Our findings indicate that HLA-B*5801 allele is significantly associated with increased risk of developing SJS/TEN in patients using allopurinol. This severe adverse event associated with allopurinol could be prevented if such genetic information is known a priori. Clinicians and policy makers may use our findings as a foundation to support the implementation of genetic testing prior to initiation of allopurinol.
These findings reveal that the risk of developing SJS/TEN among those allopurinol users with HLA-B*5801 is significantly increased by 80-97 times compared to those without the gene. The sensitivity analyses suggested that the summary odds ratios remained significant regardless of populations. These findings are suggestive of the potential of genotyping in a wide range of population.
Several strengths of our research work deserve more discussion. First, our study is the first one including all kinds of studies determining association of HLA-B*5801 and SJS/TEN development. Second, all SJS/TEN cases were in accordance with the consensus definition [16–18]. These stringent inclusion criteria lowered the risk of misclassification, resulting in increased reliability of our research findings. Third, our meta-analysis adopted the Newcastle-Ottawa scale  as a tool to evaluate quality of all case control studies. The Newcastle-Ottawa approach has been reported in several articles to have a good validity for assessing the observational study [25–28]. The average quality score of 5 represented a good quality of overall evidence.
Meta-analysis is not only pooling studies' findings, but this analysis can also determine heterogeneity occurred among the selected studies. The results from our sensitivity analysis demonstrated no significant heterogeneity among populations despite differences in their allele frequency between Asian and non-Asian. Thus, it is justified to perform such analysis despite similarity in the trend of results from the chosen reports. Our findings revealed that despite some differences in several characteristics (e.g. race, sources and selection of control), the association is still consistent and suitable for pooling using meta-analytic technique .
Despite the absence of statistical significance of publication bias tests, a possible existence of publication bias cannot be excluded. Because of a relatively small number of total sample size and limited number of studies, the power of publication bias tests might not be sufficient. Even though the meta-analysis shows a consistent significant association of HLA-B*5801 and SJS/TEN in all studies, the overall estimates should be interpreted with caution.
One interesting debate ongoing regarding association between HLA-B*5801 on the risk of developing SJS/TEN is its nature. It has been proposed by Hung et al  that HLA-B*5801 was necessary but might not be the only factor related to the risk of SJS/TEN development. Lonjou et al  stated that the gene was an important element, but it was neither necessary nor sufficient. All cases in three studies [10, 13, 15] had HLA-B*5801 allele, whereas the prevalence ranged from 40-80% among cases in the other 3 studies [11, 12, 14]. The lack of HLA-B*5801 in some cases in the latter 3 studies made it clear that such gene was not the sole factor required. We believe that SJS/TEN development is multifactorial. Other factors including other genes and environment might have a role in the development. What we can conclude was consistent with the summary of Hung et al , which was that HLA-B*5801 had a significant role in SJS/TEN occurrence. More research is necessary to further elucidate on how this gene and other factors are involved in allopurinol-induced SJS/TEN pathogenesis.
The increased risk of developing SJS/TEN in those subjects with HLA-B*5801 can be explained by the involvement of cytotoxic T-cells and amplification following cytolytic cytokine. The pathogenesis of drug hypersensitivity is postulated to have direct involvement with human leukocyte antigen (HLA) genes [9, 30]. HLA genes located on the major histocompatibility complex (MHC) region of the human chromosome 6p21.3, play a central role in the immune reaction by presenting an antigen to the T cell receptor (TCR) .
Despite several postulated mechanisms for the association of HLA-B*5801 and allopurinol-induced SJS/TEN, thus far, there is no definitive proven mechanism. Nonetheless, an immunologic mechanism might play a role in SJS/TEN development [31–33]. Several protein molecules play an important role in this complex process. For instance, MHC proteins, T-cell receptors and some cellular drug metabolizing enzymes (i.e. cytochrome P450 and Phase II metabolizing enzymes) can contribute to the hapten formation and ensuing immunological responses . Although, when the hapten is completely developed and presented at the surface of a nascent T-cell and B-cell clone, they need to proliferate and fully exert their cellular actions. Thus, the fate of these cells can determine the final outcome . These cells may divide and expand resulting in the final immune responses, or get into apoptosis and die out. To elucidate this clonal selection/expansion event among T-cells and B-cells in the SJS/TEN development, a stochastic behavior of the cells is possible . Therefore, a mathematic model describing clonal selection/expansion processes may be useful [37–39]. Nevertheless, this cellular event of the clonal selection/expansion may contribute to the low incidence of SJS/TEN observed in the general population.
Apart from the fate of the nascent antigen specific T- and B-cells, downstream events after binding between the HLA-B*5801 and its receptor may influence the T-cell stimulation. As HLA-B*5801 is presented at the surface, it requires T-cell receptor to couple with the antigen. Subsequently, the immunological system is stimulated [10, 30]. The lack of SJS/TEN development might be explained by a malfunction of the T-cell receptor, which could be due to T-cell receptor polymorphism [40, 41]. Another putative cause might be an existence of a gene exerting inhibitory effect, resulting in lower risk of SJS/TEN development. In a study of Alfirevic and colleagues  investigating the association between genes and SJS/TEN among carbamezepine (CBZ) users, despite small sample size, among those whose possess HLA-B*0702, a significant protective effect for the development of severe reaction was reported. The mechanism of this protective effect is not fully understood.
The association between HLA-B*5801 allele and allopurinol-induced SJS/TEN is consistent across different populations, both Asian and non-Asian [10–15], whereas, an association between HLA-B*1502 and CBZ-induced SJS/TEN demonstrated less consistency [12, 43, 44]. HLA-B*1502 allele, whose an association with CBZ-induced SJS/TEN is significant in most Asian populations, but not in Japanese and European population. These discrepancies might be explained by the different genetic background. Since this gene is also present in many populations (i.e. African, Caucasian, and Asian), therefore, the association of HLA-B*5801 allele with allopurinol-induced SJS/TEN can be found in various ethnic groups. On the other hand, HLA-B*1502 allele is only present in limited populations (i.e. Asian population) .
Interestingly, HLA-B*5801 has a more pronounced effect on allopurinol-induced SJS/TEN compared to those found in the case of HLA-B*1502 and CBZ-induced SJS/TEN. In the latter case, the incidence may be associated with other contributing factors (i.e. other genes) to trigger the adverse drug reaction, whereas those factors may play less role in initiating SJS/TEN in case of HLA-B*580. A study in Japan  reported that CBZ-induced SJS/TEN was associated with HLA-B*1511, a member of HLA-B75 type that also includes HLA-B*1502, HLA-B*1508, HLA-B*1515, HLA-B*1521, HLA-B*1530, and HLA-B*1531. These suggested that not only HLA-B*1502 but also other HLA-B75 members are risk factors for CBZ-induced SJS/TEN. By comparison, the strong association between HLA-B*5801 and allopurinol-induced SJS/TEN has been validated in different populations and may be a universal phenomenon since it has been identified in all Chinese, Japanese, Thai, Korean and European patients [10–15].
Notably, a main caveat in this study is the potential misinterpretation of our research findings. This meta-analysis revealed the significant association of HLA-B*5801 allele and the increased risk of allopurinol-induced SJS/TEN. This does not mean that having HLA-B*5801 test done will result in absolutely no risk of allopurinol-induced SJS/TEN. Monitoring of signs and symptoms in these patients are still needed.
Our study is also only limited to the investigation of the association between HLA-B*5801 and allopurinol-induced SJS/TEN. In fact, there has been a number of studies reporting a potential association of DRESS (Drug Rash with Eosinophillia and Systemic Symptoms) and HLA-B*5801 [10, 47]. The interpretation of our findings should be limited to SJS/TEN cases and not be generalizable to other severe cutaneous adverse reactions (SCARs).
The implications of our research findings are more likely significant among population with high prevalence of HLA-B*5801. Based on the strong association of the presence of HLA-B*5801 alleles and SJS/TEN, it is presumed that the attributable risk of SJS/TEN due to the existence of this gene is larger among those with the gene. Allele frequency was reported as high as 6-8% among Southeast Asian population and < 1% among Western European population [8, 45]. From our results, a genotypic testing of the HLA-B allele may have a benefit to the patients before receiving the drug; particularly in high risk population (e.g. Asian). Knowing the HLA-B allele status of allopurinol users may guide clinicians in determining the optimal choice in order to lower the likelihood of allopurinol-induced SJS/TEN. Presumably these events are avoidable; it might be prudent to consider whether such genetic test should be adopted into routine practice in high-risk population. In order to convince the policy makers to support such genetic testing, a formal cost-effectiveness analysis is warranted.