CASP8 -652 6N insertion/deletion polymorphism and overall cancer risk: evidence from 49 studies

The CASP8 -652 6N insertion/deletion (I/D) polymorphism reduces expression of caspase 8. We conducted a meta-analysis to clarify the relationship between this polymorphism and cancer risk. Eligible articles were retrieved from PubMed, EMBASE, CNKI, and WANFANG databases through February 2017. A total of 33 articles with 49 studies, including 33,494 cases and 36,397 controls, were analyzed. We found that the CASP8 -652 6N ins/del polymorphism was associated with decreased overall cancer risk in five genetic models [DD vs. II: odds ratio (OR)=0.76, 95% confidence interval (CI)=0.69–0.84, ID vs. II: OR=0.87, 95% CI=0.83–0.92, DD vs. ID/II: OR=0.82, 95% CI=0.75–0.89, ID/DD vs. II: OR=0.85, 95% CI=0.80–0.90, and D vs. I: OR=0.87, 95% CI=0.83–0.91]. Stratified analyses showed that the polymorphism was associated with decreased risk of colorectal, breast, esophageal, renal cell, lung, cervical, bladder, gastric, and other cancers. Overall cancer risk was reduced in Asian and Caucasian patients, both hospital- and population-based studies, and both high and low quality studies. Our results highlight the role of the CASP8 -652 6N ins/del polymorphism in decreasing cancer risk. Further studies with large-cohort populations, especially for specific cancer types and ethnic groups, are needed to confirm our findings.


INTRODUCTION
Cancer is a substantial public health burden worldwide and is the second leading cause of death in the United States. An estimated 1,688,780 new cancer cases and 600,920 cancer deaths will occur in the United States this year [1]. Approximately 14 million new cancer cases occurred worldwide in 2012, and by 2025, global cancer incidence is predicted to rise to 20 million new cases annually [2]. Although there are many cancer risk factors, genetic abnormalities play crucial roles in carcinogenesis [3][4][5][6].
Extensive epidemiological studies have assessed the association between the CASP8 -652 6N ins/del polymorphism and cancer risk. However, these studies have not produced conclusive results. The most recent previous meta-analysis of this association, conducted in 2014, assessed a relatively small number of studies. We performed this meta-analysis with a larger sample size to more precisely describe the association of interest.

Study characteristics
Our study selection workflow is shown in Figure 1. Our systematic computer-based search initially identified 108 potentially relevant articles. After scanning titles and abstracts, 67 articles about unrelated topics were excluded.
Characteristics for 33,494 cases and 36,397 controls are summarized in Table 1. Of the included studies, 12 were conducted on colorectal cancer, nine on other cancers, eight on breast cancer, three on esophageal cancer, three on renal cell carcinoma, and two on lung cancer, cervical cancer, prostate cancer, bladder cancer, lymphoma cancer, and gastric cancer, respectively. Twenty-seven studies were conducted in Asians, 20 in Caucasians, one in Africans, and one in mixed populations. Twenty-four studies were of population-based design, 22 studies were of hospital-based design, and three did not mention study design in the original data. We also classified the studies as either low quality (25 studies) or high quality (24 studies) by quality score.

Heterogeneity and sensitivity analysis
Heterogeneity was observed in all five genetic models (P<0.001, Q test). Therefore, the random-effect model was adopted to generate ORs and 95% CIs. We also conducted a sequential leave-one-out sensitivity analysis to evaluate the impact of a single study on the pooled estimates. Omission of no single study influenced the pooled ORs, indicating the statistical robustness of this meta-analysis (data not shown).

Publication bias
Begg's funnel plot shapes did not suggest any obvious asymmetry (Figure 3). Egger's test results (DD vs.

Trial sequential analysis
To minimize random errors and strengthen the robustness of our conclusions, we performed trial sequential analysis (TSA) (Figure 4). The cumulative Z-curve crossed the trial sequential monitoring boundary before the required information size was reached, suggesting that our study conclusion was convincing and no additional evidence was needed to verify said conclusion.

DISCUSSION
The present meta-analysis comprehensively evaluated the relationship between the CASP8 -652 6N ins/del polymorphism and cancer risk across 49 studies (33,494 cases and 36,397 controls). The CASP8 -652 6N ins/del polymorphism was associated with decreased cancer risk in all five genetic models, and in the following subgroups: colorectal cancer, breast cancer, esophageal cancer, renal cell carcinoma, lung cancer, cervical cancer, bladder cancer, gastric cancer, other cancers, Asian, Caucasian, mixed population, population-based controls, hospital-based controls, high quality score, and low quality score.
Human immune cells play critical roles in eliminating potentially malignant cells [58]. Caspase 8 Figure 3: Funnel plot analysis to detect publication bias for the CASP8 -652 6N ins/del polymorphism via the homozygous model. Each point represents a separate study for the indicated association. www.impactjournals.com/oncotarget protein (encoded by CASP8) maintains immune cells by mediating the activation-apoptosis balance [59]. Low caspase 8 expression or functional aberrations may decrease T lymphocyte apoptotic reactivity [13]. The CASP8 -652 6N del variant inactivates the transcription factor stimulatory protein 1 binding site, decreasing CASP8 transcription [13]. Thus, this variant may affect cancer susceptibility by influencing immune surveillance.
The first case-control study of the CASP8 -652 6N del variant-cancer association, with 4,995 cases and 4,972 controls, was conducted by Sun, et al. in 2007 [13]. The authors found that the CASP8 -652 6N deletion allele decreased susceptibility to lung, colorectal, esophageal, breast, cervical, and gastric cancers. Biochemical assays illustrated that this variant might decrease apoptotic reactivity in cancer cell-stimulated T lymphocytes. However, Umar, et al. did not detect any association between the CASP8 -652 6N polymorphism and esophageal squamous cell carcinoma (ESCC) risk in 259 patients and 259 healthy controls in an Indian population [45]. Several meta-analyses have attempted to address these contradictory conclusions. A 2012 metaanalysis by Chen, et al., including 19 case-control studies with 23,172 cases and 26,532 controls, associated the del allele, ins/del genotype, and del allele carriers with reduced overall cancer risk [16]. Similarly, in a metaanalysis incorporating 11 reports with 27,459 cases and 31,614 controls, Yin, et al. associated the CASP8 -652 5N del polymorphism with reduced overall cancer risk via homozygous, dominant, and recessive models [15]. In 2014, breast cancer-and colorectal cancer-specific meta-analyses [19,20] concluded that the CASP8 -652 6N del polymorphism reduced cancer risk. However, no association was observed between this polymorphism and prostate cancer susceptibility in a meta-analysis by Zhang, et al. [21].
To provide a more robust clarification, our metaanalysis included all eligible studies published in either the English or Chinese language. In agreement with the four previously published meta-analyses, we found that the CASP8 -652 6N ins/del polymorphism was associated with reduced overall cancer risk. In subgroup analyses, the polymorphism was associated with reduced risk of colorectal cancer, breast cancer, esophageal cancer, renal cell carcinoma, lung cancer, cervical cancer, bladder cancer, gastric cancer, and other cancers, but not prostate cancer or lymphoma. A prostate cancer-specific metaanalysis also failed to detect a significant association. This may be attributed to cancer-specific inherent heterogeneity [60,61]. Additionally, we observed an association with decreased cancer risk among Asians and Caucasians, but not Africans or mixed ethnicity populations. However, the limited number of studies in Africans and mixed ethnicity population may account for this finding, and CASP8 -652 6N ins/del polymorphism allelic distributions might vary geographically and ethnically.
Our meta-analysis of the association between the CASP8 -652 6N ins/del polymorphism and cancer risk is by far the largest such meta-analysis with the greatest statistical power published thus far. We conducted subgroup analyses to provide a more precise, cancer type-specific conclusion, and we assessed studies in both Chinese and English to minimize selection bias. However, our study had certain limitations. First, for some types of cancers, the calculated association was not robust enough due to limited numbers of original studies. Second, only one CASP8 genetic variant was considered, and confounding factors, such as other genetic mutations and environmental exposures, also influence cancer susceptibility. Third, the observed between-study heterogeneity may reduce the validity of our conclusions. Finally, publication bias, language bias, or selection bias might lead to false positive or negative findings.
The present work robustly concludes that the CASP8 -652 6N ins/del polymorphism is associated with reduced overall cancer risk. Refined studies with larger sample sizes, especially for certain cancer types and ethnic groups, are needed to fully validate this relationship.

Search strategy
We conducted a literature search in PubMed and EMBASE using the following combined terms: 'Caspase 8' or 'CASP8' and 'polymorphism' or 'polymorphisms' or 'single nucleotide polymorphism' or 'SNP' or 'variant' and 'cancer' or 'tumor' or 'carcinoma' or 'carcinogenesis' or 'neoplasm'. We also searched studies written in Chinese from two databases, WANFANG and CNKI. We searched for articles published through February 2017 without imposing language limitations. Relevant references were also collected from retrieved articles. Only the largest or the most recent study was retained if studies contained overlapping data.

Inclusion/exclusion criteria
Studies included in our analysis met the following criteria: (1) evaluated CASP8 -652 6N ins/del polymorphism with respect to cancer risk; (2) case-control design; (3) sufficient information to extract genotype frequencies for all subjects; (4) genotype frequency of controls consistent with Hardy-Weinberg equilibrium (HWE); (5) publication language was English or Chinese. Criteria for exclusion included: (1) abstract only, review, or meta-analysis; (2) case only studies; (3) no detailed genotyping data provided; (4) repeated publication.

Data extraction
Two authors (Jiarong Cai and Qingjian Ye) independently identified all eligible studies, and extracted data was included in the meta-analysis following consensus. The following items were recorded from each study: first author's name, year of publication, country, patient ethnicity, cancer type, source of controls, genotyping method, and genotype distributions of cases and controls. If reports contained more than one ethnic group or cancer type, we separated them into different studies.

Trial sequential analysis
After adopting a risk of 5% for type I errors and 30% for type II errors, the required information size (sample sizes from all included trials) was calculated. TSA monitoring boundaries were built based on required information size and risk for type I and type II errors. If the cumulative Z-curve crossed the TSA monitoring boundary before the required information size was reached (i.e. if a sufficiently small P-value was achieved), further trials were unnecessary.

Statistical analyses
We used the Chi-square test to ensure that all control genotype frequencies were in agreement with HWE. Odds ratios (ORs) with corresponding 95% confidence intervals (CIs) obtained from case and control genotype frequencies were used to assess the strength of association between the CASP8 -652 6N ins/del polymorphism and cancer risk. Pooled ORs were calculated for the following five genetic models: homozygote model (DD vs. II), heterozygote model (ID vs. II), recessive model (DD vs. ID/II), dominant model (ID/DD vs. II), and allele model (D vs. I). The Cochran's Chi-square-based Q-test and the inconsistency index (I 2 statistics) were adopted to assess heterogeneity between study results. I 2 <50% or P>0.10 indicates heterogeneity. The fixed-effects model (Mantel-Haenszel method) was used to estimate the pooled OR if no heterogeneity existed (I 2 <50% or P>0.10). Otherwise, the random-effects model (DerSimonian and Laird method) was applied. Quality assessment for each study was performed using the quality assessment criteria described previously (Supplementary Table 1) [62][63][64][65]. To decrease heterogeneity among studies, we conducted stratification analyses by ethnicity, cancer type, control source, and quality score. By adopting one-way sensitivity analysis, we recalculated the pooled ORs to assess the robustness of the results. We also conducted Begg's funnel plot and Egger's regression asymmetry test to examine potential publication bias [66][67][68][69]. STATA software v. 11.0 (Stata Corporation, College Station, TX) was used for statistical analyses [70]. P<0.05 (two-sided) was considered statistically significant.

CONFLICTS OF INTEREST
The authors declare no conflicts of interest.

GRANT SUPPORT
This study was supported by grants from the National Natural Science Foundation of China (Grant No. 31172044).