The outcome of hepatitis B virus (HBV) infection and the response to HBV surface antigen (HBsAg)–derived vaccines are determined to a major degree by the host genetic background.1, 2 Other outcomes, including progression to cirrhosis and/or liver cancer, may also be influenced by host genetics, but the evidence supporting this theory is not yet available. Neither persistent infection with HBV nor vaccine nonresponse are simple Mendelian traits. They should be considered as complex traits where viral (or vaccine), environmental, and host genetic variables contribute to the outcome. Furthermore, unlike simple Mendelian traits, several polymorphic genes will exert effects on the outcome, rather than one major gene. The polymorphic variants involved in complex traits are appropriately regarded as alleles rather than mutations, as they are found commonly in healthy populations. Possession of a susceptibility allele is neither required nor sufficient for a disease trait to develop, and we should therefore expect that the true strength of genetic associations is small. If odds ratios are used as the measure of increased susceptibility conferred by possession of an allele, then values in the range of 1.2–2 would be expected.

In an attempt to overcome the shortcomings of many genetic association studies, Nature Genetics published an editorial in 1999 that set out a list of criteria for genetic association studies.3 The criteria, listed in the box, establish a hurdle over which few studies would ever cross. While no one would criticize the requirement for rigorous phenotypic selection and accurate genotyping, it is often difficult to assemble an independent cohort with which to demonstrate reproducibility. However, two other requirements should now be given consideration: the size of the cohort and the linkage disequilibrium context of the polymorphisms that have been genotyped.

Criteria for Genetic Association Studies

  • Plausible biological context

  • Rigorous phenotypic selection (case selection)

  • Independent replication

  • Rigorous genotyping

  • Low P values

  • Appropriate statistical analysis

  • Transmission disequilibrium test

Additional Criteria

  • Study size

  • Linkage disequilibrium context (haplotype structure)

Note. Adapted from Nature Genetics3

There has been an enormous increase in the number of studies that seek to identify genes that influence susceptibility to persistent HBV infection or vaccine nonresponse since we entered the “post-genomic” era. These studies are fuelled by the unlimited availability of single-nucleotide polymorphisms, the relative ease of performing genotyping assays based on polymerase chain reaction technology, and the desire to identify major disease susceptibility gene(s). Unfortunately the literature is now littered with unreproducible genetic association studies that confuse the reader and have an understandable impact on the willingness of editors to accept further manuscripts for publication. Small studies often report strong associations with alleles that appear to exert a major effect on susceptibility to a specific disease trait. When sample size increases, the association often disappears or becomes disappointingly weak.4 If we expect to see only associations with odds ratios in the range of 1.2–2 then the number of cases required for genetic association studies should be thousands rather than the tens or hundreds that we are used to seeing.

It is often assumed that the locus that has been genotyped is the causal variation responsible for increases or decreases in susceptibility to the disease trait. However, the genotyped locus may not be causal but may be in linkage disequilibrium (LD) with the causal variant. Any two loci on the same chromosome may be inherited together as a block. The likelihood of this event's happening may be measured as the LD parameter D′. Although for many years it was believed that D' was principally a function of the distance between two loci, it is now clear that the situation is far more complicated. Hence, two loci that are seemingly distant may have strong LD, as was observed with HFE and human leucocyte antigen (HLA)A3; similarly, two adjacent loci may have low LD. The implications of this complexity is that positive genetic association studies should not be assumed to have identified a causal variant unless reported in the context of an LD or haplotype map.

In this issue of HEPATOLOGY, Deng and colleagues provide evidence that polymorphisms in the estrogen receptor influence susceptibility to persistent HBV infection.5 The strength of this association, with an odds ratio of only 1.4, will be unimpressive to many readers, but given the size of the study, it is probably a realistic assessment of the impact of this polymorphism on susceptibility to persistent infection. The low odds ratio probably reflects the functional impact of the nucleotide substitution and not the importance of the estrogen receptor in the biology of HBV infection. The study is compelling for a number of reasons. First, it meets all the criteria in Figure 1 apart from independent replication. Second, the authors have included a comprehensive analysis of linkage disequilibrium in the region. Third, the study is sufficient in size.

Vaccine nonresponse is a relatively small issue compared to the problem of persistent HBV infection. Unfortunately, no clear route to novel therapeutic interventions may be deduced from genetic associations with persistent HBV infection at present. On the other hand, studying the genetics and immunology of vaccine nonresponse may be far more tenable than studying persistent HBV infection, because age at exposure, antigen sequence, and dose are carefully controlled. If, as appears likely, there is overlap in the mechanisms of vaccine nonresponse and HBV persistence, then the results may inform our therapeutic interventions in the future.

A number of studies have sought genetic associations between vaccine nonresponse or the outcome of HBV infection and the major histocompatibility (MHC) loci. Vaccine nonresponse has been associated with a number of MHC class II alleles, including HLA-DRB1*03, HLA-DRB1*04, and HLA-DRB1*07.6 On the other hand, self-limited HBV infection has been consistently associated with HLA-DRB1*1301/2 in three different ethnic populations.7–9 In a recent issue of HEPATOLOGY, the association of HLA-DRB1*07 is reconfirmed.10 While MHC/disease associations appear to fulfil the requirement for plausible biological context, in reality the mechanism(s) underlying these associations are not readily apparent. It is likely that failure of antigen presentation to CD4+ T helper cells explains the association, but this failure has not yet been demonstrated. Furthermore, if the association relies on the inability of HLA-DRB1*07 to present HBsAg epitopes effectively, then it remains to be explained why HLA-DRB1*07 heterozygotes also have a deficient vaccine response. HLA-DRB1*07 has also been associated with persistent HBV infection, raising the worrying possibility that those individuals who fail to respond to vaccination are also at increased susceptibility to persistent infection.11 HLA-DRB1*03 has also been associated with vaccine nonresponse. It is intriguing to note that vaccination with the HBsAg–based vaccine containing the Pre-S region of the HBV envelope glycoprotein may overcome vaccine nonresponse, particularly in individuals carrying the HLA-DRB1*03 allele.12 This fact suggests that only the additional epitopes found in the Pre-S region may be presented by HLA-DRB1*03. However, HLA-DRB1*03 is part of a haplotype that frequently carries the complement component C4-null allele C4Q0, which is more strongly associated with vaccine nonresponse in some studies.13

Hohler's study of twins vaccinated for HBV demonstrate that genetic control of vaccine response is only partly explained by the MHC.2 Non-MHC candidate susceptibility genes include the cytokines because of their immunoregulatory roles. Wang provides evidence of association with allelic variants in the IL-4 and IL-12 genes.10 Unfortunately, due to study size and the heterogeneity of the cohort, these novel associations will need independent replication, as advocated by the authors, before becoming accepted. In addition, although the genes clearly fit the requirement for biological plausibility more details of the functional consequences of the polymorphisms need to be elucidated. Cytokine manipulation of HBsAg vaccine response is a realistic therapeutic advance that has already been tested in renal patients using granulocyte monozyte colony stimulating factor (GM-CSF).14

If we hope to use genetic-susceptibility studies to provide novel insights into the determinants of persistent HBV infection, we must diverge from hypothesis-driven studies and temporarily suspend the requirement for biological plausibility. Identifying novel or unexpected genes involved with HBV elimination requires that we use a hypothesis-free approach in the form of genome-wide scanning in affected sibling pairs. Using these techniques, genes become candidates for genetic association studies on the basis of their position on a chromosome rather than on the basis of our preconceived ideas about their functional role. Results from such studies will be reported in the near future.

A consequence of our strive for better-quality genetic association studies will inevitably be an increase in the number of negative or weak associations. To demonstrate that associations are reproducible and to avoid the problems of publication bias, these studies should still be published. There has never been room in the journals for negative genetic association studies, so a solution might be to create an online-only journal to publish good-quality studies irrespective of the outcome. Any volunteers?


  1. Top of page
  2. References