HLA-B*15:02 and carbamazepine: what the CPIC guideline says
The Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for carbamazepine says that carriers of HLA-B*15:02 have an increased risk of carbamazepine-induced Stevens–Johnson syndrome and toxic epidermal necrolysis (SJS/TEN). For a carrier who has not taken carbamazepine before, its recommendation is "do not use carbamazepine", classified as strong. The association is CPIC evidence level A, labelled "Testing Required".
This guide summarises the guideline and how its terms map onto HLA nomenclature. It is not prescribing advice: CPIC guidelines are intended to assist clinicians, and the treatment decision belongs to the clinician using the guideline itself. Read the guideline page before acting on anything here.
Two editions, one set of recommendations
The guideline was first published by Leckband and colleagues in 2013, for HLA-B*15:02 and carbamazepine. A 2017 update by Phillips and colleagues widened the scope "to interpret HLA-A*31:01 and HLA-B*15:02 genotyping results to guide the use of carbamazepine and oxcarbazepine", and stated that "the recommendations provided in the original guideline have not changed".
So the current guideline covers two HLA alleles and two drugs. HLA-A*31:01 with carbamazepine is labelled "Actionable PGx", and HLA-B*15:02 with oxcarbazepine "Testing Recommended", both CPIC level A. This guide concentrates on HLA-B*15:02 with carbamazepine; the summary of HLA alleles in drug guidelines covers the others.
The recommendation, with its conditions
The 2013 paper's Table 2, for HLA-B*15:02:
| Genotype | Implication | Recommendation | Classification |
|---|---|---|---|
| Noncarrier | "Normal or reduced risk of carbamazepine-induced SJS/TEN" | "Use carbamazepine per standard dosing guidelines" | Strong |
| Carrier | "Increased risk of carbamazepine-induced SJS/TEN" | "If patient is carbamazepine-naive, do not use carbamazepine" | Strong |
| Carrier | (as above) | "If patient has previously used carbamazepine for longer than 3 months without incidence of cutaneous adverse reactions, cautiously consider use of carbamazepine" | Optional |
The third row matters. The guideline explains that carbamazepine-induced SJS/TEN "usually develops within the first 3 months of therapy", so a patient who has taken it for longer without a cutaneous reaction is at low, but not zero, risk regardless of HLA-B*15:02 status. A summary that keeps only "do not use carbamazepine" drops a condition the guideline attaches to it. The current ClinPGx table words the implications as "Greater risk" and "Normal risk"; if you quote the table, say which edition you are quoting.
How results are reported
The 2013 guideline defines two results: "'positive' if one or two copies of HLA-B*15:02 are present or as 'negative' if no copies of HLA-B*15:02 are present. There is no intermediate genotype or phenotype." One copy is enough for a positive result.
It is also specific about what the allele predicts. "HLA-B*15:02 is specific for SJS and TEN; there is no evidence that it predisposes to MPEs or hypersensitivity syndrome." The 2017 update names the wider set of reactions associated with HLA-A*31:01: SJS/TEN, drug reaction with eosinophilia and systemic symptoms (DRESS), and maculopapular exanthema (MPE).
Ancestry and negative results
Two statements in the guideline shape how a result is read:
- Ancestry is not a substitute for the genotype. The 2017 update says its recommendations "hold irrespective of the patient's region of origin or ethnic group", and that population frequencies "cannot replace genotypes on an individual basis". The 2013 paper stresses considering carrier status "regardless of self-reported ethnicity".
- A negative result does not remove all risk. Because not all carbamazepine-induced SJS/TEN can be attributed to
HLA-B*15:02, the 2013 guideline says "clinicians should carefully monitor all patients as standard practice".
The two editions give population frequency figures from different sources that are not directly comparable. If you need one, quote it with its edition and its wording.
The other B75 alleles
The 2017 update goes beyond a single allele name: "Other less frequently carried members of the HLA-B75 serotype include HLA-B*15:08, HLA-B*15:11, and HLA-B*15:21", which have also been reported in association with carbamazepine-induced SJS/TEN.
It then explains how that interacts with testing. The available data focus on HLA-B*15:02, which "is the basis for the design of efficient single allele molecular typing assays. However, some labs may provide high-resolution HLA-B typing", and when that information is available the possibility of risk with HLA-B*15:08, HLA-B*15:11 and HLA-B*15:21, and with less common B75 alleles such as HLA-B*15:30 and HLA-B*15:31 where it "has yet to be described", "needs to be considered a potential risk".
All five names are valid two-field names in IPD-IMGT/HLA release 3.66.0: B*15:08 covers 3 named alleles, B*15:11 11, B*15:21 5, B*15:30 3 and B*15:31 3. B*15:02 itself covers 28.
Getting the allele name right
Two fields are needed to see the difference. B*15:02, B*15:08 and B*15:11 share a first field. A typing reported only as B*15 cannot say which, if any, is present, and the B*15 group contains many alleles that the guideline does not name.
The single-allele test answers one question. A test designed to detect B*15:02 answers whether B*15:02 is present. Whether the other B75 alleles named in the 2017 update are present is a separate question, which only higher-resolution HLA-B typing can answer. Record which kind of test produced a result.
Old spellings must be normalised. A pre-2010 report writes B*1502. It is the same two-field name as B*15:02 and should be read as such, with the spelling flagged as deprecated. The two-field guide covers why digit-based handling fails on newer names.
Keep the release. Allele names and the alleles under each two-field name change with IPD-IMGT/HLA releases; store the release with the result.
Looking it up with HLA-Verify
POST /v1/pgx/screen normalises each reported name against the pinned release and joins the two-field form to the open hla-pgx-open table of CPIC guideline rows and PharmGKB clinical annotations. Each response names the release and the table snapshot it was answered from.
The pharmacogenomic endpoints are decision support only and not a medical device. They report what the guideline says and do not decide anything about a patient. For the abacavir guideline, see HLA-B*57:01 and abacavir.
Frequently asked questions
What does the CPIC guideline say for HLA-B*15:02 carriers?
For a carbamazepine-naive carrier, "do not use carbamazepine" (strong). For a carrier who has taken carbamazepine for more than three months without a cutaneous reaction, it says to cautiously consider use (optional).
Does HLA-B*15:02 predict all carbamazepine reactions?
No. The 2013 guideline says it is specific for SJS and TEN, with no evidence that it predisposes to MPE or hypersensitivity syndrome. HLA-A*31:01 is associated with a wider range of reactions.
Are other HLA-B*15 alleles mentioned in the guideline?
Yes. The 2017 update names HLA-B*15:08, HLA-B*15:11 and HLA-B*15:21 as other B75 serotype alleles reported in association with carbamazepine-induced SJS/TEN.
Is a B*15 typing result enough for this guideline?
No. The guideline names two-field alleles, and the B*15 group includes alleles that it does not name.
Sources
- https://www.clinpgx.org/guideline/PA166251448
- https://doi.org/10.1038/clpt.2013.103
- https://files.cpicpgx.org/data/guideline/publication/carbamazepine/2013/23695185.pdf
- https://doi.org/10.1002/cpt.1004
- https://files.cpicpgx.org/data/guideline/publication/carbamazepine/2017/29392710.pdf
- https://cpicpgx.org/guidelines/
HLA-Verify is a research-and-evaluation tool for nomenclature and reference-release validation. It is not a medical device and this guide is not clinical advice. Send allele names, never patient identifiers.