Hepatocellular carcinoma (HCC) is one of the most common malignant tumors in the world and ranks third in global cancer deaths [1]. The occurrence and development of HCC varies widely among ethnic groups, regions and individuals [2], and epidemiological investigation have shown that 60% of the more than 700,000 new HCC cases annually are caused by chronic hepatitis B virus (HBV) infection and are distributed with the epidemic characteristics of HBV [3]. In hyperendemic areas such as China and Africa, chronic HBV infection contributes to at least 80% of cases of HCC [4]. The main pathogenic factors of HCC include hepatitis C virus (HCV) infection, aflatoxin food contamination and algae toxin water contamination [5]. However, exposure to these factors rarely promotes the development of HCC, even among people who are highly exposed to the pathogenic factors of HCC, which suggests that the host’s genetic predisposition plays a crucial role in the development of HCC [6].

Assuming that certain genetic variations increase the susceptibility of chronic HBV carriers to HCC development, candidate genes have been selected a priori based on this biological likelihood. Over the past decade, candidate gene association studies have found that many genes are associated with the risk of HCC, such as TNF-α [7], IGF2 [8], SPP1 [9], DNMT3B [10]. Unfortunately, certain of these studies suffered from major methodological drawbacks because of their case–control and monocentric focus; therefore, it is difficult to validate their conclusions on HCC susceptibility. Genome-wide association study (GWAS) of human age-related macular degeneration was first reported in 2005 [11], and represented a landmark for new directions and new methods for research into complex diseases or traits. GWAS is a hypothesis-free approach that can screen disease-related sequence variations from the whole genome [12]. In 2010, Zhang et al. [13] conducted a GWAS on HBV-related HCC in the Chinese population and found a significant correlation between the single nucleotide polymorphism (SNP) locus rs17401966 in the kinesin family member 1B (KIF1B) gene of the 1p36.22 region and susceptibility to HCC, and the integrated P value of the 5 study cohorts reached 3.4×10^-19. However, recent studies have yielded inconsistent or conflicting results that may have been caused by population and design differences and a small sample size. Thus, we attempted to conduct a meta-analysis of all relevant literature to provide more comprehensive and reliable associations between the rs17401966 in KIF1B gene and susceptibility to HCC with HBV infection.

In 2010, Zhang et al. [13] conducted a GWAS with samples from the primary screening stage from 355 chronic HBV carriers with HCC and 360 chronic HBV carriers without HCC from Guangxi, and they validated their results with four additional independent case-controlled populations recruited from Guangdong, Shanghai, Jiangsu and Beijing. The results showed that the SNP locus rs17401966 in the 1p36.22 region was verified in the four populations and had a merged P-value of 3.4×10^-19. The 1p36.22 region was approximately 244 kb long, including KIF1B, phosphogluconate dehydrogenase (PGD) and the 3’ terminal end of the ubiquitination factor E4B (UBE4B) genes. 1p36 is an important loss of heterozygosity (LOH). LOH has been frequently observed in tumor originated from the blood cells, nerve cells and epithelial cells [21]. In addition, Li et al. [22] found in the study of HCC patients in the group of southern China that there is high frequency of LOH in 1p36 of cancer tissues, suggesting this area is correlated with the occurrence of HCC. Thus, the newly identified UBE4B-KIF1B-PGD locus is a biologically plausible candidate for HCC susceptibility. Zhang et al. [13] also performed an immunohistochemical analysis to determine the protein expression of the KIF1B, PGD and UBE4B genes in the HCC tissues and the corresponding paracarcinoma tissues, and although the results did not indicate abnormalities in the expression of UBE4B, a significant reduction in the expression of KIF1B and PGD was observed in the cancerous tissue. A qRT-PCR analysis then showed that the expression of KIF1Bβ was correlated with the genotype of the associated SNP locus rs17401966 and indicated that the high expression of this gene was also significantly related to the protective allele G of rs17401966, which suggests that KIF1Bβ may represent a HCC-suppressor gene. KIF1B is a member of the kinesin family and an important molecule in intracellular vesicles and organelles transport [23], in combination with the absence of the adjacent gene on 1p36 (e.g., p73 and CHD5), the down-regulation of KIF1Bβ may lead to the occurrence of tumors [24,25].

In 2012, Li et al. [26] conducted a GWAS that included two independent Han cohorts at the genome-wide discovery stage, and they covered 480 HBV-positive HCC patients and 484 chronic HBV carriers from central China and 1,058 cases and 981 controls from southern China. These authors assessed the previously reported HCC-susceptible SNP rs17401966; however, the merged results for the two groups showed that the rs17401966 G allele did not reduce the risk of HCC (OR=0.90; 95% CI: 0.80-1.02). In 2013, Jiang et al. [27] conducted another GWAS with 1161 cases and 1,353 controls from Qidong at the genome-wide discovery stage, although the findings were not confirmed (OR=0.98; 95% CI: 0.87-1.11). Subsequent studies by other scholars also failed to achieve consistent results [16-19]. Because the role of the same gene can vary among different populations, among the same group and among with the same tumor, a single study may not provide enough samples for a correlation analysis or sufficient statistical validity to identify minor genes, the meta-analysis approach expands the sample size to avoid the flaws of insufficient statistical validity. In this meta-analysis, we included a total of 12 cohorts from 5 studies, and the results showed that the rs17401966 polymorphism might reduce the risk of HCC under the allele, heterozygous, homozygous, and dominant models but not under the additive and recessive models. The inconsistent results might have been caused by the lack of advantage observed in the homozygous rs17401966 GG with regard to a reduction in HCC susceptibility compared with that observed for the heterozygous rs17401966 AG. However, strong heterogeneity was observed in the merged analysis of the 12 cohorts, which might reduce the reliability of the results. The Fst statistics indicated that different degrees of differentiation occurred between the 5 cohorts of Zhang et al.’s study [13] and the other cohorts. To reduce the effect of Zhang et al.’s study [13] cohorts on the meta-analysis, we assigned them to subgroup 1, and the remaining seven cohorts were assigned to subgroup 2. Inconsistent meta-analysis results were observed between subgroup 1 and subgroup 2, which may been caused by the following reasons: (1) the 7 cohorts of subgroup 2 were from different studies that presented differences in their case selection, such as the genetic background of the chronic HBV carriers and chronic hepatitis B; (2) strong differentiation was observed between the Beijing population included in Zhang et al.’s study [13] and Chen et al.’s study [19], which might have been caused by the accelerated population migration that occurred as China’s society and economy developed or by sampling error; and (3) the HCC sampled in different areas was caused by different genes, and the KIF1B gene mutations may not have played a dominant role in the development of HCC in certain areas.

Two meta-analyses [28,29] on the association between the KIF1B rs17401966 polymorphism and HCC risk concluded that the rs17401966 G polymorphism could reduce the susceptibility to HCC; however, the source of the heterogeneous results was not evaluated in these studies. We investigated the source of heterogeneity with an Fst analysis and conducted appropriate groupings. In our meta-analysis, although publication bias was not observed and the sensitivity analysis results did not show significant changes in the merged results when any one cohort was excluded, the results were still unreliable because the cumulative meta-analysis showed that the OR value increased as the number of research cohorts increased. It is important to note that the sample size at the primary screening stage in the GWASs of Li et al. [26] and Jiang et al. [27], as well as in the GWASs of Hu et al. [16] and Chen et al. [19], were all larger than that of Zhang et al. [13]. In addition, the sample size of subgroup 2 was greater than that of subgroup 1, and the results were inconsistent. Therefore, we propose that further studies are required to verify the correlation between the KIF1B rs17401966 polymorphism and HCC risk.

Certain limitations of this meta-analysis should be acknowledged. First, for the HBV in China, priority is assigned to genotypes B and C, although most studies do not analyze these genotypes separately. Because the information was incomplete, the above factors were not considered. Second, the studies of Li et al. [26] and Jiang et al. [27] lacked genotype data; therefore, we did not include them in the meta-analysis and the results might be biased.

In conclusion, we introduced Fst statistics to evaluate the source of heterogeneity for a binary variable meta-analysis and conducted a merged analysis on 12 cohorts, and our results indicated that the rs17401966 polymorphism can reduce the susceptibility to HCC. However, the meta-analysis results were weak and the cumulative meta-analysis results were unreliable. In addition, the results of the two subgroups were inconsistent. A well-designed multicenter validation study with a larger sample size is needed to verify this correlation.