900-C-T (rs12248560, CYP2C19*17). Genome Aggregation Database. Genomes: Finnish 1,943 of 10,506 alleles (18.49%); non-Finnish European 14,949 of 67,948 (22.00%); East Asian 49 of 5,174 (0.95%). Read 10 October 2026.
- gnomAD v4 variant 10-94780653-G-A (rs4986893, CYP2C19*3). Genome Aggregation Database. Genomes: Finnish 0 of 10,550 alleles; non-Finnish European 12 of 68,004 (0.02%); East Asian 315 of 5,168 (6.10%). Read 10 October 2026.
- What your DNA file can actually read: measured on 14 real files. Aimosti, 2026. Ten public chip exports measured with the report's chip reader (aimosti/content/chip_cyp2c19_reach.json, 9 October 2026): four read *2, *3 and *17; six missed *17.
- Capodanno D, Angiolillo DJ, Lennon RJ, et al. ABCD-GENE score and clinical outcomes following percutaneous coronary intervention: insights from the TAILOR-PCI trial. Journal of the American Heart Association, 2022. doi:10.1161/JAHA.121.024156
Last reviewed . Every number on this page links to the source it comes from; if one of them has moved, tell us.