Guides › HLA pharmacogenomics

HLA alleles in drug guidelines: B*57:01, B*15:02, A*31:01, B*58:01

Updated 2026-10-29 · allele examples checked against IPD-IMGT/HLA 3.66.0 on 2026-10-29

A handful of HLA alleles appear in published drug-prescribing guidelines because carrying them is associated with a much higher risk of a serious hypersensitivity reaction to a specific drug. Four that carry CPIC level A recommendations are HLA-B*57:01 with abacavir, HLA-B*15:02 with carbamazepine, HLA-A*31:01 with carbamazepine, and HLA-B*58:01 with allopurinol.

This guide summarises what the Clinical Pharmacogenetics Implementation Consortium (CPIC) guidelines say, as published through ClinPGx, and how to make sure the HLA names you look up are the ones the guidelines mean. It is not prescribing advice. Decisions about a patient's treatment belong to the clinician, using the guideline itself.

The four alleles and their guidelines

All rows below are CPIC evidence level A, from the guideline records as of the 2026-10-26 snapshot of the open hla-pgx-open table.

The full recommendation text, including the conditions attached to each, is in the guideline pages linked under Sources. Read those, not a summary table, before acting on anything.

Abacavir and HLA-B*57:01

The CPIC abacavir guideline, first published in 2012 by Martin and colleagues, explains that HLA-B*57:01 is associated with an increased risk of a hypersensitivity reaction to the anti-HIV drug abacavir. It states that in the absence of genetic prescreening, hypersensitivity affects about 6% of patients and can be life-threatening with repeated dosing.

A 2014 update reviewed newer literature and concluded that none of it changed the therapeutic recommendations, so the original remained clinically current. The ClinPGx guideline page is the maintained version.

Carbamazepine, oxcarbazepine and phenytoin

Carbamazepine has two HLA associations in the same CPIC guideline: HLA-B*15:02 and HLA-A*31:01. The guideline records list the implications separately by gene, for example SJS/TEN for HLA-B and SJS/TEN, DRESS and MPE for HLA-A, and give recommendations for each combination of results.

The guideline also states that the latency period for these reactions is short with continuous dosing, and that its recommendations differ for patients who have taken the drug for longer than three months without a reaction. Those conditions are part of the recommendation and should not be dropped in a summary.

HLA-B*15:02 also appears in the CPIC guideline for CYP2C9 and HLA-B with phenytoin (Karnes and colleagues, 2020), where the HLA-B result is read together with the CYP2C9 metaboliser status.

Allopurinol and HLA-B*58:01

The CPIC allopurinol guideline associates HLA-B*58:01 with a significantly increased risk of allopurinol-induced severe cutaneous adverse reactions (SCAR). Its recommendation for a positive result is "Allopurinol is contraindicated", with the label Testing Recommended.

Getting the allele names right

Guideline lookups are only as good as the names fed into them. Four points come up repeatedly.

The guidelines key on two-field names. The records name B*57:01, not a specific four-field allele. In IPD-IMGT/HLA release 3.66.0, B*57:01 covers 92 named alleles, B*15:02 covers 28, A*31:01 covers 128 and B*58:01 covers 77. A typing that resolves to any of them carries the two-field name.

One field is not enough. A typing reported only as B*57 or B*15 cannot say whether B*57:01 or B*15:02 is present. Other alleles in the same groups are not the guideline alleles: in the same table snapshot, B*57:03 had no associations at all. If the result needs a two-field answer, the typing has to reach two fields.

Old spellings must be normalised first. Reports from before 2010 write B*5701. That is the same two-field name as B*57:01, and a lookup should treat it so, with a note that the spelling is deprecated. The conversion guide covers the cases where old names do not convert mechanically.

Null alleles and group names need care. A G group or P group name is a set of alleles, and a null allele is not expressed. Normalise group names to the two-field name they are written under, and keep the N on null alleles, before joining to a guideline table. The typing resolution guide explains why.

How HLA-Verify supports this

HLA-Verify carries the open hla-pgx-open table, which joins published CPIC guideline recommendations and PharmGKB clinical annotations that key on an HLA allele, with every allele name checked against the pinned IPD-IMGT/HLA release. The table is licensed CC BY-SA 4.0.

• GET /v1/pgx/table returns the whole table, free and without a key.
• POST /v1/pgx/screen takes a typing or a list of names, normalises each one first, and returns the associations for each, with the current name, the two-field name and any flags. Sent B*5701, it returns the current name B*57:01 with the flag deprecated_name and the abacavir associations.

Every response names the table snapshot and the release it was answered from. The pharmacogenomic endpoints are decision support only and not a medical device; they look up what a guideline says and do not decide anything about a patient.

Frequently asked questions

Which HLA allele is tested before abacavir?

HLA-B*57:01. The CPIC guideline labels this association "Testing Required" at evidence level A.

Is HLA-B*15:02 the only HLA allele in the carbamazepine guideline?

No. The CPIC carbamazepine and oxcarbazepine guideline also includes HLA-A*31:01 for carbamazepine, and its records give recommendations for each combination of HLA-A and HLA-B results.

Does B*5701 mean the same as B*57:01?

Yes. B*5701 is the pre-2010 spelling of the two-field name B*57:01. Normalise it before a lookup.

Can a one-field HLA-B result answer a pharmacogenomic question?

Not for these guidelines. They name two-field alleles, and a first-field result such as B*57 includes alleles the guidelines do not name.