Germline HOXB13 p.Gly84Glu mutation and cancer susceptibility: a pooled analysis of 25 epidemiological studies with 145,257 participates.

Numerous studies have investigated association between the germline HOXB13 p.Gly84Glu mutation and cancer risk. However, the results were inconsistent. Herein, we performed this meta-analysis to get a precise conclusion of the associations. A comprehensive literature search was conducted through Medline (mainly Pubmed), Embase, Cochrane Library databases. Crude odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated by STATA 12.1 software to evaluate the association of HOXB13 p.Gly84Glu mutation and cancer susceptibility. Then, 25 studies including 51,390 cases and 93,867 controls were included, and there was significant association between HOXB13 p.Gly84Glu mutation and overall cancer risk (OR = 2.872, 95% CI = 2.121-3.888, P < 0.001), particularly in prostate cancer (OR = 3.248, 95% CI = 2.313-4.560, P < 0.001), while no association was found in breast (OR = 1.424, 95% CI = 0.776-2.613, P = 0.253) and colorectal cancers (OR = 2.070, 95% CI = 0.485-8.841, P = 0.326). When we stratified analysis by ethnicity, significant association was found in Caucasians (OR = 2.673, 95%CI = 1.920-3.720, P < 0.001). Further well-designed with large samples and other various cancers should be performed to validate our results.


INTRODUCTION
Cancer is a serious problem endangering the human health and lives. Based on the reports from International Agency for Research on Cancer (IARC), cancer has become the second leading cause of mortality in developing countries, which has exceeded the mortality caused by cardiovascular incidences and become the leading cause of mortality in developed countries [1,2]. Totally, 1,658,370 new cancer cases were diagnosed and 589,430 patients died from cancer in the United States in 2015 [3], suggesting that the burden of cancer will be heavier year by year, due to the increasing number of world population and the problem of aging is getting worse [4]. Although the mechanism of carcinogenesis remains elusive, multiple environmental and lifestyle factors has been confirmed contributed to the formation of cancers. However, not all cancer patients who have been exposed to the risk factors will develop cancer, suggesting the inter-individual differences in susceptibility [5]. Therefore, genetic, environmental and life time factors were suggested to be the main determinant of individual risk for cancer [6,7]. In recent years, numerous studies have pointed out genetic factors, particularly single nucleotide polymorphisms (SNPs) of genes, plays crucial roles in tumorigenesis [8][9][10][11].
HOXB13, which encodes the transcription factor 13, belongs to the HOXB gene cluster at chromosome 17 [12], involves in embryonic development of different organs [13], regulates transcription of androgen receptor (AR) target genes [14] and is reported to function as a tumor suppressor in cancer [15]. Deregulation of HOXB13 expression has been reported in a number of malignancies, including prostate, breast, colon, lung, endometrial, renal cancers and melanoma [16][17][18][19]. Recently, a novel germline mutation, p.Gly84Glu (rs138213197), in exon one of the HOXB13gene, was suggested to have a close relationship with the risk of various cancers. Numerous studies have focused on the association between the germline HOXB13 p.Gly84Glu mutation and cancer risks, however, the results are inconsistent. Thus, we comprehensively searched all related literatures and performed present meta-analysis which has great power through polling all eligible related data to get a more precise conclusion. Figure 1 represents the process of eligible studies' identification and selection. The literature selective process was conducted rigorously according to the inclusion and exclusion criteria. Finally, 15 publications involving 25 individual studies with 51,390 cases and 93,867 controls were included in the present meta-analysis [20][21][22][23][24][25][26][27][28][29][30][31][32][33][34]. The main characteristics of included studies were summarized in Table 1. These studies included 19 prostate cancer studies, 3 breast cancer studies and 3 colorectal cancer studies. OF the 25 studies, there were 21 studies of Caucasian, and the other four studies of mixed ethnicity (both of them were the mixed population of Caucasian, Asian and African-American). As for the control source, five studies applied population-based (PB) control, 14 studies employed hospital-based (HB) control, four   Oncotarget 42315 www.impactjournals.com/oncotarget studies applied PB/HB control, one studies applied familybased (FB) control, while the other one applied HB/FB control. Simultaneously, various genotyping methods were employed of all included studies, such as, 10 studies applied TaqMan assay, two studies applied Sanger sequencing, four studies applied MassARRAY iPLEX, six studies applied Illumina SNP, and the rest three studies employed complex methods (TaqMan, MassARRAY iPLEX, Sanger sequencing). The genotype distributions of all included studies in this meta-analysis were in agreement with Hardy-Weinberg equilibrium (HWE). The estimated quality of all included studies was in the range of 7-9 scores and was listed in Table 1.

Quantitative data analyses
Finally, 25 epidemiological individual studies including 51,390 cases and 93,867 controls were enrolled in this meta-analysis. There is significant heterogeneity was found in over cancer risk estimation (I 2 = 62.8%, P heterogeneity < 0.0001). Considering that, random-effects model was used to examine the association between HOXB13 p.Gly84Glu mutation and overall cancer susceptibility (OR = 2.872, 95% CI = 2.121−3.888, P < 0.001; Figure 2). In order to detect the source of heterogeneity, subgroup analyses were conducted by cancer type, ethnicity, control source and genotyping method. When we stratified by cancer type, there is significant heterogeneity was existed (I 2 = 68.4%, P heterogeneity < 0.0001) and then random-effect models was used to verify the relationship between HOXB13 p.Gly84Glu mutation and prostate cancer risk. The results presented that HOXB13 p.Gly84Glu mutation contributed to the susceptibility of prostate cancer (OR = 3.248, 95% CI = 2.313−4.560, P < 0.001; Figure 3). While no heterogeneities (for breast cancer: I 2 = 0.0%, P heterogeneity = 0.958; and for colorectal cancer: I 2 = 36.9%, P heterogeneity = 0.205, respectively) were found  Table 2) and colorectal cancer (OR = 2.458, 95% CI = 0.98−6.177, P = 0.056; Table 2) using fixedeffect models. Moreover, further subgroup analyses were also performed by study design and genotyping method. All the results of meta-analyses were summarized in Table 2.

Sensitivity analyses
One-way sensitivity analysis of the pooled OR and 95% CIs for HOXB13 p.Gly84Glu was performed to verify if the results of our present meta-analysis were robust. The pooled ORs were calculated by means of a random effects model. To the best of our knowledge, in a sensitivity analysis, if a single study included in a metaanalysis was omitted each time, the pooled ORs were always persistent and it can be considered as the results of this meta-analysis were reliable and stable. In the present Oncotarget 42317 www.impactjournals.com/oncotarget meta-analysis, no single study was qualitatively influenced by the pooled ORs when they were sequentially omitted, as indicated by the sensitivity analyses, suggesting that the results of our present study are stable ( Figure 4).

Publication bias
Begg's tests were employed to detect the potential publication bias that may be existed in this meta-analysis, and the results suggested there was no publication bias (P = 0.815, Figure 5a). Egger's tests also confirmed the absence of publication bias in the meta-analysis (P = 0.30, Figure 5b).

DISCUSSION
Here, we conducted the largest meta-analysis to summarize the association between HOXB13 p.Gly84Glu mutation and cancer risk through pooling all candidate studies. Totally, 25 epidemiological case-control studies including 51,390 cases and 93,867 controls were enrolled this present study. In recent years, HOXB13 p.Gly84Glu mutation was found to contributed to the risk of prostate cancer [20-24, 26-31, 33, 34], especially in European countries [21]. However, no significant association was found in Asians [22] and Africans [21,22]. Additionally, there are also three studies from two publications reported the association and they got the final conclusion that this mutation was not found to have increased the susceptibility of breast cancer on both familial and sporadic patients [25,29]. Similarly, no significant association was detected among the mutation and the risk of colorectal cancer [29,32]. In our present study, we summarized all the effects of HOXB13 p.Gly84Glu mutation with the risk of various cancers. We got a conclusion that the mutation is significantly increased the risk of cancers. Then, we performed subgroup analyses according to cancer type, ethnicity, text method, et al. We found that the mutation is contributed to the risk of prostate cancer, which was in accordance with previous studies. Moreover, we failed to find a statistical association between the mutation and the susceptibility of breast cancer and colorectal cancer. As such, the results were similar with the previous studies mentioned above. Moreover, stratified analyses by ethnicity were performed, and the conclusion suggested that the mutation can increase the risk of overall cancer among Caucasians, particularly in European decedent patients.
Substantial heterogeneity between studies was existed in our meta-analysis, just as a common aspect of genetic association studies. We determined the heterogeneity by Q-test and I 2 test, and statistically  Oncotarget 42319 www.impactjournals.com/oncotarget significant heterogeneity was observed. Then, randomeffects model was used to analyze the ORs with 95% CI. Although we performed the present study according to the PRISAM strictly, including strict criteria of selection publications and meta-regression performance, there was no source of heterogeneity found in our present meta-analysis. Therefore, we carried out subgroup meta-analyses, and the results suggested that these parameters including ethnicity, cancer type, control source and genotyping method may be the main source of heterogeneities. In the stratified analysis by genotype method, the heterogeneity was significantly reduced, suggesting that genotype method may be one of source of heterogeneities.
To our knowledge, meta-analysis has great power through pooling all eligible studies, and thus gets a reliable and relative precise result. In the present study, there are several advantages existed. Above all, this is by far an analysis with the largest sample size, which can make our result more reliable and precise. What's more, the quality of each study include study was high, ranged from 7-9 score. Moreover, one-way sensitivity analysis was performed, and the result suggested that no significant influence of a single study on the pooled ORs and 95% CI. Simultaneously, no significant publication bias was detected in our present work. Of the two factors mentioned above, it demonstrated that our results were stability and reliable. In addition, subgroup analyses were conducted to explore the association of HOXB13 p.Gly84Glu mutation and susceptibility to the three types of cancers.
Some limitations existed and should be acknowledged in the meta-analysis. On the one hand, there were only three types of cancers including prostate cancer, breast cancer and colorectal cancers. Based on that, the results of this present work may not have enough power to represent all kinds of cancer. On the other hand, most of studies included in this meta-analysis were of European and USA decedents belonged to Caucasian ethnicity, which was a cause of selection bias, and other ethnicities, such as, African, and Asian ethnicities should be included in further studies. In addition, even though no sample size and language limitations were set, related studies in other languages may be ignored. What's more, only published studies were included in this meta-analysis, while other unpublished studies in different languages should be enrolled. Finally, adjusted estimations were not performed for insufficient data, such as, age, sex, smoking and drinking habits et al which can interrupt the results of present study. Simultaneously, interactions among genegene, gene-environment, and even different polymorphism loci of the same gene were not conducted for lacking of sufficient data in this work, which may regulate the gene expression, affect the function of gene product, and lead to the different OR values. Thus, further studies with same topics should consider the factors mentioned above.
In summary, the meta-analysis suggested that HOXB13 p.Gly84Glu mutation contributed to the overall cancer risk, especially for prostate cancer. Considering limitations mentioned above, further well-designed studies with larger sample size should be conducted to verify the results of the present meta-analysis.

MATERIALS AND METHODS
The present meta-analysis was performed according to the latest meta-analysis guidelines (PRISMA) [35].

Search strategy
A comprehensive computerized literature search was conducted through Medline (main Pubmed), Embase, Cochrane Library, Web of Science, Wanfang and China National Knowledge Infrastructure (CNKI) databases for related research studies reported the association between HOXB13 p.Gly84Glu mutation and cancer susceptibility. Furthermore, we also searched related studies manually from the references of reviews and articles reported the same topics. Combinations of searching terms were used as follows: "HOXB13 p.Gly84Glu'', "HOXB13 rs138213197", ''single nucleotide polymorphism, SNP or variation, mutation'' and ''cancer or carcinoma or tumor or neoplasms". In order to get a precise conclusion, no sample size and language limitations were set, hoping that we can identify all the studies that examined the association of HOXB13 p.Gly84Glu mutation and cancer risk.

Inclusion and exclusion criteria
The following criteria should be met of each included studies. (1) reported the association between HOXB13 p.Gly84Glu mutation and cancer risk; (2) used case-control design; (3) all the patients in cases group should be diagnosed by histochemical results or other gold diagnostic standers; (4) provided sufficient data of HOXB13 p.Gly84Glu mutation carriers or non-carriers, or other information such as ORs with 95% CIs for statistical analysis; (5) if there were several publications with overlapping data, only the latest one with the largest sample size was finally included in this present work. At the same time, if each of the searched studies was in accordance with the following criteria, it must be excluded. (1) human being studies; (2) review, meeting or other types of abstracts, comment, correspondence, letters or letters to the editor; case reports, or case-only studies; (3) not provided the sufficient data to extract.

Data extraction
Two independent investigators extracted the essential information according to the selection criteria mentioned above. The key data was listed as follows: