Vitamin D receptor Taq I polymorphism and the risk of prostate cancer: a meta-analysis

Numerous previous studies reported the association of Vitamin D receptor gene Taq Ipolymorphism with prostate cancer risk, however these results were controversial. In order to provide a relatively comprehensive description of this relationship, we conducted this meta-analysis by searching PubMed, Embase, and China National Knowledge Infrastructure. Finally, 36 studies with 8,423 cases and 8,887 controls were included. Taq I polymorphism was found to marginally increase the prostate cancer risk in recessive genetic model (tt/Tt vs. TT: Odds Ratio (OR) = 0.89, 95% Confidence Interval (CI) = 0.80–1.00, p = 0.05) and allele genetic model (t vs. T allele: OR = 0.91, 95% CI = 0.84–0.99, p = 0.003) in the overall analysis. Subgroup analyses showed that significant increased risk was found in Asians in homozygote model (tt vs. TT: OR = 0.63, 95% CI = 0.41–0.95, p = 0.029) and allele genetic model (t vs. T: OR = 0.78, 95% CI = 0.67–0.90, p = 0.002), and in the subgroup of population-based controls in all the genetic models. These results suggest that Taq Ipolymorphism might be a risk factor of prostate cancer risk, especially in Asians. It could be considered as a promising target to predict the prostate cancer risk for clinical practice.


INTRODUCTION
Prostate cancer (PCa) is second-most commonly diagnosed malignancy in males, and thought to be one of the leading causes of cancer-related death around the world. In 2014, approximately 233,000 newly diagnosed cases and 30,000 PCa-related deaths was reported in America [1]. Furthermore, the global incidence is rising rapidly. By 2030, the number of new PCa and PCa-related deaths annually will climb to 1,853,391 and 544,209, respectively [2]. The etiology of PCa has remained unclear. Several factors are considered to significant increase the risk of PCa, including ethnicity, hormonal status, environment, diet, aging, and genetic factors [3].
Low serum levels of vitamin D might be one of the risk factors for PCa [4]. Laboratory investigation demonstrated that vitamin D inhibits the growth and differentiation of PCa cells, decreases the invasion, metabolism and angiogenesis of tumor cell. It can also promote tumor cell apoptosis [4]. In 2007, a clinical trial suggested that calcitriol, a kind of analogue of vitamin D can significantly improve patients' survival rate by decreasing serum level of prostate special antigen (PSA) [5]. The antineoplastic effect of vitamin D is activated when binding to vitamin D receptor (VDR) [6]. 1,25-Dihydroxy vitamin D3(1,25(OH) 2 D 3 ) is the hormonally active form of vitamin D. It binds to VDR and forms a heterodimer complex, which subsequently binds to the vitamin D response element and reduces the
Recently, the relationship of several single nucleotide polymorphisms (SNPs) of VDR gene and PCa risk has been the focus of research attention [8,9]. Taq I polymorphism (rs731236) is one of the most widelystudied SNPs. It is a synonymous mutation located in exon 9 of VDR gene [10]. This mutation could reduce the mRNA stability and therefore decrease the mRNA levels of VDR gene [11]. Recently, some studies have suggested that Taq I variation might increase the susceptibility of PCa [12,13]. However, these results are debatable and inconsistent in the effect of Taq I polymorphism on PCa risk. Numerous studies in favor of the association of Taq I polymorphisms and PCa risk [14][15][16][17][18][19]. Meanwhile, some studies disapprove of the relationship [20][21][22]. The difference might be due to under-power for individual study. Moreover, previous meta-analyses [10,23,24] seem to be outdated since new data appeared [17,[25][26][27]. Therefore, we conduct this meta-analysis to get more accurate results.

Characteristics of studies
We identified 288 potentially relevant studies following the retrieval strategy. Based on the inclusion criteria, 36 studies [3, 7, 9, 12, 14-19, 22, 25-49] between 1996 to 2017 were finally included ( Figure 1). The number of cases and controls varied from 28 to 1,617, and 41 to 1,072, respectively ( Table 1). The genotype distribution frequency in the control groups was consistent with Hardy-Weinberg equilibrium (HWE) for most studies, except for four studies [12,19,25,49]. Each individual study scored more than 4 by Newcastle-Ottawa Scale (NOS), and was considered to be of high quality (Table 1). The percentages of tt, Tt and TT genotype in case group and control group were 11.9%, 40.4%, 47.7% and 12.1%, 41.3%, 46.6%, respectively in overall population.

Pooled results
As shown in Figure 2 and  Table 2).
For the stratified analysis of different ethnicities, significantly increased risk was found in Asians in T allele genotype carriers (t vs. T: OR = 0.79, 95% CI = 0.68-0.91, p = 0.002) ( Table 3 and Figure 2). However, when 15 studies performed in Caucasians and 3 studies in Africans were analyzed, no significant associations were found in any comparison models (Table 3).
Studies were stratified into TaqMan, PCR-RFLP, and SNPlex groups by genotyping methods. No significant association was found in almost subgroups, except TaqMan group in tt vs. TT/Tt comparison ( Table 5). The pooled outcome indicated that the genotyping methods included in these studies are both available and did not alter the outcomes.
As shown in Figure 4 and Table 6, we also performed a stratified analysis based on the clinical stages by Gleason Score to describe the relationship in more detail. The pooled results from 9 studies for advanced tumor group and 8 studies for localized tumor group did not reveal any association of Taq I polymorphism with the PCa risk in various genetic models. When 4 studies deviated from HWE in the controls were excluded, similar results were obtained (The results were not given).

Publication bias and sensitivity analysis
As shown in Figure 5, funnel plots did not reveal any obvious asymmetry. Moreover, the Egger's test also showed that there was no publication bias in the overall analysis (Table 2) and almost the subgroup analyses (Table  3-6). Sensitivity analyses suggested that the pooled results had not changed significantly by omitting each individual study from all the analyses ( Figure 6).

DISCUSSION
In recent years, polymorphism of VDR gene has drawn great attention, because more and more studies have shown that the mutations of VDR gene were related to the PCa risk [14][15][16][17][18][19]. However, these results have been disputable [20][21][22]. Previous meta-analyses were reported by Yin [10,23,24], in which the number of included studies was 23, 27 and 8, respectively. However, some new data was reported, which is not consistent with the results of the former three studies [17,[25][26][27]. 8,423 cases and 8,867 controls were included in our analysis from 36 independent studies. The cases included were much more than the previous meta-analyses. Therefore, our results might be more convincing and stringent.
Our meta-analysis showed that Taq I polymorphism might increase the PCa risk in overall population in recessive genetic model and allele-frequency genetic model. It is not consistent with the results of previous report by Liu et al. [10]. But for the stratified analysis of ethnicity, significant increased risk was found to be associated with Taq I polymorphism in Asians, which is consistent with the results of the report of Fei et al [24]. Ethnicity is an important biological factor for the decline of VDR function [50]. The difference in outcome among ethnicities might result from racial backgrounds and geographic discrepancies [51]. In addition, different dietary patterns could also contribute to the difference [52]. Our results suggested that the Taq I variation might be one of the valuable biomarkers for predicting the susceptibility of PCa. Further studies of Caucasian and African are required.
For the subgroup analysis by the source of controls, increased risk of PCa was found to be associated with Taq I polymorphism in population-based controls in all the comparisons. Possibly some sick population were included in the groups of HBP or hospital-based controls, these groups were special and could not represent all the population [53]. Therefore, the results of these groups might be lack of credibility. Our results revealed some discrepancies between the genotyping methods. It suggested that Taq I polymorphism in the subgroup of TaqMan, Taq I was associated with PCa risk, which may be the cause of heterogeneity. According to a report in 2004, clinical tumor stage of PCa would be accelerated by VDR gene polymorphism [54]. Hence, we performed a subgroup analysis by clinical stage. Our results indicated no association between Taq I polymorphism and susceptibility of PCa, which were different from the previous meta-analyses [24].
Although the between-study heterogeneity was detected, sensitivity analysis did not reveal any significant change in our results by omitting the studies contribute to the heterogeneity. It suggested that our results were credible and statistically robust.
Some limitations should be acknowledged. First, several studies with too little number of patients were included in our analysis, they may introduce potential bias. Second, our results were based on unadjusted parameters, Oncotarget 7139 www.impactjournals.com/oncotarget a more accurate analysis are needed, in which some related parameters should be included to adjust the outcome, including age, diet, and other important lifestyle factors.
In conclusion, our meta-analysis might be the largest meta-analysis to estimate the association of VDR gene Taq I polymorphism with the risk of PCa. Marginally increase of PCa risk was found to be related with Taq I polymorphism in overall population, especially in Asians and in population-based controls subgroup. In the future, large and well-designed researches are required to demonstrate the increased effect of Taq I polymorphism on PCa risk.

Literature and search strategy
The PubMed, Embase, and Chinese National Knowledge Infrastructure (CNKI) database searches were carried out for all the eligible papers. The following search  Oncotarget 7140 www.impactjournals.com/oncotarget terms were included: "VDR/vitamin D receptor", "prostate cancer/tumor/carcinoma" and "polymorphism/mutation/ variant". The literature search was updated to August, 2017. In addition, manually searching for the additional studies was conducted according to the references of the original and review reports.

Study selection
Retrieved studies were deemed eligible provided that they met all of the following criteria: (a) studies on human beings; (b) in a case-control or nested casecontrol design; (c) investigated the relationship of Taq I polymorphism with PCa risk; (d) distribution of genotype frequency for cases and controls could be obtained or calculated; (e) and received more than 4 points in the NOS, which was considered to be high quality; (f) the difference of baseline characters and clinical information was not significant between PCa patients and controls.

Data extraction
The studies meeting the inclusion criteria were read carefully by two investigators independently (Yansheng Zhao and Xiaofeng Liu). We collected the following information: author, year, country, ethnicity, genotyping methods, source of controls, sample size, and genotype and allele frequencies. The subjects were divided into different subgroups: Asians, Africans and Caucasians for ethnicity; hospital-based, population-based, and Benign Prostate Hyperplasia (BPH) for the source of controls; TaqMan, PCR-RFLP and SNPlex for genotyping method. The clinical stages were categorized as localized group (Gleason < 7) and advanced group (Gleason ≥ 7). In order to reach consensus on all of the items, any disagreement was resolved by a third reviewer (Lei Wang).

Statistical analysis
A χ 2 -test based on the Q statistic was conducted to evaluate the heterogeneity. The between-study heterogeneity was considered to be significant when I 2 > 50% and p < 0.05, and the random effects model was used to combine values from studies [55]. Otherwise, for homogeneous studies, the fixed effects model was chosen. The pooled odds ratios (ORs) together with its 95% confidence intervals (95% CIs) were calculated to evaluate the strength of the association. The statistical significance of ORs was determined with Z-test. To get a more reasonable result, five genetic models were adopted in our analysis: homozygote model (tt vs. TT), heterozygous model (Tt vs. TT), dominant model (tt vs. TT/Tt), recessive model (tt/Tt vs. TT) and allele genetic model (t vs. T).
To assess the potential publication bias, Begg's Funnel plot was generated based on the analysis result and database size. The more asymmetry the funnel plot looked, the more publication bias was introduced. Meanwhile, Egger's test was also performed for further investigation. For the Egger's test, the significance level was set as p value < 0.05. Moreover, HWE of controls was recalculated with the goodness-of-fit χ 2 -test, P values of > 0.05 was considered as significant equilibrium.
For each outcome, we also performed subgroup analyses according to ethnicity, source of controls, genotyping method and clinic stages. Sensitivity analysis was performed to assess the stability of pooled results.
All analyses were performed using STATA package version 12.0 (Stata Corp, College Station, TX, USA). Two-sided P values of < 0.05 was considered to be statistically significant.

CONFLICTS OF INTEREST
The authors report no conflicts of interest.

GRANT SUPPORT
None.