Avoiding severe drug hypersensitivity reactions: a case for HLA genotyping for at‐risk patients

Prescribing decisions are always personalised, and may consider a patient's age, gender, health conditions and concurrent medications. Pharmacogenetics has the potential to provide additional information to inform these decisions. In a recent MJA article, White and colleagues put forward a call for a sustainable evidence‐based pharmacogenomic screening program for Australia, highlighting DPYD genotyping before fluoropyrimidine chemotherapy as an urgent need.1 DPYD, TPMT and UGT1A1 genotyping are performed to avoid severe toxicities associated with drug over‐exposure in the setting of antineoplastic therapy. With the exception of TPMT, these tests are not typically requested in the community setting. The abbreviations of these and other metabolic enzymes used throughout this article are defined in Box 1.

We agree with White and colleagues’ call for national action, and put forward human leucocyte antigen (HLA) genotyping to prevent severe drug hypersensitivity reactions as another important use case in Australia, particularly in the community setting. Since the initial association between HLA‐B*57:01 for abacavir hypersensitivity was reported in 2002, growing evidence supports the causal link between specific HLA alleles and drug hypersensitivity reactions. For severe manifestations, such as drug reaction with eosinophilia and systemic symptoms (DRESS) or life‐threatening mucosal and epidermal blistering (Stevens–Johnson syndrome and toxic epidermal necrolysis [SJS/TEN]), the mechanism involves direct binding of a drug (or metabolite) to a specific HLA allele, resulting in a complex targeted by T cells. To date, the best characterised associations are for allopurinol (HLA‐B*58:01), carbamazepine (HLA‐B*15:02, HLA‐A*31:01) and phenytoin (HLA‐B*15:02); these are widely prescribed drugs and are often commenced in the community setting.

Provision of rebated testing by Medicare considers cost–benefit aspects on a population level, including diagnostic test characteristics such as positive predictive value (PPV) and number needed to test (NNT). These depend on the population‐specific allele frequencies and incidence of the hypersensitivity reaction in that population;2 although both are estimates, the latter is especially difficult to determine.

The diagnostic test characteristics for genotyping HLA‐B*57:01 for abacavir hypersensitivity support pre‐emptive testing (PPV of around 50% and NNT of 20–40 in populations of European descent).3 Medicare‐rebated testing for this allele has been available since 2008. Test characteristics for the other alleles mentioned are less favourable. Allopurinol–HLA‐B*58:01 and carbamazepine–HLA‐B*15:02 each exhibit a PPV of about 3% and an NNT of about 500 for SJS/TEN; for DRESS and allopurinol–HLA‐B*58:01, estimates are 1% and 2000, respectively.3 Carbamazepine–HLA‐A*31:01 has a PPV of less than 1% for SJS/TEN, but the NNT ranges widely in the populations that have been studied; NNT is about 500 in Japanese people where the allele is prevalent.3 Although characteristics for lamotrigine and HLA‐B*1502 are weaker for SJS/TENS (PPV of about 1% and NNT>3000), we include it for illustrative purposes as an increasing proportion of cases in Australia involve this drug.

Lower PPV means that fewer patients with a positive test are likely to develop a reaction and more people need to be tested to prevent one. This is not fully understood but may be related to the presence of concomitant disease states or other genetic factors, for example, renal impairment in the context of allopurinol treatment, and CYP2C9 metaboliser status in the context of phenytoin treatment.4 Of note, hypersensitivity reactions can be dose‐dependent but do not exhibit a typical dose–response relationship.4 A specific Medicare rebate for testing these alleles is currently not available, although results may be obtained through HLA class I typing through transplantation and immunogenetics laboratories.

Clinical validity and population impact are also influenced by the odds ratio of people having the allele and experiencing the adverse event compared to those without that allele experiencing the event; however, a higher odds ratio does not always coincide with higher values for diagnostic test characteristics.2 For example, compared with taking the drug in question and not having the specific alleles, the odds ratio for SJS/TENS and carbamazepine–HLA‐B*15:02 is >1000, whereas for abacavir hypersensitivity and abacavir–HLA‐B*57:01 it is >500, and for SJS/TENS and allopurinol–HLA‐B*58:01 it is >800.5

The Database of Adverse Event Notifications (DAEN), maintained by the Therapeutic Goods Administration, provides a record of notifications of adverse drug reactions.6 Over the past 20 years (the period since the first publication involving abacavir), 340 cases of DRESS and SJS/TEN in people taking allopurinol, carbamazepine, phenytoin or lamotrigine were reported (Box 2). We expect this to be an underestimate due to voluntary reporting. By comparison, 334 SJS/TEN cases were identified through census of New South Wales hospital admission data in the period 2000–2012, 40% (134) of which were attributed to allopurinol, carbamazepine, phenytoin or their derivatives.8 Case numbers reported to the DAEN over the 20‐year period have been stable and there is no evidence of a reduction in the incidence of reactions. Carbamazepine and lamotrigine exhibit the highest number of cases, and both are used increasingly (Box 2). Ancestry is not recorded in the DAEN.

The prevalence of these HLA risk alleles varies in different populations; HLA‐B*58:01 and HLA‐B*15:02 are particularly common in Asian populations. This is important in Australia, where self‐reported Asian ancestry represented more than 17% of the total population in 2021.9 Box 3 presents the population sizes of Asian Australian groups with data available for the HLA alleles presented, and the proportions of carriers of the individual alleles.3,10 There are no allele frequency data for populations from Cambodia, Burma or Laos, collectively representing a population size of about 110000 Australians indicating these ancestries in 2021.

Although these severe reactions are rare, determination of HLA alleles raises the possibility of avoiding such outcomes by identifying patients for whom alternatives might be safer. National pre‐emptive testing in Taiwan in people initiating allopurinol or carbamazepine spared seven and ten cases of severe drug hypersensitivity, respectively, in the 5‐ and 3‐year respective recruitment periods compared with expected case numbers based on historical incidence. The prevalence of each allele was 19.6% for HLA‐B*58:01 and 7.7% for HLA‐B*15:02 in the two cohorts.11,12 The United States, like Australia, has an ethnically diverse population, and pre‐emptive HLA screening before initiating allopurinol or carbamazepine was found to be cost‐effective if targeted to people of Asian ancestry.13,14

International guidelines are available to inform prescribing decisions with regard to HLA related risks for allopurinol, carbamazepine and phenytoin.15,16,17,18 The Australian medicines handbook likewise provides guidance for these drugs.19 Testing is recommended before initiating these drugs as severe hypersensitivity reactions typically occur within the first months of treatment. For allopurinol, the Australian medicines handbook recommendation is to consider testing and use an alternative drug, or seek specialist advice, for a positive result. Routine HLA genotyping to avoid phenytoin hypersensitivity is not recommended by the Australian medicines handbook at present, although avoidance of the drug is suggested for patients known to carry HLA‐B*15:02. Similarly, HLA genotyping is not recommended for lamotrigine, presumably due to weak test performance for predicting lamotrigine hypersensitivity and the possibility that other, as yet unstudied, alleles have a mechanistic role.3,20

Polymerase chain reaction (PCR) based genotyping tests for HLA‐B*58:01, HLA‐B*15:02 and HLA‐A*31:01 are available in the community setting through pathology providers and transplantation and immunogenetics laboratories.21 Targeted PCR based tests typically cost around $80 per allele, with a turnaround time of 2–5 days. Class I HLA typing by sequence‐specific oligonucleotide based technologies can also identify carrier status at these sites. Infrequently, sequencing via specialised laboratories may be required for inconclusive tests. Array‐based pharmacogenetic panels, including direct‐to‐consumer tests, do not include these alleles at present.

The purpose of this article is to remind prescribers about the availability of HLA genotyping tests; we suspect these are under‐requested, potentially related to the lack of a specific Medicare rebate. We have presented allelic frequencies in source populations and have noted that a considerable proportion of the Australian population has Asian ancestry and is therefore at relatively high risk of carrying such alleles. Considerations for practice are summarised in Box 4. We also encourage the reporting of all severe adverse drug reactions to the DAEN. We welcome the recent initiative for a national registry for severe cutaneous adverse reactions that will no doubt serve to answer remaining questions in this space.2