| BACKGROUND:Adverse drug reactions (ADRs), especially severe adverse drug reactions (SADRs), have become an intractable problem in both clinical practice and for pharmaceutical companies due to its dose-independent, rare and unpredictable characteristics. A growing number of studies have found that many SADRs are strongly associated with human leukocyte antigen alleles (HLA). For instance, Han Chinese individuals who carry HLA-B* 15:02 have a 2505-fold risk of developing Stevens-Johnson syndrome/toxic epidermal necrolysis (SJS/TEN) after taking carbamazepine (CBZ) when compared with HLA-B* 15:02 negatives. In another case, the risk of abacavir-induced hypersensitivity syndrome (ABC HSS) is 960-fold higher in HLA-B*57:01 positive Caucasian population than the negatives. The discovery of these pharmacogenomics biomarkers promises that SADRs can be predicted in advance by pre-screening susceptible HLA alleles before making medical decisions, thus facilitating personalized medicine and averting SADRs.Although many associations between ADRs and HLA have been reported, results from these studies are scattered and results are incomparable among studies. How to integrate these drug-HLA associations and how to make independent results comparable become a very important issue. Additionally, accurate diagnostic criteria of ADRs can significantly affect both the discovery and the clinical application of biomarkers but presently receive little attention. For example, two diagnostic criteria are very popular in identifying ABC HSS, clinically suspected hypersensitivity (CSH) and immunologically confirmed hypersensitivity (ICH); however, it remains uncertain whether these criteria have an influence on the strength of the ABC HSS and HLA-B*57:01 association.OBJECTIVES:Based on published drug-HLA associations, this study aims to construct a comprehensive and detailed drug-HLA association database, HLADR, which is a pharmacogenomics database with curated associations between ADRs and HLA and to statistically analyze the HLADR data. Besides, this study also aims to make a meta-analysis of the association between ABC HSS and HLA-B*57:01 to evaluate the influence of diagnostic criterion on the strength of the association.METHODS:The HLADR database was built through an extensive literature search, review, and curation of reports of drug-HLA associations and was implemented in MySQL for data storage and PHP for web functions. R program, Cytoscape software and RevMan software were applied for HLADR data analysis and meta-analysis.RESULTS:With a total of 1513 records (i.e., drug-HLA pairs associated with ADRs), the HLADR database, which curated almost all published drug-HLA associations, was constructed (http://pgx.fudan.edu.cn/hladr/) to serve as a comprehensive drug-HLA associations sharing platform to early identify and prevent ADRs. Statistical analysis showed that the drug-HLA association matrices mainly corresponded to severe cutaneous adverse drug reactions (SCARs), including HSS, SJS, and TEN, which were mostly induced by nervous system drugs and anti-infective drugs, and the HLA class I and class II alleles were the most susceptible alleles. In addition, the drug-HLA associations revealed a strong ethnicity specificity. The meta-analysis of ABC HSS-HLA-B*57:01 associations showed that HLA-B*57:01 was related to ABC CSH with a pooled odds ratio (OR) of 29.11 (P<0.00001,95%CI=17.79-47.65), whereas the pooled OR for ABC ICH was as high as 1216.06 (P<0.00001, 95%CI=299.24-4941.92). Moreover, HLA-B*57:01 showed a higher sensitivity in predicting the risk of ABC ICH than in predicting ABC CSH (0.960 vs 0.441).CONCLUSIONS:HLADR is the first database that is dedicated to the reported drug-HLA pairs associated with ADRs and could be accessed freely. It represents a comprehensive collection of almost all drug-HLA associations with data quality checked and comparable performance parameters calculated and normalized. It could serve as an important central portal for researchers to acquire integrated information about drug-HLA associations with ADRs. Mining the rich information in HLADR may lead to new research hypotheses and opportunities. In addition, a clear ethnicity dependency of the drug-HLA associations in the drug-HLA relationship network implied that selecting study samples from a pure ethnic origin should be emphasized in order to realize personalized medicine. Finally, the significantly different association strengths between HLA-B*57:01 and ABC CSH and ABC ICH suggest that diagnostic criteria of ADRs in pharmacogenomics studies should be standardized in order to improve the predictable power of HLA biomarkers. In conclusion, this study will have a positive impact on the development of pharmacogenomics and personalized medicine. |
PDF Full Text Request