HLA-B*58:01 genotyping prevalence and the association with allopurinol-induced severe cutaneous adverse reactions: a living systematic review and meta-analysis

Abstract

Evidence on the prevalence of HLA-B*58:01 genotyping and its association with allopurinol-induced severe cutaneous adverse reactions (SCARs) is lacking, especially in low-resource settings. We addressed these gaps by conducting comprehensive and race/ethnic origin-specific evaluations. We conducted a systematic search from inception to 31 December 2022 using databases (PubMed, Embase, and medRxiv), Google (for Vietnamese articles), and manual searching. We included original studies that investigated the association between HLA-B*58:01 genotyping and allopurinol-induced SCARs. We excluded studies on: (1) animals; (2) pharmacokinetics/pharmacodynamics; (3) genetic markers or genetic testing methods; (4) single group; and (5) cost-effectiveness of screening. Risk of bias was assessed using Newcastle–Ottawa Scale. We used random-effects model to report the summary estimates and 95% confidence interval (95% CI) in the meta-analysis. We included 13,719 patients from 24 case–control studies. The prevalences of HLA-B*58:01 genotyping (overall 5.8%; 95% CI 2.9% to 11.5%; I² = 98%) varied by races (Asian [7.7%; 95% CI 3.4% to 16.8%; I² = 98%] and White in Eastern/Western Europe [2.3%; 95% CI 1.2% to 4.3%; I² = 86%]) and ethnic origins (East and Central Asia [5.5%; 95% CI 1.5% to 17.8%; I² = 98%] and South and Southeast Asia [12.9%; 95% CI 9.5% to 17.3%; I² = 79%]). HLA-B*58:01 genotyping was associated with substantially increasing risk of allopurinol-induced SCARs (odds ratio 117.6; 95% CI 70.3 to 196.8; I² = 45%) regardless of the subgroups. We found a higher prevalence of HLA-B*58:01 genotyping in some Asian populations compared with the Whites. There is evidence to confirm a strong association between this allele and allopurinol-induced SCARs.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-16062-w.

Introduction

Allopurinol is widely prescribed as a first-line treatment for gout and hyperuricaemia due to its urate-lowering efficacy. However, its use has been associated with the development of severe cutaneous adverse reactions (SCARs), which can pose significant risks of morbidity and mortality^1–4. SCARs can present with extensive skin and mucosal involvement (e.g., rash, erythematous or purpuric macules, blistering, epidermal detachment), systemic symptoms (e.g., fever, organ failure), or laboratory abnormalities (e.g., eosinophilia, neutrophilia)⁵.

The HLA-B*58:01 allele has been identified as a strong genetic risk factor for the development of allopurinol-induced SCARs⁶. The association between this allele and allopurinol-induced SCARs was reportedly observed in some Western and Asian populations, including Han Chinese, Japanese, Korean, or Thai^7–16. Mechanistically, allopurinol is metabolised into oxypurinol, its active metabolite. In individuals carrying the HLA-B*58:01 allele, oxypurinol can bind non-covalently to the peptide-binding groove of the HLA-B*58:01 molecule on antigen-presenting cells. This alters the peptide repertoire or presents a novel complex that is recognised as foreign by CD8⁺ T cells, triggering a cytotoxic T-cell response^17,18. These findings have led some countries to recommend HLA-B*58:01 screening before initiating allopurinol therapy in populations with a high prevalence of this allele⁶, which was shown to be cost-effective in some settings^19–21.

Despite the growing body of evidence, there remains an inconsistency among findings from different settings and a lack of race/ethnic origin-specific prevalence of HLA-B*58:01 genotyping^19–23. These issues are likely due to: (1) biases from studies with lower levels of evidence (case–control study-based findings) and (2) new data not being consolidated. This gap could limit the HLA-B*58:01 screening and other preventive measures in high-risk populations, especially in low- and middle-income Asian countries.

Given the lack of high-quality evidence in these low-resource settings, we planned to continuously consolidate new data on the prevalence of HLA-B*58:01 genotyping and its association with allopurinol-induced SCARs. We aimed to provide more precise and race/ethnic origin-specific estimates to better address the genetic variation related to allopurinol-induced SCARs. Our goal was to improve patient safety by identifying other populations at high risk for allopurinol-induced SCARs and providing evidence to implement preventive measures in these groups efficiently.

Methods

Study design and reporting standards

We used the methods and recommendations in the Cochrane Handbook for Systematic Reviews of Interventions²⁴, with some adjustments for our exposure of interest. The protocol for this review was registered and stored at the Open Science Framework (OSF) Registries²⁵. We followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Statement to report this review²⁶. The PRISMA checklist was presented in eTables 1 and 2 (Supplementary Data).

Selection criteria

Types of studies

We included all original studies that investigated the association between HLA-B*58:01 genotyping and allopurinol-induced SCARs. The following formats were considered for inclusion: full-text publications, preprints, and conference abstracts. We did not set any restrictions for languages of publication, provided that there was an English abstract. We excluded reports that were: (1) studies on animals; (2) pharmacokinetic/pharmacodynamic studies; (3) studies that investigated genetic markers or genetic testing methods; (4) non-comparative epidemiological studies (e.g., case reports or case series); and (5) cost-effectiveness of screening studies that did not address the association between or prevalences of HLA-B*58:01 and allopurinol-induced SCARs.

Types of patients

We included 2 cohorts: (1) hospital-based adults who were treated with allopurinol and were tested for the HLA-B*58:01 genotyping (cohort 1, including SCAR group and tolerant group); and (2) population-based adults who were not treated with allopurinol but were tested for the HLA-B*58:01 genotyping (cohort 2, population group). The settings of these cohorts (time and place) were from the individual studies. Where being approved, patients taking oxypurinol—a metabolite of allopurinol that has been made as a medication—were also considered. Unless provided in the demographic information, the races and ethnic origins of the patients were identified based on the study settings. Races and ethnic origins (or ethnicities) were defined using conventional concepts²⁷. We excluded studies that evaluated patients who: (1) received allopurinol for ≤ 2 days; (2) were taking any new medications during the data collection timeframe of the studies; or (3) had a history of SCARs.

Types of outcome measures

We investigated different syndromes of SCARs, including acute generalised exanthematous pustulosis (AGEP), drug reaction with eosinophilia and systemic symptoms (DRESS), Stevens–Johnson syndrome (SJS), toxic epidermal necrolysis (TEN), and Stevens–Johnson/toxic epidermal necrolysis overlap syndrome (SJS/TEN). Studies reporting severe symptoms of maculopapular eruption/exanthema (MPE) or erythema multiforme major (EMM) were also included. We excluded studies using patient-reported outcomes for SCAR diagnosis or where SCAR diagnosis was not verified by a dermatologist.

Data sources and search strategy

We conducted systematic searches on PubMed (pubmed.ncbi.nlm.nih.gov), Embase (embase.com), and medRxiv (medrxiv.org) databases from inception. With a living systematic review/meta-analysis design, 5 repeated searches were performed between 1 June 2020 and 31 December 2022 (the latter was the date of the last search for 3 databases). At the last search, we also checked if any of the included preprints were published to retrieve the most updated information. We searched for articles that were published in Vietnamese using Google (google.com). Additionally, a manual search for references in relevant review articles was performed. Details of the search strategy were provided in eTable 3 (Supplementary Data). We detected and removed duplicate references using EndNote (version 19, The EndNote Team, Clarivate, Pennsylvania, United States of America).

Data collection

Selection of studies

Initially, 2 authors (HTP and T-VMH) independently reviewed the results of the systematic searches for inclusion. After screening the abstracts, the 2 reviewing authors classified the studies as “included” or “excluded”. If there was any disagreement after the first round of screening, the 2 reviewing authors re-assessed the full-text version in a discussion. When they could not reach a consensus, 1 other author (M-HT) was consulted for the final decision.

Risk of bias assessment

According to our last search, no randomised controlled trials were relevant or eligible for inclusion. Thus, we decided to use the Newcastle–Ottawa Scale (NOS) to assess the risk of bias of the included studies²⁸, which was conducted independently by 2 authors (HTP and MHT). Any discrepancies after the first round of quality assessment were discussed among the 2 initial evaluators and a third author (M-HT) before reaching the final decisions. As all included studies had a case–control design, we assessed the risk of bias based on the following categories: selection (4 items, maximum score of 4), comparability (1 item, maximum score of 2), and exposure (3 items, maximum score of 3). Regarding the NOS, there is currently no established threshold for the total score to distinguish studies with low and high risk of bias²⁸. In this review, using a conventional cut-off²⁹, we proposed that studies with NOS score ≤ 3 implied a high risk of bias. We presented the scores of the NOS for all included studies in eTable 4 (Supplementary Data).

Data extraction and management

Initially, 2 authors (T-VMH and A-HN) independently conducted data extraction based on the recommendations of Cochrane²⁴. We used a piloted and standardised form to extract the following information, where appropriate: (1) general information (author; title; publication date, format, and language; country); (2) study design (type; setting; date; inclusion/exclusion criteria; comparability of study groups; outcome measurement); (3) patient characteristics (age; sex; race; ethnic origin; number of patients recruited/investigated; HLA-B*58:01 genotyping); and (4) outcomes (syndromes of SCARs). Any disagreements, if not being solved in discussions, were determined by a third author (M-HT). If the studies were not reported in English or Vietnamese, we sent the copies to a certified agency for translation. Missing data were obtained either by computing from available information or contacting the primary investigators, where appropriate.

Data analysis

Exposure and outcome measurement

For the measurement of the HLA-B*58:01 allele (dichotomous exposure) and SCARs (dichotomous outcomes), we recorded the number of exposed patients, number of events, and total number of patients of the included studies. The pooled prevalence of HLA-B*58:01 genotyping was reported with the 95% confidence interval (95% CI). For the association between HLA-B*58:01 genotyping and SCARs, we reported the pooled odds ratio (OR) with the 95% CI.

Heterogeneity assessment

We assessed the heterogeneity among the studies both visually (using forest plots) and statistically (using Chi-squared tests with a significance level of 0.05). The levels of heterogeneity were investigated with the I² statistic, based on the following convention: I² statistic > 75% implying considerable heterogeneity²⁴. We pre-specified subgroup analyses for all measurements to explore the patterns of heterogeneity. In case of high likelihood of heterogeneity, we decided to conduct the meta-analysis, as the pooled results were still more beneficial than individual reports from case–control studies. However, from our perspectives, these findings were considered exploratory only.

Reporting bias assessment

We clarified the potential reporting bias visually with the funnel plot and statistically with Egger’s test at a significance level of 0.05³⁰. As assessing the funnel plot was subjective and the low power of Egger’s test cannot reliably rule out reporting bias³¹, all 5 authors (HTP, T-VMH, M-HT, A-HN, and MHT) independently assessed both of these results to reach a final decision. If the funnel plots revealed asymmetry, we attempted to differentiate reporting bias from other potential sources of asymmetry. In these cases, we used contour-enhanced funnel plots with significance levels of 0.01, 0.05, and 0.1³².

Data synthesis

We pooled the data in the meta-analysis using random-effects model with the inverse variance approach, as we had expected some degrees of heterogeneity. We conducted all data syntheses and statistical analyses using R software (version 4.2.1, R Foundation for Statistical Computing, Vienna, Austria). We used the metafor package for the meta-analysis and forestploter to draw the forest plots of the pooled results^33,34.

We estimated the pooled prevalence of HLA-B*58:01 genotyping in 3 populations: (1) allopurinol-induced SCAR group (from cohort 1); (2) allopurinol-tolerant group (from cohort 1); and (3) population group (from cohort 2). The pooled ORs for the exposure (HLA-B*58:01 positive or negative) and outcome of interest (SCARs) were reported for cohort 1. The subgroup analysis was pre-specified after data extraction and was conducted based on the following factors: race (Asian or White) and ethnic origin (Eastern and Western Europe, East and Central Asia, or South and Southeast Asia). For studies reporting multi-racial or multi-ethnic information, we classified them based on the dominant race and ethnic origin. For exploratory purposes, the prevalences of HLA-B*58:01 genotyping in 5 Asian ethnicities (Han Chinese, Japanese, Korean, Thai, and Vietnamese) were estimated a posteriori. We performed the sensitivity analyses to investigate: (1) the differences between fixed-effect and random-effects models and (2) the pooled results after excluding studies with a high risk of bias (using the conventional cut-off of NOS score ≤ 3²⁹).

Results

Searching results

We searched and identified 528 records from the databases/registers and 44 records from citation searching, websites, and Vietnamese medical journals (Ho Chi Minh City Journal of Medicine [ISSN 1859-1779], Journal of 108-Clinical Medicine and Pharmacy [ISSN 1859-2872], Journal of Clinical Medicine [ISSN 1859-3593], Journal of Medical Research [ISSN 2354-080X], Journal of Medicine and Pharmacy [ISSN 2734-9209], MedPharmRes [eISSN 2615-9139], Vietnam Journal of Preventive Medicine [ISSN 0868-2836], and Vietnam Medical Journal [ISSN 1859-1868]). After screening for eligibility, we included 24 reports (from 24 studies) in this review (Table 1). The specific number of records identified, included and excluded, and the reasons for exclusions are presented in the PRISMA flow diagram (Fig. 1).

Table 1.

Characteristics of included studies.

Study	Characteristics
Bardin, 2018
General information	Publication format: journal publication (abstract)
	Publication year (language): 2018 (English)
	Country: Vietnam
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 10 in SCAR group, 112 in tolerant group
	Age (years): mean 55 (SD 12) in SCAR group, mean 46 (SD 12) in tolerant group
	Sex (female): 2/10 (20%) in SCAR group, 0/112 (0%) in tolerant group
	Race: Asian
	Ethnic origin: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): mean 345 (SD 142) in SCAR group, mean 356 (SD 114) in tolerant group
	Duration of allopurinol (days): not reported
Outcomes	SJS, SJS/TEN, TEN
Cao, 2012
General information	Publication format: journal publication (full-text)
	Publication year (language): 2012 (English)
	Country: China
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 38 in SCAR group, 63 in tolerant group, 572 in population group
	Age (years): mean 63.8 (SD 14.4) in SCAR group, mean 60.9 (SD 15.1) in tolerant group, not reported in population group
	Sex (female): 12/38 (31.6%) in SCAR group, 6/63 (9.5%) in tolerant group, not reported in population group
	Race: Asian
	Ethnic origin: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): median 300 (IQR 100–600) in SCAR group, not clearly reported in tolerant group
	Duration of allopurinol (days): mean 26.2 (SD 13.5) in SCAR group, mean 122.7 (SD 11.6) in tolerant group
Outcomes	DRESS, MPE, SJS/TEN
Cheng, 2015
General information	Publication format: journal publication (full-text)
	Publication year (language): 2015 (English)
	Country: China
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: (1) were Han Chinese; (2) used allopurinol
	Exclusion criteria: had history of bone marrow transplantation, chemotherapy or cancer
Patient demographics	Number of patients investigated: 92 in SCAR group, 75 in tolerant group, 99 in population group
	Age (years): mean 59.1 (SD 1.6) in SCAR group, mean 53.4 (SD 1.5) in tolerant group, mean 49.3 (SD 0.6) in population group
	Sex (female): 32/92 (34.8%) in SCAR group, 4/75 (5.3%) in tolerant group, 35/99 (35.4%) in population group
	Race: Asian
	Ethnic origin: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): mean 186.5 (SD 13.7) in SCAR group, mean 289.8 (SD 15.2) in tolerant group
	Duration of allopurinol (days): mean 22.0 (SD 1.5) in SCAR group, mean 1153.4 (SD 303.0) in tolerant group
Outcomes	DRESS, SJS, SJS/TEN, TEN
Chiu, 2012
General information	Publication format: journal publication (full-text)
	Publication date (language): 2012 (English)
	Country: China
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: were not Han Chinese
Patient demographics	Number of patients investigated: 20 in SCAR group, 30 in tolerant group
	Age (years): median 68.5 (IQR 33–96) in SCAR group, median 71.5 (IQR 41–97) in tolerant group
	Sex (female): 9/20 (45.0%) in SCAR group, 4/30 (13.3%) in tolerant group
	Race: Asian
	Ethnic origin: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): median 250 (IQR 100–600) in SCAR group, median 200 (IQR 100–300) in tolerant group
	Duration of allopurinol (days): median 24.5 (IQR 10–56) in SCAR group, median 1620 (365–5400) in tolerant group
Outcomes	DRESS, EMM, SJS, TEN
Cristallo, 2011
General information	Publication format: journal publication (full-text)
	Publication date (language): 2011 (English)
	Country: Italy
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: White (Northern Italy)
	Exclusion criteria: not clearly reported
Patient demographics	Number of patients investigated: 7 in SCAR group, 115 in population group
	Age: mean 64.0 (SD 9.7) in SCAR group, not reported in population group
	Sex (female): 4/7 (57.1%) in SCAR group, not reported in population group
	Race: White
	Ethnic origin: Eastern/Western Europe
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	SJS, TEN
Do, 2015
General information	Publication format: journal publication (full-text)
	Publication date (language): 2015 (Vietnamese)
	Country: Vietnam
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 22 in SCAR group, 75 in population group
	Age (> 60 years): 10/22 (45.5%) in SCAR group, not reported in population group
	Sex (female): 11/22 (50.0%) in SCAR group, not reported in population group
	Race: Asian
	Ethnic origin: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): median 300 (IQR 300–300) in SCAR group
	Duration of allopurinol (days): not reported
Outcomes	DRESS, MPE, SJS/TEN
Do, 2019
General information	Publication format: journal publication (full-text)
	Publication date (language): 2019 (Vietnamese)
	Country: Vietnam
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: (1) had gout; (2) used allopurinol
	Exclusion criteria: (1) using medications with high risk of allergy; (2) prior intermittent use of allopurinol
Patient demographics	Number of patients investigated: 7 in SCAR group, 128 in tolerant group
	Age: mean 49.0 in SCAR group, mean 46.3 in tolerant group
	Sex (female): 2/7 (28.6%) in SCAR group, 0/128 (0.0%) in tolerant group
	Race: Asian
	Ethnicity: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): median 300 in SCAR group, median 300 in tolerant group
	Duration of allopurinol (days): median 18 in SCAR group, median 260.5 in tolerant group
Outcomes	SJS, SJS/TEN
Do, 2020
General information	Publication format: journal publication (full-text)
	Publication date (language): 2020 (English)
	Country: Vietnam
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 31 in SCAR group, 395 in tolerant group
	Age (years): mean 60.0 (SD 15.9) in SCAR group, mean 45.4 (SD 10.3) in tolerant group
	Sex (female): 9/31 (29.0%) in SCAR group, 0/395 (0.0%) in tolerant group
	Race: Asian
	Ethnicity: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): mean 303.4 (SD 97.2) in SCAR group, mean 369.1 (SD 103.8) in tolerant group
	Duration of allopurinol (days): not reported
Outcomes	DRESS, SJS/TEN
Gonçalo, 2013
General information	Publication format: journal publication (full-text)
	Publication date (language): 2013 (English)
	Country: Portugal
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 25 in SCAR group, 23 in tolerant group, 3200 in population group
	Age (years): mean 67.4 (SD 16.0) in SCAR group, mean 62.0 in tolerant group, mean 32.0 in population group
	Sex (female): 14/25 (56.0%) in SCAR group, 7/23 (30.4%) in tolerant group, 65% in population group
	Race: White
	Ethnicity: Eastern/Western Europe
Exposure	Dosing of allopurinol (mg/day): median 300 in SCAR and tolerant group
	Duration of allopurinol (days): > 4 months in tolerant group
Outcomes	DRESS, SJS, SJS/TEN, TEN
Hung, 2005
General information	Publication format: journal publication (full-text)
	Publication date (language): 2005 (English)
	Country: Taiwan
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 51 in SCAR group, 135 in tolerant group, 93 in population group
	Age (years): median 66 (IQR 18–91) in SCAR group, median 56 (IQR 21–84) in tolerant group, not reported in population group
	Sex (female): 27/51 (52.9%) in SCAR group, 10/135 (7.4%) in tolerant group, not reported in population group
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): median 100 (IQR 50–300) in SCAR group, median 150 (IQR 100–400) in tolerant group
	Duration of allopurinol (days): median 26 (IQR 1–56) in SCAR group, median 22 (IQR 6–107) in tolerant group
Outcomes	DRESS, SJS, SJS/TEN, TEN
Jung, 2011
General information	Publication format: journal publication (full-text)
	Publication date (language): 2011 (English)
	Country: South Korea
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: (1) had chronic kidney disease stage 3–5 or post-kidney transplantation; (2) used allopurinol
	Exclusion criteria: used allopurinol for < 60 days in tolerant group
Patient demographics	Number of patients investigated: 9 in SCAR group, 432 in tolerant group
	Age (years): mean 41.6 (SD 15.9) in SCAR group, mean 35.9 (SD 18.1) in tolerant group
	Sex (female): 3/9 (33.3%) in SCAR group, 114/432 (26.4%) in tolerant group
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): mean 112.5 (SD 38.7) in SCAR group, mean 100 (SD 50.1) in tolerant group
	Duration of allopurinol (days): mean 59.1 (SD 45.5) in SCAR group, mean 887.1 (SD 821.3) in tolerant group
Outcomes	DRESS, SJS
Kang, 2011
General information	Publication format: journal publication (full-text)
	Publication date (language): 2011 (English)
	Country: South Korea
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: not clearly reported (for tolerant group: chronic kidney disease)
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 25 in SCAR group, 57 in tolerant group, 485 in population group
	Age (years): median 58 (IQR 35–80) in SCAR group, median 51 (IQR 20–76) in tolerant group, not reported in population group
	Sex (female): 11/25 (44.0%) in SCAR group, 20/57 (35.1%) in tolerant group, not reported in population group
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): median 150 (IQR 100–600) in SCAR group, median 100 (IQR 50–200) in tolerant group
	Duration of allopurinol (months): median 1.0 (IQR 0.2–5.3) in SCAR group, median 29.1 (IQR 6–72) in tolerant group
Outcomes	DRESS, SJS/TEN
Kaniwa, 2008
General information	Publication format: journal publication (full-text)
	Publication date (language): 2008 (English)
	Country: Japan
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: not reported
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 10 in SCAR group, 493 in population group
	Age (years): median 74 (IQR 67–76.5) in SCAR group, not reported in other groups
	Sex (female): 2/10 (20.0%) in SCAR group, not reported in other groups
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	SJS, TEN
Lonjou, 2008
General information	Publication format: journal publication (full-text)
	Publication date (language): 2008 (English)
	Country: European countries
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 31 in SCAR group, 1822 in population group
	Age (years): median 56 (IQR 40–71.5) in SCAR group, not reported in other groups
	Sex (female): 13/31 (41.9%) in SCAR group, not reported in other groups
	Race: White (87.1%), Asian (6.5%), African (3.2%), South American (3.2%)
	Ethnicity: Eastern/Western Europe
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	SJS, SJS/TEN, TEN
Low, 2020
General information	Publication format: journal publication (full-text)
	Publication date (language): 2020 (English)
	Country: Malaysia
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: (1) aged ≥ 18; (2) used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 55 in SCAR group, 42 in tolerant group
	Age (years): mean 64.9 in SCAR group, mean 59.9 in tolerant group
	Sex (female): not reported
	Race: Asian
	Ethnicity: South and Southeast Asia (59.8%), East and Central Asia (40.2%)
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	AGEP, DRESS, SJS/TEN
Ng, 2016
General information	Publication format: journal publication (full-text)
	Publication date (language): 2016 (English)
	Country: Taiwan
Design	Type: case–control study
	Setting: hopsital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 146 in SCAR group, 285 in tolerant group
	Age (years): mean 65 (SD 16) in SCAR group, mean 59 (SD 14) in tolerant group
	Sex (female): 66/146 (45.2%) in SCAR group, 17/285 (6.0%) in tolerant group
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	DRESS, DRESS/SJS/TEN, MPE, SJS, SJS/TEN
Nguyen, 2021
General information	Publication format: journal publication (full-text)
	Publication date (language): 2021 (English)
	Country: Vietnam
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 81 in SCAR group, 74 in tolerant group
	Age (years): mean 57.6 (SD 14.6) in SCAR group, mean 60.3 (SD 11.4) in tolerant group
	Sex (female): 20/81 (24.7%) in SCAR group, 9/74 (12.2%) in tolerant group
	Race: Asian
	Ethnicity: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): mean 324.7 (SD 121.0) in SCAR group, mean 281.8 (SD ) in tolerant group
	Duration of allopurinol (days): not reported
Outcomes	AGEP/SJS, DRESS, DRESS/SJS/TEN, SJS/TEN
Niihara, 2013
General information	Publication format: journal publication (full-text)
	Publication date (language): 2013 (English)
	Country: Japan
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 7 in SCAR group, 25 in tolerant group
	Age (years): median 72 (IQR 68.5–78.5) in SCAR group, not reported in tolerant group
	Sex (female): 2/7 (28.6%) in SCAR group, not reported in tolerant group
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	SJS, EMM
Park, 2016
General information	Publication format: journal publication (full-text)
	Publication date (language): 2016 (English)
	Country: South Korea
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 9 in SCAR group, 949 in tolerant group
	Age (years): not reported
	Sex (female): not reported
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): not reported
Outcomes	SJS/TEN
Pham, 2022
General information	Publication format: journal publication (full-text)
	Publication date (language): 2022 (English)
	Country: Vietnam
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: (1) bullous autoimmune skin diseases; (2) erythema multiforme/erythema infectiosum; (3) Kawasaki disease; (4) bone marrow transplantation; (5) acropustulosis; (6) Von Zumbusch psoriasis; and (7) concomitant use of another drug of high allergy potential at SCARs onset
Patient demographics	Number of patients investigated: 100 in SCAR group, 183 in tolerant group, 810 in population group
	Age (years): mean 58.5 (SD 14.2) in SCAR group, mean 55.2 (SD 12.2) in tolerant group, not reported in population group
	Sex (female): 24/100 (24.0%) in SCAR group, 29/183 (15.8%) in tolerant group, not reported in population group
	Race: Asian
	Ethnicity: South and Southeast Asia
Exposure	Dosing of allopurinol (> 150 mg/day): 96/100 (96.0%) in SCAR group, 164/183 (89.6%) in tolerant group
	Duration of allopurinol (days): not reported
Outcomes	DRESS, DRESS/SJS/TEN, SJS, SJS/TEN, TEN
Saksit, 2017
General information	Publication format: journal publication (full-text)
	Publication date (language): 2017 (English)
	Country: Thailand
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 86 in SCAR group, 182 in tolerant group
	Age (years): median 68 (IQR 28–84) in SCAR group, median 66 (IQR 29–91) in tolerant group
	Sex (female): 45/86 (52.3%) in SCAR group, 39 (21.4%) in tolerant group
	Race: Asian
	Ethnicity: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): median 100 (IQR 50–300) in SCAR group, median 100 (IQR 50–400) in tolerant group
	Duration of allopurinol (days): median 19 (IQR 2–60) in SCAR group, median not reported in tolerant group (mean > 180)
Outcomes	DRESS, SJS/TEN
Sukasem, 2016
General information	Publication format: journal publication (full-text)
	Publication date (language): 2016 (English)
	Country: Thailand
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 30 in SCAR group, 100 in tolerant group, 1095 in population group
	Age (years): median 73 (IQR 30–88) in SCAR group, not reported in other groups
	Sex (female): 13/30 (43.3%) in SCAR group, not reported in other groups
	Race: Asian
	Ethnicity: South and Southeast Asia
Exposure	Dosing of allopurinol (mg/day): mean 239.9 (SD 87.0) in SCAR group, not reported in other groups
	Duration of allopurinol (days): mean 16.4 (SD 14.3), not reported in other groups
Outcomes	DRESS, MPE, SJS/TEN
Tassaneeyakul, 2009
General information	Publication format: journal publication (full-text)
	Publication year (language): 2009 (English)
	Country: Thailand
Design	Type: case–control study
	Setting: hospital
	Inclusion criteria: (1) used allopurinol; (2) agreed to participate
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 27 in SCAR group, 54 in tolerant group
	Age (years): median 65 (IQR 38–81) in SCAR group, median 63.5 (IQR 46–90) in tolerant group
	Sex (female): 12/27 (44.4%) in SCAR group, 11/54 (20.4%) in tolerant group
	Race: Asian
	Ethnicity: South and Southeast Asian
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): median 14 (IQR 3–50) in SCAR group, median 780 (IQR 90–18,000) in tolerant group
Outcomes	SJS, SJS/TEN, TEN
Tohkin, 2013
General information	Publication format: journal publication (full-text)
	Publication year (language): 2013 (English)
	Country: Japan
Design	Type: case–control study
	Setting: hospital and population
	Inclusion criteria: for hospital setting: used allopurinol
	Exclusion criteria: not reported
Patient demographics	Number of patients investigated: 18 in SCAR group, 986 in population group
	Age (years): mean 72.3 (SD 10.0) in SCAR group, not reported in population group
	Sex (female): 6/18 (33.3%) in SCAR group, not reported in population group
	Race: Asian
	Ethnicity: East and Central Asia
Exposure	Dosing of allopurinol (mg/day): not reported
	Duration of allopurinol (days): mean 21.7 (SD 11.9), not reported in population group
Outcomes	SJS, TEN

AGEP: acute generalised exanthematous pustulosis; DRESS: drug reaction with eosinophilia and systemic symptoms; EMM: erythema multiforme major; IQR: interquartile range; MPE: maculopapular eruption; SCAR: severe cutaneous adverse reaction; SD: standard deviation; SJS: Stevens–Johnson syndrome; SJS/TEN: Stevens–Johnson/toxic epidermal necrolysis overlap syndrome; and TEN: toxic epidermal necrolysis (TEN).

In the context of SCARs, DRESS can also be referred as drug hypersensitivity syndrome (DHS), drug-induced hypersensitivity syndrome (DIHS), or hypersensitivity syndrome (HSS). To ensure consistency, outcomes that were reported as DHS, DIHS, or HSS were classified as DRESS.

Fig. 1.

PRISMA flow diagram.

Descriptive characteristics

Study design and setting

All 24 reports were case–control studies. Among these, 5 studies did not have well-defined study bases for population-based control selection^{10,13,16,35,36}, 12 studies inappropriately compared cases with population-based or genomic database controls without prior exposure to allopurinol^{7,8,10–14,16,25,35–37}. No studies used matching to control for potential confounders, while 5 studies used clearly defined multivariable regression modelling to report confounder-adjusted OR^9,25,38–40. There were 3 studies on the White population from Europe, 11 studies on Asians from East and Central Asia and 10 studies on Asians from South and Southeast Asia. For Asia-based studies, 6 were conducted in Vietnam, 5 in China/Taiwan, 3 in Japan, 3 in South Korea, and 3 in Thailand. Most studies were conducted at multiple centres.

Patient characteristics

We included 13,719 observations from 24 studies (937 in SCAR group, 3344 in tolerant group, and 9352 in population group). People in the tolerant group (2 studies with mean/median age of 65 or more) were probably younger than those in the SCAR group (10 studies with mean/median age of 65 or more). Compared with males, females were less likely to experience SCAR or be included as tolerant controls in most studies. The mean/median dose of ≥ 200 mg of allopurinol was reported in most studies (9 out of 14 studies). Demographic information was not adequately reported in the population group.

Exposure

All studies used polymerase chain reaction (PCR) for HLA-B*58:01 genotyping. In particular, there were 3 primary techniques being used to determine the HLA-B*58:01 genotype, i.e., sequence-specific primers, sequence-specific oligonucleotide, and sequence-based typing. However, for most studies investigating the population-based control group, genotyping techniques were not clearly reported.

Outcomes

Most studies used RegiSCAR, Roujeau, or Bastuji-Garin criteria to confirm SCAR diagnoses. The most frequently investigated SCAR was SJS/TEN (reported in 18 studies), followed by SJS and DRESS (reported in 15 and 14 studies, respectively). We also recorded 10 studies reporting TEN, 4 studies reporting severe MPE, 2 studies reporting EMM, and 1 study reporting AGEP. Among 937 cases, the prevalences of 5 primary syndromes of SCARs were as follows (in decreasing order): DRESS (36.4%), SJS (29.2%), SJS/TEN (18.4%), TEN (6.9%), and AGEP (0.1%). Besides SJS/TEN, 2 other overlap syndromes were also reported, including AGEP/SJS⁴¹ and DRESS/SJS/TEN^39,41,42.

Prevalence of HLA-B*58:01

In allopurinol-induced SCAR group

All 24 studies reported the prevalence of HLA-B*58:01 genotyping in the allopurinol-induced SCAR group^{7–16,35–48}. Among 937 patients with allopurinol-induced SCARs, 837 tested positive for HLA-B*58:01. The pooled prevalence of HLA-B*58:01 genotyping in this group was 88.5% (95% CI 81.3% to 93.1%; I² = 80%; Fig. 2). We also found some discrepancies among different races and ethnic origins. We detected a higher prevalence of HLA-B*58:01 genotyping in the Asian sub-population (91.1%; 95% CI 85.4% to 94.8%; I² = 73%; Fig. 2A) than that in the White from Eastern or Western Europe (60.3%; 95% CI 47.7% to 71.6%; I² = 0%; Fig. 2). In Asian settings, 87.7% (95% CI 73.3% to 94.9%; I² = 82%; Fig. 2B) of people with allopurinol-induced SCARs in East and Central Asia tested positive for HLA-B*58:01, whereas the estimate for people in South and Southeast Asia was 93.4% (95% CI 90.5% to 95.4%; I² = 0%; Fig. 2B). Results of the sensitivity analyses were comparable to that of the primary analysis (eFig. 1–4, Supplementary Data).

Fig. 2.

Pooled prevalence of HLA-B*58:01 genotyping in the allopurinol-induced SCAR group. Abbreviations: 95% CI, 95% confidence interval; IV, inverse variance; NOS, Newcastle–Ottawa Scale (C, comparability; E, exposure; S, selection; T, total); SCAR, severe cutaneous adverse reaction. The pooled prevalence of HLA-B*58:01 was stratified by race (Panel A) and by ethnic origin (Panel B). We fitted the normal-normal random-effects model with restricted maximum likelihood estimator, using logit transformation and inverse logit back-transformation.

In allopurinol-tolerant group

We identified 19 studies reporting the prevalence of HLA-B*58:01 genotyping in the allopurinol-tolerant group^{7–9,11,12,14,15,37–48}. Among 3344 allopurinol-tolerant people, 360 tested positive for HLA-B*58:01. The pooled prevalence of HLA-B*58:01 genotyping in this group was 10.5% (95% CI 8.8% to 12.4%; I² = 53%; Fig. 3). There was no significant difference among races and ethnic origins (Fig. 3). The sensitivity analyses also showed similar findings (eFig. 5–8, Supplementary Data).

Fig. 3.

Pooled prevalence of HLA-B*58:01 genotyping in the allopurinol-tolerant group. Abbreviations: 95% CI, 95% confidence interval; IV, inverse variance; NOS, Newcastle–Ottawa Scale (C, comparability; E, exposure; S, selection; T, total). The pooled prevalence of HLA-B*58:01 was stratified by race (Panel A) and by ethnic origin (Panel B). We fitted the normal-normal random-effects model with restricted maximum likelihood estimator, using logit transformation and inverse logit back-transformation.

In population group

We identified 12 studies reporting the prevalence of HLA-B*58:01 genotyping in the population group^{7,8,10–14,16,35–37,42}. Among 9352 healthy people, 505 tested positive for HLA-B*58:01. The pooled prevalence of HLA-B*58:01 genotyping in this group was 5.8% (95% CI 2.9% to 11.5%; I² = 98%; Fig. 4). There was a high degree of heterogeneity between races (Asian and White; Fig. 4A), which could be due to the heterogeneity in East and Central Asia populations (Fig. 4B). The prevalence of HLA-B*58:01 genotyping in South and Southeast Asia (12.9%; 95% CI 9.5% to 17.3%; I² = 79%; Fig. 4B) was higher than that in Eastern and Western Europe (2.3%; 95% CI 1.2% to 4.3%; I² = 86%; Fig. 4B). We also found similar results in the sensitivity analyses (eFig 10–13, Supplementary Data). For Asian population, HLA-B*58:01 genotyping was more frequent in people of Han Chinese, Korean, Thai, and Vietnamese descent than in the Japanese (eFig. 9, Supplementary Data).

Fig. 4.

Pooled prevalence of HLA-B*58:01 genotyping in the population group. Abbreviations: 95% CI, 95% confidence interval; IV, inverse variance; NOS, Newcastle–Ottawa Scale (C, comparability; E, exposure; S, selection; T, total). The pooled prevalence of HLA-B*58:01 was stratified by race (Panel A) and by ethnic origin (Panel B). We fitted the normal-normal random-effects model with restricted maximum likelihood estimator, using logit transformation and inverse logit back-transformation.

Association between HLA-B*58:01 and allopurinol-induced SCARs

We identified 19 studies investigating the association between HLA-B*58:01 genotyping and allopurinol-induced SCARs^{7–9,11,12,14,15,37–48}. Having HLA-B*58:01 genotyping was associated with substantially increasing risk of allopurinol-induced SCARs (pooled OR 117.6; 95% CI 70.3 to 196.8; I² = 45%; Fig. 5). No subgroup differences were detected (Fig. 5). Result of Egger’s test implied asymmetry (p < 0.001), and this was likely due to publication bias (eFig. 18, Supplementary Data). The sensitivity analyses implied similar findings (eFig. 14–17, Supplementary Data).

Fig. 5.

Association between HLA-B*58:01 genotyping and allopurinol-induced SCARs. Abbreviations: 95% CI, 95% confidence interval; IV, inverse variance; NOS, Newcastle–Ottawa Scale (C, comparability; E, exposure; S, selection; T, total); OR: odds ratio. The pooled OR was stratified by race (Panel A) and by ethnic origin (Panel B). We fitted the normal-normal random-effects model with restricted maximum likelihood estimator, using logit transformation and inverse logit back-transformation.

Discussion

Summary of main results

We included 24 studies, with a total of 937 individuals with allopurinol-induced SCARs, 3344 allopurinol-tolerant individuals, and 9352 healthy individuals from the populations. The pooled prevalences of HLA-B*58:01 genotyping in the allopurinol-induced SCAR group, allopurinol-tolerant group, and population group were 88.5%, 10.5%, and 5.8%, respectively. Compared with the Whites in Eastern and Western Europe, Asian populations had higher prevalences of HLA-B*58:01 genotyping and had a greater risk for allopurinol-induced SCARs. Within the Asian settings, HLA-B*58:01 genotyping was more frequently detected in people of Han Chinese, Korean, Thai, and Vietnamese descent than in the Japanese.

Overall completeness and applicability of evidence

We identified a lack of evidence for the following commonly-reported races and ethnic origins: (1) races: Pacific Islander, Black, Hispanic or Latino/a/x, Indigenous, Middle Eastern or North African; (2) ethnic origins: North Africa, Sub-Saharan Africa, West Asia or Middle East, Pacific or Oceania, North America, Central America and Caribbean, and South America. There is also a gap in the prevalence (or risk) of allopurinol-induced SCARs in patients who tested positive for HLA-B*58:01 genotyping. Although the results of the EuroSCAR study showed an increasing risk of SCARs in patients receiving ≥ 200 mg of allopurinol per day, our data did not facilitate investigating the association between allopurinol dosage and SCARs¹. Our goal is to address these in the updated versions of this review.

Findings from this review can support HLA-B*58:01 screening and other preventive measures in people of Han Chinese, Korean, Thai, and Vietnamese descent due to the high prevalence of HLA-B*58:01 genotyping. However, the decision to implement these policies may highly depend on the feasibility and relevant costs. A lack of alternative pharmacologic options to allopurinol can also mitigate the applicability of evidence from this review, as high-risk populations with hyperuricaemia or gout will either be undertreated or exposed to allopurinol.

Potential biases in the review process

We followed a systematic approach to ensure replicability and mitigate biases in the review process^24,26. We searched and retrieved the studies from multiple sources with no restrictions on the language of publication. Each study was independently assessed for eligibility and risk of bias by at least 2 authors. Data were extracted using a standardised form. We pre-specified all analyses except for subgroup analysis of ethnicities, as this was not our primary objective. Since we did not have patient-specific data, we had to classify the subgroups using the dominant races and ethnic origins, which may have introduced some levels of bias from studies reporting multi-racial or multi-ethnic data. The threshold for excluding studies with a high risk of bias in the sensitivity analyses was also not widely recognised²⁸. The variability in HLA-B*58:01 genotyping techniques might influence the detection of this allele in the original studies. The detection of SCARs can also involve score-based diagnostic criteria with some levels of uncertainty, which might cause misclassification bias.

Agreements and disagreements with other studies or reviews

Although there are published and ongoing reviews of HLA-B*58:01 genotyping and allopurinol-induced SCARs^49–51, only a few systematically investigated the epidemiology of this allele worldwide. There has been no meta-analysis to identify the prevalence of HLA-B*58:01 genotyping in different sub-populations. The association between HLA-B*58:01 genotyping and allopurinol-induced SCARs was consistent with other reviews^50,51. However, our findings revealed a stronger magnitude, especially in the Asian populations.

Conclusion

We found a higher prevalence of HLA-B*58:01 genotyping in some Asian populations compared with the Whites. There is evidence to confirm a strong association between this allele and allopurinol-induced SCARs. Before initiating allopurinol for hyperuricaemia or gout treatment, people of Han Chinese, Korean, Thai, and Vietnamese descent should be screened for HLA-B*58:01 genotyping. This screening measure should also be applied to other populations with evidence of a high prevalence of HLA-B*58:01 genotyping. Further studies should target the under-reported races and ethnic origins to explore the prevalence of HLA-B*58:01 genotyping. The risk of developing allopurinol-induced SCARs in patients who tested positive also needs to be addressed. To facilitate the adoption of screening and other preventive measures, cost-effectiveness analysis should be conducted whenever and wherever possible.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

AGEP

Acute generalised exanthematous pustulosis

CI

Confidence interval

DHS

Drug hypersensitivity syndrome

DIHS

Drug-induced hypersensitivity syndrome

DRESS

Drug reaction with eosinophilia and systemic symptoms

EMM

Erythema multiforme major

HLA

Human leukocyte antigen

HSS

Hypersensitivity syndrome

IQR

Interquartile range

ISSN

International Standard Serial Number

MPE

Maculopapular eruption

NOS

Newcastle–Ottawa Scale

OR

Odds ratio

PRISMA

Preferred reporting items for systematic reviews and meta-analyses

SCAR

Severe cutaneous adverse reaction

SD

Standard deviation

SJS

Stevens–Johnson syndrome

SJS/TEN

Stevens–Johnson/toxic epidermal necrolysis overlap syndrome

TEN

Toxic epidermal necrolysis

Author contributions

H.T.P.: Conceptualisation, Methodology, Validation, Formal analysis, Investigation, Writing – original draft, Writing – review & editing. M.H.T.: Conceptualisation, Methodology, Formal analysis, Investigation, Writing – original draft, Writing – review & editing, Supervision, Project administration. T.-V.M.H.: Validation, Formal analysis, Investigation, Data, Writing – original draft, Writing – review & editing. A.-H.N.: Formal analysis, Investigation, Data, Writing – original draft, Writing – review & editing, Visualisation. M.-H.T.: Conceptualisation, Methodology, Validation, Formal analysis, Investigation, Data, Writing – original draft, Writing – review & editing, Visualisation, Supervision, Project administration.

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

HTP reported receiving speaking fees and travel reimbursement from Servier Vietnam Ltd and Pfizer Vietnam Ltd, grants from Servier Vietnam Ltd outside the submitted work. M-HT reported receiving travel reimbursement from Pfizer Vietnam Ltd, speaking fees and grants from Servier Vietnam Ltd outside the submitted work. Other authors declare no competing interests.