|Home | About | Journals | Submit | Contact Us | Français|
Sex hormones, in particular the androgens, are important for the growth of the prostate gland and have been implicated in prostate cancer carcinogenesis, yet the determinants of endogenous steroid hormone levels remain poorly understood. Twin studies suggest a heritable component for circulating concentrations of sex hormones, although epidemiological evidence linking steroid hormone gene variants to prostate cancer is limited. Here we report on findings from a comprehensive study of genetic variation at the CYP19A1 locus in relation to prostate cancer risk and to circulating steroid hormone concentrations in men by the Breast and Prostate Cancer Cohort Consortium (BPC3), a large collaborative prospective study. The BPC3 systematically characterised variation in CYP19A1 by targeted resequencing and dense genotyping; selected haplotype-tagging single nucleotide polymorphisms (htSNPs) that efficiently predict common variants in U.S. and European whites, Latinos, Japanese Americans, and Native Hawaiians; and genotyped these htSNPs in 8,166 prostate cancer cases and 9,079 study-, age-, and ethnicity-matched controls. CYP19A1 htSNPs, two common missense variants and common haplotypes were not significantly associated with risk of prostate cancer. However, several htSNPs in linkage disequilibrium blocks 3 and 4 were significantly associated with a 5–10% difference in estradiol concentrations in men (association per copy of the two-SNP haplotype rs749292–rs727479 (A–A) versus noncarriers; P=1 × 10−5), and withinverse, although less marked changes, in free testosterone concentrations. These results suggest that although germline variation in CYP19A1 characterised by the htSNPs produces measurable differences in sex hormone concentrations in men, they do not substantially influence risk for prostate cancer.
Prostate cancer is the most commonly diagnosed cancer in males in developed countries yet aetiological risk factors for the disease are not well understood. The only established risk factors for the disease are age, family history of prostate cancer and ethnicity. There has been considerable interest in the potential role of sex hormones in prostate cancer carcinogenesis, with a particular focus on androgens, which are important for the development, growth and maintenance of the prostate gland. Finasteride, which blocks the metabolism of testosterone within the prostate, has been found to reduce the risk of prostate cancer (although the increase incidence of high grade tumors on biopsy has led to controversy) (1) and prostate tumours can be induced when testosterone, either alone or together with estradiol, is administered to laboratory animals (2). Estrogens have also been implicated in prostate biology and in the development of prostate cancer, via direct estrogen-receptor mediated effects and indirect effects, although data suggest the role of estrogens may vary with disease progression (3, 4). However, a recent re-analysis of the worldwide prospective data found no large associations between circulating androgen and estrogen concentrations in humans and prostate cancer risk (5).
Twin studies suggest a heritable component for circulating concentrations of sex hormones in men (6). The CYP19A1 gene has been identified as a candidate locus that may influence circulating sex hormone concentrations and risk for hormone-related cancers (7). CYP19A1 encodes aromatase, an enzyme which catalyses the conversion of the C19 androgens, androstenedione and testosterone, to the C18 estrogens, estrone and estradiol, respectively. This cytochrome P450 enzyme is expressed primarily in the gonads, as well as in peripheral sites including the prostate (8). In postmenopausal women, common variants in the CYP19A1 gene have been found to be associated with a 10% to 20% difference in circulating estrogen levels and a number of studies have assessed the relationship of variants in CYP19A1 with risk for cancers of the breast and endometrium (9–12). However, published data on CYP19A1 in men in relation to sex hormones and prostate cancer are relatively limited, partly due to incomplete characterisation of genetic variation at the locus of interest and small sample sizes (13–23).
In the present study, we examined the contribution of common genetic variation at the CYP19A1 locus to prostate cancer risk and to concentrations of serum sex hormones and sex hormone binding globulin (SHBG) in a large, collaborative investigation (The Breast and Prostate Cancer Cohort Consortium, (BPC3)) (24).
The BPC3 has been described in detail elsewhere (24). Briefly, the consortium includes large well-established cohorts assembled in the United States and Europe that have DNA for genotyping and extensive questionnaire data from cohort members. For prostate cancer, analyses include men from seven cohort studies: the American Cancer Society Cancer Prevention Study II (ACS CPS-II) (25), the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) Study (26), the European Prospective Investigation into Cancer and Nutrition (EPIC) Cohort (itself comprising cohorts from Denmark, Germany, Great Britain, Greece, Italy, the Netherlands, Spain, and Sweden) (27), the Health Professionals Follow-up Study (HPFS) (28), the Hawaii/Los Angeles Multi-ethnic Cohort Study (MEC) (29), the Physicians' Health Study (PHS) (30), and the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial (31). With the exception of the MEC and PLCO, these cohorts are composed predominantly of whites of European descent. The MEC contributes African American, Latino, Japanese American and Native Hawaiian cases and controls recruited from Los Angeles and Hawaii. The PLCO includes over 400 African American participants. Cases of prostate cancer were identified through population-based cancer registries or self report confirmed by medical records. The BPC3 data for prostate cancer consist of a series of matched nested case-control studies from each cohort; controls were matched within each cohort to cases on a number of potential confounding factors, including age (within five years), ethnicity, and in some cohorts, additional criteria, such as region of recruitment in EPIC (for further details on the selection of controls in each cohort see Supplementary Methods). The study protocol was approved by all institutional review boards.
We used a multi-stage approach to characterize genetic variation by initially resequencing the coding exons and exon-intron boundaries of CYP19A1 in a multiethnic panel of 95 advanced prostate cancer cases to identify putative functional alleles directly. The linkage disequilibrium (LD) structure of the locus was determined empirically in a multiethnic panel of 349 individuals from the MEC by genotyping a dense network of SNPs, including SNPs selected from dbSNP (www.ncbi.nlm.nih.gov/SNP/) and the Celera SNP database (Celera SNP database (www.Celera.com) and common missense SNPs found during resequencing (although not novel and can also be found in dbSNP) using the Sequenom and Illumina genotyping platforms at the Broad Institute/MIT Center for Genome Research. Haplotype-tagging SNPs (htSNPs) for each haplotype block, determined by the confidence interval method of Gabriel et al.(32, 33), were chosen based on Rh2, a measure of the correlation between observed and predicted haplotypes based on the htSNP genotypes(34), to select a minimum set of SNPs that would achieve an Rh2≥0.7 for all common haplotypes among whites with an estimated frequency of ≥5%.
The LD pattern across CYP19A1 has been previously shown (9). Briefly, the genomic structure of the region consists of four LD blocks spanning 181kb which was determined by 107 SNPs with MAF ≥ 0.05 among whites, and 2 missense SNPs with a frequency of >1% among whites in the multiethnic panel. In total 19 htSNPs (Supplementary Table 1) were selected to provide high predictability (R2H ≥ 0.7) (35) of 27 common haplotypes (≥ 0.05 frequency in at least one ethnic group among the five ethnic groups in the multiethnic panel) across the 4 LD block regions (33, 36) with inter-block distances < 6 kb. Haplotype frequencies were similar for whites across cohorts (data not shown), while some differences in haplotype frequencies were seen between whites, African Americans, Japanese Americans, and Native Hawaiians.
Genotyping of the htSNPs and two missense SNPs in the 17,245 case patient and control participant samples was conducted using the TaqMan assay (Applied Biosystems) in four BPC3 laboratories. Initial quality control checks of the SNP assays were performed at the manufacturer (ABI); an additional 500 test reactions were run by the BPC3 on the multiethnic reference panel; greater than 99.5% concordance was observed across genotyping platforms. Assay characteristics for the htSNPs for CYP19A1 are available on the public website http://www.uscnorris.com/mecgenetics/CohortGCKView.aspx. Sequence validation for each SNP assay was performed and 100% concordance observed (http://snp500cancer.nci.nih.gov) (37). To assess inter-laboratory variation, each centre ran assays on a designated set of 94 samples from the SNP 500 cancer panel, showing completion and concordance rates of greater than 99% (37). The internal quality of genotype data at each center was assessed by 5–10% blinded samples in duplicate or triplicate (depending on study) and intra-laboratory concordance rates of greater than 99.5% were observed. Empty water wells were also included on each plate and positioned according to the individual laboratory's protocols. Hardy-Weinberg Equilibrium (HWE) checks have been performed among the controls in each study and stratified by ethnicity in the MEC and by country for EPIC. No deviation in HWE was observed (p<0.01) across more than one study for any given assay.
Of the 8,248 prostate cancer cases and 9,312 controls sent for genotyping, at least 1 SNP was successfully genotyped for 8,166 (94%) cases and 9,079 (94%) controls. This study therefore comprises 8,166 case patients and 9,079 control participants. Among these men, we evaluated the relationship of prostate cancer risk to the htSNPs, to the common CYP19A1 haplotypes predicted among Whites by the 19 htSNPs, and to the two common missense SNPs at the CYP19A1 locus, including SNPs R264C (rs700519) and T201M (rs28757184). In a subset of control participants from ATBC, EPIC, HPFS, PHS and PLCO, we also investigated whether variation at this locus contributes to inter-individual differences in circulating levels of sex hormones and SHBG, using measurements of steroid hormones made previously by individual cohorts (38–41).
We used conditional logistic regression to estimate odds ratios for prostate cancer associated with carrying either 1 or 2 versus 0 copies of the minor allele for each SNP. We estimated haplotype-specific odds ratios using an expectation-substitution approach to account for haplotype uncertainty given unphased genotype data (42, 43). Haplotype frequencies and subject-specific expected haplotype indicators were calculated separately for each cohort (and country within EPIC or ethnicity in the MEC and PLCO). To test the global null hypothesis of no association between variation in common CYP19A1 haplotypes and risk of prostate cancer and to control for type-I error over all the SNPs and haplotypes considered, we used a likelihood ratio test comparing a model with additive effects for each common haplotype (treating the most common haplotype as the referent) to the intercept-only model. For all cancer risk analyses, we test for significance associations at the 0.01 level to minimize the chance of both false positive and false negative results.
We tested for heterogeneity in odds ratio estimates across cohorts among white participants and by ethnicity. In addition, we calculated risk-stratum-specific odds ratios and tested for departures from a multiplicative interaction model to assess whether other risk factors for prostate cancer modify the association with htSNPs, missense SNPs or common haplotypes, including age (at diagnosis), body mass index and family history of prostate cancer. To assess the influence of genetic variation in CYP19A1 on prostate cancer diagnosed with an aggressive phenotype, we calculated stratum specific odds ratios for high grade prostate cancer (defined as poorly differentiated or Gleason score>=8), advanced stage prostate cancer (Stage C or D), aggressive disease (defined as high grade, advanced stage prostate cancer or death from prostate cancer).
We used fixed effect models to evaluate associations of circulating steroid hormone and sex hormone binding globulin (SHBG) concentrations with CYP19A1 htSNPs, missense SNPs and common haplotypes in cases and controls, with study cohort included as a fixed effect with a suitable nesting of batches within study where appropriate, and adjustment for age (in five-year age-groups). For these analyses hormone and SHBG concentrations were logarithmically transformed for statistical analyses to approximately normalise their distributions. We report the P-value from the test of trend across genotype groups.
The demographic and other characteristics of cases and controls from the seven cohorts are shown in Table 1. Most study participants were U.S. or European whites (75%), followed by African Americans (11%), Latinos (7%), Japanese Americans (5%), and Native Hawaiians (1%). Among participants, 14% of cases and 9% of controls reported a father or a brother with prostate cancer. Cases and controls were similar with respect to age, BMI, and height. Data on tumour stage and grade were available from six of the seven participating cohorts (all but PLCO). Stage information was available on 70% of genotyped prostate cancer cases, and of these, 19% had advanced disease (defined as stage C or D disease at diagnosis). Gleason score was recorded for 65% of genotyped cases, and 18% of cases had a Gleason sum of 8 or greater.
Genotype specific odds ratios for the 21 SNPs tested, including the two missense SNPs, are shown in Table 2 for analyses pooling all participants. There was no evidence of an association of the 2 missense SNPs with risk for prostate cancer (P > 0.50 for both for analyses). None of the htSNPs showed a significant association with prostate cancer risk at the 0.01 level. The SNP-specific odds ratios did not show significant heterogeneity across ethnicities or across cohorts among whites at the 0.01 level. Global tests of association between CYP19A1 common haplotypes and prostate cancer were also not significant at the 0.01 level (see Table 3), and no statistically significant associations were observed between individual common haplotypes in each block and risk for prostate cancer (P > 0.01 for all).
We calculated odds ratios for risk of advanced disease by CYP19A1 htSNPs, missense SNPs and common haplotypes, classifying disease severity by prostate cancer stage, grade or a combined score of prostate tumour stage and histological grade. For aggressive disease in relation to CYP19A1 htSNPs, rs2445762 in LD block 1 was significantly associated with high grade disease with a significant decrease in risk being observed in heterozygotes (P 2 d.f.test = 0.0002, OR in heterozygotes = 0.76, 99% CI = 0.62–0.93, OR in homozygotes = 1.15, 99% CI = 0.83–1.59), and similarly with the composite variable for aggressive disease (P = 0.0003). We found no significant association between other CYP19A1 htSNPs or missense SNPs and risk for tumour stage or grade or the composite score of aggressive disease at the 0.01 level. There was no evidence of any association between CYP19A1 common haplotypes and risk for aggressive prostate cancer, defined as high grade, advanced stage or a composite variable.
Tests for departures from multiplicative interaction models were null when we examined statistical interaction between CYP19A1 htSNPs, missense SNPs or common haplotypes and prostate cancer risk with the following risk factors: family history (at least one first degree relative diagnosed with prostate cancer versus none), age at diagnosis (≤ 65, > 65 years), and BMI (< 25, ≥ 25 < 30, ≥ 30 kg/m2).
In analyses of genetic variation in CYP19A1 in relation to circulating sex hormone and SHBG concentrations, there were significant associations at the 0.01 level for a number of htSNPs with concentrations of estradiol, free estradiol, free testosterone and androstanediol glucuronide. No significant associations were observed with concentrations of testosterone or SHBG, and neither of the missense SNPs were associated with concentrations of sex hormones or SHBG.
For estradiol, the most significant associations were observed for htSNPs in LD blocks 3 and 4: Ptrend <0.005 for every htSNP in block 3, and P≤0.006 for rs727479 and rs10046 in block 4 (Table 4). These associations did not differ significantly among the three cohorts that measured estradiol, nor did they differ between cases and controls or by ethnicity, although only approximately 5% of the hormone data were from non-white participants. Percentage change in estradiol between homozygotes for the wild-type and the variant allele for these SNPs ranged from approximately 5 to 10%. There is a high degree of LD between previously defined LD blocks 3 and 4, and therefore many of the htSNPs are highly correlated (r2 ≥ 0.83 for SNP pairs rs749292 and rs6493494, and, rs2414096 and rs10046) (9). Haplotypes in blocks 3 and 4 were also strongly associated with estrogen levels and the magnitude of the associations was similar to the independent tagging SNPs in these blocks (data not shown).
The two SNPs most significantly associated with estradiol concentrations, rs749292 and rs727479, were only modestly correlated with each other (r2 = 0.46) and each remained significantly associated with circulating levels when modelled concurrently (Ptrend = 0.002; Ptrend = 0.006, respectively). A 2-SNP haplotype (A–A) comprised of these SNPs was found to be a more significant predictor of estradiol concentrations (Ptrend = 1 × 10−5) (Table 5); however, it accounted for only 0.75% of the variation in estradiol concentrations and when we examined risk for prostate cancer in relation to this two snp-haplotype we observed no significant association with risk for disease at the 0.01 significance level (P = 0.028, OR in men with one copy of the A–A haplotype = 0.96, 99% CI = 0.88–1.06, OR for two copies = 1.08, 99% CI = 0.96–1.22).
Full results for all sex hormones and SHBG in relation to CYP19A1 htSNPs are shown in Supplementary Table 2. Findings for free estradiol were broadly similar to those observed for estradiol, both with respect to level of statistical significance and the percentage change in hormone level by genotype. Significant associations were also observed between CYP19A1 htSNPs in LD blocks 3 and 4 and the ratio of free estradiol to free testosterone, as an index of aromatase activity, with trends mirroring those observed for estradiol concentrations.
For testosterone, there were no significant associations with htSNPs at the 0.01 significance level, although weak associations that did not reach statistical significance were observed with the 3 htSNPs in LD block 3 (Ptrend <0.09), and with rs727479 in LD block 4 (P = 0.04); changes in testosterone concentrations by genotype were the inverse of those seen for estradiol concentrations and percentage changes in hormone concentrations were approximately 2–3% for htSNPs in LD block 3 and 4% for rs727479. For free testosterone concentrations, however, significant or borderline significant associations were observed with a number of htSNPs in LD blocks 3 and 4 (P < 0.01 for rs6493494, rs727479 and rs10046), with trends in free testosterone concentrations being the inverse of those observed for estradiol and free estradiol (Table 4) and percentage changes in free testosterone being approximately 3–4%.
For androstranediol glucuronide, significant associations were observed between androstanediol glucuronide and htSNPs rs28566535/CV1664178 and rs1902586 in LD block 2 (P trend = 3 × 10−4 and 0.001, respectively) and with rs727479 in LD block 4 (P trend = 0.009) (Supplementary Table 2). The percentage changes in androstranediol glucuronide concentration between the wild-type and the variant allele were 18.7%, 13.9% and 6.3%, respectively. We observed no significant associations between CYP19A1 htSNPs and concentrations of SHBG (Supplementary Table 2).
We investigated common genetic variation at the CYP19A1 locus in relation to prostate cancer risk and hormone levels in a large, collaborative investigation (BPC3) (24). We evaluated both haplotype patterns in four well-characterized LD blocks and two common missense variants and found no evidence that this gene harbours a prostate cancer susceptibility allele. In addition, this is the first large study that has comprehensively assessed genetic variation in CYP19A1 in relation to sex hormones and SHBG concentrations in men. Our results provide evidence for significant associations of a number of htSNPs in CYP19A1 with circulating concentrations of estradiol, free estradiol, free testosterone and androstranediol glucuronide.
Ten studies of the relationship between genetic variation in CYP19A1 and risk for prostate cancer have been published to date (13–18, 20–23), with inconsistent findings. These studies only considered a limited number of variants across the locus, and with few exceptions (15, 20, 21) were small, being underpowered to detect the modest magnitude of effect anticipated for a common low-penetrance susceptibility allele (RR<1.5. These variants include a tetranucleotide repeat (14, 16, 21–23), 2 nonsynonymous mutations (13, 15, 17, 18, 20), a G/A SNP in the 3`untranslated region (rs10046) (20), and 6 other genetic variants identified in screening of prostate cancer patients (15). The null association with the missense mutation R264C in the current study contrasts with those from three small previous studies (13, 17, 18), but is in agreement with null results from two large recent studies (15, 20), as is our null finding for rs10046 (20). Our results suggesting no strong association with T201M, however, contrast with those from the one previous study, which reported an increase in risk, especially in association with low grade organ confined disease (15). The genome-wide association studies of prostate cancer also provide data on the CYP19A1 locus but have not implicated the region. For example, in a genome-wide association study in the Cancer Genetic Markers of Susceptibility (CGEMS) project of 1,172 prostate cases and 1,157 controls, 46 SNPs that met quality control metrics were evaluated across CYP19A1 (from 30 kb upstream of the ATG and 30 kb downstream of the polyA tail)(44). Using HapMap Phase 2 CEU samples, the 46 SNPs genotyped tagged (using r2 > 0.8) a total 164 SNPs in 40 bins, which represents 74.9% of all CEU HapMap SNPs with MAF > 5%. In CGEMS, no SNPs had a p value < 0.01 in the adjusted trend test; rs2124873 displayed the smallest p value (p = > 0.03) in CGEMS.
Findings from the current study for prostate cancer risk in relation to genetic variation in CYP19A1 provide no evidence that risks differ by ethnicity, family history, age or BMI. Our findings also did not support a strong association with severity of disease, with one exception. We observed a statistically significant association between rs2445762 and risk for high grade disease and a composite score of aggressive disease, with a decreased risk being observed in heterozygotes, but not in homozygotes, for the variant allele. Given the multiple-testing in this study and the small number of cases with two copies of the rs2445762 variant allele (91 and 191 cases with high grade and aggressive disease, respectively), these associations with aggressive prostate cancer are likely due to chance; replication is warranted in further large prostate cancer studies which have detailed information on prostate cancer phenotype.
Despite the lack of evidence for an association between genetic variants and prostate cancer risk, we found evidence for an association between CYP19A1 variants and circulating sex hormone concentrations. The strongest associations between CYP19A1 variation and hormone concentrations were observed for htSNPs in LD blocks 3 and 4. These findings are consistent with those from a recent report on 5531 men from Sweden and the US that found SNP rs2470152 in intron 1 of CYP19A1 was associated with serum estradiol (8% to 13% difference between AA and GG homozygotes in the three cohorts studied; P = 2 × 10−14 for all cohorts combined) (45). For the two most strongly associated SNPs in the current study, the LD relationship with rs2470152 was strong for rs749292 (D' = 0.96, r2 = 0.65) and weaker LD for rs727479 (D' = 0.60, r2 = 0.19), and given these allelic associations, the findings with estradiol for these SNPs (rs749292, 7% difference between homozygotes, P = 2 × 10−4; rs727479, 9% difference, P = 1 × 10−5) are comparable with the Swedish data. Other published studies on sex hormones and CYP19A1 variation in men have had small sample sizes and results have been conflicting (19, 46–48). Data from the current study for estradiol and free estradiol in relation to CYP19A1 variants are also broadly consistent with those previously reported for women (9–11, 49, 50), including findings from 3,400 postmenopausal women participating in BPC3 (9). For five out of seven variants in CYP19A1 LD blocks 3 and 4 found to be significantly associated with estradiol concentrations in women in BPC3, we observed significant associations in men. However, the strength and the magnitude of these associations among men in the current study is somewhat weaker than that observed among postmenopausal women; in men we found variants to be significantly associated with a 5% to 10% difference in estradiol concentrations, whereas in postmenopausal women in BPC3 these variants were significantly associated with a 10% to 20% difference in endogenous estrogen levels (9). It is possible that while CYP19A1 variants influence aromatase activity, the impact on circulating hormone concentrations in men might be relatively small because of homeostasis of free testosterone by the hypothalamic-pituitary-gonadal feedback loop.
We found no evidence for a significant association of total testosterone with CYP19A1 variation; however, we found several variants in LD blocks 3 and 4 to be significantly associated with a 2–4% difference in levels of free testosterone, with the trend in hormone concentration being the reverse of that seen for estradiol. An explanation for these findings may lie in the fact that free testosterone is the substrate for conversion to estradiol by aromatase. The association found between androstanediol glucuronide and genetic variation in LD blocks 2 and 4 of CYP19A1 is also compatible with an influence of CYP19A1 genetic variants on aromatase activity as androstanediol glucuronide is an end metabolite of testosterone (51).
Overall, our findings that variants in CYP19A1 were associated with small differences in circulating hormone concentrations but no detectable effect on risk for prostate cancer are consistent with what is known on the association of hormones with risk. A recent collaborative re-analysis of the world-wide data reported no strong association between serum concentrations of sex hormones and risk of prostate cancer (5), with differences in serum sex hormone concentrations between the highest and lowest fourths of the distribution being in the order of two-fold (38). Thus, the small differences (5–10%) in sex hormones in relation to CYP19A1 variants may be of insufficient magnitude to have a detectable influence on risk for prostate cancer. The relationship between circulating hormone concentrations and intraprostatic hormone levels, however, remains unclear, and the authors of the collaborative reanalysis concluded that any biological interpretation of their results for serum hormones must be viewed with caution (5). Aromatase is expressed in the prostate and local estrogen production may lead to intraprostatic levels exceeding those in the circulation, with prostate cells being exposed to estradiol from the blood plus locally produced estradiol (3, 52, 53). However, the null association in the current study between CYP19A1 variants and prostate cancer risk does not lend support to the hypothesis that intraprostatic sex hormone metabolism by aromatase is strongly associated with risk since it would be anticipated that CYP19A1 variation might have a similar effect on intraprostatic sex hormone concentrations as on circulating hormones.
Strengths of this study lie in the scale of the analysis, both with respect to the comprehensive screening for prostate cancer susceptibility alleles across the entire gene region facilitated by the consortium's cost-effective haplotype-tagging approach, and with respect to the large sample size, which makes it possible to conduct adequately powered subgroup analyses by potential prostate cancer risk factors and by tumour characteristics. In this study, with over 8,000 case patients and 9,000 control participants, we have greater than 95% power to detect a dominant effect or log-additive odds ratio of 1.3 for an allele with 5% minor allele frequency at the 0.001 level. Furthermore, with respect to subgroup analyses we still have, for example, greater than 95% power to detect a stratum-specific dominant odds ratio of 1.7 for a 5% frequency variant at the 0.001 level when the stratum consists of only 20% of the sample. The current study, however, had less power to assess whether variation in CYP19A1 is associated with risk for prostate cancer in non-white ethnic groups. Further analyses may clarify associations in non-white populations as sample sizes are increased with longer follow-up or the addition of new cohorts.
We have studied the importance of variation of CYP19A1 to steroid hormone concentrations and to prostate cancer risk. Ultimately, however, the importance of variation in CYP19A1 on endogenous hormone concentrations should be assessed in the context of variation in the many other genes involved in steroid hormone biosynthesis and metabolism pathways, as well as in genes involved in steroid hormone-binding and -receptor pathways. Whilst variation in individual genes that encode enzymes in the hormone metabolism pathways may individually only result in modest differences in hormone concentrations, a combination of several genes with functional variants could result in a cumulative large effect which would endure over the course of a human lifespan (54). Indeed, twin studies suggest that additive genetic factors may account for up to 57% and 25% of the total variation in circulating testosterone and estradiol concentrations, respectively, in men (6). Thus, the development of a multi-factorial score for the prediction of hormone concentrations within large-scale studies, such as is planned within BPC3, may provide important insights into the determinants of endogenous hormone concentrations and risk for the diseases and disorders that they influence. Such conditions are not limited to cancer and may also include, for example, vascular disease (55), diabetes (56) and those related to cognitive decline (57, 58), and to bone metabolism (19, 45, 46).
In summary, results from this study suggest that although germline mutations in CYP19A1 produce measurable differences in steroid hormone concentrations in men, they do not substantially influence risk for prostate cancer. These findings are consistent with a reanalysis of worldwide data which found no large associations between serum concentrations of sex hormones and prostate cancer risk. Our results for steroid hormone concentrations in relation to variants in CYP19A1, however, may have wider relevance for other conditions of public health importance in men.
The authors gratefully acknowledge the participants in the component cohort studies.
Grant Support: National Cancer Institute cooperative agreements UO1-CA98233, UO1-CA98710, UO1-CA98216, and UO1-CA98758 and Cancer Research UK (R.C. Travis).