Introduction

Cancer is projected to become the leading cause of death worldwide according to a new edition of the World Cancer Report from the International Agency for Research on Cancer, which has become a major public health challenge (Siegel et al. 2013). Although a recently published paper indicated that cancer prevention and management is moving in the right direction (Karim-Kos et al. 2008). Due to the combination of earlier detection, better access to care and improved treatment, survival is increased and mortality is decreased. Still, cancer prevention efforts have much to attain (Lutz et al. 2003; Soerjomataram et al. 2007). New markers for identifying high-risk populations as well as novel strategies for early detection and preventive care are urgently needed. The mechanism of carcinogenesis is still not fully understood.

CYP1B1 gene is located on chr2p22-p21, which is involved in the metabolic activation of polycyclic aromatic hydrocarbons (PAHs) including benzo(a)pyrene and dimethylbenz(a)anthracene (DMBA), but with a product distribution that is distinct from CYP1A1 (Buters et al. 1999; Shimada et al. 1996). CYP1B1 is commonly overexpressed in human malignancies (Murray et al. 1997) and activates a variety of carcinogens. For example, CYP1B1 catalyzes both the formation of dihydrodiols of specific PAHs and their subsequent oxidation to carcinogenic dihydrodiol epoxides (Kim et al. 1998). CYP1B1 is transcriptionally induced by compounds such as 2,3,7,8-tetrachlorodibenzo-p-dioxin or dioxin, and regulated by several key transcriptional factors including oestrogen receptor and aryl hydrocarbon receptor (Sutter et al. 1994). Apart from its role in xenobiotic metabolism, CYP1B1 is implicated in the bioactivation of pro-carcinogens (Hayes et al. 1996; Martin et al. 2010; Spink et al. 1997). The enzyme also appears to play a role in the metabolism of certain anticancer agents used in the treatment of hormone-induced cancers (Sissung et al. 2008). In humans, CYP1B1 is genetically polymorphic and more than 50 single nucleotide polymorphisms (SNPs) have been reported so far, of which certain deleterious mutations are associated with primary congenital glaucoma (Stoilov et al. 1998). Of the most common SNPs of CYP1B1 gene, four have been reported to result in amino acid substitutions including Arg by Gly at codon 48(rs10012), Ala by Ser at codon 119 (rs1056827), Leu by Val at codon 432 (rs1056836) and Asn by Ser at codon 453 (rs1800440).

In the past decade, a number of molecular epidemiological studies have been done to evaluate the association between CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms and different types of cancer risk in diverse populations (Catsburg et al. 2012; Reding et al. 2012; Jang et al. 2012; Salinas-Sánchez et al. 2012a, b; Cerne et al. 2011a,b ; dos Santos et al. 2011; Wang et al. 2011a, 2012; Zhu et al. 2011; Rudolph et al. 2011; Timofeeva et al. 2009; Sainz et al. 2011; Cleary et al. 2010; Hlavata et al. 2010; Trubicka et al. 2010; Ozbek et al. 2010; Soucek et al. 2010; Sachse et al. 2002; Sliwinski et al. 2010; Ashton et al. 2010; Delort et al. 2010; Church et al. 2010; Yadav et al. 2009; Tai et al. 2010; Vrana et al. 2010; Li et al. 2002, 2005, 2013; Kato et al. 2009; Beuten et al. 2009; Rotunno et al. 2009; Fontana et al. 2009; MARIE-GENICA et al. 2010; Sangrajrang et al. 2009; Reding et al. 2009; Kilfoy et al. 2009; Huang et al. 2009; Shimada et al. 2009; Listgarten et al. 2004; Sigurdson et al. 2009; Okobia et al. 2009; Cote et al. 2009; Yuan et al. 2008; Delort et al. 2008; Freedman et al. 2009; Cotterchio et al. 2008; Figueroa et al. 2008; Diergaarde et al. 2008; Shah et al. 2008; Harth et al. 2008; Beuten et al. 2008; Gulyaeva et al. 2008; Van Emburgh et al. 2008; Hirata et al. 2008; Zienolddiny et al. 2008; Singh et al. 2008; Yoon et al. 2008; Justenhoven et al. 2008; Cussenot et al. 2007; Küry et al. 2007; Bethke et al. 2007; Matyjasik et al. 2007; Holt et al. 2007; Berndt et al. 2007; Cote et al. 2007; Tao et al. 2006; Sillanpää et al. 2007; De Roos et al. 2006a, b; Gaudet et al. 2006; Gallicchio et al. 2006; Sobti et al. 2006; Sellers et al. 2005; Huber et al. 2005; Cicek et al. 2005; Le Marchand et al. 2005; Wenzlaff et al. 2005; Landi et al. 2005; Liang et al. 2005; Sørensen et al. 2005; Doherty et al. 2005; Wen et al. 2005; Rylander-Rudqvist et al. 2004; Fukatsu et al. 2004; Zimarina et al. 2004; Cecchin et al. 2004; Dunning et al. 2004; Hung et al. 2004; Ahsan et al. 2004; Wu et al. 2004; Thyagarajan et al. 2004; Sasaki et al. 2003, 2004; McGrath et al. 2004; Chang et al. 2003; Rylander-Rudqvist et al. 2003; Lee et al. 2003; Kocabaş et al. 2002; Tanaka et al. 2002; De Vivo et al. 2002; Ko et al. 2001; Goodman et al. 2001; Watanabe et al. 2000; Zheng et al. 2000; Bailey et al. 1998; Berber et al. 2013; Maurya et al. 2014; Martínez-Ramírez et al. 2013; Rebbeck et al. 2006; Lundin et al. 2012; Tang et al. 2000; Rodrigues et al. 2011; Holt et al. 2013). However, the results were inconsistent or even contradictory. Partially because of the possible small effect of the polymorphism on cancer risk and the relatively small sample size in each of published studies. In addition, some recent meta-analyses analyzed such an association only for single cancer such as breast cancer, lung cancer, prostate cancer, endometrial cancer, and so on (Economopoulos and Sergentanis 2010; Chen et al. 2010; Wang et al. 2011b; Xu et al. 2012; Cui et al. 2012; Yang et al. 2012). Hence, the correlation of these polymorphic genes remains unknown. Every single study may be underpowered to achieve a comprehensive and reliable conclusion. Hence, in order to explore the association we performed a comprehensive meta-analysis by including the most recent and relevant articles to identify statistical evidence of the association between CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms and risk of all cancers that have been investigated.

Materials and methods

Identification and eligibility of relevant studies

A comprehensive literature search was performed using the PubMed, Cochrane Library, and EMBASE database for relevant articles published (the last search update was Feb 18, 2014) with the following key words “CYP1B1’’, ‘‘cytochrome P-450 1B1’’, ‘‘cytochrome P450 1B1’’, ‘‘polymorphism’’, “Variant”, or “Mutation”, and “Cancer” or Carcinoma”. In addition, studies were identified by a manual search of the reference lists of reviews and retrieved studies. We included all the case–control studies and cohort studies that investigated the association between CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms and cancer risk with genotyping data. All eligible studies were retrieved, and their bibliographies were checked for other relevant publications.

Inclusion and exclusion criteria

The included studies have to meet the following criteria: For inclusion, the study outcome had to be cancer, there had to be at least two comparison groups (case group vs. control group), and the investigation had to provide data on the CYP1B1 genotype distribution. We excluded review articles, editorials, case reports, studies with preliminary results not on CYP1B1 polymorphism or outcome, and investigations of the role of CYP1B1 expression related to disease. When the same sample was used in several publications, only the most complete study was included following careful examination.

Data extraction

Information was carefully extracted from all eligible studies independently by two investigators according to the inclusion criteria listed above. The following data were collected from each study: first author’s name, year of publication, country of origin, ethnicity, source of controls (population-based controls, hospital-based controls, and family-based controls), sample size, and numbers of cases and controls in the CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser genotypes whenever possible. Ethnicity was categorized as ‘‘Caucasian’’, ‘‘Asian’’, and “African”. When one study did not state which ethnic group was included or if it was impossible to separate participants according to phenotype, the sample was termed as ‘‘mixed population’’. Meanwhile, studies investigating more than one kind of cancer were counted as individual data set only in subgroup analyses by cancer type. We did not define any minimum number of patients to include in this meta-analysis. Articles that reported different ethnic groups and different countries or locations, we considered them different study samples for each category cited above.

Statistical analysis

Crude odds ratios (ORs) together with their corresponding 95 % CIs were used to assess the strength of association between the CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms and the risk of cancer. The pooled ORs were performed for co-dominant model (Leu432Val: Val/Val versus Leu/Leu and Leu/Val versus Leu/Leu, Asn453Ser: Ser/Ser versus Asn/Asn and Asn/Ser versus Asn/Asn, Arg48Gly: Gly/Gly versus Arg/Arg and Arg/Gly versus Arg/Arg, and Ala119Ser: Ser/Ser versus Ala/Ala and Ala/Ser versus Ala/Ala); dominant model (Leu432Val: Leu/Val + Val/Val versus Leu/Leu, Asn453Ser: Asn/Ser + Ser/Ser versus Asn/Asn, Arg48Gly: Arg/Gly + Gly/Gly versus Arg/Arg, and Ala119Ser: Ala/Ser + Ser/Ser versus Ala/Ala); recessive model (Leu432Val: Val/Val versus Leu/Val + Leu/Leu, Asn453Ser: Ser/Ser versus Asn/Ser + Asn/Asn, Arg48Gly: Gly/Gly versus Arg/Gly + Arg/Arg, and Ala119Ser: Ser/Ser versus Ala/Ser + Ala/Ala); and additive model (Leu432Val: Val versus Leu, Asn453Ser: Ser versus Asn, Arg48Gly: Gly versus Arg, and Ala119Ser: Ser versus Ala), respectively. Between-study heterogeneity was assessed by calculating Q-statistic (heterogeneity was considered statistically significant if P < 0.10) (Davey and Egger 1997) and quantified using the I 2 value, a value that describes the percentage of variation across studies that are due to heterogeneity rather than chance, where I 2 = 0 % indicates no observed heterogeneity, with 25 % regarded as low, 50 % as moderate, and 75 % as high (Higgins et al. 2003). If results were not heterogeneous, the pooled ORs were calculated by the fixed-effect model (we used the Q-statistic, which represents the magnitude of heterogeneity between-studies) (Mantel and Haenszel 1959). Otherwise, a random-effect model was used (when the heterogeneity between-studies was significant) (DerSimonian and Laird 1986). In addition to the comparison among all subjects, we also performed stratification analyses by cancer type (if one cancer type contained less than three individual studies, it was combined into the ‘‘other cancers’’ group), Moreover, the extent to which the combined risk estimate might be affected by individual studies was assessed by consecutively omitting every study from the meta-analysis (leave-one-out sensitivity analysis). This approach would also capture the effect of the oldest or first positive study (first study effect). In addition, we also ranked studies according to sample size, and then repeated this meta-analysis. Sample size was classified according to a minimum of 200 participants and those with fewer than 200 participants. The cite criteria were previously described (Klug et al. 2009). Last, sensitivity analysis was also performed, excluding studies whose allele frequencies in controls exhibited significant deviation from the Hardy–Weinberg equilibrium (HWE), given that the deviation may denote bias. HWE was calculated by using the goodness-of-fit test, and deviation was considered when P < 0.05. Begg’s funnel plots (Begg and Mazumdar 1994) and Egger’s linear regression test (Egger et al. 1997) were used to assess publication bias. If publication bias existed, the Duval and Tweedie nonparametric “trim and fill” method was used to adjust for it (Dual and Tweedie 2000). A meta-regression analysis was carried out to identify the major sources of between-studies variation in the results, using the log of the ORs from each study as dependent variables, and cancer type, ethnicity, sample size, HWE, and source of controls as the possible sources of heterogeneity. All of the calculations were performed using STATA version 10.0 (STATA Corporation, College Station, TX).

Results

Literature search and meta-analysis databases

Figure 1 graphically illustrates the trial flow chart. A total of 548 articles regarding CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms with respect to cancer were identified. After screening the titles and abstracts, 198 articles were excluded because they were duplicated. In addition, 235 articles were excluded because they were review articles, case reports, or other polymorphisms of CYP1B1. Last, of these published articles, four articles (24, 50, 73, 120) were excluded because their populations overlapped with other four included studies (17, 20, 48, 98). As summarized in Supplemental Table 1, 111 publications with 235 case–control studies were selected among the meta-analyses, including 54,124 cases and 62,932 controls for Leu432Val (115 studies from 107 publications), 30,532 cases and 39,193 controls for Asn453Ser (48 studies from 46 publications), 23,494 cases and 27,083 controls for Arg48Gly (38 studies from 34 publications), and 17,796 cases and 19,891 controls for Ala119Ser (34 studies from 30 publications). Among these studies, for Leu432Val polymorphism, there were 5 bladder cancer studies, 38 breast cancer studies, 12 colorectal cancer studies, 12 endometrial cancer studies, 7 head and neck cancer studies, 13 lung cancer studies, 7 ovarian cancer studies, 11 prostate cancer studies, and 10 studies with the “other cancer”. For Asn453Ser polymorphism, there were 15 breast cancer studies, 8 colorectal cancer studies, 5 endometrial cancer studies, 5 lung cancer studies, 3 ovarian cancer studies, 4 prostate cancer studies, and 8 studies with the “other cancer”. For Arg48Gly polymorphism, there were 12 breast cancer studies, 6 colorectal cancer studies, 5 endometrial cancer studies, 4 ovarian cancer studies, 5 prostate cancer studies, and 6 studies with the “other cancer”. For Ala119Ser polymorphism, there were 11 breast cancer studies, 3 colorectal cancer studies, 3 endometrial cancer studies, 4 lung cancer studies, 4 ovarian cancer studies, 4 prostate cancer studies, and 5 studies with the “other cancer”. All of the cases were pathologically confirmed.

Fig. 1
figure 1

Study flow chart explaining the selection of the 111 eligible articles included in the meta-analysis

Full size image

Meta-analysis results

CYP1B1 Leu432Val

The evaluations of the association of CYP1B1 Leu432Val polymorphism with cancer risk are shown in Table 1. Overall, significantly increased cancer risk was found (dominant model: OR = 1.07, 95 % CI = 1.03–1.12, P h < 0.001, I 2 = 54.2 %; homozygous model: OR = 1.09, 95 % CI = 1.02–1.17, P h < 0.001, I 2 = 56.7 %; heterozygous model: OR = 1.06, 95 % CI = 1.01–1.10, P h < 0.001, I 2 = 41.5 %; additive model: OR = 1.05, 95 % CI = 1.02–1.09, P h < 0.001, I 2 = 65.0 %) when all the eligible studies were pooled into the meta-analysis. Then we performed subgroup analysis by cancer type. We found that individuals with the minor variant genotypes had a higher risk of endometrial cancer (dominant model: OR = 1.16, 95 % CI = 1.03–1.31, P h = 0.096, I 2 = 36.9 %), lung cancer (dominant model: OR = 1.25, 95 % CI = 1.07–1.48, P h = 0.061, I 2 = 40.9 %; recessive model: OR = 1.19, 95 % CI = 1.03–1.37, P h = 0.359, I 2 = 8.8 %; homozygous model: OR = 1.36, 95 % CI = 1.15–1.61, P h = 0.124, I 2 = 33.2 %; heterozygous model: OR = 1.19, 95 % CI = 1.06–1.34, P h = 0.153, I 2 = 29.9 %; additive model: OR = 1.19, 95 % CI = 1.05–1.35, P h = 0.015, I 2 = 53.0 %), and prostate cancer (heterozygous model: OR = 1.09, 95 % CI = 1.00–1.19, P h = 0.147, I 2 = 33.9 %) and had a lower risk of ovarian cancer (recessive model: OR = 0.82, 95 % CI = 0.68–1.00, P h = 0.278, I 2 = 19.8 %). We further examined the association of the CYP1B1 Leu432Val polymorphism and cancer risk according to cancer type and ethnicity (Table 1). For samples of Asians, significant association was found among head and neck cancer (recessive model: OR = 1.94, 95 % CI = 1.44–2.61, P h = 0.110, I 2 = 54.7 %; heterozygous model: OR = 1.31, 95 % CI = 1.10–1.57, P h = 0.198, I 2 = 38.2 %), prostate cancer (dominant model: OR = 1.35, 95 % CI = 1.03–1.78, P h = 0.218, I 2 = 34.4 %; homozygous model: OR = 1.85, 95 % CI = 1.03–3.32, P h = 0.401, I 2 = 0.0 %; additive model: OR = 1.36, 95 % CI = 1.07–1.71, P h = 0.160, I 2 = 45.4 %), and other cancer (dominant model: OR = 1.58, 95 % CI = 1.24–2.01, P h = 0.326, I 2 = 10.7 %; recessive model: OR = 2.23, 95 % CI = 1.31–3.79, P h = 0.800, I 2 = 0.0 %; homozygous model: OR = 2.57, 95 % CI = 1.50–4.40, P h = 0.866, I 2 = 0.0 %; heterozygous model: OR = 1.45, 95 % CI = 1.12–1.87, P h = 0.394, I 2 = 0.0 %; additive model: OR = 1.60, 95 % CI = 1.30–1.97, P h = 0.439, I 2 = 0.0 %). For samples of Caucasians, significant association was found among lung cancer (dominant model: OR = 1.37, 95 % CI = 1.04–1.81, P h = 0.018, I 2 = 66.6 %; heterozygous model: OR = 1.26, 95 % CI = 1.08–1.47, P h = 0.106, I 2 = 47.6 %), ovarian cancer (recessive model: OR = 0.77, 95 % CI = 0.62–0.95, P h = 0.575, I 2 = 0.0 %; homozygous model: OR = 0.77, 95 % CI = 0.60–0.97, P h = 0.425, I 2 = 0.0 %), and prostate cancer (dominant model: OR = 1.17, 95 % CI = 1.05–1.30, P h = 0.313, I 2 = 15.8 %; homozygous model: OR = 1.26, 95 % CI = 1.08–1.46, P h = 0.149, I 2 = 47.4 %; heterozygous model: OR = 1.16, 95 % CI = 1.04–1.31, P h = 0.571, I 2 = 0.0 %; additive model: OR = 1.13, 95 % CI = 1.05–1.22, P h = 0.170, I 2 = 43.5 %). We also examined the association of the CYP1B1 Leu432Val polymorphism and cancer risk according to cancer type and source of controls (Table 1). For the population-based studies, significant association was only found among lung cancer (dominant model: OR = 1.26, 95 % CI = 1.00–1.58, P h = 0.017, I 2 = 57.2 %; recessive model: OR = 1.17, 95 % CI = 1.01–1.36, P h = 0.207, I 2 = 26.7 %; homozygous model: OR = 1.40, 95 % CI = 1.05–1.86, P h = 0.050, I 2 = 48.4 %; additive model: OR = 1.17, 95 % CI = 1.00–1.37, P h = 0.006, I 2 = 62.6 %). For the hospital-based studies, significant association was found among breast cancer (homozygous model: OR = 1.32, 95 % CI = 1.13–1.54, P h = 0.182, I 2 = 23.0 %), endometrial cancer (dominant model: OR = 1.33, 95 % CI = 1.03–1.70, P h = 0.030, I 2 = 57.0 %), and prostate cancer (dominant model: OR = 1.14, 95 % CI = 1.00–1.29, P h = 0.177, I 2 = 34.6 %; homozygous model: OR = 1.22, 95 % CI = 1.00–1.50, P h = 0.167, I 2 = 38.1 %).

Table 1 Stratified analysis of CYP1B1 Leu432Val polymorphism on cancer risk
Full size table
Table 2 Stratified analysis of CYP1B1 Asn453Ser polymorphism on cancer risk
Full size table
Table 3 Stratified analysis of CYP1B1 Arg48Gly polymorphism on cancer risk
Full size table
Table 4 Stratified analysis of CYP1B1 Ala119Ser polymorphism on cancer risk
Full size table
Table 5 Meta-regression analysis of the main characteristics
Full size table
Table 6 Summary ORs (95 % CI) and value of the heterogeneity of CYP1B1 polymorphisms under different genetic models according to studies with HWE on cancer risk
Full size table
Table 7 Summary ORs (95 % CI) and value of the heterogeneity of CYP1B1 polymorphisms under different genetic models according to studies with a minimum of 200 participants on cancer risk
Full size table

Significant heterogeneity was observed among these studies for dominant model comparison (P h < 0.001), recessive model comparison (P h < 0.0001), additive model comparison (P h < 0.001), homozygous model comparison (P h < 0.001), and heterozygous model comparison (P h < 0.001). Then, we assessed the source of heterogeneity by ethnicity, cancer type, source of controls, HWE, and sample size. Table 5 lists the results of meta-regression analysis. The results indicated that cancer type (heterozygous model: P = 0.021), ethnicity (heterozygous model: P = 0.044), source of controls (homozygous model: P = 0.013; additive model: P = 0.014; recessive model: P = 0.021; dominant model: P = 0.034), and sample size (heterozygous model: P = 0.004; dominant model: P = 0.010) but not HWE (dominant model: P = 0.262; recessive model: P = 0.607; heterozygous model: P = 0.345; homozygous model: P = 0.273; additive model: P = 0.344) contributed to substantial heterogeneity among the meta-analysis. Examining genotype frequencies in the controls, significant deviation from HWE was detected in the thirteen studies (Catsburg et al. 2012; Wang et al. 2011a,b; Sliwinski et al. 2010; Yadav et al. 2009; Vrana et al. 2010; Shimada et al. 2009; Sellers et al. 2005; Cicek et al. 2005; Le Marchand et al. 2005; Sasaki et al. 2004; Tanaka et al. 2002; Bailey et al. 1998; Maurya et al. 2014). When these studies were excluded, the results were changed among prostate cancer (heterozygous model: OR = 1.09, 95 % CI = 0.93–1.27), Asians of head and neck cancer (recessive model: OR = 1.17, 95 % CI = 0.62–2.24; heterozygous model: OR = 1.09, 95 % CI = 0.81–1.47), and hospital-based studies of other cancer (dominant model: OR = 1.30, 95 % CI = 1.08–1.57; homozygous model: OR = 1.36, 95 % CI = 1.00–1.83; heterozygous model: OR = 1.26, 95 % CI = 1.04–1.54; additive model: OR = 1.20, 95 % CI = 1.05–1.38), as shown in Table 6. In addition, when our meta-analysis was performed excluding studies with small sample sizes, the results were changed among prostate cancer (heterozygous model: OR = 1.08, 95 % CI = 0.99–1.18), Asians of prostate cancer (dominant model: OR = 1.21, 95 % CI = 0.88–1.65; homozygous model: OR = 1.38, 95 % CI = 0.67–2.83; additive model: OR = 1.19, 95 % CI = 0.91–1.57), hospital-based studies of endometrial cancer (dominant model: OR = 1.25, 95 % CI = 0.95–1.63), and hospital-based studies of prostate cancer (dominant model: OR = 1.10, 95 % CI = 0.97–1.26; homozygous model: OR = 1.18, 95 % CI = 0.96–1.44), as shown in Table 7. Last, a single study involved in the meta–analysis was deleted each time to reflect the influence of individual data set to the pooled ORs, the results did not be changed among overall analysis and any subgroup analysis.

We performed Begg’s funnel plot and Egger’s test to assess the publication bias of literatures. Begg’s funnel plots and Egger’s test suggested that there might be publication bias in any genetic model (dominant model: P = 0.004; heterozygous model: P = 0.006; recessive model: P = 0.009; additive model: P = 0.003; homozygous model: P = 0.001). This might be a limitation for the meta-analysis because studies with null findings, especially those with small sample size, are less likely to be published. Adjusting for possible publication bias using the Duval and Tweedie nonparametric ‘‘trim and fill’’ method for overall studies, the results did not change between CYP1B1 Leu432Val polymorphism with the risk of cancer. Figure 2 lists the Duval and Tweedie nonparametric “trim and fill’’ methods funnel plot in any genetic model.

Fig. 2
figure 2

The Duval and Tweedie nonparametric “trim and fill” method’s funnel plot of the meta-analysis of cancer risk and CYP1B1 Leu432Val polymorphism (homozygous model: A, heterozygous model: B, additive model: C, recessive model: D, and dominant model: E)

Full size image

CYP1B1 Asn453Ser

The evaluations of the association of CYP1B1 Asn453Ser polymorphism with cancer risk are shown in Table 2. Overall, no significant association was found among any genetic model (dominant model: OR = 1.01, 95 % CI = 0.96–1.05, P h = 0.012, I 2 = 34.6 %; recessive model: OR = 0.98, 95 % CI = 0.90–1.07, P h = 0.818, I 2 = 0.0 %; homozygous model: OR = 0.98, 95 % CI = 0.90–1.08, P h = 0.787, I 2 = 0.0 %; heterozygous model: OR = 1.00, 95 % CI = 0.95–1.05, P h = 0.016, I 2 = 34.5 %; additive model: OR = 0.99, 95 % CI = 0.95–1.03, P h = 0.050, I 2 = 28.0 %) when all the eligible studies were pooled into the meta-analysis. Then we performed subgroup analysis by cancer type. We found that individuals with the minor variant genotypes had a higher risk of prostate cancer (dominant model: OR = 1.19, 95 % CI = 1.02–1.39, P h = 0.368, I 2 = 0.0 %) and had a lower risk of endometrial cancer (heterozygous model: OR = 0.81, 95 % CI = 0.69–0.95, P h = 0.668, I 2 = 0.0 %; additive model: OR = 0.82, 95 % CI = 0.71–0.94, P h = 0.379, I 2 = 0.0 %). We further examined the association of the CYP1B1 Asn453Ser polymorphism and cancer risk according to cancer type and ethnicity (Table 2). For samples of Caucasians, significant decreased cancer risk was found among endometrial cancer (dominant model: OR = 0.82, 95 % CI = 0.71–0.94, P h = 0.700, I 2 = 0.0 %; heterozygous model: OR = 0.81, 95 % CI = 0.69–0.95, P h = 0.668, I 2 = 0.0 %; additive model: OR = 0.82, 95 % CI = 0.71–0.94, P h = 0.379, I 2 = 0.0 %). We also examined the association of the CYP1B1 Asn453Ser polymorphism and cancer risk according to cancer type and source of controls (Table 2). For the population-based studies, significant association was only found among endometrial cancer (dominant model: OR = 0.80, 95 % CI = 0.68–0.94, P h = 0.564, I 2 = 0.0 %; heterozygous model: OR = 0.81, 95 % CI = 0.69–0.95, P h = 0.668, I 2 = 0.0 %; additive model: OR = 0.82, 95 % CI = 0.71–0.94, P h = 0.379, I 2 = 0.0 %).

Significant heterogeneity was observed among these studies for dominant model comparison (P h = 0.012), additive model comparison (P h = 0.050), and heterozygous model comparison (P h = 0.016). Then, we assessed the source of heterogeneity by ethnicity, cancer type, source of controls, HWE, and sample size. Table 5 lists the results of meta-regression analysis. The results indicated that cancer type (dominant model: P = 0.766; heterozygous model: P = 0.880; additive model: P = 0.717), ethnicity (dominant model: P = 0.847; heterozygous model: P = 0.897; additive model: P = 0.555), source of controls (dominant model: P = 0.635; heterozygous model: P = 0.743; additive model: P = 0.845), sample size (dominant model: P = 0.576; heterozygous model: P = 0.988; additive model: P = 0.291), and HWE (dominant model: P = 0.399; heterozygous model: P = 0.118; additive model: P = 0.618) did not contribute to substantial heterogeneity among the meta-analysis. Examining genotype frequencies in the controls, significant deviation from HWE was detected in the one study (Figueroa et al. 2008). When this study was excluded, the results did not be changed among overall analysis and other subgroup analysis, as shown in Table 6. In addition, when our meta-analysis was performed excluding studies with small sample sizes, the results did not also change in the meta-analysis, as shown in Table 7. Last, a single study involved in the meta-analysis was deleted each time to reflect the influence of individual data set to the pooled ORs, the results were changed among prostate cancer (OR = 1.15, 95 % CI = 0.94–1.40).

Both Begg’s funnel plot and Egger’s test were performed to assess the publication bias of literatures. The Egger’s test results (dominant model: P = 0.497; recessive model: P = 0.078; additive model: P = 0.917; Homozygous model: P = 0.098; Heterozygous model: P = 0.539) and Begg’s funnel plot (Fig. 3) suggested no evidence of publication bias in the meta-analysis.

Fig. 3
figure 3

Begg’s funnel plot of the meta-analysis of thyroid cancer risk and CYP1B1 Asn453Ser polymorphism (additive model and dominant model)

Full size image

CYP1B1 Arg48Gly

The evaluations of the association of CYP1B1 Arg48Gly polymorphism with cancer risk are shown in Table 3. Overall, no significant association was found (dominant model: OR = 0.99, 95 % CI = 0.91–1.07, P h < 0.001, I 2 = 71.9 %; recessive model: OR = 0.93, 95 % CI = 0.83–1.04, P h < 0.001, I 2 = 61.7 %; homozygous model: OR = 0.91, 95 % CI = 0.79–1.06, P h < 0.001, I 2 = 74.3 %; heterozygous model: OR = 1.01, 95 % CI = 0.94–1.09, P h < 0.001, I 2 = 65.0 %) when all the eligible studies were pooled into the meta-analysis. In addition, high between-studies heterogeneity was observed among overall analysis (additive model: I 2 = 77.2). Then we performed subgroup analysis by cancer type. We found that individuals with the minor variant genotypes had a lower risk of endometrial cancer (recessive model: OR = 0.55, 95 % CI = 0.42–0.73, P h = 0.176, I 2 = 36.8 %). We further examined the association of the CYP1B1 Arg48Gly polymorphism and cancer risk according to cancer type and ethnicity (Table 3). For samples of Caucasians, significant decreased cancer risk was found among endometrial cancer (recessive model: OR = 0.42, 95 % CI = 0.28–0.61, P h = 0.419, I 2 = 0.0 %; homozygous model: OR = 0.29, 95 % CI = 0.12–0.71, P h = 0.020, I 2 = 74.3 %). We also examined the association of the CYP1B1 Arg48Gly polymorphism and cancer risk according to cancer type and source of controls (Table 3). For the population-based studies, significant association was found among other cancer (heterozygous model: OR = 1.25, 95 % CI = 1.04–1.50, P h = 0.152, I 2 = 46.9 %). For the hospital-based studies, significant decreased cancer risk was found among endometrial cancer (recessive model: OR = 0.46, 95 % CI = 0.31–0.67, P h = 0.114, I 2 = 54.0 %).

Significant heterogeneity was observed among these studies for dominant model comparison (P h < 0.001), recessive model comparison (P h < 0.0001), additive model comparison (P h < 0.001), homozygous model comparison (P h < 0.001), and heterozygous model comparison (P h < 0.001). Then, we assessed the source of heterogeneity by ethnicity, cancer type, source of controls, HWE, and sample size. Table 5 lists the results of meta-regression analysis. The results indicated that source of controls (dominant model: P = 0.005; homozygous model: P = 0.022; heterozygous model: P = 0.009; additive model: P = 0.036) and sample size (recessive model: P = 0.008; homozygous model: P = 0.044) but not cancer type (dominant model: P = 0.137; recessive model: P = 0.102; homozygous model: P = 0.092; heterozygous model: P = 0.164; additive model: P = 0.063), ethnicity (dominant model: P = 0.166; recessive model: P = 0.454; homozygous model: P = 0.303; heterozygous model: P = 0.168; additive model: P = 0.135), and HWE (dominant model: P = 0.703; recessive model: P = 0.759; heterozygous model: P = 0.903; homozygous model: P = 0.505; additive model: P = 0.672) contributed to substantial heterogeneity among the meta-analysis. Examining genotype frequencies in the controls, significant deviation from HWE was detected in the eight studies (Reding et al. 2012; Salinas-Sánchez et al. 2012a; Beuten et al. 2009; Figueroa et al. 2008; Holt et al. 2007; Sasaki et al. 2003, 2004; Tanaka et al. 2002). When these studies were excluded, the results were changed among other cancer (heterozygous model: OR = 1.25, 95 % CI = 1.04–1.50) and hospital-based studies of endometrial cancer (dominant model: OR = 0.33, 95 % CI = 0.16–0.69; homozygous model: OR = 0.19, 95 % CI = 0.10–0.33; heterozygous model: OR = 0.42, 95 % CI = 0.27–0.66; additive model: OR = 0.45, 95 % CI = 0.35–0.59), as shown in Table 6. In addition, when our meta-analysis was performed excluding studies with small sample sizes, the results were changed among hospital-based studies of endometrial cancer (recessive model: OR = 0.53, 95 % CI = 0.22–1.26), as shown in Table 7. Last, a single study involved in the meta–analysis was deleted each time to reflect the influence of individual data set to the pooled ORs, the results did not changed among the meta-analysis.

High between-studies heterogeneity was found among breast cancer, endometrial cancer, and other cancer. When the study of Zimarina et al. (2004) was excluded, the high between-studies heterogeneity were deleted among breast cancer (dominant model: I 2 = 26.2; homozygous model: I 2 = 8.1; heterozygous model: I 2 = 40.9; additive model: I 2 = 16.1), Caucasian of breast cancer (dominant model: I 2 = 0.0; recessive model: I 2 = 17.9; homozygous model: I 2 = 32.1; heterozygous model: I 2 = 0.0; additive model: I 2 = 35.1), hospital-based studies of breast cancer (dominant model: I 2 = 0.0; recessive model: I 2 = 3.0; homozygous model: I 2 = 0.0; heterozygous model: I 2 = 21.3; additive model: I 2 = 0.0), endometrial cancer (dominant model: I 2 = 58.1; homozygous model: I 2 = 42.2; heterozygous model: I 2 = 27.7; additive model: I 2 = 68.8). When the study of Zienolddiny et al. (2008) was excluded, the high between-studies heterogeneity were deleted among other cancer (dominant model: I 2 = 73.5; recessive model: I 2 = 21.1; homozygous model: I 2 = 29.8; heterozygous model: I 2 = 73.2; additive model: I 2 = 68.1).

We performed Begg’s funnel plot and Egger’s test to assess the publication bias of literatures. Begg’s funnel plots and Egger’s test suggested that there might be publication bias in recessive model (P = 0.033) and homozygous model (P = 0.047). Adjusting for possible publication bias using the Duval and Tweedie nonparametric ‘‘trim and fill’’ method for overall studies, the results did not change between CYP1B1 Arg48Gly polymorphism with the risk of cancer.

CYP1B1 Ala119Ser

The evaluations of the association of CYP1B1 Ala119Ser polymorphism with cancer risk are shown in Table 4. Overall, no significant association was found (recessive model: OR = 1.04, 95 % CI = 0.89–1.20, P h < 0.001, I 2 = 66.2 %; homozygous model: OR = 1.03, 95 % CI = 0.87–1.24, P h < 0.001, I 2 = 74.5 %) when all the eligible studies were pooled into the meta-analysis. In addition, high between-studies heterogeneity was observed among overall analysis (dominant model: I 2 = 79.7; heterozygous model: I 2 = 75.3; additive model: I 2 = 81.9). Then we performed subgroup analysis by cancer type. We found that individuals with the minor variant genotypes had a high risk of prostate cancer (recessive model: OR = 1.45, 95 % CI = 1.07–1.97, P h = 0.160, I 2 = 45.4 %; homozygous model: OR = 1.88, 95 % CI = 1.08–3.28, P h = 0.081, I 2 = 60.2 %) and other cancer (recessive model: OR = 1.78, 95 % CI = 1.34–2.37, P h = 0.293, I 2 = 19.2 %). We further examined the association of the CYP1B1 Ala119Ser polymorphism and cancer risk according to cancer type and ethnicity (Table 4). For samples of Caucasians, significant decreased cancer risk was found among other cancer (heterozygous model: OR = 0.62, 95 % CI = 0.38–1.00, P h = 0.045, I 2 = 67.7 %). We also examined the association of the CYP1B1 Arg48Gly polymorphism and cancer risk according to cancer type and source of controls (Table 4). For the population-based studies, significant association was only observed among breast cancer (homozygous model: OR = 1.12, 95 % CI = 1.01–1.25, P h = 0.151, I 2 = 36.3 %). For the hospital-based studies, significant association was found among breast cancer (dominant model: OR = 0.60, 95 % CI = 0.37–0.97, P h = 0.016, I 2 = 70.9 %; heterozygous model: OR = 0.73, 95 % CI = 0.58–0.92, P h = 0.100, I 2 = 52.1 %), prostate cancer (dominant model: OR = 1.45, 95 % CI = 1.07–1.97, P h = 0.160, I 2 = 45.4 %; homozygous model: OR = 1.88, 95 % CI = 1.08–3.28, P h = 0.081, I 2 = 60.2 %), and other cancer (dominant model: OR = 1.78, 95 % CI = 1.34–2.37, P h = 0.293, I 2 = 19.2 %).

Significant heterogeneity was observed among these studies for dominant model comparison (P h < 0.001), recessive model comparison (P h < 0.0001), additive model comparison (P h < 0.001), homozygous model comparison (P h < 0.001), and heterozygous model comparison (P h < 0.001). Table 5 lists the results of meta-regression analysis. Then, we assessed the source of heterogeneity by ethnicity, cancer type, source of controls, HWE, and sample size. The results indicated that cancer type (recessive model: P = 0.008; homozygous model: P = 0.007; additive model: P = 0.038) and ethnicity (dominant model: P = 0.031; heterozygous model: P = 0.043; additive model: P = 0.016) but not source of control (dominant model: P = 0.201; recessive model: P = 0.564; homozygous model: P = 0.269; heterozygous model: P = 0.248; additive model: P = 0.392), sample size (dominant model: P = 0.382; recessive model: P = 0.117; homozygous model: P = 0.084; heterozygous model: P = 0.541; additive model: P = 0.142), and HWE (dominant model: P = 0.676; recessive model: P = 0.586; heterozygous model: P = 0.990; homozygous model: P = 0.644; additive model: P = 0.860) contributed to substantial heterogeneity among the meta-analysis. Examining genotype frequencies in the controls, significant deviation from HWE was detected in the nine studies (Reding et al. 2012; Salinas-Sánchez et al. 2012b; Salinas-Sánchez et al. 2012a; Gulyaeva et al. 2008; Zienolddiny et al. 2008; Sasaki et al. 2003; 2004; Maurya et al. 2014; Rodrigues et al. 2011). When these studies were excluded, the results were changed among other cancer (recessive model: OR = 1.11, 95 % CI = 0.45–2.73) and endometrial cancer (dominant model: OR = 0.38, 95 % CI = 0.24–0.60), as shown in Table 6. In addition, when our meta-analysis was performed excluding studies with small sample sizes, the results did not changed among this meta-analysis, as shown in Table 7. Last, a single study involved in the meta–analysis was deleted each time to reflect the influence of individual data set to the pooled ORs. When the study of Zimarina et al. (2004) was excluded, significant association was found between CYP1B1 Ala119Ser polymorphism and breast cancer susceptibility in Caucasians (homozygous model: OR = 1.15, 95 % CI = 1.03–1.29; recessive model: OR = 1.16, 95 % CI = 1.04–1.29). In addition, when the study of Zimarina et al. (2004) was excluded, the results were also changed among hospital-based studies of breast cancer (dominant model: OR = 0.82, 95 % CI = 0.64–1.03; heterozygous model: OR = 0.82, 95 % CI = 0.63–1.05).

High between-studies heterogeneity was found among breast cancer, endometrial cancer, prostate cancer, and other cancer. When the study of Zimarina et al. (2004) was excluded, the high between-studies heterogeneity were deleted among breast cancer (additive model: I 2 = 51.0 %), Caucasian of breast cancer (dominant model: I 2 = 45.9 %; recessive model: I 2 = 27.3 %; homozygous model: I 2 = 44.4 %; heterozygous model: I 2 = 17.3 %; additive model: I 2 = 58.8 %), hospital-based studies of breast cancer (homozygous model: I 2 = 20.9; additive model: I 2 = 0.0). When the study of Tanaka et al. (2002) was excluded, the high between-studies heterogeneity were deleted among prostate cancer (dominant model: I 2 = 63.7; heterozygous model: I 2 = 41.4; additive model: I 2 = 0.0) and hospital-based studies of prostate cancer (dominant model: I 2 = 63.7; heterozygous model: I 2 = 41.4; additive model: I 2 = 0.0).

Both Begg’s funnel plot and Egger’s test were performed to assess the publication bias of literatures. The Egger’s test results (dominant model: P = 0.949; recessive model: P = 0.271; additive model: P = 0.911; Homozygous model: P = 0.241; Heterozygous model: P = 0.535) and Begg’s funnel plot (Figure not shown) suggested no evidence of publication bias in the meta-analysis.

Discussion

CYP1B1 is commonly over-expressed inhumanmalignancies and activates a variety of carcinogens. For example, CYP1B1 catalyzes both the formation of dihydrodiols of specific PAHs and their subsequent oxidation to carcinogenic dihydrodiol epoxides. The importance of CYP1B1 in chemical carcinogens is well illustrated in animal models in which metabolites of CYP1B1 were shown to induce PCa (Williams et al. 2000; Cavalieri et al. 2002). Furthermore, CYP1B1-null mice, created by targeted gene disruption in embryonic stem cells, were protected from 7,12-dimethylbenz(a)anthracene-induced malignant lymphomas (Buters et al. 1999). Many studies have reported the role of CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser with all cancers risk, but the results remained controversial. Some recent meta-analyses analyzed such an association only for single cancer such as breast cancer, lung cancer, prostate cancer, endometrial cancer, and so on. Importantly, several published studies were not included in the previous meta-analysis and additional original studies with larger sample sizes have been published since then. Hence, the correlation of these polymorphic genes remains unknown. In order to derive a more precise estimation of association, we performed the meta-analysis of CYP1B1 Leu432Val (54,124 cases and 62,932 controls), Asn453Ser (30,532 cases and 39,193 controls), Arg48Gly polymorphisms (23,494 cases and 27,083), and Ala119Ser (17,796 cases and 19,891 controls) with cancer risk.

For Asn453Ser and Arg48Gly polymorphisms, significant decreased endometrial cancer was observed among Caucasians. For Ala119Ser polymorphism, we found that individuals with the minor variant genotypes had a high risk of prostate cancer (recessive model: OR = 1.45, 95 % CI = 1.07–1.97; homozygous model: OR = 1.88, 95 % CI = 1.08–3.28) and Caucasians of breast cancer (homozygous model: OR = 1.15, 95 % CI = 1.03–1.29; recessive model: OR = 1.16, 95 % CI = 1.04–1.29). For Leu432Val polymorphisms, We found that individuals with the minor variant genotypes had a higher risk of endometrial cancer (dominant model: OR = 1.16, 95 % CI = 1.03–1.31), lung cancer (dominant model: OR = 1.25, 95 % CI = 1.07–1.48; recessive model: OR = 1.19, 95 % CI = 1.03–1.37; homozygous model: OR = 1.36, 95 % CI = 1.15–1.61; heterozygous model: OR = 1.19, 95 % CI = 1.06–1.34; additive model: OR = 1.19, 95 % CI = 1.05–1.35) and had a lower risk of ovarian cancer (recessive model: OR = 0.82, 95 % CI = 0.68–1.00).

For Asn453Ser and Arg48Gly polymorphisms, significant decreased endometrial cancer risk was observed among Caucasians, but not Asians and Africans. For Ala119Ser polymorphism, significant increased breast cancer risk was observed among Caucasians, but not Asians and Africans. For Leu432Val polymorphisms, significant increased lung cancer among Caucasians, but not Asians and Africans. The results suggested a possible role of ethnic difference in genetic background and the environment they lived in. It should be considered that the apparent inconsistency of these results may underlie differences in ethnicity, lifestyle and disease prevalence as well as possible limitations due to the relatively small sample size. The current knowledge of carcinogenesis indicates a multi-factorial and multistep process that involves various genetic alterations and several biological pathways. Thus, it is unlikely that risk factors of cancer work in isolation from each other. And the same polymorphisms may play different roles in cancer susceptibility, because cancer is a complicated multi-genetic disease, and different genetic backgrounds may contribute to the discrepancy. And even more importantly, the low penetrance genetic effects of single polymorphism may largely depend on interaction with other polymorphisms and/or a particular environmental exposure.

Based on biochemical properties described for CYP1B1 polymorphism, we would expect that the allele would be associated with higher susceptibility for all types of cancer. However, our results showed that such association was observed among ovarian cancer, lung cancer, and endometrial cancer for CYP1B1 Leu432Val, endometrial cancer for Asn453Ser and Arg48Gly, and prostate cancer and breast cancer for CYP1B1 Ala119Ser, suggesting that other factors may be modulating the CYP1B1 polymorphisms functionality. Several previous studies assessed the effect of CYP1B1 Leu432Val, Arg48Gly, and Ala119Ser polymorphisms on these cancers risk, which finding is consistent with our results. However, the exact mechanism for association between different tumor sites and CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms was not clear, carcinogenetic mechanism may differ by different tumor sites and the CYP1B1 genetic variants may exert varying effects in different cancers.

In the present meta-analysis, between-studies heterogeneity was observed between CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms and cancer of risk. Meta-regression analysis indicated that cancer type, ethnicity, source of controls, and sample size contributed to substantial heterogeneity among the meta-analysis for Leu432Val polymorphism, source of controls and sample size contributed to substantial heterogeneity among the meta-analysis for Arg48Gly, and cancer type and ethnicity contributed to substantial heterogeneity among the meta-analysis for Ala119Ser. The hospital-based studies may have some biases because such controls may contain certain benign diseases which are prone to develop malignancy and may not be very representative of the general population. The small number studies hinder the ability of drawing more definite conclusions. Thus, the use of a proper and representative cancer-free control subjects and the large sample size studies are very important in reducing biases in such genotype association studies. And this indicates that it may be not appropriate to use an overall estimation of the relationship between CYP1B1 polymorphism and cancer risk.

The current meta-analysis has some strength compared with individual studies and previous meta-analyses. First, differently from previous meta-analyses, we explored the impact of CYP1B1 Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser on a great diversity of cancer sites, allowing for a general view of its influence on cancer susceptibility. Second, our meta-analysis explores and analyzes the sources of heterogeneity between studies about CYP1B1 in cancer. Third, a systematic review of the association of Leu432Val, Asn453Ser, Arg48Gly, and Ala119Ser polymorphisms with the risk of cancer is statistically more powerful than any single study. Fourth, the quality of eligible studies included in current meta-analysis was satisfactory and met our inclusion criterion.

There are also still some limitations inherited from the published studies. First, our results were based on single-factor estimates without adjustment for other risk factors including alcohol usage, environmental factors and other lifestyle. Second, in the subgroup analysis may have had insufficient statistical power to check an association. Third, the controls were not uniformly defined. Fourth, a potential limitation of our results is the small number of studies for some tumor sites and subgroups, which hinders the ability of drawing more definite conclusions for some results. For these cases, the interpretation of the results should be taken carefully.

In summary, this meta-analysis suggests that Leu432Val polymorphism is associated with ovarian cancer, lung cancer, and endometrial cancer risk, Asn453Ser and Arg48Gly polymorphisms are associated with endometrial cancer risk among Caucasians, and Ala119Ser polymorphism is associated with prostate cancer risk and Caucasians of breast cancer risk. In addition, our work also points out the importance of new studies for Ala119Ser polymorphism in endometrial cancer, because high heterogeneity was observed (I 2 > 75 %).