The Association between NQO1 Pro187Ser Polymorphism and Bladder Cancer Susceptibility: A Meta-Analysis of 15 Studies

Background NAD(P)H:quinone oxidoreductase 1 (NQO1), an obligate two-electron reductase, plays an important role in reducing reactive quinones to less reactive and less toxic hydroquinones. Genetic variations in NQO1 gene that impede its enzyme function may be considered as putative risk factor for cancer. Numerous studies have been performed to investigate the association between NQO1 Pro187Ser polymorphism and bladder cancer risk; nevertheless, the results remain controversial. Methods We indentified eligible publications from PubMed, Embase and CBM databases. Pooled odds ratios (ORs) and 95% confidence intervals (CIs) were used to access the strength of the associations. False-positive report probability (FPRP) analysis was also performed for all statistically significant findings. Results We collected a total of 15 studies including 4298 cases and 4275 controls in the final meta-analysis. Overall, the NQO1 187Ser carriers were associated with an increased bladder cancer risk (homozygous: OR = 1.43, 95% CI = 1.08-1.90; recessive: OR = 1.33, 95% CI = 1.03-1.72; dominant: OR = 1.19, 95% CI = 1.04-1.37, and allele comparing: OR = 1.18, 95% CI = 1.06-1.33). Stratification analyses showed a statistically significant association among Asians (homozygous: OR = 1.82, 95% CI = 1.39-2.38; recessive: OR = 1.52, 95% CI = 1.20-1.93, dominant: OR = 1.40, 95% CI = 1.05-1.88, and allele comparing: OR = 1.35, 95% CI = 1.15-1.58), never smokers (homozygous: OR = 2.30, 95% CI = 1.14-4.65; heterozygous: OR = 2.26, 95% CI = 1.43-3.56; dominant model: OR = 1.59, 95% CI = 1.14-2.21, and allele comparing: OR = 1.72, 95% CI = 1.27-2.33), hospital-based studies (homozygous: OR = 1.46, 95% CI = 1.09-1.94; recessive: OR = 1.32, 95% CI = 1.02-1.69; dominant: OR = 1.28, 95% CI = 1.05-1.56, and allele comparing: OR = 1.24, 95% CI = 1.07-1.43), studies with genotyping performed by PCR-RFLP under all genetic models, and studies with minor allele frequency >0.30 (homozygous: OR = 1.69, 95% CI = 1.25-2.27; recessive: OR = 1.46, 95% CI = 1.10-1.95, and allele comparing: OR = 1.25, 95% CI = 1.04-1.51), respectively. Conclusions Despite some limitations, our meta-analysis provides sufficient evidence that NQO1 Pro187Ser polymorphism may contribute to bladder cancer risk. These findings need further validation in well-designed prospective studies with larger sample size and different ethnicities, especially for Asians.


Introduction
Cancer is a major public health burden worldwide. Approximately 12.7 million new cancer cases and 7.6 million cancer deaths were reported based on GLOBOCAN 2008 [1]. Bladder cancer is the seventh most frequently diagnosed cancer among men, with about 386300 new cancer cases and 150200 cancer deaths in 2008 [1]. It is clear that cancer is a multistep process involving complex interactions between environmental and genetic factors [2,3]. Cigarette smoking, occupational exposure to chemical carcinogens, drugs consuming (e.g., phenacetin and cyclophosphamide), as well as Schistosoma hematobium infection are well-known etiologic factors for bladder cancer [4]. Despite the environmental factors play important roles in the development of cancer, host genetic factors are also closely related to the pathophysiology of many human cancers including bladder cancer [5]. NAD(P)H:quinone oxidoreductase 1 (NQO1), also known as diphtheria toxin diaphorase (DT-diaphorase), is a cytoplasmic 2-electron reductase, belonging to the NAD(P)H dehydrogenase (quinone) family. NQO1 plays an important role in the aromatic amine metabolism pathway [6]. It acts to reduce and detoxifies quinines and their derivative, thereby protecting cells against oxidative stress and carcinogenesis. Polymorphisms that lead to aberrant alteration of NQO1 activity may have profound impacts on cancer disposition.
The NQO1 gene is mapped to chromosome 16q22.1 and comprised of six exons and five introns [7,8]. Among all the identified single nucleotide polymorphisms (SNPs) in this gene, a non-synonymous NQO1 rs1800566 SNP is of particular interest and widely investigated in molecular epidemiology studies. This SNP is located at nucleotide position 609, causing a C-to-T transition and consequential proline to serine amino acid substitution at codon 187 in exon six [8]. Genotype-phenotype studies have demonstrated that the rs1800566 C>T (Pro187Ser) polymorphism is associated with a decreased activity of NQO1 enzymatic activity [9,10]. Compared with the wild-type CC genotype carriers, the homozygous TT genotype carriers have only 2-4% of the quinone reductase activity [9,10]. As expected, it has been reported that the polymorphism in the xenobiotic metabolizing enzyme gene, NQO1 Pro187Ser, is associated with bladder cancer risk [11]. Besides this study, numerous studies have investigated the association between NQO1 Pro187Ser polymorphism and bladder cancer risk , however, the conclusions remain controversial rather than conclusive. With this in mind, we carried out the current meta-analysis and aimed to investigate the association between this polymorphism and bladder cancer susceptibility, and to draw a more precise conclusion. The false-positive report probability (FPRP) analysis was also conducted to preclude false association resulting from multiple calculations.

Literature search strategy
Literatures examining the association between NQO1 Pro187Ser polymorphism and bladder cancer risk were searched from MEDLINE and EMBASE databases using the following key words: "NQO1 or NAD(P)H:quinone oxidoreductase 1 or DT-diaphorase or DTD or quinone reductase or rs1800566 C>T or Pro187Ser", "polymorphism or variant or variation" and "bladder" (prior to May 16, 2014). We also searched additional related publications on this topic from the literature cited by each article identified from the mentioned database. We only included the latest or the largest study in our final meta-analysis when the studies with overlapping data published by same authors or from the same institutions. As native-Chinese speakers, we also searched publications written in Chinese from Chinese Biomedical (CBM) database with the combinations of "NQO1" and "bladder cancer" to maximize the coverage of our investigation and minimize the selection bias.

Inclusion and exclusion criteria
Studies included should meet the following inclusion criteria: (1) evaluating the association between NQO1 Pro187Ser polymorphism and bladder cancer risk; (2) case-control design; (3) sufficient data including allele frequency and distribution of genotypes provided to estimate odds ratios (ORs) and their 95% confidence intervals (CIs); (4) genotype distribution in the controls was consistent with Hardy-Weinberg equilibrium (HWE); (5) independent from other studies.
Studies would be excluded if they were review articles, case only or non-cancer subjects only studies, conference abstracts, case reports, studies deviated from HWE in controls, and duplicate publications.

Data extraction
Two authors (Sen Yang and Tao Jin) extracted the following information independently from each publication with the inclusion criteria mentioned above: the first author's surname, publication year, country of origin, ethnicity, method of matching controls to cases, source of controls, smoking status, genotyping methods, total number of cases and controls, numbers of cases and controls with the CC (Pro/Pro), CT (Pro/Ser) and TT (Ser/Ser) genotypes, HWE and minor allele frequency (MAF) for controls. We conducted stratification analyses by ethnicity (categorized as Caucasians, Asians and Africans), source of control (hospital-based and population-based), genotyping methods [restriction fragment length polymorphism PCR (PCR-RFLP), mass spectrometry and TaqMan], MAF (<0.20, 0.20-0.30 and >0.30 according to the frequency and number of each investigation), and smoking status (never smokers and ever smokers). Any disagreements were resolved by discussions between the two authors until consensus reached.

Statistical methods
The association between NQO1 Pro187Ser polymorphism and bladder cancer risk was evaluated by homozygous model ( . Goodness-of-fit chi-square test was used to test deviation from HWE and P<0.05 was considered as departure from HWE. Chi squarebased Q-test was used to evaluate the between-study heterogeneity. Effect of heterogeneity was also quantified with the I 2 statistic which measures the degree of between-study inconsistency. I 2 lies between 0 to 100% with higher values indicating a greater degree of heterogeneity [33]. Fixed-effects model (Mantel-Haenszel method) were chosen when the P value of the heterogeneity test was 0.10 or I 2 < 50% [34]; otherwise, the random-effects model (DerSimonian and Laird method) were selected which tends to provide wider 95% CIs [35]. Potential publication bias was verified by standard error of log (OR) for each study plotted against their consequential log (OR). Egger's linear regression test was used to assess the Funnel plot asymmetry [36,37]. Sensitivity analyses were calculated by removing a single study individually and recalculating the corresponding ORs and 95% CIs.
It has been acknowledged that many reported associations between genetic variation and susceptibility to diseases might be false positive, given determination of statistical significance largely relies on P value alone. To rule out false associations resulted from multiple tests, falsepositive report probability (FPRP) analysis was also performed for all statistically significant findings as described previously [38][39][40]. Briefly, FPRP value for a given association between SNP of interest and bladder cancer risk was calculated with FPRP threshold of 0.2 and prior probability of 0.1. Statistical power was used to detect an OR of 1.50/0.67 for alleles with a risk/ protective effect. An association with FPRP value below 0.2 was declared noteworthy. Metaregression analysis was performed to investigate the major sources of heterogeneity across the studies in the current study.
STATA (version 11.0; Stata Corporation, College Station, TX) and SAS software (version 9.1; SAS Institute, Cary, NC) were used for statistical analyses. All P values were two-sided, and P<0.05 was considered as statistically significant finding.

Characteristics of study
A total of 39 publications examining the association between NQO1 Pro187Ser and bladder cancer risk were identified from MEDLINE and EMBASE, and two additional publications from CBM (Fig. 1). Initial screening led to the removal of 20 articles for failure to meet eligibility requirements. As a result, 21 studies that fulfilled the crude inclusion criteria were chosen for further analysis . Among them, two [28,29] were excluded, one was overlapped with a study published more recently [24], and the other duplicated the use of subjects that were reported in a study with much larger sample size [25]. Additional three studies were excluded from our final analysis due to reporting case only studies [30][31][32]. In the last, one study was also excluded for the deviation from HWE (P<0.001) in controls [23]. Ultimately, 15 publications with a total of 4298 bladder cancer cases and 4275 controls were included in the final meta-analysis (Table 1). In the included investigations, sample sizes ranged from 61 to 1128 for bladder cancer cases and from 100 to 1123 for controls. There were eight studies conducted on Caucasians, six studies on Asians, and only one study on Africans. In term of study design, five studies were population-based, and ten were hospital-based. Moreover, eight studies provided the genotype information for ever smokers and never smokers, respectively (S1 Table). Genotyping methods used in the studies included PCR-RFLP (11 studies), TaqMan (two studies) and mass spectrometry (two studies).

Meta-analysis results
As shown in Table 2 Table 3. For a prior probability of 0.1 and OR of 1.5, the FPRP analysis indicated that the significant association was noteworthy for all subjects (homozygous:
With the purpose to find the source of heterogeneity, we also performed meta-regression by ethnicity, source of controls, genotyping methods and MAF. As shown in Table 4, we found that source of controls might contributed to the heterogeneity in the current meta-analysis to some extent although P values were not statistical significant (heterozygous model: P = 0.067; recessive model: P = 0.096). In contrast, other covariates were not likely to cause heterogeneity across studies as indicated by

Publication bias
We performed both Begg's and Egger's tests to evaluate the publication bias for the studies included in the final meta-analysis (Fig. 3). The shapes of the funnel plots seemed symmetrical, indicating no significantly statistical evidence of publication bias for the association between NQO1 Pro187Ser polymorphism and bladder cancer risk (data not shown).

Discussion
In the current updated meta-analysis of 15 studies including a total of 4298 cases and 4275 controls, we found that the NQO1 Pro187Ser polymorphism was associated with increased bladder cancer susceptibility. Further stratified analyses revealed that this association was more predominant among Asians, hospital-based studies, studies adopting PCR-RFLP method for genotyping, studies with MAF >0.30, as well as never smokers. Noteworthily, these results were further confirmed by FPRP analysis. To our knowledge, several meta-analyses have been performed to investigate the role of NQO1 Pro187Ser polymorphism in the bladder cancer risk [41][42][43][44][45]. In a meta-analysis of eight studies including 1410 cases and 1485 controls [41], Chao et al. observed increased bladder cancer risk with homozygous rare Ser/Ser genotype for the white. After having pooled together 12 studies with 3041 cases and 3128 controls [42], Gao et al. found the significant association only under the allele comparing, but not other genetics models. The study carried out by Zhang et al. [43], including nine publications with 2661 cases and 2738 controls, observed no association under all genetics models. Similarly, Gong et al. also failed to detect any association, while analyzing 11 studies with 2937 cases and 3008 controls [44]. A meta-analysis by Lajin et al. [45], pertaining to all types of cancer, observed a significantly increased overall cancer risk with variant Ser/Ser genotype, as compared to the homozygous wild type (Pro/Pro). When stratified by cancer type, the same study revealed a significantly increased risk of bladder under all five genetics models, involving 10 studies included 1814 bladder cancer cases and 1905 controls. In a meta-analysis carried out by Mandal et al. [27], 12 studies with a total of 2286 cases and 2294 controls were included, they found the Pro187Ser polymorphism was associated with bladder cancer risk under all genetic models. The controversial results in single studies and even meta-analyses urged us to reevaluate the association with inclusion of the latest studies. Some of the previous significant findings were not replicated in our meta-analysis. For example, unlike Chao et al. [41], we did not observe a significant increased risk among Caucasians. We also did not replicate the association between Pro187Ser and bladder cancer risk in heterozygous model reported by Lajin et al. and Mandal et al. [27,45]. However, we did validate some findings from previous publication, i.e. the polymorphism was significantly associated with bladder cancer risk under allele comparison model [42,45], and the association was significant among the Asian group [42]. More importantly, some novel findings were discovered in the current meta-analysis besides. For instance, we observed significantly increased risk of developing bladder cancer among never smokers who carried variant alleles of NQO1 Pro187Ser polymorphism.
To preclude the possibility that the findings may be false positive due to the limited sample size, FPRP analysis was performed in the current meta-analysis. The associations among all subjects, Asians, studies with hospital-based design, PCR-RFLP genotyping method, MAF >0.30, and never smokers were validated. It is well known that association studies solely depending on a P value may sometime lead to false positive findings. To address this issue, FPRP analysis has been introduced, in which the observed P value coupled with both the prior probability and the statistical power of the test to validate the significance of a finding [38]. Thus, it is important to perform FPRP analysis to further validate the positive findings and calculated statistical power in a meta-analysis [46]. Theoretically, FPRP analysis with a smaller prior probability would be more likely to determine a real positive finding, Nevertheless, in that case, we might miss some real positive findings due to the limited sample size and weak effect of selected SNPs. Therefore, in the current study, we selected the prior probability of 0.1, which was proper to detect relatively weak effect of SNPs.
NQO1 is an obligate two-electron reductase which can reduce reactive quinones to hydroquinones. Additionally, NQO1 also shows superoxide scavenging activity as well as protective activity against procarcinogenic benzenes, thus it has been considered as an important defense of host against cancer [47,48]. NQO1 plays an important role in protecting cells against chemically induced oxidative stress, cytotoxicity, mutagenicity, and carcinogenicity [14]. It is also implicated in bioactivation of certain environmental procarcinogens (e.g., nitroaromatic compounds and heterocyclic amines) that are present in tobacco smoke and certain processed foods [49,50]. Many molecular epidemiology studies have investigated the association between NQO1 Pro187Ser polymorphism and bladder cancer risk, but the conclusions were controversial. Some studies reported that the Ser/Ser genotype of Pro187Ser polymorphism was significantly associated with increased bladder cancer risk, when Pro/Pro genotype was used as the reference group [12,21,24], whereas others failed to confirm this finding. Some studies also found the Pro187Ser polymorphism was associated with bladder cancer risk under the dominant model [21,22,26], or under the recessive model [12,18,24]. Overall, the Pro187Ser polymorphism has been shown to significantly associate with an increased bladder cancer risk. The association was also observed among Asians subjects, hospital-based studies, and never smokers. These significant findings may be ascribed to a much larger sample size in the current study than previous meta-analyses. In the current study, we found that the NQO1 Pro187Ser polymorphism was shown to associate with bladder cancer in hospital-based subgroups but not population-based subgroups. Since the hospital-based subjects are usually less representative and more prone to bias than population-based studies, these results should be interpreted cautiously. We also found the never smokers harboring the variant genotype were associated with increased bladder cancer risk. The main reason may be ascribed to genetic susceptibility. In other words, low penetrance polymorphism, like NQO1 Pro187Ser, commonly only has weak effect on bladder cancer risk. Therefore, the effect of this SNP among smokers might be superseded by the much stronger effects of smoking. Other reasons for this positive association among never smokers could be the reduced sample sizes in the subgroups, and the fact that the never smokers may also be exposed to second-hand smoke and other risk factors. Furthermore, sensitivity analyses demonstrated that successively excluding one study at a time did not lead to obvious change in the pooled ORs, which strengthens the results of this meta-analysis. Meta-regression suggested that source of controls but not other covariates might partially explain the heterogeneity in the current meta-analysis. In addition, the absence of publication bias further supports the validity of our finding.
Though we have included the latest investigations, some limitations should be acknowledged, while interpreting the results. First, due to lack of original data, our conclusions were based on unadjusted estimates of ORs without adjustment for age, gender, Schistosoma hematobium infection status and other risk factors (e.g., smoking and drinking status), which may lead some confounding bias. Second, the controls were not uniformly defined; some studies used healthy subjects as the control group, whereas others used cancer-free patients as the reference group. Third, sample size for each case-control study was relative small, with most studies including less than 500 cases, except for two [20,25]. Therefore, pooled data was not sufficient enough to perform a comprehensive analysis, particularly for subgroup analysis by cancer type and smoking status. Fourth, there was only one study conducted in African population [25], and five studies in Asian subjects. Further studies with large sample size are needed urgently verify our finding, especially those regarding Asians. Fourth, some degree of heterogeneity was found in overall comparison under all genetic models. Heterogeneity is a potential problem when interpreting the results of all meta-analyses [51]. Meta-analysis is a statistical technique to integrate the findings from independent comparable studies on the same topic. A meta-analysis is necessary, if each single study is too small or too restricted in scope to yield conclusive results on a given topic. However, since each study was carried out independently, inter-study variation might be unavoidable, including the source of control, ethnicity, genotyping method, etc. As a result, combining the findings across such studies may suffer from between-study heterogeneity. To overcome this obstacle in integrating various studies, DerSimonian et al. introduced random-effects model in 1986, which can tolerate variations across studies and use an objective method for weighting that can be generalized by including study characteristics into the analysis [35]. In this study, random-effects model was chosen due to heterogeneity. In the subgroup analysis, we found that the heterogeneity might not result from each of factors individually, but from combined resources. For example, as to the ethnicity, Caucasians exhibited heterogeneity under homozygous and recessive models but not other models, whereas Asians showed heterogeneity only under heterozygous and dominant models. Finally, common genetic analysis challenges were only addressed to some extent, including direction of effect of alleles, MAF cutoffs for common and rare variants, technical quality control challenges, and related measures when taking data from lab-generated output to genetic statistical analysis.
In conclusion, the current meta-analysis indicates that the NQO1 Pro187Ser polymorphism may confer host increased genetic susceptibility to the risk of bladder cancer. However, large and well-designed studies are warranted to validate our findings.