Individuals were enrolled in southern China from 2009 to 2017, and treated at the Second Affiliated Hospital of Guangzhou Medical University, Guangzhou First People's Hospital, the First Affiliated Hospital of Jinan University, and Guangdong 999 Brain Hospital in Guangzhou. The study was approved by the hospital ethics committee. All participants or their parents provided written informed consent in accordance with the Declaration of Helsinki.

Cases with cross-reactivity were defined as sequential rashes in response to two or more AEDs in the same individual. For example, when a patient experienced a skin rash induced by CBZ, CBZ was discontinued, and 2 weeks later the patient was treated with LTG, then rashes appeared again. The diagnosis of AEDs-induced cross-reactivity was confirmed by a dermatologist based on the clinical and morphological characteristics of the patients' skin. MPE was defined as skin eruptions without mucosal or systemic involvement. SJS and TEN were diagnosed according to Roujeau's diagnostic criteria. Hypersensitivity syndrome was defined as a fever, rash, eosinophilia, and systemic manifestation (e.g., hepatitis and nephritis).

Five-hundred healthy volunteers who have not taken AEDs were recruited as normal controls. To verify the significant allele, we further increased the sample size of normal controls from 500 to 5,270. All healthy volunteers were recruited from the Healthy Physical Examination Center in the Second Affiliated Hospital of Guangzhou Medical University, and had not taken AEDs. Because most of the patients had taken CBZ, we recruited another cohort comprising of individuals with CBZ-induced MPE and CBZ-tolerant to clarify whether AEDs-induced cross-reactivity and CBZ-induced MPE have the same risk HLA allele. The cases with CBZ-induced MPE and CBZ tolerant controls were recruited from the Second Affiliated Hospital of Guangzhou Medical University between 2007 and 2018. CBZ tolerant controls were participants with epilepsy who took CBZ for at least 3 months without evidence of cutaneous adverse reactions.

All cases and controls were Southern Han Chinese.

Sequence based high-resolution HLA-A/B/C/DRB1 genotyping was performed by Shanghai Tissue Bank Biotechnology Co., Ltd. (Shanghai, China).

We performed statistical analyses using SPSS version 17.0 (SPSS, Inc., Chicago, IL, USA). Fisher's exact tests were used to assess the difference in the presence of HLA alleles between the case and control groups. Because CBZ and LTG were the most common culprit drugs, we conducted an independent t-test to compare the mean age, latency to cADRs, and initial and maximum dosages between one group of CBZ/LTG as the first culprit drug and the other group of CBZ/LTG as the second culprit drug. P < 0.05 (two-sided) were considered significantly different. The corrected P (Pc) values were estimated by Bonferroni's correction for multiple comparisons (n = 12, 16, and 14 for HLA-A, HLA -B, and HLA-C alleles, respectively).

We performed meta-analyses on data obtained from other studies to investigate the relationship between HLA-B^*13:01 allele and cADRs induced by AEDs or/and non-AEDs. A complete search of online databases, including MEDLINE, EMBASE, Google Scholar, was conducted.

The following terms were used in our searches: “HLA-B^*13:01” or “human leukocyte antigen B^*13:01,” “Stevens–Johnson syndrome” or “SJS,” “toxic epidermal necrolysis” or “TEN,” “cutaneous adverse drug reactions” or “cADRs,” “maculopapular eruption” or “MPE,” “Antiepileptic drugs” or “AEDs.” The latest search was conducted on April1, 2018.

Criteria for the selection of studies were: (1) the report was of a case-control study on association between HLA-B^*13:01 and cADRs induced by AEDs or non-AEDs; (2) the genotyping method and ethnicity were provided; (3) the presence of HLA-B^*13:01 in the cases, either the population controls or tolerant controls was reported or could be obtained from the authors or other sources; (4) the most recent publication with the largest number of samples was selected when duplicate publications were identified. Exclusion criteria were: (1) reports were not of case-control studies; (2) repeated studies; (3) studies did not indicate the presence of HLA-B^*13:01 in the case group and control group; (4) abstracts and reviews; (5) non-human studies.

The following information were extracted: the first author of the study, publication year, ethnicity of the study population, presence of HLA-B^*13:01 allele among cADRs cases and controls, total number of cADRs cases and controls, and main results. The methodological quality was assessed according to the recommendations of the Cochrane Collaboration Handbook (https://www.cochrane.de).

Data managements and analyses were conducted using STATA (Version 10.1 Stata Corp LP, College Station, TX, USA). Odds ratios (ORs) with corresponding 95% confidence intervals (CIs) were calculated to verify the association between the HLA-B^*13:01 allele and drugs-induced cADRs. Begg's test was used to evaluate publication bias (16). Statistical heterogeneity among studies was assessed via the Q statistic and I2 tests (17). A P < 0.1 and an I²-value >50% were defined as evidence of statistical heterogeneity, under which the association was assessed using the random model. Analyses were also performed separately on studies using different types of control patients (e.g., drug-tolerant or normal controls) and different drugs induced cADRs.

Seven articles met the criteria. Data from 8 studies, including the present study, were used for meta-analysis. The data included the populations from Thailand and China.

A total of 23 patients with AEDs-induced cross-reactivity and 500 healthy volunteers were recruited in this study. The demographic variables and clinical manifestations of cases are summarized in Table 1. There were 14 females and 9 males. The average age was 24 years (range 6–53 years). All cases occurred as sequential rashes induced by two of the following drugs: CBZ, LTG, OXC, levetiracetam, topiramate, valproic acid, PB, and PHT. The phenotype of AED-induced cross-reactivity in this cohort was MPE without mucosal or systemic involvement.

CBZ and LTG were the most common culprit drugs. With respect to the sex ratio, mean age, latency to cADRs, and initial and maximum dosages, no significant difference was found between one group of CBZ/LTG as the first culprit drug and the other group of CBZ/LTG as the second culprit drug.

The HLA genotypes of patients with AEDs-induced cross-reactivity are shown in Table 2. The distribution of HLA alleles in the 23 cases and 500 normal controls is shown in Table 3. Three alleles were significantly associated with AEDs-induced cross-reactivity. The HLA-B^*13:01 allele was found in 8 (34.78%) of the 23 AEDs-induced cross-reactivity patients and 73 (14.60%) of the 500 normal controls (P = 0.02; OR: 3.12; 95% CI: 1.28–7.62; Pc = 0.32, n = 16 for HLA-B^*13:01 correction). The HLA-A^*11:02 allele was found in two (8.70%) of the 23 AEDs-induced cross-reactivity patients, and in none (0%) of the 500 normal controls (P = 0.002; OR: 68.32; 95% CI: 6.83–683.53; Pc = 0.02, n = 12 for HLA-A^*11:02 correction). The HLA-C^*04:03 allele was found in three (13.64%) of the 22 AED-induced cross-reactivity patients and 10 (2.01%) of the 498 normal controls (P = 0.01; OR: 7.71; 95% CI: 1.96-30.3; Pc = 0.14, n = 14 for HLA-C^*04:03 correction). Furthermore, one case with HLA-A^*11:02 and one case with HLA-C^*04:03 were positive for HLA-B^*13:01. If excluding these cases, the presence of HLA-A^*11:02 and C^*04:03 in the patients was 4.35% (1/23) and 9.09% (2/22), respectively. The presence of the two alleles HLA-A^*11:02 and C^*04:03 is very low in the normal Chinese population (http://www.allelefrequencies.net, such as 0 and 2.01% in this cohort). These results suggested HLA-B^*13:01 should be considered further. To exclude the possibility of lost significance, we compared the presence of HLA-B^*13:01 between the 23 patients of AEDs-induced cross-reactivity and a larger control cohort containing 5,270 normal individuals, which revealed a significant association between HLA-B^*13:01 and AEDs-induced cross-reactivity (8/23, 34.78%, vs. 763/5,270, 14.48%; P = 0.014, OR:3.15, 95%CI: 1.33–7.46).

Since 18 subjects (18/23, 78.26%) showed cross-reactivity induced by two aromatic AEDs, we compared the presence of HLA-B^*13:01 between the 18 patients of aromatic AEDs-induced cross-reactivity (6/18, 33.33%) and the two normal control cohorts, which showed no significant difference (6/18 vs. 73/500, P = 0.066; 6/18 vs. 763/5,270, P = 0.054, respectively).

Because most of the AEDs-induced cross-reactivity patients (14/23, 60.9%) had taken CBZ, we recruited another cohort comprising of individuals with CBZ-induced MPE (145) and CBZ-tolerant controls (179) to clarify whether AEDs-induced cross-reactivity and CBZ-induced MPE have the same risk HLA allele. The HLA-B^*13:01 presence was significantly higher in AEDs-induced cross-reactivity individuals than in CBZ-tolerant individuals (8/23, 34.78%, vs. 28/179, 15.64%; P = 0.049, OR: 2.88, 95%CI: 1.11–7.42), while its presence in CBZ-induced MPE was numerically higher than that in CBZ-tolerant individuals without statistical significance (33/145, 22.76%, vs. 28/179, 15.64%; P = 0.103) (Supplementary Table 1).

In addition, in this cohort, two of 23 AEDs-induced cross-reactivity patients (2/23) were positive for HLA-B^*15:02, three of 23 patients (3/23) were positive for HLA-A^*24:02, and one patient (1/23) was positive for HLA-A^*31:01. Compared to the normal controls, there was no significant difference in the presence of HLA-B^*15:02, HLA-A^*24:02, or HLA-A^*31:01 (8.7% vs. 12.80%, P = 0.80; 13.04% vs. 29.60%, P = 0.09; 4.35% vs. 4.20%, P = 1.0, respectively) (Table 3).

To clarify the relationship between HLA-B^*13:01 and cADRs, a meta-analysis was performed using data from different populations. Seven case-control studies including Thailand and China population met the criteria (Supplementary Table 2). The meta-analysis analyzed 8 studies, including the present study. The data covered two aspects, i.e., HLA-B^*13:01 is associated with cADRs induced by single AED, and HLA-B^*13:01 is associated with cADRs induced by single non-AED. There was a total of 254 cases with cADRs induced by single AED, 446 AEDs tolerant controls, and 6,678 normal controls, and there was a significant association between HLA-B^*13:01 and cADRs induced by single AED (P = 0.000), compared to tolerant controls and normal controls (Figure 1A). There was a total of 117 cases with cADRs induced by single non-AED, 1,195 non-AEDs tolerant controls, and 3,309 normal controls, and there was a significant association between HLA-B^*13:01 and cADRs induced by single non-AED (P = 0.000), compared to tolerant controls and normal controls (Figure 1B). We further combined the cases of cADRs induced by AEDs with the case of cADRs induced by non-AEDs, and there was a significant association between HLA-B^*13:01 and cADRs in Thai and Chinese populations (P = 0.000), compared to tolerant controls and normal controls (Figure 1C).

This study had several limitations. First, we did not divide the cases into subgroups according to cross-reactivity induced by the same order of drug prescribed due to the limit sample. Second, we did not exactly match the drug tolerant subgroup. To decrease deviation, we investigated the significance of the association by comparison with a large number of normal individuals. A further study with a larger samples size and cases from multiple centers along with exactly matched tolerant controls should be conducted to confirm that the HLA-B^*13:01 allele is a risk factor for AEDs-induced cross-reactivity, and a common risk factor for cADRs.

Z-ZZ was employed by Shanghai Tissue Bank Biotechnology Co. Ltd, China. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer PK declared a past co-authorship with several of the authors, BQ, JW, and NH, to the handling editor.