Research Article For reprint orders, please contact:
[email protected]
HLADR: a database system for enhancing the discovery of biomarkers for predicting human leukocyte antigen-mediated idiosyncratic adverse drug reactions Aim: To establish a database for the associations between idiosyncratic drug reactions (IDRs) and human leukocyte antigens (HLAs) and to systematically assess the characteristics of the drug–HLA associations. Materials & methods: Electronic databases were searched to extensively identify drug–HLA association studies from 1966 to present. Results: A drug-HLA-IDR database, HLADR, was created. The drug–HLA relationship network clearly reflected an ethnicity dependency of the associations. The positive predictive values and the negative predictive values demonstrated that other potential factors may also regulate the occurrence of HLAspecific IDRs. Conclusions: Constructing studies with samples from homogeneous ethnic groups and identifying cofactors that affect negative predictive values and positive predictive values will become necessary to enhance the predictability of HLA biomarkers for future research on IDRs. Keywords:association • biomarkers • human leukocyte antigen • idiosyncratic drug reactions • pharmacogenomics
Idiosyncratic drug reactions (IDRs), known as type B adverse drug reactions (ADRs) [1,2] , are generally regarded as rare, dose-independent and unpredictable [2,3] . They are usually not identified until after the drug is marketed and taken by large populations of patients due to the rarity of the adverse events. Although such drugs may prove valuable to many patients, IDR occurrences usually result in restricted use or withdrawal of the drugs; therefore, intensive research has aimed at identifying risk factors including genetic ones. Genetic variants of the human leukocyte antigens (HLAs), which are highly polymorphic proteins encoded by the human major histocompatibility complex region on chromosome 6p21.3, have been reported to be strongly associated with various IDRs [4– 10] , demonstrating the potential to inform the drug discovery process and clinical utilities of using HLA alleles as biomarkers to predict and prevent IDRs. Indeed, several drug–HLA-induced IDR associations are so strong that they led to a label change as
10.2217/bmm.15.98 © 2015 Future Medicine Ltd
approved by the US FDA [47] . For example, HLA-B*57:01 and HLA-B*15:02 are the risk alleles for abacavir-induced hypersensitivity and carbamazepine-induced Stevens–Johnson syndrome (SJS), respectively, and screening patients for the risk HLA alleles prior to initiating therapy with abacavir or carbamazepine is recommended. However, the predictive power of the existing HLA biomarkers is not always satisfactory. For example, although B*57:01 is recommended to predict hypersensitivity to abacavir, 52% of patients carrying B*57:01 are tolerant [6] , indicating that other concurrent, patient-specific factors may be required for IDR pathogenesis. In addition, while B*57:01 was reported as a genetic determinant of flucloxacillin-induced liver injury, 12% of the cases do not carry B*57:01 [4] , implying that other factors independent of B*57:01 may be essential to cause liver injury. Although recent studies have highlighted the fact that abacavir can trigger specific toxic T-cell responses by altering the peptide rep-
Biomark. Med. (Epub ahead of print)
Tingting Du‡,1, Lun Yang*,‡,2, Heng Luo‡,2,3, Peng Zhou2, Hu Mei2, Jiekun Xuan1,2, Qinghe Xing4, Baitang Ning2, Donna L Mendrick*,2 & Leming Shi*,1,2,3 Center for Pharmacogenomics & State Key Laboratory of Genetic Engineering, School of Life Sciences & School of Pharmacy, Fudan University, 826 Zhangheng Road, Shanghai 201203, China 2 National Center for Toxicological Research, US Food & Drug Administration, 3900 NCTR Road, Jefferson, AR 72079, USA 3 University of Arkansas at Little Rock/ University of Arkansas for Medical Sciences Joint Bioinformatics Graduate Program, 2801 South University Avenue, Little Rock, AR 72204, USA 4 Institutes of Biomedical Science, Fudan University, 138 Shanghai Medical School Road, Shanghai 200032, China *Authors for correspondence:
[email protected] [email protected] lun.yang@ gmail.com ‡ These authors contributed equally to this study 1
part of
ISSN 1752-0363
Research Article Du, Yang, Luo et al. ertoire loaded on the surface of the HLA-B*57:01 protein [11–13] , this mechanism is insufficient to explain all immune-mediated IDRs. Further investigation into the molecular mechanisms of IDRs involving HLAs is highly warranted and a thorough review of existing knowledge is required. Unfortunately, there is a lack of a central information portal dedicated to the published drug-HLA-IDR association pairs, making a systematic comparison of different studies on associations between drug-induced adverse reactions and HLA alleles difficult, if not impossible. Thus, a manual collection of almost all published studies on HLA alleles associated with IDRs was assembled into an in-house database, HLADR (open access at [48]). A 2 by 2 confusion matrix was then extracted from each and every case–control study focusing on genetic information of IDR-related HLAs. A comprehensive analysis of the contents in the HLADR database and a summary of mechanistic studies on drugHLA-IDR associations were performed. This led to the hypothesis that potential cofactors may contribute to adverse reactions and enabled some proposed tentative solutions for enhancing the predictability of HLA biomarkers for IDRs. Materials & methods Literature search & data integration
PubMed, EBSCO, Ovid MEDLINE and Web of Science were systematically searched by terms of ‘HLA, IDRs, ADRs, pharmacogenomics, association or hypersensitivity [Title/Abstract]’ from the year 1966 to January 15 2015. Reference lists of identified articles were also reviewed in case potential articles were missed. Included studies should meet the following criteria: study design: case-healthy control, case-tolerant control or case reports; the article type should be original research articles, case reports/series, conference scenes or clinical trials and reviews including meta-analysis were excluded; associations between drug-induced rather than chemicals-induced adverse reactions and HLAs were focused, but a fraction of important chemicals that could result in occupational diseases such as asthma were also included; meeting abstracts, whether English or not were also included if they met the above criteria. Duplicates were removed by sorting the records, by their references and manually inspection. Any information related to the association between IDRs and HLAs was extracted. The IDR names were recorded in consistency with PharmGKB. Not only basic information such as drugs, IDRs, HLA alleles and ethnic origin was recorded but also a 2 by 2 contingency table reflecting the association strengths (p-value, odds ratio [OR], sensitivity, specificity, pos-
10.2217/bmm.15.98
Biomark. Med. (Epub ahead of print)
itive-predictive values [PPVs] and negative-predictive values [NPVs]) for the specific drug-HLA-IDR pair was constructed. However, the association strengths across different studies were incomparable since different statistical methods were applied in the original reports. To ensure the comparability for these variables, Fisher’s exact test and Haldane’s modification were applied to recalculate the p-value and OR for every drug-HLA-IDR pair, respectively. Additionally, names of IDRs and HLA alleles were also standardized based on authoritative databases PharmGKB [49] and IMGT/HLA [50] . The ethnical and geographical origins of the study subjects as well as the molecular structural information and FDA drug label changes resulting from the drug–HLA associations were also provided. HLADR database construction
MySQL [51] for data storage and PHP (version 5.6.7, download at [52]) for web functions were applied to build the HLADR database system. Statistical analysis
The OR was recalculated by applying Haldane’s modification, which added 0.5 to all cells to accommodate a possible zero count in a 2 by 2 contingency table. p-value was recalculated by using Fisher’s exact twotailed test. A p-value less than 0.05 can be considered as statistically significant. OR >1 means patients with a specific HLA allele have the tendency of developing IDRs, while OR
