A Comprehensive Analysis of Steroid Hormones and Progression of Localized High-Risk Prostate Cancer

Prostate cancer is the most commonly diagnosed cancer and the second most common cause of cancer-related deaths among North American men (1). Currently, approximately 85% of newly diagnosed cases of prostate cancer are localized within the prostate. Despite the fact that sex steroid hormones play a central role in prostate biology, a collaborative analysis of 18 prospective studies failed to demonstrate an association between six circulating steroids—testosterone, dihydrotestosterone (DHT), dehydroepiandrosterone-sulfate (DHEA-S), androstenedione (4-Dione or AD), androstanediol-glucuronide (3α-diol-G) and estradiol (E₂)—and the risk of developing prostate cancer (2). Also, another large study did not find any association between prediagnostic circulating hormone levels (testosterone, DHT, E₂, and 3α-diol-G) and lethal prostate cancer (3). In the latter study, the authors did, however, find an association between an increased risk of lethal prostate cancer with higher prediagnostic levels of DHT in men with aggressive Gleason score (GS) ≥8 and stage T3 disease (3).

Results from phase III chemoprevention trials further support the role of sex steroid hormones in prostate cancer carcinogenesis (4, 5). Indeed, finasteride (a 5α-reductase type I inhibitor) and dutasteride (a dual 5-AR inhibitor targeting 5α-reductase type I and type II), both of which block the conversion of testosterone into DHT in target cells, reduced the occurrence of prostate cancer by approximately 25% (4, 5). However, this reduction in prostate cancer risk comes at the expense of a controversial but potentially increased risk of developing higher-grade prostate cancer (4, 5). In agreement with the finasteride and dutasteride findings, in men with a confirmed diagnosis of prostate cancer, previous studies have suggested that low pretreatment levels of serum testosterone are associated with disease aggressiveness and clinical outcomes (6, 7).

However, other circulating hormones may contribute to clinical outcomes in men with localized disease. This is reinforced by the fact that cancer cells possess the intracrine biosynthesis machinery to locally synthesize active androgens from precursors such as DHEA-S, dehydroepiandrosterone (DHEA), 4-Dione or AD, and androstenediol (A5diol) originating from extragonadal peripheral sources, such as the adrenal glands (Fig. 1; refs. 8, 9). To our knowledge, a comprehensive assessment of the circulating hormonal environment including precursors associated with disease aggressiveness and clinical outcomes in men with newly diagnosed localized prostate cancer has not been assessed. This lack of information is mostly due to the fact that the majorities of previous studies conducted in localized disease were mostly focused on testosterone and relied on low specificity and accuracy radioimmunoassays measuring a single hormone at a time. The use of gold standard mass spectrometry (MS) methods for accurate and specific assessment of various steroids is thus critical.

In this study, we analyzed the preoperative plasma levels of 15 steroids using sensitive and specific MS assays in a prospective cohort of 1,766 men undergoing radical prostatectomy (RP) for newly diagnosed prostate cancer. We then evaluated the effect of the association between these steroids and clinical and pathologic prognostic factors and also the independent prognostic value of individual circulating hormones.

The study included 1,908 patients with localized prostate cancer who were recruited between 2007 and 2012 at four university hospital centers in the Province of Quebec in Canada (Montreal, McGill, Quebec, and Sherbrooke). All patients had localized prostate cancer at the time of diagnosis and underwent RP. Lymph-node dissection was performed at the discretion of the surgeons. Serial prostate-specific antigen (PSA) measurements and clinical data were gathered during follow-up. A total of 1,908 single preoperative blood samples were available for steroid analysis. One hundred and forty-two patients were documented to have a prior exposure to a 5α-reductase inhibitor and, therefore, were excluded from the analyses leaving a total of 1,766 patients for this study. No other forms of hormonal manipulation were administered before blood sampling. The objective of the study was the assessment of associations of steroid hormone levels with prostate cancer prognostic factors and progression. After prostatectomy, patients were seen regularly and PSA was measured every 3 months for 2 years, every 4–6 months for 2 years, and then every 6–12 months or at the discretion of the physicians. Disease-free survival (DFS) was defined as the occurrence of biochemical recurrence (BCR), metastasis, and/or death (all-causes). BCR was defined as (i) the occurrence of a first PSA >0.2 ng/mL any time after surgery or (ii) a detectable PSA of <0.2 ng/mL that triggered initiation of salvage radiation or androgen ablation therapy (10–13). Before surgery, each patient provided written informed consent for research and the protocol was evaluated and approved by the Centre Hospitalier Universitaire (CHU) de Québec (Québec, Canada) and local Ethical Research Committees.

Blood samples were collected and processed at a preoperative visit and banked at −80°C until analyzed. LC/MS-MS and GC-MS were used to measure plasma steroid levels as described previously (Fig. 1; ref. 14). Ten unconjugated steroids were measured in a single assay using 250 μL of plasma, whereas sulfates and glucuronides were measured in two independent assays using 20 μL and 100 μL of plasma, respectively. Analyses were performed in a blinded fashion. Reference steroids were purchased from Steraloids. Internal standards (deuterated steroids) were added to samples and quality controls were included in each run. The measured steroids and their limits of quantification were as follows: DHEA, 100 pg/mL; progesterone (Prog), 50 pg/mL; A5diol, 50 pg/mL; testosterone, 30 pg/mL; DHT, 10 pg/mL; androsterone (AST), 50 pg/mL; androstane-3β, 17β-diol (3β-Diol), 10 pg/mL; estrone (E₁), 5 pg/mL; estradiol (E₂), 1 pg/mL; 4-Dione or AD, 50 pg/mL; AST-glucuronide (AST-G), 1 ng/mL; androstane-3α, 17β-diol-3-glucuronide (3α-diol-3G), 0.25 ng/mL; androstane-3α, 17β-diol-17-glucuronide (3α-diol-17-G), 0.25 ng/mL; DHEA-S, 0.075 mg/mL; estrone-sulfate (E₁-S), 0.075 ng/mL. Three low and three high hormone concentration quality control replicates were included in each run and all metabolite coefficients of variation were <10%.

Quantitative variables are described as mean, SD, 95% confidence interval (CI), median, and range, and categorical variables as frequencies and percentages. Parametric or nonparametric (F test or Wilcoxon–Mann–Whitney test) tests were used to compare continuous data by groups after normality verification; χ² or exact tests were used for categorical data comparisons. Log-transformed data were used to account for the abnormal hormonal distribution and were used in subsequent analyses as performed in studies similar in scope (13, 15–18). Data generated with untransformed hormone concentration are also provided for completeness in the Supplementary section. Spearman correlations were estimated between serum concentrations to evaluate correlations between steroids. Cox proportional hazard models (HR) were used to estimate the association between hormone levels and DFS. In multivariable analyses, the statistical model included age, PSA level (continuous), the pathologic GS (GS6, GS3+4, GS4+3, GS ≥8), pathologic T staging [extracapsular extension (pT3a) and seminal vesicle invasion (pT3b)], margin, and nodal status (19, 20). On the basis of the high recurrence rate in high-risk patients, the clinical impact of the steroid metabolome on progression was also specifically evaluated in this subgroup. Interactions between high-risk disease and each hormone were tested in separate Cox proportional hazard models, which included risk group, hormones, and interaction term as independent factors. Because we studied 15 hormones, FDRs (q-values) were calculated to determine the degree to which the results were prone to false positives with the use of the RQVALUE package (http://genomics.princeton.edu/storeylab/qvalue/; ref. 21). The q value represents the rate at which hormones data are null. All analyses were performed done using SAS 9.4 by the biostatistician involved in the study (DS).

We studied 1,766 men with localized prostate cancer who were undergoing RP as part of the multi-institutional PROCURE Prostate Cancer Biobank project (22). The median follow-up time was 73.2 months after surgery. Five hundred twenty-four patients experienced BCR (29.6%), occurring at a median time of 19.0 months after surgery. Seventy-five patients developed metastasis and 111 deaths (all-cause) were documented. Table 1 depicts characteristics of the studied cohort. Clinicopathologic features incorporated in multivariable models and the associated risk of recurrence after radical prostatectomy are showed in Supplementary Fig. S1.

Mean hormone values of the PROCURE cohort are depicted in Supplementary Table S1. All steroids and their relationships with PSA and GS are depicted in Supplementary Tables S2 and S3. Higher GS was associated with a stepwise increase (>30%) in AST-G levels (P = 0.036; q = 0.180), 3α-diol-3G (P = 0.036; q = 0.180) and DHEA-S (P = 0.023; q = 0.180; Supplementary Table S3). None of the steroid hormones examined were associated with pathologic T staging (Supplementary Table S4). The 323 patients in the lower risk category (GS 6/PSA < 10/≤ pT2c) displayed higher mean preoperative testosterone levels (4.10 ng/mL, 95% CI, 3.74–4.45) compared to the 312 patients in the high-risk category (GS ≥ 8 and/or PSA > 20 and/or > pT3a; 3.76 ng/mL; 95% CI, 3.59–3.93; P = 0.049; q = 0.367). A similar observation was made for AST (P = 0.040; q = 0.367; Supplementary Table S5).

In multivariable Cox analyses adjusted for established prognostic factors shown in Supplementary Fig. S1, patients with increasing levels of the adrenal precursor A5diol had an increased risk of progression (Fig. 2; Supplementary Table S6). Results for regression analyses with A5diol modeled as quartiles (Q) also exposed an increased risk of recurrence (Supplementary Table S7). Compared with men in the lowest quartile (Q1), those in highest quartiles (Q4) had increased risk of progression (Q4/Q1: HR = 1.34; 95% CI, 1.06–1.70; P = 0.015; q = 0.120). Restricting the analyses of quartiles to BCR/metastasis generate similar data (not shown).

Similarly, A5diol levels (as a continuous variable) were associated with DFS. Indeed, a one-unit increment in log-transformed A5diol levels was associated with a HR of 1.19 (95% CI, 1.03–1.37; P = 0.017; q = 0.255; Supplementary Table S6). Results for regression analyses with DHEA-S modeled as quartile also suggested an increased risk of recurrence (Q4/Q1: HR = 1.34; 95% CI, 1.06–1.69; P = 0.016; q = 0.120; Supplementary Table S7). Trends for a shorter DFS were observed with higher levels (continuous) of the adrenal precursor DHEA-S (HR = 1.12; 95% CI, 0.99–1.27; P = 0.071; q = 0.410) as well as of the DHT metabolite 3α-diol-3G (HR = 1.13; 95% CI, 0.99–1.29; P = 0.082; q = 0.410; Fig. 2; Supplementary Table S6). Restricting the analyses to BCR/metastasis generate similar data (Supplementary Table S8). Performing the analysis with untransformed hormone values identify similar associations. Indeed, the associated change in HR per SD variation in untransformed hormone concentration is depicted in Supplementary Table S9. Correlation between all measured hormones is shown in Supplementary Fig. S2.

Analyses were then conducted in 312 patients with high-risk disease comprising ≥1 high-risk features: GS ≥ 8, PSA > 20, or > pT3a. A positive interaction between A5diol and disease risk was observed (P = 0.022). In these patients, a one-unit increment in log-transformed A5diol levels was linked to DFS with a HR of 1.47 (95% CI, 1.15–1.87; P = 0.0017; q = 0.026; Fig. 3A). When analyzed as quartiles, A5diol levels were also associated with DFS (log-rank P = 0.036; Fig. 3B). Results obtained for DHEA-S displayed a HR of 1.24 (95% CI, 1.01–1.52; P = 0.043; q = 0.323; log-rank P = 0.099; Fig. 3C; Supplementary Table S10). Restricting the analyses to BCR/metastasis generate similar data (Supplementary Table S11). Performing the analysis with untransformed hormone values in this subgroup identify additional associations with DHEA and 3α-diol-17G (Supplementary Table S12). Finally, trends were observed for associations between A5diol, DHEA-S levels, and metastasis-free survival. Increment in log-transformed A5diol levels was associated with a HR of 1.51 (95% CI, 0.98–2.30; P = 0.057; q = 0.335) and DHEA-S with a HR of 1.43 (95% CI, 0.99–2.05; P = 0.054; q = 0.335; Supplementary Table S13). The association with testosterone did not reach significance with a HR of 1.41 (95% CI, 0.84–2.38; P = 0.196; q = 0.490). The association between metastasis-free survival and untransformed hormone values is depicted in Supplementary Table S14.

The evaluation of circulating steroids in relation to prostate cancer progression has been largely focused on the potent androgens testosterone and DHT. To our knowledge, a comprehensive assessment of preoperative circulating steroids from gonadal, adrenal, and peripheral sources in relation to prognostic factors and DFS in men with localized disease in a sizeable prospective cohort like PROCURE has never been evaluated, and especially not with an accurate and specific MS method.

Several studies have demonstrated a correlation between testosterone levels and prognostic factors in men with localized prostate cancer. Indeed, low pretreatment testosterone levels were previously associated with adverse pathologic factors defining an aggressive disease phenotype (6, 7, 23–26). In agreement, in this analysis of the PROCURE cohort of patients with prostate cancer with localized disease, we also observed an association between lower levels of testosterone and AST and tumor aggressiveness based on the combination of PSA values at diagnosis, GS, and pT staging. In addition, we showed that DHEA-S and inactive androgen glucuronide end products progressively increased with higher GS. For instance, high circulating levels of inactive AST-G, which is the major DHT metabolite found in circulation, is observed in patients with a GS ≥8. This suggests that proficient intracellular formation of 5α-reduced androgens and their glucuronides occurs in these patients, the latter being proposed to better reflect androgen exposure (27). Intraprostatic androgen measurements of tumor cells and their microenvironment will be required to establish whether circulating glucuronide derivatives reflect increased local androgen exposure (27, 28) and/or the increased androgen inactivation capacity of cancer cells previously reported in aggressive disease (9, 29, 30).

Beyond few associations observed with prognostic factors, our data expose an added prognostic value particularly for the adrenal precursors A5diol and DHEA-S on progression in this clinical context. For instance, increment in A5diol levels was associated with an independent additional risk of progression. This association is clinically apparent in high-risk disease (P_interaction = 0.022; Fig. 3). This observation is of biologic significance, as A5diol, a major metabolite of DHEA in prostate homogenates (31, 32) has been shown to have dual actions in cancer cells binding both the androgen and estrogen receptors (33, 34). We also observed a trend for an association between A5diol levels and the development of metastatic disease. Besides, the commonly used drug, bicalutamide, fails to block A5diol androgen receptor transactivation in prostate cancer cells in vitro (33), and A5diol levels in the prostate are unaffected (35, 36) or only partially reduced (32) by androgen-deprivation therapy. Therefore, it is plausible that this adrenal precursor contributes to progression after prostatectomy, supports development of metastasis, and thus progression to the lethal castration-resistant state (37). The ongoing follow-up of the PROCURE cohort will help clarify this possibility. Results obtained with DHEA-S suggest a tendency toward an increased risk of progression with higher preoperative levels (Fig. 3A and C). DHEA-S, present in the micromolar range, is the most abundant circulating steroid in men and acts as a main steroid reservoir for cancer cells (38, 39). Indeed, tumor cells are capable of converting DHEA-S to active steroids following the successive action of multiple enzymes including steroid sulfatase, 3β-hydroxysteroid dehydrogenase, 17β-hydroxysteroid dehydrogenase, and 5α-reductase enzymes. The associations of DHEA-S with GS, and the trend observed with progression of high-risk disease in multivariable models, suggest that peripheral sources of steroids originating from the adrenals may contribute to progression in those men (39). Because A5diol and DHEA-S originate essentially from the adrenals, it is unlikely that the variations observed are the consequence of altered steroid metabolism associated with disease aggressiveness, perhaps except for testosterone, as previously suggested (3), and the inactive androgen glucuronides known to be formed in peripheral tissues including the prostate (40).

The strengths of this study include a large and multi-institutional prospective cohort of patients, with pathologically confirmed localized prostate cancer, all of whom were treated by radical prostatectomy and the ongoing clinical follow-up. In addition, the measurement of steroid hormones was performed using sensitive, specific, and validated MS assays. To our knowledge, this study is one of the largest studies to assess a comprehensive relationship between circulating steroid hormones and progression in men with localized prostate cancer. Limitations include the exploratory nature of the study and the limited number of metastatic events. Longer follow-up periods will be required to more adequately address the associations with metastatic disease.

In this large prospective cohort of men with localized prostate cancer, we showed that the steroid metabolome is associated with prognostic factors. The importance of the steroid metabolome is clinically apparent in high-risk disease where the circulating pool of adrenal precursors may support disease progression, and may help further stratify this population of patients. Further research is necessary to confirm the associations observed between A5diol, DHEA-S, and clinical outcomes. Indeed, large studies are required in men with localized prostate cancer, especially in those at high risk of recurrence, where circulating levels of these hormones might (i) improve prognostication/stratification, (ii) help refine the circulating hormonal milieu associated with adverse clinical outcomes, and (iii) advance hormonal strategies of patients with progressive disease. On the basis of the natural history of prostate cancer progression following biochemical recurrence (41), and knowing that intracellular levels of A5diol and DHEA-S remain at significant levels in patients on androgen-deprivation therapy (32, 35, 36), the ongoing follow-up of the PROCURE cohort will ultimately allow, in an approximately 5–10-year timeframe horizon, to assess the impact of these hormones on the development of metastasis, castration-resistant disease, and prostate cancer mortality.

E. Lévesque reports receiving a commercial research grant from Janssen Inc. and Astellas Inc., and has received speakers bureau honoraria from and has provided expert testimony for Pfizer Inc. M. Carmel reports receiving a commercial research grant from Abbvie and Janssen and is consultant/advisory board member for Bayer. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E. Lévesque, C. Guillemette

Development of methodology: C. Guillemette

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): P. Caron, L. Lacombe, V. Turcotte, Y. Fradet, A. Aprikian, F. Saad, M. Carmel, S. Chevalier, C. Guillemette

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E. Lévesque, L. Lacombe, D. Simonyan, F. Saad, C. Guillemette

Writing, review, and/or revision of the manuscript: E. Lévesque, L. Lacombe, D. Simonyan, Y. Fradet, A. Aprikian, F. Saad, S. Chevalier, C. Guillemette

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Lévesque, L. Lacombe, V. Turcotte, Y. Fradet, A. Aprikian, C. Guillemette

Study supervision: E. Lévesque, C. Guillemette

The authors are thankful to all participating patients and staff at each site associated with the PROCURE Biobank who have made this scientific contribution possible. The personnel involved in the Biobank at each site are the employees of their Centre or Research Institute and are not PROCURE employees. This work was supported by research grants from the Fonds de Recherche du Québec-Santé (FRQ-S) Innovation fund to the CHU Research Centre (grant no. 26678, to C. Guillemette, E. Lévesque, L. Lacombe, Y. Fradet) and Prostate Cancer Canada (grant no. DS2013-55, to E. Lévesque), the Cancer Research Society (grant no. 16181, to C. Guillemette), and the Canada Research Chair Program (grant no. 950-203962, to C. Guillemette). E. Lévesque holds a CIHR Clinician-Scientist Award. C. Guillemette holds the Canada Research Chair in Pharmacogenomics (Tier I). Results are based on samples and patient data obtained from the PROCURE Biobank, supported by donations in a partnership with the Cancer Research Society of Canada.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.