Genetic polymorphisms in C-reactive protein increase cancer susceptibility

Elevated levels of C-reactive protein (CRP) partially induced by polymorphisms in the CRP gene have been associated with human cancer. The purpose of this study was to test the hypothesis that CRP gene polymorphisms (+942G>C, 1846C>T) modify inherited susceptibility to cancer. We systematically identified the publications addressing the association of CRP gene polymorphisms with cancer susceptibility. Studies that fulfilled all inclusion criteria were considered eligible in this meta-analysis. We analyzed a total of 8 case-control studies. Individuals with the CC genotype were found to have an almost 4 fold higher risk of cancer than those with the GG or GC and GG genotypes. A significant association was also indicated in subgroup of colorectal cancer. Meta-analysis of 1846C>T polymorphism showed increased cancer risk in relation to the 1846 TT genotype (TT vs. CC: OR = 1.15, 95% CI = 1.01–1.31; TT vs. CT + CC; OR = 1.17, 95% CI = 1.03–1.32). Similar results were suggested in Caucasian populations and colorectal cancer. These data suggest that both +942G>C and 1846C>T polymorphisms in the CRP gene may influence cancer susceptibility.

Scientific RepoRts | 5:17161 | DOI: 10.1038/srep17161 Methods Publication search. The Web of Science, Embase and PubMed were searched exhaustively using search terminology ((polymorphism) OR (polymorphisms)) AND ((C-reactive protein) OR (CRP)) AND (cancer) and their synonyms (variants, carcinoma, tumor, neoplasm) to identify the publications reporting on CRP polymorphisms and cancer risk. The electronic search lasted eight months. Additional usable data were obtained by hand searching the bibliographies of genetic association studies on the subject in this analysis. We used no restrictions on the number of samples and language to minimize publication bias.
Inclusion criteria and exclusion criteria. Studies were considered in this analysis if the following conditions were fulfilled: (1) a case-control study with cancer patients investigated; (2) the relationship between CRP polymorphisms and cancer risk was assessed; (3) genotype frequency of the same polymorphism must be available in at least four studies; (4) the study must be unique without any subsequent update. We excluded the studies where the controls were cancer patients and genotype data were unaccessible even after having contacted corresponding authors.

Data extraction.
For the studies included, two investigators collected the first author's surname, publication year, study country, ethnicity, cancer type, number of genotyped cases and controls, source of controls, genotyping methods and genotype frequency. Ethnicity was categorized as East Asian or Caucasian. Samples from the USA were grouped into Caucasian ethnicity and those from China and Japan were considered as East Asian ethnicity. We counted the different cancer types and ethnic populations reported in the same article as separate studies that were appropriately classified into the category described above.
Statistical methods. Cancer risk in relation to CRP polymorphisms was estimated by crude ORs and 95% CIs (OR, odds ratio; 95% CI, confidence interval). We calculated the pooled ORs using multiple genetic models (Table 1). Subgroup analyses by cancer type was performed for CRP SNP + 942G> C, while for CRP SNP 1846C> T, data were stratified by ethnicity in addition to cancer type.
Heterogeneity across studies was evaluated by the Chi square-based Q-test, and a P value more than .10 indicated the effect size was homogeneous. We combined OR for the single studies using the Mantel-Haenszel method unless little heterogeneity was indicated, or else the DerSimonian and Laird method was used 16,17 . Hardy-Weinberg equilibrium (HWE) was examined by using the χ 2 test in the control group of each study. Sensitivity analysis by sequentially omitting the single studies and recounting the pooled ORs and 95% CIs was performed to estimate the effect of individual studies on overall risk of cancer. The funnel plot was utilized to test the publication bias and Egger's test (linear regression analysis) was used to check the symmetry of funnel plots 18 .
STATA software (version 12.0, Stata Corporation, College Station, TX) was performed to analyze statistical data. All tests were two-tailed and the significance level was fixed at .10.

Results
Selection of studies. As showed in Fig. 1 and abstract evaluation, 127 articles were removed due to non-cancer studies or polymorphism studies irrelevant to CRP SNPs. Of the 31 remaining articles, 24 were eventually excluded because of unavailable raw data or case-only design. As a result, 7 articles consisting of 7 case-control studies of + 942G> C polymorphism and 7 studies of 1846C> T polymorphism were considered in the final analysis 13-15,19-22 . Characteristics of studies. Summary characteristics of the studies included are described in Table 2 Association between cancer risk and 1846C>T polymorphism. A total of 3 543 cases and 4 263 controls were analyzed in this meta-analysis. On the whole, the 1846 TT genotype was found to increase    Fig. 3; TT vs. CT + CC; OR = 1.17, 95% CI = 1.03-1.32, P heterogeneity = 0.656). When we performed stratification analyses by ethnicity, significantly elevated risk of cancer was found in Caucasians using the TT vs. CC (OR = 1.19, 95% CI = 1.02-1.39, P heterogeneity = 0.630) and TT vs. CT + CC genetic model (OR = 1.21, 95% CI = 1.04-1.40, P heterogeneity = 0.355). The same two genetic models showed an association with colorectal cancer when analysis was constrained to cancer type (Table 1).

Sensitivity analysis.
To evaluate the influence of individual studies on risk of overall cancer, we performed leave-one-out sensitivity analysis and recomputed the pooled ORs. The ORs calculated after excluding a single study did not show any differences from the primary values. This process assured the stability of overall results (Figure not shown).

Publication bias. Evaluation of publication bias was performed using both Begg's funnel plot and
Egger's test. The studies of CRP polymorphisms were symmetrically distributed (Fig. 4, + 942G> C, C vs. G; Fig. 5, 1846C> T, T vs. C), which was confirmed by the Egger's test (P = 0.352; P = 0.628). Therefore, our meta-analysis results are not affected by publication bias and worthy of trust.

Discussion
To the best of our knowledge, this is the first quantitative assessment of the genetic association studies reporting on the relationship between CRP polymorphisms and cancer susceptibility. This meta-analysis summarized a total of 8 case-control studies, providing evidence that supported a significant role of CRP polymorphisms in cancer. More specifically, the CC genotype of + 942G> C polymorphism was associated with significantly increased risk of cancer, particularly colorectal cancer. We also noted that the carriage of 1846 TT genotype had higher risk to develop cancer. Subgroup analysis by ethnicity and cancer type showed a similar trend towards an increased risk in Caucasians and colorectal cancer. These data suggest that genetic polymorphisms in the CRP gene may have effects on the development of cancer. As inflammation is important in the progression of human cancer, much attention has been directed to the CRP, an inflammation-related gene. Several lines of work have connected the functional polymorphisms at CRP locus with cancer. For SNP + 942G> C, Wen and workmates investigated the association with endometrial cancer in a relatively large study, suggesting CRP + 942G> C alone was not associated with this cancer in Chinese patients 13 . Such an insignificant association was seen in most of the published studies representing distinct ethnicities 14,15,20 . These observations are in disagreement with those suggested in the current study, where increased susceptibility of cancer was revealed. As most individually published studies, even in different populations, shared the same finding that CRP + 942G> C was not an independent risk factor for cancers, the most persuasive explanation to the discrepancy is the inadequate  statistical power caused by the small number of subjects in these studies. In addition, it should be noted that CRP + 942G> C itself may not modify cancer risk, but it influences cancer development by interacting with body mass index (BMI) and family history of cancer 15 , implicating the biological functions of CRP + 942G> C are possibly determined by physical conditions of individuals themselves.
In terms of CRP 1846C> T, previous reports have generated controversial results. A nested case-control study of lung cancer in 1 262 samples of Caucasian descent suggested that the association between 1846C> T and lung cancer was not statistically significant 20 . In contrast, a Caucasian study demonstrated a decreased risk of colorectal cancer in relation to the C allele 21 , an observation contradicted a later colorectal cancer study in which the same allele was found to have a fixed 30% increased colorectal cancer risk 15 . Again, inadequacy of study sample may be responsible a large part for the existing inconsistency. An alternative explanation may relate to the differences in cancer type, because etiology of various human cancers is heterogeneous and complex, studies of a single polymorphism seem impossible to determine the association with cancer.
Our findings are supported by previous mechanic studies. The study by Heikkila and workmates suggested that elevated expression of CRP was linked to increased overall risk of cancer (RR = 1.10, 95% CI = 1.02-1.18); the association was more pronounced in lung cancer (RR = 1.32, 95% CI = 1.08-1.61) 23 . Likewise, a subsequent meta-analysis analyzed 1 918 lung cancer cases and revealed significantly increased risk of lung cancer associated with higher CRP levels among men (RR = 1.18, 95% CI = 1.09-1.28) 24 . According to epidemiologic data, the increased CRP levels should be attributable to the genetic variations in the CRP gene 25 , including + 942G> C and 1846C> T polymorphisms 26,27 . Therefore, it is plausible that CRP polymorphisms correlate with cancer development. For the significant association observed in Caucasians, but not in Asians, there are several possibilities, one of which relates to the remarkably different expression patterns of CRP in non-homogeneous ethnic populations 25 .
Our findings should be interpreted with caution due to a few potential limitations. First, we may have missed some studies containing usable data, even though an exhaustive literature search was undertaken in the PubMed, a database we put more emphasis on. To maximize the case panel in this analysis, we additionally searched Web of Science and Embase in which a large number of medical research papers are collected. Second, selection bias may have been introduced, as we included both population-based studies randomly selecting controls from individuals with no lesions and hospital-based studies using healthy controls ascertained through routine health check. Third, despite a significant association with overall risk of cancer and colorectal cancer was suggested in this meta-analysis, we only can infer but cannot conclude that CRP polymorphisms are susceptibility loci of other types of cancer, highlighting the necessity for further investigation.
In conclusion, we found that genetic polymorphisms in the CRP gene, + 942G> C and 1846C> T, are associated with an increased overall risk of cancer. Subgroup analyses by ethnicity and cancer type also showed an significant association in Caucasians and colorectal cancer. Future research is quite necessary to provide compelling evidence of the association between CRP gene polymorphisms and cancer risk.