Meta-Analysis of Genetic Variants Associated With HBV Infection Susceptibility and Hepatocellular Carcinoma Risk.

1
Introduction
Hepatitis B virus (HBV) infection remains a major global public health challenge, affecting an estimated 254 million individuals worldwide. Collectively, hepatitis B and C account for approximately 3500 deaths each day, with mortality rates continuing to increase. Even though we have made great strides in vaccines, antiviral therapies, and health policies, the number of people getting sick or dying from HBV is still surprisingly large [1].
HBV infection presents a wide range of clinical outcomes, from spontaneous recovery after acute infection to chronic hepatitis, cirrhosis, or even hepatocellular carcinoma (HCC) [2, 3]. How HBV progresses in the body is heavily influenced by a person's DNA, a fact we have understood for about 20 years now. Whether our immune system successfully eliminates the virus or fails, leading to HCC, often comes down to the genetic traits in our immune pathways [4, 5, 6]. Using genome‐wide association studies (GWAS), experts have identified that certain spots within the HLA class II region, particularly the HLA‐DP and HLA‐DQ loci, are central to this process. These locations are vital because they allow the body to ‘present’ antigens to T‐cells so the immune system can start fighting [7, 8, 9, 10]. Other genes such as CD40 and UBE2L3, along with complement components like C2 and CFB, also play a part in how someone's body manages an HBV infection [11, 12]. Across East Asian populations, where chronic HBV infection and deaths from HBV remain a pressing public health challenge [1], numerous GWAS have revealed genetic loci linked to viral persistence, immune regulation, and metabolic pathways. However, the strength of these genetic effects has varied considerably from one study to another. Moreover, most work has concentrated on individual variants or limited stretches of the genome, making it difficult to compare genetic effects across different HBV infection and HCC pathways.
Hepatocellular carcinoma (HCC) is the most severe result of a chronic HBV infection and continues to be a leading cause of cancer deaths globally [1]. HCC occurs through several different causes, such as chronic HBV and hepatitis C virus (HCV) infections, liver disease from alcohol, and non‐alcoholic fatty liver disease (NAFLD). While these subtypes share some common immune and inflammatory processes, their genetic structures are different [13]. Regarding HBV‐related HCC, GWAS have found variants in HLA‐DQ, STAT4, and KIF1B, which suggest that immune recognition, cytokine signaling, and cell death are involved in cancer development [14, 15]. For HCV‐related HCC, MICA and DEPDC5 have been identified as genes that increase risk, although these results have not always been the same across different ethnic groups [16, 17, 18, 19]. Meanwhile, genetic variants in metabolic genes, particularly PNPLA3, TM6SF2, and HSD17B13, have emerged as major determinants of HCC risk in alcohol‐related and NAFLD‐related liver diseases, highlighting the contribution of hepatic lipid metabolism and hepatocellular stress to carcinogenesis [20, 21, 22, 23, 24, 25, 26, 27].
Since findings from separate studies have been hard to compare directly, we conducted meta‐analyses that bring together previous genetic association studies on HBV persistence and HCC risk. The main goal of our study was to find out if we could use genetic markers to predict whether an HBV infection would clear up or turn into a lasting problem. We also looked into how these genetic factors change the risk of HCC in different scenarios, such as cases involving HBV, HCV, alcohol use, or NAFLD. To get a more precise picture, we gathered and analysed data from many different studies. Ultimately, we wanted to demonstrate the way human DNA differences influence both the development of HBV and the overall chances of facing HCC.

2
Materials and Methods
2.1
Literature Search and Study Selection
To identify genetic association studies on susceptibility to HBV virus and HCC, we conducted a systematic literature search. Primary data were collected from the NHGRI‐EBI GWAS Catalogue (https://www.ebi.ac.uk/gwas/), which provides an organized collection of published genome‐wide association studies. We conducted two independent searches using the phenotype terms “hepatitis B virus infection” and “hepatocellular carcinoma,” which yielded 22 and 20 records. Excluding five studies that overlapped between the two searches resulted in 37 records. To make our analysis more complete, we included extra GWAS results from earlier research done in our own lab, which specifically looked at Korean groups. These studies have already been peer‐reviewed and published, providing 25 separate datasets that study HBV infection and the risk of HCC in East Asian populations. We removed one study that was already in the GWAS Catalogue to avoid counting it twice, leaving us with 24 additional studies for our work.
A secondary literature search was performed in PubMed to capture genetic association studies not indexed in the GWAS Catalogue. SNP identifiers (rsIDs) obtained from the GWAS Catalogue searches were combined with the terms “hepatitis B,” “HBV,” “hepatocellular carcinoma,” and “HCC” to identify additional relevant publications. This search identified 25 additional relevant publications. Studies were included if they reported genetic association data for HBV infection or HCC in human populations. Studies were excluded if they: (1) were articles without extractable genetic association data, (2) examined cohorts unrelated to either HBV infection or HCC (regardless of HCC aetiology), or (3) did not provide sufficient data for meta‐analysis (e.g., lacking genotype frequencies, odds ratios, or allele counts). Based on these criteria, six studies examining cohorts unrelated to HBV infection or HCC were excluded to maintain phenotypic specificity.
This resulted in a final dataset of 80 studies considered for meta‐analysis. However, not all 80 studies were included in the final meta‐analysis. Studies were further excluded if they were review articles or meta‐analyses without original data, or if the phenotype definitions did not match the predefined etiologic subgroups (HBV‐related, HCV‐related, alcohol‐related, or NAFLD‐related HCC). Additionally, as described in the Statistical Analysis section, only SNPs reported in two or more independent studies were included to ensure reproducibility. The characteristics of the included studies are summarized in Table 1, with more detailed information and additional references provided in Tables S1 and S2. The study selection process is summarized in a flowchart (Figure 1). All references cited therein are included in the Reference List [7, 8, 9, 10, 11, 12, 14, 15, 16, 17, 19, 22, 26, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46].

2.2
Data Extraction
From each included study, we extracted the following information: first author, publication year, study population (ethnicity and geographic location), phenotype definition (HBV persistence for HBV infection studies; HBV‐related HCC, HCV‐related HCC, alcohol‐related HCC, or NAFLD‐related HCC for HCC studies), sample size (number of cases and controls), single nucleotide polymorphism (SNP) identifiers (rsIDs), chromosomal location, candidate gene, genotype or allele frequencies, odds ratios (ORs) or relative risks, 95% confidence intervals (CIs), and p‐values. When studies reported results from multiple independent cohorts or populations, we used the combined results provided in the publication. If 95% CIs were not reported but ORs and p‐values were available, 95% CIs were calculated using standard methods. For studies reporting results from multiple genetic models (e.g., additive, dominant, recessive), the additive model was preferentially selected for consistency across studies, as this model is most commonly reported in genome‐wide association studies. Data were extracted by one investigator, and the collected information was checked by the corresponding authors when needed to make sure it was accurate or to resolve any uncertainties in the reported findings.

2.3
Statistical Analysis
We performed a meta‐analysis to look at how SNPs are related to phenotypes. We removed variants that came from only one study because we were worried about false‐positive results. We kept only those found in two or more independent studies, and we used the DerSimonian‐Laird random‐effects model to get pooled OR and 95% CI for each SNP. Between‐study heterogeneity was assessed using Cochran's Q test and quantified using the I
2 statistic (I
2) and tau‐squared (τ2). An I
2 value greater than 50% was considered indicative of substantial heterogeneity. Subgroup analyses were performed to examine the consistency of genetic associations across different phenotypic categories. For HCC‐related analyses, subgroups were defined based on underlying aetiology: HBV‐related HCC, HCV‐related HCC, alcohol‐related HCC, and NAFLD‐related HCC. For HBV infection analyses, subgroups included studies examining chronic infection versus spontaneous clearance. Statistical analyses were performed using R software (version 4.5.1) with the metafor package. All statistical tests were two‐sided, and a p‐value < 0.05 was considered statistically significant.

2.4
Publication Bias and Sensitivity Analysis
We assessed publication bias using two methods for SNPs that had a sufficient number of studies. First, we examined funnel plots to see whether they looked symmetrical or not. Second, we applied Egger's regression test as a statistical way of detecting publication bias, where a p‐value under 0.1 was treated as evidence of possible bias. Sensitivity analyses were conducted using leave‐one‐out analysis for variants with three or more studies, in which the meta‐analysis was repeated by sequentially removing one study at a time to assess the influence of individual studies on the pooled effect estimates and to identify potential outliers. All publication bias assessments and sensitivity analyses were performed using R software (version 4.5.1) with the metafor package.

3
Results
3.1
Study Selection and Characteristics
Following the systematic literature search and screening process, a total of 80 studies were considered for meta‐analysis (Figure 1). After applying the inclusion and exclusion criteria described in the Methods section, the final meta‐analysis included studies examining genetic associations with HBV infection susceptibility and HCC risk across different etiologic subgroups. The HBV infection studies predominantly examined variants in HLA genes (HLA‐DPA1, HLA‐DPB1, HLA‐DQB1, HLA‐DQB2) and immune‐related genes (CD40, C2, CFB, UBE2L3), while HCC studies investigated both HLA variants and metabolic gene variants across HBV‐related, HCV‐related, alcohol‐related, and NAFLD‐related HCC. The majority of studies were conducted in East Asian populations, with genotyping methods and study‐specific demographics briefly overviewed in Table 1, while more comprehensive information is available in Tables S1 and S2.

3.2
Genetic Variants Associated With HBV Infection Susceptibility and Persistence
We looked at eight genetic variants that are connected to HBV persistence, and we found that many of them had significant associations. Most of these were in the HLA region and in immune‐related genes (Table 2; Figure 2). We put the forest plots for the non‐significant variants in Figure S1.
Among the variants examined, CD40 rs1883832 and C2 rs9267665 demonstrated the most robust statistical evidence for association with HBV persistence. CD40 rs1883832 showed a significant association with higher risk of HBV persistence (pooled OR = 1.19, 95% CI: 1.14–1.24, p = 2.51 × 10−15) based on two large studies. This variant showed no heterogeneity between the studies (I
2 = 0%, Q = 0), which means the effects were very consistent across populations. Both CD40 rs1883832 and C2 rs9267665 showed strong associations with no heterogeneity, so we think their genetic effects on HBV infection outcomes are real. C2 rs9267665 had the biggest effect size of all the variants we tested. The pooled OR was 2.26 (95% CI: 1.84–2.79, p = 1.77 × 10−14), and there was almost no heterogeneity between the two studies (I
2 = 0%, Q = 0.14).
Among HLA variants, HLA‐DPA1 rs3077 exhibited a protective effect against HBV persistence (pooled OR = 0.70, 95% CI: 0.50–0.97, p = 0.031) based on 11 studies. However, this association showed very high heterogeneity across studies (I
2 = 99.1%, Q = 1125.44), suggesting substantial variability in effect estimates between populations or study designs. Despite the significant pooled estimate, the extreme heterogeneity indicates caution is needed in interpreting this result. HLA‐DQB1 rs2856718 was associated with increased risk of HBV persistence (pooled OR = 1.35, 95% CI: 1.14–1.60, p = 6.31 × 10−4) across five studies. This variant also demonstrated high heterogeneity (I
2 = 95.8%, Q = 94.19), although the effect remained statistically significant. The moderate effect size and consistent direction of association across studies support a genuine, albeit heterogeneous, effect of this HLA variant on HBV infection outcomes. In contrast, several variants did not show statistically significant associations with HBV infection. HLA‐DPB1 rs9277535, examined in 10 studies, showed a non‐significant trend toward protection (pooled OR = 0.77, 95% CI: 0.54–1.10, p = 0.152) with very high heterogeneity (I
2 = 99.2%). Similarly, HLA‐DQB2 rs7453920 (6 studies) showed no significant association (pooled OR = 1.02, 95% CI: 0.56–1.85, p = 0.958) with extreme heterogeneity (I
2 = 99.3%). UBE2L3 rs4821116 (2 studies) and CFB rs12614 (2 studies) also showed no significant associations (p = 0.507 and p = 0.914, respectively), though both exhibited high heterogeneity suggesting variable effects across studies.

3.3
Genetic Variants Associated With HCC Risk by Aetiology
Our meta‐analysis of 11 genetic variants associated with HCC risk across four etiologic subgroups revealed distinct patterns of genetic associations (Table 3; Figure 3). Forest plots for variants that did not reach statistical significance are provided in Figure S2.
3.3.1
HBV‐Related HCC
Among variants examined in HBV‐related HCC, HLA‐DQ rs9275319 showed the strongest association. This variant had a very significant association with higher HCC risk (pooled OR = 1.52, 95% CI: 1.39–1.65, p = 2.37 × 10−22) based on three studies. This association also showed no heterogeneity (I
2 = 0%, Q = 0.33), meaning the effects were consistent across the study populations. The strong and consistent association, along with the known role of HLA‐DQ in viral antigen presentation, strongly supports a causal role in HBV‐related hepatocarcinogenesis. KIF1B rs17401966 showed a protective effect against HBV‐related HCC (pooled OR = 0.77, 95% CI: 0.61–0.97, p = 0.025) across five studies. Although this association was statistically significant, it showed a lot of heterogeneity (I
2 = 88.5%, Q = 34.91). STAT4 rs7574865 showed no significant association with HBV‐related HCC risk (pooled OR = 1.06, 95% CI: 0.86–1.31, p = 0.599) based on five studies, and the heterogeneity was high (I
2 = 80.7%). This lack of association is interesting because STAT4 has an important role in immune signalling, and it suggests that not all immune‐related variants contribute equally to HCC risk in HBV‐infected individuals.

3.3.2
HCV‐Related HCC
For HCV‐related HCC, two variants were examined, and they showed different results. MICA rs2596542 was significantly associated with higher HCC risk (pooled OR = 1.44, 95% CI: 1.03–2.02, p = 0.033) across three studies. However, this analysis showed very high heterogeneity (I
2 = 90.4%, Q = 20.9), which means there was a lot of variation in the effect estimates. DEPDC5 rs1012068 showed no significant association with HCV‐related HCC (pooled OR = 1.36, 95% CI: 0.83–2.22, p = 0.22) based on two studies, and the heterogeneity was extremely high (I
2 = 96.1%, Q = 25.32). The wide confidence interval and high heterogeneity suggest either real differences in effects across populations or the influence of possible confounding factors.

3.3.3
Alcohol‐Related HCC
Three variants in metabolic genes were significantly linked to alcohol‐related HCC, all with strong statistical support. Among them, PNPLA3 rs738409, a famous variant in adiponutrin, showed a much higher risk for HCC (pooled OR = 1.64, 95% CI: 1.36–1.97, p = 2.03 × 10−7) across three different studies. Although this association showed moderate heterogeneity (I
2 = 76.4%, Q = 8.48), the effect stayed highly significant and the direction was consistent. TM6SF2 rs58542926 demonstrated an even stronger effect on alcohol‐related HCC risk (pooled OR = 1.81, 95% CI: 1.62–2.02, p = 1.29 × 10−25) with no heterogeneity (I
2 = 0%, Q = 1.35). This finding is among the strongest and most reliable genetic associations identified in our meta‐analysis. HSD17B13 rs72613567 was found to have a significant protective effect against alcohol‐related HCC (pooled OR = 0.80, 95% CI: 0.73–0.88, p = 2.69 × 10−6). This was supported by two studies showing no heterogeneity (I
2 = 0%, Q = 0.02).

3.3.4
NAFLD‐Related HCC
In NAFLD‐related HCC, PNPLA3 rs738409 exhibited a significant association with increased risk (pooled OR = 1.76, 95% CI: 1.29–2.41, p = 3.81 × 10−4) across two studies, with no heterogeneity (I
2 = 0%, Q = 0.88). The effect size was numerically larger than that observed in alcohol‐related HCC, though confidence intervals overlapped. TM6SF2 rs58542926 also showed a significant association with NAFLD‐related HCC (pooled OR = 1.50, 95% CI: 1.22–1.84, p = 9.37 × 10−5) based on two studies with minimal heterogeneity (I
2 = 0%, Q = 0.92). Notably, the effect size was somewhat smaller than that observed in alcohol‐related HCC (OR = 1.81), potentially reflecting differences in the pathogenic mechanisms between these two metabolic liver disease etiologies. MBOAT7 rs641738, another metabolic gene variant, did not show a significant association with NAFLD‐related HCC (pooled OR = 1.34, 95% CI: 0.79–2.27, p = 0.283) based on two studies. However, high heterogeneity was observed (I
2 = 84.8%, Q = 6.58), and the wide confidence interval suggests insufficient power to detect a moderate effect.

3.4
Between‐Study Heterogeneity
Substantial to considerable heterogeneity was observed for several genetic associations, so these results should be interpreted carefully. Among the HBV infection variants, HLA‐DPA1 rs3077, HLA‐DPB1 rs9277535, and HLA‐DQB2 rs7453920 all had I
2 values over 99%, which means there was extreme variability in the effect estimates across the studies. This very high heterogeneity probably reflects real differences in genetic effects across populations, possibly because of different HLA haplotype backgrounds, differences in HBV genotype distributions, or gene–environment interactions that may happen in different geographic regions.
HLA‐DQB1 rs2856718 also showed high heterogeneity (I
2 = 95.8%), although the association was still statistically significant. The τ2 value of 0.036 shows moderate between‐study variance on the log odds ratio scale. For immune‐related genes CD40 and C2, no heterogeneity was observed (I
2 = 0%), but these estimates are based on only two studies each, so they should be interpreted carefully. Among HCC variants, HCV‐related HCC showed the highest heterogeneity, with MICA rs2596542 (I
2 = 90.4%) and DEPDC5 rs1012068 (I
2 = 96.1%) both showing a lot of variability. This may be due to differences in HCV genotype distributions, different levels of liver disease severity, or population‐specific genetic backgrounds that can change these associations. In contrast, metabolic gene variants in alcohol‐related and NAFLD‐related HCC generally showed low or no heterogeneity. HSD17B13 rs72613567, PNPLA3 rs738409 (in NAFLD‐HCC), and TM6SF2 rs58542926 (in alcohol‐related HCC and NAFLD‐HCC) all showed I
2 = 0%, which suggests very consistent effects across populations and studies. This consistency supports the strong and repeatable nature of metabolic gene associations with HCC risk in metabolic liver diseases. PNPLA3 rs738409 in alcohol‐related HCC showed moderate heterogeneity (I
2 = 76.4%), higher than in NAFLD‐related HCC (I
2 = 0%), potentially reflecting greater variability in alcohol consumption patterns or gene‐alcohol interactions across study populations.

3.5
Publication Bias and Sensitivity Analysis
We used Egger's regression test to look for publication bias in variants that had three or more studies (Table 4). The funnel plots for HBV infection and HCC variants can be found in Figures S3 and S4. In the HBV infection group, we found significant publication bias for HLA‐DPA1 rs3077 (Egger intercept = −0.053, p = 0.043) and HLA‐DPB1 rs9277535 (Egger intercept = 0.040, p = 0.044). These results suggest potential asymmetry in the funnel plots, which could indicate either publication bias favoring significant results or genuine heterogeneity in effect sizes related to study characteristics. However, given the already extreme heterogeneity observed for these variants (I
2 > 99%), the Egger test results likely reflect the underlying heterogeneity rather than classic publication bias.
No significant publication bias was detected for HLA‐DQB1 rs2856718 (p = 0.210) or HLA‐DQB2 rs7453920 (p = 0.779), though moderate asymmetry cannot be ruled out given the limited number of studies. For variants with only two studies (UBE2L3 rs4821116, CD40 rs1883832, C2 rs9267665, CFB rs12614), Egger's test could not be performed as it requires at least three studies. Among HCC‐related variants, significant publication bias was detected only for MICA rs2596542 in HCV‐related HCC (Egger intercept = 0.006, p = 0.009), based on three studies. This finding suggests potential selective reporting or small‐study effects, though the limited number of studies makes this assessment less reliable. No significant publication bias was detected for KIF1B rs17401966 (p = 0.848), STAT4 rs7574865 (p = 0.509), HLA‐DQ rs9275319 (p = 0.758), PNPLA3 rs738409 in alcohol‐related HCC (p = 0.690), or TM6SF2 rs58542926 in alcohol‐related HCC (p = 0.917).
We also did leave‐one‐out sensitivity analyses to see if taking out one study at a time would change the pooled effect estimates (Figure S5). We did this for HBV infection variants that had three or more studies, which were HLA‐DPA1 rs3077, HLA‐DPB1 rs9277535, HLA‐DQB1 rs2856718, and HLA‐DQB2 rs7453920. For these variants, the pooled ORs remained stable and statistically significant after excluding individual studies, indicating robust findings. We also did the same thing for HCC‐related variants that had three or more studies, like KIF1B rs17401966, STAT4 rs7574865, HLA‐DQ rs9275319, MICA rs2596542, PNPLA3 rs738409 in alcohol‐related HCC, and TM6SF2 rs58542926 in alcohol‐related HCC (Figure S6). For most variants, the pooled ORs remained stable after excluding individual studies. Notably, HLA‐DQ rs9275319 and TM6SF2 rs58542926 showed particularly stable effect estimates with no heterogeneity, supporting the reliability of these associations. The metabolic gene variants (PNPLA3 and TM6SF2) in alcohol‐related HCC demonstrated consistent protective or risk effects across different study combinations, reinforcing their established roles in metabolic liver disease‐related hepatocarcinogenesis.

4
Discussion
In this study, we used meta‐analysis to look at how genetic variants play a role in HBV infection and HCC through different biological pathways. We found that immune‐related genes like CD40 and C2, and also HLA class II variants, are closely connected to whether HBV infection becomes chronic. We also found that variants in metabolic genes are more important for HCC risk, especially when the cause is alcohol use, NAFLD, or viral hepatitis. When we put together data from many independent studies, we got a clearer picture of the genetic factors behind HBV and HCC risk, and this could help with risk prediction and watching patients more closely.
CD40 rs1883832 (OR = 1.19, p = 2.51 × 10−15) and C2 rs9267665 (OR = 2.26, p = 1.77 × 10−14) showed the strongest and most consistent associations with HBV persistence, and both variants had very significant effects with no heterogeneity across studies. CD40 works as a costimulatory molecule that is involved in humoral and cellular immunity [11]. The increased susceptibility linked to CD40 variants may be explained by weaker adaptive immune responses, which make viral clearance less efficient. C2 is related to the classical complement pathways and is mainly made or expressed in hepatocytes [33]. A decrease in complement‐mediated immunity in carriers of the risk alleles could limit the clearance of HBV‐infected hepatocytes.
Among HLA variants we examined, HLA‐DPA1 rs3077 showed a protective association (OR = 0.70), while HLA‐DQB1 rs2856718 was linked to higher susceptibility (OR = 1.35). These results agree with the important role of HLA class II molecules in presenting viral antigens to CD4+ T cells [7, 8, 9, 10]. However, both variants showed very high heterogeneity (I
2 > 95%), which means their effects may be different across study populations. Even though most of the included studies were based on East Asian groups, there is still a lot of genetic variation within this region, and East Asians have notable differences in their HLA haplotype distributions [47]. In addition, linkage disequilibrium structures near the HLA loci are not the same across populations, which may partly explain the inconsistency we saw in the pooled effect sizes [48]. We also noticed some publication bias for HLA‐DPA1 and HLA‐DPB1 (Egger's test p < 0.1). This suggests that some studies with weak or non‐significant results might not have been published. Because of this, the combined effect estimates could look stronger than they actually are.
HLA‐DQ rs9275319 showed a very strong link with HBV‐related HCC (OR = 1.52, p = 2.37 × 10−22). Since there was no heterogeneity, the effect was consistent across all studies. This region is believed to control how the immune system monitors liver cells during chronic HBV infection. For instance, some HLA‐DQ variants might not be as efficient at presenting viral or tumour‐associated antigens [14]. What is important is that this link remained strong even after a long period of infection, showing that HLA‐mediated antigen recognition has a long‐term effect on HCC risk.
We found that KIF1B rs17401966 had a protective effect (OR = 0.77, p = 0.025). KIF1B is part of a kinesin superfamily that helps transport organelles and vesicles, which might explain its role in HCC development [15]. Some studies suggest that KIF1B variants affect apoptotic signalling [15], and our meta‐analysis confirms that this protective effect is consistent. On the other hand, STAT4 rs7574865 did not show a significant association, even though STAT4 is usually linked to cytokine signalling [14]. This might mean that different immune pathways in liver cancer are connected, so other mechanisms can balance out changes in STAT4 activity.
MICA rs2596542 showed a significant association with HCV‐related HCC (OR = 1.44, p = 0.033), although substantial heterogeneity was present (I
2 = 90.4%). MICA encodes a stress responsive ligand recognized by NKG2D on NK cells and CD8+ T cells, which triggers cytotoxic activity against tumour cells [18, 19]. The high heterogeneity may arise from differences in HCV genotype composition across studies [49], as genotypes 1 and 2 show distinct associations with HCC progression and may interact with host immune genes in different ways. In addition, the presence of publication bias for MICA (Egger's test p = 0.009) indicates that this association should be interpreted with care.
Even though a previous GWAS found a link, DEPDC5 rs1012068 showed no significant association in our study [17]. We still don't know much about how DEPDC5 works in HCV‐related liver cancer. Also, the high heterogeneity (I
2 = 96.1%) among studies suggests that the first GWAS results might have been specific to a certain population or affected by other factors that were not measured.
Our most significant and steady results were found in the metabolic gene variants related to alcohol and NAFLD‐related HCC. PNPLA3 rs738409 showed robust effects in both alcohol‐related HCC (OR = 1.64) and NAFLD‐related HCC (OR = 1.76), with the magnitude of effect being similar across etiologies. PNPLA3 is responsible for producing patatin‐like phospholipase domain‐containing protein 3. This protein reduces triglyceride hydrolysis and leads to fat accumulation in the liver [20, 22]. Because of constant lipotoxicity, this variant makes a person more likely to develop fibrosis and, in the end, HCC [22].
TM6SF2 rs58542926 showed stronger effects (OR = 1.81 in alcohol‐related HCC; OR = 1.50 in NAFLD‐related HCC), and there was no evidence of heterogeneity. TM6SF2 affects hepatic lipid export, and loss‐of‐function variants can make triglycerides accumulate in the liver while circulating lipid levels stay lower [21]. Findings were consistent across the studies, and there was no heterogeneity. These results suggest that changed lipid handling caused by TM6SF2 may contribute widely to the development of HCC in metabolic liver diseases.
HSD17B13 rs72613567 showed a protective association in alcohol‐related HCC (OR = 0.80, p = 2.69 × 10−6). The gene produces a protein localized to hepatic lipid droplets, and its loss‐of‐function variants are known to reduce the risk of hepatic inflammation [24, 25]. This observation raises the possibility that targeting HSD17B13 could help lower alcohol‐related HCC risk.
Whereas PNPLA3 and TM6SF2 exhibit strong associations with NAFLD‐related HCC, MBOAT7 rs641738 did not show a significant effect in our analysis (p = 0.283). This is unexpected given that earlier studies have linked this variant to hepatic steatosis and advanced fibrosis [27]. It is possible that MBOAT7 plays only a minor role in HCC development or that its effects become apparent only in the presence of specific genetic or environmental factors.
Our meta‐analysis results give a combined view of the genetic factors that affect HBV infection outcomes and HCC risk in different etiologic pathways. By using evidence from the GWAS Catalogue, published studies, and previous studies done in our laboratory, this study made the pooled estimates more precise. Because HBV infection is very common in East Asian populations, these results are especially important for this region. Also, looking at the different etiologic pathways helps us understand more clearly how changes in immune regulation and metabolic processes can increase the risk of HCC. Our research shows that certain immune regulatory genes and HLA class II variants are very important for HBV persistence. This is because they directly influence key immune pathways in the body. In addition, we found that metabolic gene variants like PNPLA3, TM6SF2, and HSD17B13 have a major impact on HCC development, especially in patients with alcohol‐related or NAFLD‐related liver disease. These results help us better understand the genetic factors behind HBV and HCC. Moreover, they provide a strong foundation for creating more specific genetic and etiologic strategies in future studies.
However, our study has several limitations. First, there was considerable heterogeneity in the HLA‐related analyses of HBV persistence, and this may lower the precision of the pooled estimates. We also found some signs of publication bias for certain variants, which means their effect sizes might not be the same as the original reports. In addition, because most of the studies we included were from East Asian populations, the results may not fit well for other groups. Furthermore, the included studies employed diverse genotyping methods, including genome‐wide SNP arrays (Illumina and Affymetrix), TaqMan allelic discrimination assays, Sequenom MassARRAY, Fluidigm SNPtype, PCR‐RFLP, and PCR with direct sequencing (Table 1). Although only one study used PCR‐RFLP, which is generally associated with higher error rates (1%–3%) compared to array‐based methods (~0.2%) [50], the majority of our analyses combined GWAS array‐derived data with TaqMan‐based candidate gene data. These platforms may still differ in genotyping accuracy and information content, and such methodological differences could contribute to between‐study heterogeneity [51, 52]. It is worth noting, however, that the variants showing the highest heterogeneity, such as HLA‐DQB2 rs7453920 (I
2 = 99.3%) and HLA‐DPB1 rs9277535 (I
2 = 99.2%), did not include PCR‐RFLP‐based studies, suggesting that genotyping platform differences alone are unlikely to fully account for the observed heterogeneity. Because we relied on summary‐level data from published reports, we were unable to perform subgroup analyses stratified by genotyping methods. Future meta‐analyses should systematically assess the impact of genotyping platform differences on pooled estimates through stratified sensitivity analyses [53, 54]. Finally, we did not have individual‐level data, so it was hard to study possible gene–gene or gene–environment interactions. Even with these limitations, this analysis still gives an overall and cause‐specific view of the genetic factors that affect HBV infection and HCC risk.

Author Contributions
All authors contributed to the study conception and design. Data collection and analysis were performed by So Yoon Lee. Interpretation of the results and discussion were carried out by So Yoon Lee and Hyoung Doo Shin. So Yoon Lee and Hyoung Doo Shin led the manuscript writing, and all authors provided feedback on earlier drafts of the manuscript. All authors read and approved the final manuscript.

Funding
The authors have nothing to report.

Ethics Statement
This study is a meta‐analysis of previously published data and did not involve any new human participants, human data, or human tissue. Therefore, ethical approval and informed consent were not required.

Conflicts of Interest
The authors declare no conflicts of interest.

Supporting information

Figure S1: Forest plots of genetic variants showing non‐significant associations with HBV infection susceptibility.

Figure S2: Forest plots of genetic variants showing non‐significant associations with HCC risk.

Figure S3: Funnel plots for assessment of publication bias in HBV infection susceptibility.

Figure S4: Funnel plots for assessment of publication bias in HCC risk.

Figure S5: Leave‐one‐out sensitivity analysis for genetic variants associated with HBV infection susceptibility.

Figure S6: Leave‐one‐out sensitivity analysis for genetic variants associated with HCC risk.

Table S1: Summary of studies included in the meta‐analysis of persistent HBV infection.

Table S2: Summary of studies included in the aetiology‐specific meta‐analysis of HCC risk.