Association Between HLA genotypes and Oxcarbazepine-induced Cutaneous Adverse Drug Reactions: A Systematic Review and Meta-Analysis

- PURPOSE: To systematically review and quantitatively synthesize associations between HLA genotypes and oxcarbazepine-induced cutaneous adverse drug reactions (OXC-cADRs), including Stevens– Johnson syndrome (SJS) and maculopapular rash. METHODS: Studies investigating associations between HLA genotypes and OXC-cADRs were systematically searched irrespective of language, in PubMed, HuGENet (Human Genome Epidemiology Network), and the Cochrane Library from their inception until January, 2017. Inclusion criteria were studies investigating associations between HLA genotypes and OXC-cADRs that reported sufficient data for calculating the frequency of HLA genotype carriers among cases and controls. Overall odds ratios (ORs) with corresponding 95%CIs were calculated using a random-effects model to determine the association between HLA genotypes and OXC-cADRs. RESULTS: The initial searches identified 91 articles, of which 6 studies met the selection criteria. The studies included 229 patients with OXC-cADRs, 251 OXC-tolerant patients, and 2,358 participants from general populations of Han Chinese, Korean, and Thai ethnicities . Associations between HLA-B*1502 and OXC-induced SJS were found in both the general population [OR=30.2 (95%CI=3.45-264)] and in OXC-tolerant individuals [OR=26.4 (95%CI=7.98-87.6)]. An association between the HLA-B*1502 and OXC-induced maculopapular rash was found in the general population [OR=5.67 (95%CI=2.03-15.9)] while HLA-A*3101 also associated with OXC-induced maculopapular rash [overall OR=29.2 (95%CI=6.70-128)]. CONCLUSIONS: Strong associations between the HLA-B*1502 and OXC-cADRs (SJS and maculopapular rash) were found in both controls from general population and OXC-tolerant groups. There was also an association between HLA-B*3101 and OXC-induced maculopapular rash. For patient safety, genetic screening especially for HLA-B*1502 prior to OXC therapy at least in these closely related ethnicities is warranted. Further studies need to better define other ethnicities at risk and a wider range of MHC gene subtypes.


INTRODUCTION
Oxcarbazepine (OXC) is a keto-analogue of carbamazepine (CBZ) and approved as a monotherapy or adjunctive therapy to treat partial seizures in adults and children (1). It has less severe cutaneous adverse drug reactions (cADRs) and thus a safer choice for CBZ-intolerant patients (2). OXC-cADRs vary from mild maculopapular rash to severe reactions (i.e., Stevens-Johnson syndrome/toxic epidermal necrolysis, SJS/TEN), and drug rashes with eosinophilia and systemic symptoms with eosinophilia and systemic symptoms, DRESS). However, the most common OXC-cADR is maculopapular rash (3). 25-78% of CBZ-hypersensitive patients develop a cross reaction with OXC, while 29-40% of OXC-induced hypersensitive patients cross react with CBZ (4)(5)(6).
Several studies suggested that most cADRs caused by epileptic drugs (e.g., CBZ, phenytoin, lamotrigine) are drug-specific immune responses through human leukocyte antigens (HLAs) (7,8). HLA is a gene group encoding the major histocompatibility complex (MHC) located on chromosome 6 in humans (9, 10). MHCs are cellsurface receptors that capture and present self-and pathogen-derived peptides to T-cell receptors. The associations among MHC class I and II and OXC-____________________________________________ ABBREVIATIONS: BFDE = bullous fixed drug eruption; cADRs = cutaneous adverse drug reactions; CBZ = carbamazepine; CI = confidence intervals; DRESS = drug rash with eosinophilia and systemic symptoms; HLAs = human leukocyte antigens; HuGENet = Human Genome Epidemiology Network; HWE = Hardy-Weinberg equilibrium; MHC = major histocompatibility complex; NOS = The Newcastle-Ottawa scale; OR = odds ratio; OXC = oxcarbazepine; OXC-cADRs = oxcarbazepine-induced cutaneous adverse drug reactions; PCR= polymerase chain reaction; SJS = Stevens-Johnson syndrome; TEN = Toxic epidermal necrolysis cADRs have been studied in several epidemiological studies (11-14). However, these previous studies have shown a wide range of HLA genotypes (i.e., HLA-A*3101, HLA-B*1502), OXC-cADRs (i.e., maculopapular rash, SJS/TEN), and the magnitude of associations. A major limitation of the individual studies is the low incidence of OXC-cADRs. Small sample sizes among those studies may contribute observed variations. An important part of reducing cADRs is an understanding of population risks as well as HLA genotyping before drug treatment. Therefore, to better assess these risks, we aimed to review all relevant studies and to quantitatively synthesize the magnitude of the associations using a systematic review and meta-analysis technique.

SEARCH STRATEGY AND SELECTION CRITERIA
PubMed, Human Genome Epidemiology Network (HuGENet) and the Cochrane Library were systematically searched from their inception until January 2017 using keyword combinations or synonyms for "HLA genotypes" and "oxcarbazepine" without language or study design restrictions. Only human studies were included. Additional studies were retrieved from bibliographies of the included articles. Two reviewers (WT, ML) independently screened titles and/or abstracts for relevance followed by full-text article assessments for inclusion. Studies were included if: (1) HLA genotypes/OXC-cADRs associations were investigated; (2) all patients received OXC before HLA genotypes screening, and; (3) sufficient data for calculating the frequency of HLA genotypes carriers were reported. When studies shared the same population, the one reporting most data and patients was selected. Where data was insufficient for meta-analysis, additional data was sought from corresponding authors.
Two reviewers (WT, ML) extracted data by study design, eligibility criteria, definition, and diagnostic criteria for cases and controls, patient demographics, dose and duration of OXC exposure, the HLA genotyping technique and Hardy-Weinberg equilibrium (HWE) information. The genotype frequencies were examined by the HWE to determine whether the patients from the selected studies were representative of the population (15, 16). Study quality used the Newcastle-Ottawa scale (NOS) comprising three domains: selection, comparability, and outcome or exposure (17). All disagreements throughout were resolved by discussion between the reviewers until consensus was made.

DATA ANALYSIS
The included studies demonstrating an association between HLA genotypes and OXC-cADRs were characterized and summarized based on the most recent data. The overall odds ratios (ORs) with 95% confidence intervals (CIs) were calculated to determine associations between HLA genotypes and OXC-cADRs. All analyses were performed using the DerSimonian and Laird method under a random-effects model (18). The analyses were also performed separately on studies using different types of control (e.g., general population or OXCtolerant control), different design, different HLA genotypes, and different type of OXC-cADRs.
Information regarding HLA genotypes was obtained from the Allele Frequency Net Database, a genetic database, and from studies reporting allele frequencies of the genes; this group was defined as a general population control. Whereas OXC-tolerant control was defined by information of HLA genotypes in the control group obtained from patients who received OXC without any history of cADRs.
Statistical heterogeneity was assessed via the Q-statistics and I-squared tests. P-values≤0.10 indicated heterogeneity between studies (19). Isquared values of 25%, 50%, 75% denote a low, moderate, and high degree of heterogeneity across studies (20). All statistical analyses were performed using the R program (version 3.4.0) (R foundation for statistical computing, 2017).

HLA-A genotypes
The associations between HLA-A genotypes and OXC-cADRs of the included studies are summarized in Table 3. Two studies (21, 24) investigated associations between HLA-A genotypes (i.e., HLA-A*3101 and HLA-A*3201) and several types of OXC-cADRs (i.e., maculopapular rash, SJS, DRESS and BFDE). However, there was only sufficient data to assess the association between HLA-A*3101 and OXCinduced maculopapular rash that could be metaanalysed. Associations between HLA-A*3201 and OXC-induced maculopapular rash were determined by Moon et al (21) in general population control and OXC-tolerant groups. An association between HLA-A*3201 and OXCinduced maculopapular rash was found in general population controls (overall OR=8.46, 95%CI=1.37-52.2) ( Table 3). Moon et al (21) and Chen et al (23) found 8 out of 53 HLA-A*3101 carriers in their cases general population controls, the numbers of and controls were 5 out of 579. We found an association between HLA-A*3101 and OXC-induced maculopapular rash in the general population controls (overall OR=29.2, 95%CI=6.70-128; I 2 =0.0%, p=0.45) ( Figure 2). In OXC-tolerant controls, the number of HLA-A*3101 carriers in cases was 6 out of 35 and 15 out of 156 for controls. However, there was no statistically significant association between HLA-A*3101 and OXC-induced maculopapular rash in OXC-tolerant group ( Figure 2).

HLA-B genotypes
The associations between HLA-B genotypes and OXC-cADRs of the included studies are summarized in Table 3 Table 3). In OXC-tolerant controls, there were 13 HLA-B*1502 carriers out of 19 cases and 8 out of 109 for controls. The overall OR was 26.4 (95%CI=7.98-87.6; I 2 =0.0%, p=0.64) ( Figure 3A and Table 3).
In addition, we combined numbers of OXCinduced maculopapular rash and OXC-induced SJS events in to further investigate their associations between HLA-B*1502 and OXCinduced maculopapular rash and SJS. In general population controls only, HLA-B*1502 was associated with OXC-induced maculopapular rash and SJS; overall OR was 7.15 (95%CI=2.64-19.4; I 2 =0.0%, p=0.50) ( Figure 3C and Table 3

MHC class II and OXC-cADRs
The associations between MHC class II and OXC-cADRs are summarized in Table 3 (Table 3).

DISCUSSION
To our knowledge, this is the first systematic review and meta-analysis study to identify the associations between HLA genotypes and OXC-cADRs. In our study, 49 different HLA genotypes were identified as risks of OXC-cADRs but metaanalysis could be applied to only 4 of these (i.e.,  Table 3. Notably Other OXC-induced pathologies such as TEN are not well studied and, to our knowledge, are confined to studies where SJS and TEN data were pooled (7,11,25). Thus, based on current findings and the previous studies, subjects harboring the allele might develop TEN when given OXC.
All of the included studies were from Han-Chinese, Thai and Korean populations (12-14, 21-23). Whether these associations with OXC-cADRs are more widespread need large-scale studies in more ethnically diverse populations.
We found two studies (21, 23) that found associations between HLA-A*3101 and OXCinduced cADRs while for CBZ-induced cADRs (i.e., maculopapular rash) other ethicities including Caucasian and Japanese are susceptible (26-30). Furthermore, associations between HLA-A*3101 and CBZ-induced SJS/TEN were also reported in some studies (26, 30) which suggests that HLA-A*3101 could also induce OXC-cADRs. Thus in further studies that we suggest using wider ethnic groups, a wider range of MHC genes should be included.
Recently, 3D molecular coupling models of HLA protein molecules and carbamazepine, oxcarbazepine and abacavir were developed (31, 32). These models shed an understanding of how HLA molecules binds specifically to their ligands and potentially causes those adverse drug reactions (i.e. Steven-Johnson syndrome, toxic epidermal necrolysis, DRESS) (31). Despite the structural similarities of OXC and CBZ and their reported cross-reactivity, OXC does not share the HLArelated risk factors with CBZ-induced SJS/TEN which are 30-40 folds more than for OXC in Han Chinese (24). In addition, some patients carrying HLA-B*1502 with a history of CBZ-induced SJS could tolerate OXC (23). To understand the biological basis of this, the mechanism of the cross reactivity needs further studies.

CONCLUSION
Strong associations between the HLA-B*1502 and OXC-cADRs (SJS and maculopapular rash) were found in both controls from cases using either control from either the general population or and OXC-tolerant groups. In general population controls, OXC-induced maculopapular rash was associated with HLA-A*3101. These strong links were detected in Korean, Han-Chinese and Thai ethnicities. Therefore, a genetic screening in these ethnicities should precede an OXC treatment.   Abbreviations: BFDE= bullous fixed drug eruption; cADRs = cutaneous adverse drug reactions; DRESS = drug rash with eosinophilia and systemic symptoms; HLA = human leukocyte antigen; MP rash = maculopapular rash; ND = Not determined; OXC = oxcarbazepine; SJS = Stevens-Johnson syndrome. red or heavy pigmented eruptions with blisters mostly on the same sites with re-exposure to the causative agents. Abbreviations: ALDEN = Algorithm of drug causality for epidermal necrolysis; BFDE = bullous fixed drug eruption; cADRs = cutaneous adverse drug reactions; DRESS = drug rash with eosinophilia and systemic symptoms; HLA = human leukocyte antigen; MP rash = maculopapular rash; NA = Not applicable; NOS = The Newcastle-Ottawa scale; NR = Not report; OXC = oxcarbazepine; SJS = Stevens-Johnson syndrome