Association between the Cytotoxic T-Lymphocyte Antigen 4 +49G > A polymorphism and cancer risk: a meta-analysis

We carried out a literature search using the PubMed database (between January 2000 and February 2010) to identify all papers that investigated the association between the CTLA-4 +49 A > G polymorphism and cancer risk in all ethnic groups, using combinations of the search phrases "CTLA-4 and polymorphism and cancer". We evaluated the titles and abstracts of all relevant publications, but excluded abstracts, case reports, editorials, and review articles. Studies included in the current meta-analysis had to meet the following criteria: the study must have used a case-control study design; the report must have included cancer diagnoses and sources for the cases and controls; the report must have included genotype frequencies; the authors must have given size of their samples, the odds ratios (Ors) used, and their 95% confidence intervals (CIs); definitions of exposure or risk genotypes must have been similar in all reports; and the methods of data collection and analysis must have been statistically acceptable.

Data were collected on the CTLA-4 +49 A > G genotype from studies of different types of cancer. The first author, year of publication, country, ethnicity of the study population, the number of cases and controls, and the type of study were described (Table 1). In the stratification analyses for ethnicity, there were: 7 Caucasian populations, including Spanish, Polish, German, American, Sardinian, Macedonian and Italian, and 3 Asian populations, including Chinese (containing Taiwanese), Trukese, and Irani. In the stratification analyses for cancer type, we compared epithelial tumors to non-epithelial tumors, and solid tumors to non-solid tumors. There were 12 types of solid tumors, including those caused by renal cell cancer, colon carcinoma, cervical squamous cell carcinoma, breast cancer, lung cancer, esophageal cancer, gastric cancer, oral squamous cell carcinoma, thymoma, nasopharyngeal carcinoma, HBV-related HCC and melanoma. There were 3 types of non-solid tumors, including those caused by CLL, non-Hodgkin's lymphoma and MALT lymphoma. The 4 types of epithelial tumors were those caused by oral squamous cell carcinoma, cervical squamous cell carcinoma, melanoma, and breast cancer. The 8 non-epithelial tumors included those caused by gastric cancer, colon carcinoma, HBV-related HCC, nasopharyngeal carcinoma, thymoma, renal cell cancer, lung cancer and esophageal cancer.

The selection of published studies we used for meta-analysis were further evaluated using sensitivity analyses. Odds ratios (ORs) and 95% confident interval (CIs) from each case-control study were used to assess the strength of association between the CTLA-4 +49A > G genotypes and the risk of cancer in dominant (GA+AA vs GG) genetic models. A combined OR was calculated according to Woolf's method [20]. A χ2-based Q statistic test was performed to assess between-study heterogeneity [21]. If the P value of the heterogeneity test was ≥ 0.05, then a fixed effect model using the Mantel-Haenszel method was used to calculate the combined OR, which assumed the same homogeneity of effect size across all studies. If the P value of the heterogeneity test was <0.05, it showed that the between-study heterogeneity was statistically significant. A random effects model, using the DerSimonian and Laird method, was performed to calculate the combined OR [22]. If there were no between-study heterogeneity, the combined OR calculated by those two methods would be identical. The significance of the combined OR was determined using a Z-test, in which P < 0.05 was considered significant. Finally, the combined ORs and their 95% CIs were presented. Stratification analyses for different types of cancers were conducted for colorectal cancer, breast cancer, gastric cancer, lymphoma, skin cancer, and other cancers (lung cancer, nasopharyngeal carcinoma, cervical squamous cell carcinoma, esophageal cancer, oral squamous cell carcinoma, HBV-related hepatocellular carcinoma or renal cell cancer) to estimate cancer-specific ORs. Stratification analyses by ethnicity were also conducted for Caucasian, Chinese and other Asian race populations to estimate ethnicity-specific ORs. Publication bias was assessed using a funnel plot, in which the standard error of log (OR) of each study was plotted against its OR value. The resulting asymmetrical plot suggested, according to Egger's linear regression test, that there was possible publication bias [23]. The significance of the intercept was determined using a Student t-test, as suggested by Egger. If the P-value of Egger's linear regression test was less than 0.05, it meant that there was a publication bias in the meta-analysis. The SAS/Genetics software program (Version 9.1, SAS Institute, Inc., Cary, NC, USA) was used to determine the LD of SNP pairs and the Hardy-Weinberg equilibrium. Other statistical software used included SPSS12.0 for Windows (SPSS Inc., Chicago, USA), Stata Version 10.0, and Review Manager (Version 4.2, the Cochrane Collaboration). All P-values were two-sided.

We searched NCBI PubMed using the key terms 'CTLA-4' 'cancer' and 'polymorphism' and found 36 epidemiological studies. Of these 36 studies, 14 studies were excluded either because they were not case-control studies or because the +49A/G polymorphism or its genotype frequency was not reported. The remaining 22 case-control studies contained 32 data sets (Sun's study had six datasets, and Pavkovic's, Welsh's, Hadinia's, Solerio's, and Chuang's studies all had two datasets each) [16, 24‐28]. We created a database of the information extracted from each article. The essential information, including first author, cancer type, year of publication, numbers of cases and controls, and the frequencies of CTLA-4 +49A/G for all studies are listed in Table 1. There were four studies concerning breast cancer [16, 17, 29], six concerning colorectal cancer [25, 27, 30‐32], six concerning non-solid tumors [26, 33‐36], three concerning gastric cancer [16, 25, 37], three concerning skin cancer [28, 38], two concerning Thymoma [24], two concerning lung cancer [16], one concerning nasopharyngeal carcinoma [39], one concerning cervical squamous cell carcinoma [40], one concerning Esophagus cancer [16], one concerning Oral squamous cell carcinoma [41], one concerning HBV-related hepatocellular carcinoma [42], and one concerning renal cell cancer [30]. Among the 32 datasets included in the final analysis, there were 14 concerning Caucasians and 18 concerning Asians (14 concerning Chinese, one concerning Trukese and three concerning Irani). Additional information is listed in the forest plots in our meta-analyses. We performed a sensitivity analysis on the selection of published studies in this meta-analysis.

The results of our Q test show that there was no between-study heterogeneity in the dominant genetic models (GA+AA VS GG) for all 32 datasets (P = 0.12). In the subgroup analyses for cancer type and ethnicity (results shown in Figures 1, 2, 3, 4 and 5), the heterogeneity test did not reveal any significant differences between the dominant genetic models (GA+AA VS GG): six colorectal cancer studies (P = 0.40), four breast cancer studies (P = 0.11), six non-solid tumor studies (P = 0.23), three skin cancer studies (P = 0.30) (Figure 1), and twenty-six solid tumor studies (P = 0.15) (Figure 3), seventeen non-epithelial tumor studies (P = 0.16), nine epithelial tumor studies (P = 0.25) (Figure 5), fourteen Caucasian population studies (P = 0.51), fourteen Chinese population studies (P = 0.13), and eighteen Asian studies (P = 0.05) (Figure 2 and 4). However, the results of our heterogeneity test did indicate significant differences (P = 0.01) for gastric cancer (Figure 1).

Bild vergrößern

Bild vergrößern

Bild vergrößern

Bild vergrößern

Bild vergrößern

For the CTLA-4 +49G > A polymorphism, we obtained our meta-analysis data from 32 datasets consisting of 11,273 cases and 13,179 controls. The associations between the CTLA-4 +49G > A genotype and cancer risks were estimated using dominant (GA+AA vs GG) genetic models in either fixed or random effect models according to the heterogeneity Q test. We used 32 datasets in these comparisons. Compared with the wild-type +49G > A GG genotype, the carriers of variant genotypes (GA/AA) had a 1.24-fold elevated risk of cancer (95% CI = 1.18-1.32, P < 0.05) as estimated by a fixed effect model for dominant genetic effects (Figure 6).

Bild vergrößern

The effect of the CTLA-4 +49G > A polymorphism was further evaluated using stratification analysis. In the six colorectal cancer studies, which included 699 cases and 1,411 controls, subjects with variant genotypes (597 cases and 1,144 controls) had a non-significant increased risk of colorectal cancer (OR = 1.03, 95% CI = 0.78-1.35, P = 0.86) as estimated using a fixed effect model (Figure 1). Similar results were found in the three gastric cancer studies, with 778 cases and 982 controls (OR = 1.17, 95% CI = 0.96-1.43, P = 0.12), and the six non solid tumour studies, with 506 cases and 878 controls (OR = 1.08, 95% CI = 0.79-1.48, P = 0.62) (Figure 1). However, in the four breast cancer studies, consisting of 2,411 cases and 2,439 controls, the variant genotypes (1,436 cases and 1,292 controls) were associated with a significantly increased risk of breast cancer (OR = 1.31, 95% CI = 1.17-1.48, P < 0.00001) (Figure 1). Similar results were found in the three skin cancer studies, which included 2,342 cases and 2,372 controls (OR = 1.30, 95% CI = 1.10-1.52, P = 0.001), and in the 26 solid tumour studies, which had 10,767 cases and 12,301 controls (OR = 1.25, 95% CI = 1.18-1.33, P < 0.00001) (Figure 1 and 3). We also found that patients with the variant genotypes had significantly increased risks for developing either non-epithelial tumors or epithelial tumors, with ORs of 1.23 (95% CI = 1.14-1.32, P < 0.00001) and 1.29 (95% CI = 1.17-1.41, P < 0.00001), respectively (Figure 5).

In the stratification analyses for ethnicity, we found that the increased risk of cancer associated with +49G > A variant genotypes was more pronounced in Caucasians (OR = 1.29, 95% CI = 1.13-1.47, P = 0.0002), Asians (OR = 1.23, 95% CI = 1.16-1.32, P < 0.00001), and Chinese (OR = 1.23, 95% CI = 1.15-1.31, P < 0.00001) (Figure 2 and 4).

To evaluate publication bias, the CTLA-4 +49G > A genotypes were plotted against the precision ones using a funnel plot. The result was approximately symmetrical, so Egger's test suggests that there is no publication bias in the current meta-analysis (P = 0.998). This indicates that biases from publications and other factors may not have had a significant influence on the results of our meta-analysis on the association between CTLA-4 +49G > A polymorphism and cancer risk (Figure 7).

Bild vergrößern