Carcinogenesis is a complex process that involves both genetic and environmental factors and their interactions . Since the genetic factors such as single nucleotide substitutions and gross chromosomal aberrations were considered to make more contribution to tumorigenesis, numerous studies focused on the candidate-polymorphism approach notably increased the number of associations between polymorphism and cancer risk . Recently, increasing attention has been paid to the relationships between genetic variants such as SNPs and cancer risk. Plentiful epidemiological evidence demonstrated that the process of detoxification and elimination of xenobiotics was involved in the development of cancers [14, 16, 17].
The ABCG2 gene (also known as BCRP), located on chromosomal locus 4q22, encodes an ABC half-transporter protein that localizes to the plasma membrane. ABCG2 works as a multidrug efflux pump, influencing the metabolism of multiple substances including anticancer drugs and carcinogenic xenobiotics [10–13]. The abnormal distribution of carcinogenic xenobiotics may increase the local carcinogen burden of specific cells and organelles  and cause tumorigenesis. To date, there have been a great deal of epidemiological studies to investigate the association between the C421 polymorphism in ABCG2 and the risk of various types of cancers, however, the exact relationship between cancer susceptibility and the ABCG2 C421A polymorphism remains unclear. To address this issue, we performed a synthetic analysis based on data collected from all studies that have investigated the relationship between this polymorphism and cancer risk.
Generally speaking, our meta-analysis, which included 3593 cases and 5875 controls, revealed a significant association between the ABCG2 C421A allele and decreased cancer risk in different genetic model (heterozygote model-AC versus CC: OR = 0.759, 95%CI = 0.620-0.930; dominant effects model-AA/AC versus CC: OR = 0.771, 95%CI = 0.634-0.938; additive effects model-A allele versus C allele: OR = 0.809, 95%CI = 0.687-0.952). This result provides convincing evidence that the C421A polymorphism in ABCG2 might protect against cancer development. Moreover, this effect persisted when the studies were stratified by ethnicity classification. When stratified by cancer, source of controls and ethnicity classification, the C421A polymorphism of ABCG2 was also an important protective factor against cancer development in the subgroups studied.
However, the studies published by Hu and Korenaga [33, 34] were contradictory to ours, indicating that carriers of the A allele of ABCG2 C421A had an increased risk of cancer. Three explanations may have contributed to this disparity in results. One possible explanation could be that the different environmental factors or different sample size of the two studies may influence the function of the ABCG2 C421A for developing cancer. Another possible explanation for this observation was that gene-gene interactions and gene-environment interactions may be responsible for this discrepancy. Moreover, the overlapping function of the other ABC transporters may be involved in. However, the precise mechanism of the contradictory effect remains unsure, further studies may help to clarify this issue.
It is well known that four capital factors of genome-wide association studies (GWAS) consisting of models of the allelic architecture of common diseases, sample size, map density and sample-collection biases need to be taken into account in order to optimize the cost efficiency of identifying precise disease-susceptibility loci . To validate the strong associations between the C421A polymorphism in ABCG2 and the cancer risk, Daniele Campa and his colleagues extracted date from two previous GWAS on CLL (chronic lymphocytic leukemia) [40, 41]. Based on the genotyping information from them, allelic odds ratios were calculated to confirm the relationship between the genetic polymorphisms and CLL risk. Their results identified a statistically significant association between the risk of CLL and the ABCG2 C421A genotype . Furthermore, these authors also performed a meta-analysis of 3 studies [30, 40, 41], which provided evidence of a significant association between decreased cancer risk and the ABCG2 C421A polymorphism. Our findings were consistent with these results .
Heterogeneity is a potential problem which might influence the interpretation of the results. In our meta-analysis, significant heterogeneity between studies was present in additive model, heterozygote model and dominant model (Table 2). The heterogeneity reduced or disappeared when the studies were stratified by cancer type, source of controls and ethnicity, however, we did not find adequate evidence to determine which of them contributed most to the substantial heterogeneity. The publication bias for the association between this polymorphism and cancer risk was not observed in our meta-analysis.
Some possible limitations of our meta-analysis should be acknowledged and taken into consideration. First, detailed information, such as the mean age and sex of the case and control populations, was not available in all of the selected studies, which limited further analyses. Second, the results may be influenced by the lack of observations regarding gene-gene and gene-environment interactions even different polymorphic loci of the same gene. Third, the conclusions had the possibility to be disturbed due to the existence of overlapping function of the other ABC transporters. Fourth, the numbers of published studies were not sufficiently enough for a comprehensive analysis on different types of cancer. For example, there were no published data for gastric cancer, nervous system neoplasm and lung cancer with association of ABCG2 C421A published up to now; we did not posses enough statistical power to detect the precise association. More studies are needed to explore the relationship between C421A polymorphism in ABCG2 and cancer risk. In spite of these potential limitations, our meta-analysis also has many advantages. Firstly, sufficient date was extracted form well-selected studies, providing good statistical power for this meta-analysis. Secondly, studies included in our meta-analysis contained available genotype frequency and the distribution of the genotypes in the control population of all the studies were consistent with Hardy-Weinberg equilibrium. Thirdly, no publication bias was detected among the pooled results.