This is a case-control study with the control group defined by patients tolerant with CBZ. The study was conducted in the Department of Neurology at NDGD Hospital from 1 January 2019 to 31 December 2020. The analysis included a total of 7 cases of CBZ-induced SJS, 12 cases of CBZ-induced MCARs, and 18 cases of CBZ-tolerant control as shown in Figure 1.

All patients were recruited retrospectively with the definitions based on the information written in the medical records. The definition of CBZ tolerant is the patients who administered CBZ and the attending physician did not write any diagnosis of SJS or other skin manifestations of allergy. The definition of CBZ-induced SJS is the patients who administered CBZ and the attending physician wrote the diagnosis of SJS. The definition of CBZ-induced MCARs is the patients who administered CBZ and the attending physician wrote the diagnosis of itching skin, hypersensitivity reaction, CBZ allergy, anaphylaxis level 1, skin reaction, or symptomatic allergy. The patients who administered both CBZ and PHT were excluded from this study, only epileptic patients treated with monotherapy CBZ were included. Patients without information of genotype were also excluded from this study. The diagnosis of SJS/TEN was performed by certified dermatologists in accordance with the hospital’s guidelines, which included the following criteria: 1) Onset of symptoms within 2 months of initiating CBZ. 2) Presence of a rash affecting the face, upper body, limbs, or spreading extensively across the entire body. 3) Mucosal damage observed in at least two natural cavities, such as the eyes, nose, mouth, vagina, or anus. 4) Skin detachment (10% for SJS, 30% for TEN) or the presence of Nikolsky’s sign.

Within 76 patients with available DNA, 30 DNA samples were sent to the genotyping service at the laboratory of Pharmacogenomic and Personalize Medicine (PPM) of Ramathibodi Hospital Mahidol University, Bangkok, Thailand, 46 DNA samples were sent to the genotyping service at University Medical Center (UMC), University of Medicine and Pharmacy at Ho Chi Minh city. The genotyping method used at PPM was Luminex™ flow cytometry. In brief, the sample DNA binds complementarily to a panel of probes, which was designed with known nucleic sequences. The fluorescence detection technology was used to identify successful binding complexes, and hence, identify the sequence and genotype of the sample. The genotyping method used at UMC was real-time Polymerase Chain Reaction (real-time PCR), using Tagman™ genotyping assay.

All statistical analyses were done by R software version 4.2.3 (R Foundation for Statistical Computing, Vienna, Austria). Student’s t-test was used to compare the differences between 2 independent groups. Chi-squared tests were used to compare the ratio of males and females between 2 groups; Fisher’s exact test was employed to compare the risk of alleles between the cases and the controls.

This analysis aimed to compare our conclusion with the Vietnamese population in the both the North and the South by evaluating the impact of HLA-B*15:02 in Vietnamese epileptic patients treated with CBZ, utilizing various databases such as Pubmed, Google Scholar, Cochrane Library, and ScienceDirect (Figure 2). The keywords used were “HLA-B*15:02 Vietnamese”, “HLA-B*15:02 Vietnam’’, “SJS”, “Steven-Johnson Syndrome”. To select the suitable studies, the selection criteria were: 1) Containing the case-control analysis of HLA-B*15:02 and SJS, 2) Targeting Vietnamese epileptic patients, 3) Including the clear definitions of case and control with corresponding numbers, and the exclusion criteria were: 1) Duplicating studies, 2) Researching about the development of genotyping methods in Vietnam, 3) Being a case report or cross-sectional study. Data analysis and visualization were performed on Rstudio, using random effect with a p-value lower than 0.05 is considered statistically significant, and with I² higher than 50% was defined as heterogeneous. The estimation method used for the random effect is restricted maximum likelihood.

This study was approved by the Ethics Committee of NDGD Hospital, Ho Chi Minh City, Vietnam, under approval number 23-2015/CN-HĐĐĐ, on 13 October 2015. As we used only retrospective data from health records while also maintaining patient confidentiality, no informed consent was required for this study.

Table 1 describes the characteristic of the participants. The age of patients was widely distributed from 25 to 85 years old. The separation of gender was fairly equal between males (21 patients) and females (16 patients) in total participants. However, there was a higher proportion of males in the SJS group. The percentage of males respectively was 50% and 55.6% in the group of CBZ-induced MCARs and CBZ-tolerant control. All three groups contained patients with chronic diseases, such as hypertension or dyslipidemia. Nearly 30% of epileptic patients had a medical history of cerebral infarction or brain hemorrhage.

Among twelve carbamazepine-induced MCARs patients, five patients were diagnosed with itching skin and hypersensitivity reactions, six patients were diagnosed with symptomatic allergy and skin reactions of allergy, and one patient was diagnosed with anaphylaxis level 1 due to CBZ allergy. It is important to note that none of the twelve patients exhibited symptoms specific enough to be classified as SJS/TEN.

Table 2 shows the HLA allele polymorphism in the epileptic southern population of Vietnam, where a total of 22 alleles were observed. The most popular HLA-B alleles are 15:02 (20%), 38:02 (10%), and 46:01 (10%). There is a 5% of frequency in each of the following alleles: 7:05; 18:01, 37:01, 57:01. The rest 15 alleles account for around 1.7%–6.7% of the whole population.

Table 3 illustrates that in the Southern Vietnamese population, the HLA-B*15:02 allele is a statistically significant risk factor for SJS (p = 0.017), but not for MCARs (p = 0.660). The odd ratio of the HLA-B*15:02 allele is 12.5. In other words, the carriers of this allele have 12.5 times higher risk than non-carriers, in terms of SJS. The sensitivity and specificity are 71.4% and 83.3% respectively, meaning that the hospital’s protocol can correctly identify 71.4% of SJS patients, and correctly identify 83.3% of non-SJS patients. Besides, the PPV and NPV are 62.5% and 88.2% respectively. These two parameters are interpreted in the context that the results of HLA-B*15:02 are available. In these cases, if the result of a patient is positive, the possibility for this patient to have SJS is 62.5%, and if the result of a patient is negative, the possibility for this patient to not have SJS is 88.3%. Moreover, the NNT and NNG are respectively 2 and 9, which confirms that to protect 1 patient from SJS, the intervention of replacing CBZ with an alternative drug has to be done in 2 patients, or the genetic test has to be done in 9 patients.

Out of the twelve cases of allopurinol-induced MCARs, three patients tested positive for HLA-B*15:02. While the strength of evidence is not large enough to be statistically significant, it is worth noting that the rate of HLA-B*15:02 in the MCARs group is higher compared to the group of tolerant controls.

A meta-analysis was performed about the risk of HLA-B*15:02 and SCARs in Vietnamese with a total of 6 studies (Van Nguyen et al., 2015; Van Nguyen et al., 2017; Huyen et al., 2020; van Nguyen et al., 2021; Chu, 2022; Bui et al., 2023). All these 6 studies were conducted and recruited patients exclusively in the northern region of Vietnam, specifically at Bach Mai Hospital, Tam Anh Hospital (Ha Noi branch), and the National Hospital of Dermatology and Venerology. Surprisingly, no eligible studies conducted in the southern region of Vietnam were identified or included in this meta-analysis (Supplementary Table S1). Interestingly, a recent study by Bui TP et al. (Bui et al., 2023) reported a negative association between HLA-*B15:02 and SCARs in Vietnamese in January 2023. However, our analysis using pooled data from all six studies showed a strong positive association between HLA-B*15:02 and SCARs, as determined by both common effect and random effects models (Figure 3). Carriers of HLA-B*15:02 had 12.82 times higher odds of developing SCARs compared to non-carriers (p-value <0.01). The odd ratio calculated from previous studies in the meta-analysis was consistent with the odd ratio in the current study.