A reaction, not a disease: a medicine most people tolerate takes the skin off a few. The genome-wide scan found the risk in one place only — HLA — and the odds ratio among people who had taken allopurinol was 7.77.
Stevens–Johnson syndrome and toxic epidermal necrolysis are the same reaction at two severities. A medicine that millions of people take without incident provokes, in a very small number, an immune attack on the skin and mucous membranes. The epidermis detaches. It is treated like a burn, in intensive care, and it kills people.
The medicines that most often cause it are ordinary: allopurinol for gout, some antiseizure drugs, sulfonamide antibiotics, nevirapine.
A study of 424 European cases against 1,881 controls scanned the whole genome. Six positions reached significance, with odds ratios between 1.53 and 1.74.
All six were in the HLA region. Nothing else in the genome reached genome-wide significance at all.
rs9469003, in HCP5, sits in that region. Taken together, the risk alleles formed a haplotype more strongly associated with the reaction than any single position — and that haplotype is in linkage disequilibrium with HLA-B*58:01.
Among the patients who had been exposed to allopurinol, the odds ratio for that haplotype was 7.77 (95% CI 4.66–12.98).
The same haplotype. A four-to-five-fold jump in effect, produced not by finding a better marker but by asking about the right drug.
That is the whole shape of this subject. There is no gene for "reacting badly to medicines". There is a specific HLA type that presents a specific drug to the immune system in a way that starts a fight — and outside the presence of that drug, it does nothing at all.
Two guidelines, both about testing before a first dose rather than about interpreting a reaction afterwards:
Allele frequencies differ enormously between populations, which is why these recommendations name populations rather than applying flatly to everyone. That is not a caveat bolted onto the guidance; it is the guidance.
It is not a test you should run on a raw data file. HLA types are typed in a clinical laboratory, not read off a consumer chip, and a marker that tags an allele well in one population can tag it poorly in another.
And if you are having a reaction to a medicine right now — a rash with fever, blistering, sore eyes or mouth — this page is not the right place to be. That is an emergency, and it is treated on what is happening, not on a genotype.
Source. Wolf et al. (Orphanet J Rare Dis 2011) performed a genome-wide association study on 424 European SJS/TEN cases and 1,881 controls selected from a reference control panel. Six SNPs in the HLA region showed significant evidence of association, with odds ratios ranging from 1.53 to 1.74. The haplotype formed by their risk alleles was more strongly associated with disease than any single SNP, and was much stronger in patients exposed to allopurinol (OR 7.77, 95% CI 4.66–12.98). The associated haplotype is in linkage disequilibrium with HLA-B*58:01, already known to be associated with allopurinol-induced SJS/TEN in Asian populations. No locus outside the HLA region reached genome-wide significance in this sample.
Position listed here. rs9469003 in HCP5, within the MHC. HCP5 lies close to HLA-B and variants in it are used elsewhere as proxies for HLA-B alleles; proxy accuracy is population-dependent and this page does not treat the SNP as equivalent to an allele call.
Prescribing guidelines. CPIC recommends HLA-B genotype be considered before allopurinol (2013 guideline, 2015 update unchanged) and has a separate guideline for HLA genotype and carbamazepine or oxcarbazepine (2017 update): HLA-B*15:02 is strongly associated with carbamazepine- and oxcarbazepine-induced SJS/TEN, and HLA-A*31:01 with maculopapular exanthema, DRESS and SJS/TEN on carbamazepine. Both are pre-prescription safety tests performed in clinical laboratories.
Clinical course. SJS and TEN are distinguished by the percentage of body surface area with epidermal detachment. Management is supportive and centres on immediate withdrawal of the causative drug, fluid and electrolyte management, wound care and specialist ophthalmological review; mortality rises steeply with detached surface area and with age. No genotype has a role once the reaction has begun.
What a 23andMe/AncestryDNA export or raw VCF can and can't tell you about Drug-Induced Stevens–Johnson Syndrome and Toxic Epidermal Necrolysis comes down to these specific, well-studied positions — not a diagnosis.
HCP5 · rs9469003
See detailed info →Research-derived gene–drug associations only — not a prescription, dosing guide, or medical advice. Always follow your prescriber's guidance.
| Gene | Drug | What the research shows |
|---|---|---|
| HLA-B | Carbamazepine and oxcarbazepine | HLA-B*15:02 is strongly associated with a greater risk of Stevens–Johnson syndrome and toxic epidermal necrolysis in people treated with carbamazepine or oxcarbazepine. A separate allele, HLA-A*31:01, is associated with a greater risk of maculopapular exanthema, DRESS and SJS/TEN with carbamazepine. CPIC issued recommendations for the use of both drugs based on HLA genotype in its 2017 update. Allele frequencies differ substantially between populations, and this is a test ordered by a prescriber before a first dose — not something to act on from a consumer raw data file. (CPIC Guideline for HLA Genotype and Use of Carbamazepine and Oxcarbazepine: 2017 Update (Clin Pharmacol Ther 2018, PMID 29392710).) |
| CYP2C9 | Phenytoin | Phenytoin has a narrow therapeutic index and large variability in blood level between people, partly because CYP2C9 clears it and CYP2C9 activity varies. Separately, HLA-B*15:02 is associated with an increased risk of Stevens–Johnson syndrome and toxic epidermal necrolysis on phenytoin — the same allele that carries that risk with carbamazepine. CPIC issued combined guidance in 2021 covering both genes: one governs how much drug there is, the other governs whether it is safe to start at all. (CPIC Guideline for CYP2C9 and HLA-B Genotypes and Phenytoin Dosing (Clin Pharmacol Ther 2021, PMID 32779747).) |
Because the risk is not to medicines in general. An HLA molecule presents a particular drug to the immune system, and without that drug present there is nothing to present. The same haplotype carried an odds ratio between 1.53 and 1.74 across all cases and 7.77 among those exposed to allopurinol.
No, and it is worth being firm about it. HLA alleles are typed in a clinical laboratory. A SNP used as a proxy for an allele can tag it well in one population and poorly in another, so a consumer file can be wrong in both directions — falsely reassuring and falsely alarming. Where testing is recommended, the prescriber orders it.
The recommendations are population-specific because the allele frequencies are. That is a real difference rather than a hedge: an allele that is common in one population and rare in another produces a completely different expected benefit from testing everyone. Your prescriber applies the local guidance.
Contact a doctor now rather than reading about genotypes. A rash with fever, blistering, or sores in the mouth or eyes shortly after starting a new drug is treated as an emergency, and the first step is stopping the drug under medical supervision. Nothing on this page changes that or waits for it.
Free to reuse. This page's text is original writing from freely-available research, licensed CC BY 4.0 — reuse it, including commercially, with attribution to MyGeneLog. It's general research-derived information, not medical advice or a diagnosis — see Terms of Use.