CCDC26 rs4295627 polymorphisms associated with an increased risk of glioma: A meta-analysis

,

Gliomas are the most common primary brain tumors that occur mainly in adults. They include astrocytic, oligodendroglial, oligoastrocytic, ependymal and choroid plexus, and other neuroepithelial tumors. [1][2][3] Gliomas account for about 80% of all primary malignant brain tumors, with an incidence of 5-10 cases per 100,000. 3,4 Despite advances in neurosurgery and chemotherapy, the prognosis for most glioma patients remains dismal. 5,6 Prevention of glioma progression has therefore become an important strategy for fighting against the disease.
The etiology of gliomas is still not well understood. Previous studies have reported that some genetic loci, such as 5p15.33 (rs2736100, TERT) and 8q24.21 (rs4295627, CCDC26) , may be associated with the risk of glioma. 7,8 The coiled-coil domain containing 26 (CCDC26) gene encodes a retinoic acid modulator of cell differentiation and death. 9 Retinoid acid induces caspase-8 transcription through phosphorylation of cAMP response element-binding, and increases apoptosis induced by death stimuli in neuroblastoma cells and in glioblastoma cells by downregulation of telomerase activity. 10,11 Genetic variants of CCDC26 are associated with a number of common tumors, including glioma. 7,8,12,13 The SNP rs4295627 maps to intron 3 of the CCDC26 gene. Several studies have been published exploring the relationship between the rs4295627 polymorphism and the risk of glioma. 7,8 However, the results of those studies are inconsistent. [14][15][16] Therefore, the aim of the present study was to investigate the effect of CCDC26 rs4295627 polymorphism on the risk of glioma by summarizing it quantitatively, using a meta-analysis approach.

Data sources
A comprehensive search was performed for available articles published in English, using the databases of PubMed, EMBASE and Google Scholar up to December 2015, and by hand-searching the reference lists of the computer-retrieved articles. The literature search was conducted using the following terms: "glioma", "gliomas" or "glioblastoma"; "rs4295627", "CCDC26" or "coiled-coil domain containing 26"; and "polymorphism", "genotype" or "mutation". The search was limited to human studies and references from the retrieved publications were checked to find additional articles on the topic.

Inclusion and exclusion criteria
All relevant studies reporting the association between the rs4295627 polymorphism and glioma risk were considered for inclusion. The inclusion criteria were as follows: (1) use of a case-control or cohort design; (2) the exposure of interest was rs4295627 polymorphism; (3) the outcome of interest was glioma; (4) sufficient raw data for evaluating odds ratios (OR) and their 95% confidence interval (CI); or, if the raw data was not available, the OR and 95% CI for specific genetic models were included. The following exclusion criteria were also used: (1) articles only having an abstract, review articles and comments; (2) studies overlapping with other studies.

Data extraction
Two authors independently reviewed all the articles and extracted data in separate databases. Conflicts were resolved by discussion and consensus. The following information was extracted from each study: the name of the 1 st author, the year of publication, ethnicity, the number of cases and controls, the sources of the controls, the genotyping method, and either the raw data on genotype frequency or ORs with corresponding 95% CIs for specific genetic models.

Statistical analysis
Comprehensive meta-analysis software (v. 2; Biostat Inc., Englewood, USA) was used for all the statistical analyses. 17 The pooled odds ratios (ORs) and 95% confidence intervals (CIs) were taken to calculate the effects of the CCDC26 rs4295627 polymorphism on the risk of glioma. Heterogeneity between studies was estimated by the I 2 test and heterogeneity Q statistic test. P-value < 0.10 and I 2 > 50% were considered the level of statistical significance. The random-effects model was used for pooled effects; otherwise, the fixed-effects model was chosen for analysis since heterogeneity was not obvious. The genetic models assessed for the pooled ORs of the polymorphism were the additive model (G vs T), dominant model (GG + GT vs TT), recessive model (GG vs GT + TT), homozygote comparison (GG vs TT) and heterozygote comparison (GT vs TT). Subgroup analyses by ethnicity (American Caucasians, European Caucasians and Asian Han) were also conducted to further explore the details of CCDC26 rs4295627 polymorphisms and the risk of glioma. In addition, possible publication biases were investigated using the funnel plot and Begg's linear regression test. For Egger's tests, p-values < 0.1 suggest significant publication bias. 17

Study characteristics
A total of 99 articles were identified using the search strategies from the 3 electronic libraries. Following deduplication and screening of the titles and abstracts, 43 out of the 99 records were excluded for showing no information related to rs4295627 polymorphisms and glioma risk. Another 46 articles were then removed owing to insufficient data after reviewing the full text. Out of the 10 candidate studies, 7,8,[14][15][16][18][19][20][21][22] 4 articles 16,[18][19][20] and part of the data in 1 article 21 were rejected because of repeated or overlapping data. Finally, 6 articles 7,8,14,15,21,22 with 15 studies about rs4295627 polymorphisms and glioma risk, involving 8292 glioma cases and 12,419 controls, were used in this meta-analysis.
The main information from the 15 studies is summarized in Table 1. Six of the selected studies were carried out in Europe, 7 in the USA and 2 in China. The number of cases in these studies varied substantially (ranging from 18 to 1374 individuals). Twelve of the 15 studies provided detailed raw genotype data for calculating the ORs and 95% CIs (Table 2), the other 3 studies provided the ORs and 95% CIs for specific genetic models.

Quantitative synthesis
The results of this meta-analysis are presented in Fig Fig. 3 right). Between-study heterogeneity was found in most of the 5 genetic models. Accordingly, the fixed-effects model was used only in heterozygote    comparison (GT vs TT). For the other genetic models, the random-effects model was utilized in the meta-analysis, since heterogeneity was significant. The results of the subgroup analyses are presented in Table 3. Racial differences were found among the subgroups of ethnicity. A significant association was discovered between the rs4295627 polymorphisms and the risk of glioma in the American Caucasians, and between-study heterogeneity was no longer significant in this group. The Han Chinese population showed no significant association between the rs4295627 polymorphisms and the risk of glioma; however, the small number of studies included for analysis may lack sufficient power to draw accurate conclusions. European Caucasians showed the highest effect size compared with the other samples.

Publication bias
The funnel plots (Fig. 4) of all 5 genetic models were roughly symmetric and suggested no publication bias. Egger's test also showed no significant evidence of publication bias (p = 0.74 for G vs T; p = 0.97 for GG vs TT; p = 0.23 for GT vs TT; p = 0.30 for GG + GT vs TT; and p = 0.87 for GG vs GT + TT).

Discussion
The present meta-analysis used the data from 15 casecontrol studies (with a total of over 20,000 participants) to evaluate the effect of the CCDC26 rs4295627 polymorphisms on the risk of glioma. The pooled results for different genetic models were all significant, which indicated that a person who carries the G allele of rs4295627 polymorphisms has an increased likelihood of glioma. Individuals with the homozygous variant have the highest risk of glioma (OR = 1.72), and the effect size of the recessive genetic model is higher than that of the dominant model (OR = 1.65 vs OR = 1.36, adjusted for heterogeneity). Different ethnicities have different frequencies of alleles; therefore, subgroup analyses according to ethnicity were performed to decrease biases. The subgroup analyses showed that the rs4295627 polymorphism in the CCDC26 gene increased the risk of glioma in Caucasians, but similar associations are not observed in Han Chinese individuals. However, since only 2 studies involved Asian samples, the results may have insufficient power to reveal a reliable association. More studies are needed to confirm the results for the Chinese population.
Gliomas are the most common adults tumors of the central brain, and have high mortality and morbidity. 3 A better understanding of the mechanism of gliomas will contribute to finding better ways to prevent, diagnose and treat them. Confirmation of genetic biomarkers could help in making early diagnosis, predicting patient outcomes or carrying out personalized therapy. However, previously published articles have found that many gene variations may be associated with the risk of glioma, such as CCDC26 gene mutations. The CCDC26 gene plays an important role in cell differentiation and apoptosis. CCDC26 gene mutations are present in most glioma samples, but are absent in normal brain tissues. Theoretically, CCDC26 gene mutations may change the intrinsic regulatory mechanism in organisms and thus increase the occurrence of glioma. This study showed that mutations of the G allele of the rs4295627 polymorphism, located in intron 3 of the CCDC26 gene, increases the risk of glioma, which is consistent with the theoretical assumption.
Between-study heterogeneity is common in meta-analyses, and exploring potential sources of heterogeneity is an essential component of any meta-analysis. 23 Evidence of heterogeneity was observed in most of the genetic models for this meta-analysis. In fact, the effect size of the rs2736100 polymorphism on glioma risk varied greatly in the 15 studies included. Although most of the studies found that mutations of the rs2736100 polymorphism may increase the risk of glioma, some studies reported inconsistent results. The between-study heterogeneity found in the meta-analysis may have several potential causes. The most likely sources of heterogeneity are the lack of a standardized classification for the different types or severity of glioma. Rs4295627 polymorphism has been proved to be strongly associated with oligodendro-glial tumor risk, but not glioblastoma risk. 24 Differences in the participants' characteristics, including differences in age and ethnicity, are another potential source of heterogeneity. Other possible sources of the observed heterogeneity, such as false-positive or false-negative associations, may result from survival bias and recall errors involved in cross-sectional studies. 25,26 The present study included subgroup analyses by ethnicity and found that ethnicity may be one of the sources of heterogeneity. However, since most of the 15 studies lacked data for other factors (research design, interventions and outcome measures), this hypothesis is still speculative and needs further testing.
The present study has some limitations that should be acknowledged. First, precise information about the type of glioma and the participants' details were absent from most of the 15 studies included. No relevant published or unpublished studies with null results was identified, which may bias the results of the present meta-analysis. Ideally, to avoid bias, the effect size should be adjusted for all factors known to contribute to glioma. However, because some of the studies used in the current metaanalysis did not include all the relevant data, the crude effect size was calculated using only tabular data. Second, although a meta-analysis is a good method of obtaining a large sample size and increasing statistical power, the heterogeneity bias may bring in some "noise". A random-effects model may minimize the influence of heterogeneity by assuming that different studies show substantial diversity and assessing both within-study sampling error and between-study variance. However, the influence of heterogeneity in the present analysis should be noted.
Despite these limitations, this study has a large sample size and sufficient statistical power to estimate the effect of the rs2736100 polymorphism on the risk of glioma. In contrast, a small sample size (<100 cases and controls) can overestimate a true association due to deficiencies in statistical power. In addition, although the number of studies included in this meta-analysis was small, no evidence of publication bias was found.
In conclusion, this meta-analysis confirms that the CCDC26 rs4295627 polymorphism contributes to individual susceptibility to glioma, and provides a more accurate estimate of the effect of the rs4295627 polymorphism on the risk of glioma. Because of the previously mentioned limitations of this meta-analysis, further studies are necessary to resolve the existing controversies over the effect of the rs4295627 polymorphism on the risk of glioma.