Selection and characteristics of eligible studies
According to the inclusion and exclusion criteria, a total of 14 studies [14, 16-19, 24-32] were deemed eligible for this meta-analysis (Figure 1 detailed the study selection process). Among them, 8 studies investigated only the Arg264Cys polymorphism [14, 16-18, 24-26, 28], 5 studies explored only the (TTTA)n repeat polymorphism[19, 29-32], and 1 study focused on both polymorphisms. The main characteristics of these studies are listed in Supplementary Table 1 (see Additional file 1) and Supplementary Table 2 (see Additional file 2). With the exception of the work of Fukastu et al (2004) with a score of 5, all studies were considered of high quality base on the NOS≥6. Association between the CYP19A1 Arg264Cys polymorphism and PCa risk
Nine studies (7 case-control studies and 2 nested case-control studies), which recruited a total of 11,824 patients and 11,368 control subjects, addressed the relationship between the Arg264Cys polymorphism and PCa (see Supplementary Table 1)[14, 16-18, 24-28]. These studies enrolled Caucasians (4 studies), Asians (2 studies), Indians (1 study), African-Americans (1 study), and subjects of mixed ethnicities (1 study). One study employed intervention measures (finasteride) that might confer PCa risk and might conceal the real situation of this polymorphism in PCa; therefore, only data in the placebo arm in this study were considered . In 2 studies, genotype distributions in the control groups deviated from HWE [14, 18]. Controls were hospital-based (HB) in 3 studies, population-based (PB) in 5 studies, and both HB and PB in 1 study. One study merely recruited cases with a family history of PCa in a first-degree relative , 1 study described the ratio of familial cases in all participants , while the remaining 7 studies did not provide information about family history. Peripheral blood were used to detect this polymorphism in 7 studies[14, 16, 17, 25-28], while peripheral blood or frozen prostate tissues were used in the other 2 studies[18, 24].
The allele and genotype frequencies in the overall population and all subgroups were summarized in Table 1.The overall pooled results and subgroup analyses are listed in Table 2. Neither the T allele nor the TT genotype was associated with PCa in the overall population, Caucasians, or Asians (Figure 2 displayed non-significant association between the T allele and overall PCa risk). Subgroup analyses could not be performed for African-Americans, Indians, or mixed ethnicities since each of these ethnicities was included only in one study. Subgroup analysis stratified by the source of controls was also conducted. When PB controls were considered, no evident association was found between the Arg264Cys polymorphism and PCa in any genetic model. Nevertheless, an association between the Arg264Cys polymorphism and PCa was observed under the dominant model (CT+TT vs. CC: OR=1.35, 95%CI=1.02-1.78, P=0.04) and heterozygous model (CT vs. TT: OR=1.40, 95%CI=1.04-1.88, P=0.03) in the HB subgroup. Subgroup analysis could not be accomplished for the PB+HB subgroup since only one study included PB+HB controls. We further conducted subgroup analyses based on the sample acquisition and genotyping method, respectively. No significant association was detected in these subgroups (Table 2).
Association between the CYP19A1 (TTTA)n repeat polymorphism and PCa risk
Six studies (5 case-control studies and 1 nested case-control study) which recruited a total of 1488 cases and 1621 controls were analyzed (see Supplementary Table 2)[19, 27, 29-32]. Peripheral blood were used in all these studies to detect the (TTTA)n repeat polymorphism. Three of them were conducted in Asians, 1 each in Caucasians,Indians and subjects of mixed ethnicities. The repeat numbers ranged from 7 to 14, but the 9-repeat allele was not detected in 4 studies [29-32] , and the 14-repeat allele was identified only in 1 study . Thus, the 7-, 8-, and 10-13-repeat alleles were considered in this meta-analysis. Soni and coworkers  did not find 10, 11, and 13 repeats either in patients or in controls, so that their study could not be included in the analysis of these three alleles. For the same reason, the study of Tang  was not used in the analysis of the 13-repeat allele.
The pooled results are listed in Table 3. Only the 8-repeat allele was significantly associated with the risk of PCa in the overall population (OR=1.34, 95%CI=1.14-1.58, P=0.001) (Figure 3), while all the other alleles appeared not related to the overall PCa risk. Among the 6 studies addressing the 8-repeat allele, 1 was performed in Caucasians, 3 in Asians, 1 in Indians, and 1 in subjects of mixed ethnicities. Therefore, subgroup analysis only applied to Asians and yielded a non-significant result (OR=1.27, 95%CI=0.88-1.85, P=0.17). . In the subgroup analysis based on the source of controls, the 8-repeat allele was a risk factor of PCa in the PB subgroup (OR=1.41, 95%CI=1.13-1.74, P=0.002). Although in the subgroup analysis stratified by the genotyping method, the 8-repeat allele increased PCa susceptibility in the subgroup of studies using capillary electrophoresis to investigate the (TTTA)n repeat polymorphism (OR=1.34, 95%=1.09-1.65, P=0.006), the results should be treated cautiously because there were only 2 studies using this genotyping method.
Heterogeneity and Sensitivity Analysis
Although there was considerable heterogeneity among the studies in certain subgroups in this meta-analysis (Table 2 and Table 3), sensitivity analysis showed that the synthetic results were not materially altered by omitting any single study, indicating the stability of the results (Figure 4 and Figure 5).
Begg’s funnel plots did not reveal any evidence of obvious asymmetry in studies on the Arg264Cys polymorphism (Figure 6), and Egger’s test with P>0.05 further verified the absence of publication bias (Table 2). For the (TTTA)n repeat polymorphism, Begg’s funnel plots and Egger’s test indicated the absence of publication bias except for the 7-, 8- and 13-repeat alleles (P<0.05; Table 3).