Associations of Serum Sex Steroid Hormone and 5α-Androstane-3α,17β-Diol Glucuronide Concentrations with Prostate Cancer Risk Among Men Treated with Finasteride

There is general consensus that normal, physiologic variability in blood concentrations of sex steroid hormones is not associated with the risk of prostate cancer (1). However, both finasteride, a type II steroid 5α-reductase inhibitor, and dutasteride, a dual type I and type II steroid 5α-reductase inhibitor, dramatically change the intraprostatic hormonal milieu. By inhibiting intraprostatic conversion of testosterone (T) to dihydrotestosterone (DHT), these drugs substantially lower the concentration of intraprostatic DHT (2) and modestly increase the concentrations of blood T (3); whether estradiol (E₂) and estrone (E₁) are affected remains uncertain. In the Prostate Cancer Prevention Trial (PCPT), which tested whether finasteride could prevent prostate cancer, finasteride reduced the risk of total prostate cancer by 25% but also increased the risk of high-grade cancer (4). This paradoxical finding remains unexplained. It could be attributable to increasing the detection of high-grade cancer, as finasteride improves the sensitivities of screening tests [digital rectal examination (DRE)] (5) and prostate-specific antigen (PSA; ref. 6) and diagnostic biopsies (7); however, it is also possible that low intraprostatic DHT provides a growth advantage for aggressive tumors (8).

Here, we investigate whether, among men who were compliant with finasteride treatment, the pre- and posttreatment concentrations of steroid hormones and their treatment-associated changes are associated with prostate cancer risk. We use serum 5α-androstane-3α,17β-diol glucuronide (3α-dG), a distal metabolite of DHT, as a surrogate measure of intraprostatic DHT (9, 10), because direct assay of prostate tissue was not feasible. Our primary hypothesis is that among men treated with finasteride, low posttreatment concentration of 3α-dG, reflecting a larger reduction in intraprostatic DHT, would be associated with lower overall prostate cancer risk. Secondarily, given the association of finasteride with high-grade disease, we also consider whether low posttreatment 3α-dG was associated with an increased risk of high-grade disease. In more exploratory analyses, we examine whether the pre- and posttreatment concentrations of T, E₁, and E₂ were associated with prostate cancer risk. Findings from this study could make it feasible to identify men who would maximally benefit from the use of finasteride for cancer prevention, and could provide insight into the etiology of the increased risk of high-grade cancer among men treated with finasteride.

Data are from the PCPT, a randomized, placebo-controlled trial that tested whether finasteride, a steroid 5α-reductase type II inhibitor, could reduce the 7-year period prevalence of prostate cancer. Details about study design and participant characteristics have been described previously (4). Briefly, 18,880 men age 55 years and older with normal DRE and PSA levels of 3 ng/mL or below, as well as no history of prostate cancer, severe lower urinary tract symptoms or clinically significant coexisting conditions, were randomized to receive finasteride (5 mg/d) or placebo. During the PCPT, men underwent DRE and PSA determinations annually, and a prostate biopsy was recommended for participants with an abnormal DRE or if a PSA adjusted for the effect of finasteride was 4.0 ng/mL or greater. At the final study visit at year 7, all men not previously diagnosed with prostate cancer were requested to undergo an end-of-study prostate biopsy. All biopsies consisted of a minimum of 6 cores collected under transrectal ultrasonographic guidance and were reviewed for adenocarcinoma by both the pathologist at the local study site and a central pathology laboratory with concordance achieved in all cases. Clinical stage was assigned locally and tumors were graded centrally using the Gleason scoring system. All men gave informed consent and study procedures were approved by Institutional Review Boards at each study center, the Southwest Oncology Group (SWOG, San Antonio, TX), and the Fred Hutchinson Cancer Research Center (Seattle, WA).

The study reported here is part of a large nested case–control study designed to examine multiple hypotheses about prostate cancer biology and risk (11). Cancer cases and controls in this report were from the finasteride treated study arm. Cases (n = 676) had biopsy-confirmed cancer identified before study unblinding, and blood samples both at baseline and before cancer diagnosis. Controls (n = 759) were disease-free at the end-of-study biopsy and had both baseline and follow-up blood samples. Controls were frequency matched to cases on distributions of age (±5 years) and having a first-degree relative with prostate cancer, and included all non-Whites. There were more controls than cases because men diagnosed with cancer in the first 2 years, or before a follow-up blood was collected, were excluded. Men who were not compliant with the study treatment, defined as either (i) reporting not using the drug at the time of the posttreatment serum collection (n = 155) or (ii) having a posttreatment finasteride blood concentration of zero (n = 33), were excluded, leaving 553 cases and 694 control participants in the study.

Information on age, race, diabetes status, family history of prostate cancer in first-degree relatives, and history of smoking were collected at baseline using self-administered questionnaires. Participants' height and weight were measured at baseline, and body mass index (BMI) was calculated as weight (kg)/height (m²).

Nonfasting blood was collected approximately 3 months before randomization and annually thereafter until diagnosis or the study end. Venous blood was drawn into glass collection tubes without anticoagulant, refrigerated, and shipped to a central repository where they were centrifuged, aliquoted, and stored at −70°C. Concentrations of T, 3α-dG, E₁, E₂, and sex hormone binding globulin (SHBG) were measured at baseline and at year 3 postbaseline. For the approximately 5% of men missing a year 3 blood sample, the sample closest in time was used (range years 1–7).

Total T, 3α-dG, E₁, E₂, and SHBG were quantified in serum by highly specific immunoassays at the Reproductive Endocrine Research Laboratory, University of Southern California Keck School of Medicine (F.Z. Stanczyk). Total T and SHBG were measured by a direct solid-phase, competitive chemiluminescent enzyme immunoassay and a direct solid-phase, 2-site chemiluminescent immunometric assay, respectively, using the Immulite 2000 analyzer (Siemens Healthcare Diagnostics, Deerfield, IL). The sensitivity of the T assay is 20 ng/dL and that of the SHBG assay is 1 nmol/L. The interassay CVs for T were 11.9%, 7.6%, and 9.1% at concentrations of 124, 539, and 1,058 ng/dL, respectively, and for SHBG were 5.2%, 5.2%, and 6.6% at 21, 63, and 80 nmol/L, respectively. 3α-dG was measured manually by direct competitive radioimmunoassay, using kits obtained from Beckman-Coulter). This assay was validated extensively and measures the predominant form of 3α-dG, which contains the glucuronide at carbon 17 instead of carbon 3 (12). The 3α-dG assay sensitivity is 0.5 ng/mL and the interassay CVs were 2.7% and 9.0% at concentrations of 4.5 and 6.4 ng/mL, respectively. Estrogens were measured by radioimmunoassay after organic solvent extraction and Celite column partition chromatography, as described previously (13). The E₁ and E₂ assay sensitivities are 2 and 4 pg/mL, respectively. The interassay CVs were 11%, 12%, and 9% at concentrations of 24, 61, and 159 pg/mL, respectively, for E₁, and 10% at concentrations of 22, 66, and 183 pg/ml for E₂.

Free and bioavailable (non-SHBG-bound) T and E₂ were calculated using a validated method (14) based on measured total T and E₂ levels, respectively, and SHBG, assuming an average concentration for albumin (15, 16). This method has been found to have high validity (14). Assays were not successful for small numbers of samples, and thus the numbers with baseline, follow-up, and change measures differs slightly for each analyte. Quality control (QC) samples from pooled serum from healthy volunteers, split into 6 pools, were also included in each analytical batch. Between 1 and 6 samples from the same pool were placed randomly within each box of samples. QC data were monitored regularly and laboratory personnel were blinded to sample type. Coefficients of variation for 3α-dG, T, E₁, E₂, and SHBG were 14.0%, 10.5%, 15.2%, 14.9%, and 12.2%, respectively.

Paired t tests were used to test whether the absolute or percentage differences in 3α-dG, T (total and free), E₁, E₂ (total and free), and SHBG concentrations between baseline and posttreatment were significantly different from zero. Baseline-adjusted change was calculated from a linear regression using change as the dependent variable and baseline as the independent variable; the residual from this model was added to the population mean change. Spearman rank-order correlations were used to assess the associations among treatment-associated changes in all measures, and were computed both unadjusted and, using residuals from linear regression models, adjusted for SHBG. Unconditional and polytomous logistic regression analyses were used to estimate ORs and 95% confidence intervals (CI) for the associations of steroid concentrations, and changes in these concentrations, with total, low- and high-grade cancer. Each measure was categorized into quartiles based on distributions in controls. Models were adjusted for SHBG, as well as frequency matching variables and variables associated with prostate cancer risk in this cohort, including age (continuous), race (Caucasian, other), family history of prostate cancer in first-degree relatives (yes, no), body mass index (continuous), and serum cholesterol. Diabetes, although associated with cancer risk, could not be included in stratified models because of its low frequency; however, findings did not change when men with diabetes were excluded. Low-grade cancer was defined as Gleason Score 2 to 6, and high-grade cancer was defined as Gleason Score 7 to 10 and, to capture a more rare but more phenotypically uniform group of highly aggressive cancers (17) as Gleason Score 8 to 10. There were no significant differences between findings for Gleason 7 to 10 and 8 to 10 cancers, in part because of the relatively small number of Gleason 8 to 10 cancers, and the results for both are described together in the text as “high-grade” cancer. All statistical tests were 2-sided, with a statistical significance level set at P = 0.05. All statistical analyses were carried out using SAS statistical software (version 9.2; SAS Corporation).

Table 1 gives baseline demographic and health-related characteristics in case and control participants. Cases and controls did not differ significantly by BMI, smoking status, alcohol consumption, or physical activity, nor did they differ by age or family history of prostate cancer because of matching. Compared with cases, PSA levels and years of education were significantly lower in controls. The much higher proportion of non-White men in the control group was because of purposeful oversampling.

Table 2 gives mean and median levels of serum androgens, estrogens, and SHBG at baseline and follow-up, along with the absolute, percentage, and baseline-adjusted differences between these 2 time points. For all measures, the changes between baseline and follow-up were statistically significant (P < 0.001). As expected, the largest change was a mean 74% reduction in serum 3α-dG. There were small increases in median concentrations of serum T, SHBG, E₁, and E₂, which ranged from 6.0% to 11.2% and were attenuated for free compared with total T and E₂.

Table 3 gives the correlations among changes in 3α-dG, serum steroids, and SHBG. Changes in 3α-dG were not correlated with changes in other steroids or SHBG. Changes in T were moderately and positively correlated with changes in E₁, E₂, and SHBG, and changes in E₁ and E₂ were strongly correlated with each other. Correlations were similar when adjusted for SHBG (data not shown).

In this subset of men in the PCPT who were finasteride-compliant approximately 3 years postrandomization, neither absolute, percentage, nor baseline-adjusted changes in sex steroids or 3α-dG were associated with risks of total, low-, or high-grade cancer (data not shown).

Table 4 gives the covariate-adjusted associations of pretreatment steroid concentrations with prostate cancer risk. Compared with men in the lowest quartile of T, those in the highest quartile had a 36% (95% CI, 57%–7%) reduced risk of total cancer. This association was similar for low- and high-grade cancer but was not linear; reduced risks were limited to those in the highest quartile only. The association of free T with cancer risk was substantially attenuated and not significant. There was also a 38% (−1%−93%) increased risk of cancer, comparing men in the highest to lowest quartiles of E₁. The association was limited to low-grade cases.

Table 5 gives the covariate-adjusted associations of posttreatment steroid and SHBG concentrations with prostate cancer risk. Neither T, free T, nor 3α-dG was associated with the risk of total, low-, or high-grade cancer. Concentrations of E₁, E₂, and free E₂ were positively associated with cancer risk: comparing the fourth to first quartiles (Q4 vs. Q1) risks were increased by 54% (9%–117%), 49% (7%–107%), and 34% (−4%–87%), respectively. For all associations, trends were significant (all P_trend < 0.03) and similar for low- and high-grade disease.

In additional analyses not shown, the results for pre- and posttreatment T, free T, SHBG, and 3α-dG did not change when models were further adjusted for other steroids. Results for pre- and posttreatment E₁, E₂, and free E₂ did not change when additionally adjusted for T and 3α-dG. Associations of pretreatment E₁ and E₂ with cancer risk did not change when mutually adjusted for each other, but posttreatment associations for both were attenuated and no longer significant.

In this nested case–control study among treatment-compliant men in the PCPT, there was a 74% reduction in serum 3α-dG after approximately 3 years of finasteride treatment. Neither pretreatment, posttreatment, nor the magnitude of change in 3α-dG were associated with the risk of total or Gleason Score 2 to 6, 7 to 10, or 8 to 10 cancer. Finasteride treatment also modestly increased serum T, free T, E₁, E₂, and free E_2. Pretreatment, men in the highest quartiles of T and E₁ had a 36% lower and a 38% higher risk of prostate cancer, respectively. Posttreatment, men in the highest quartiles of E₁, E₂, and free E₂ had 54%, 47%, and 34% increased risks of prostate cancer, respectively. The magnitudes of change in steroid concentrations were not associated with cancer risk.

The lack of association between posttreatment 3α-dG and cancer risk was unexpected. We had hypothesized that larger decreases in 3α-dG, reflecting larger reductions in intraprostatic DHT after treatment, would indicate response to finasteride treatment and thereby be associated with larger reductions in prostate cancer risk. This finding suggests that, at least among treatment-compliant men, the concentration of intraprostatic DHT was reduced below a threshold, beyond which its further reduction did not affect cancer risk.

One of the main controversies about the use of steroid 5-α reductase inhibitors for the primary prevention of prostate cancer is whether or not the observed increased risk of high-grade cancer after treatment in 2 clinical trials was causal or because of diagnostic bias (4, 18). Some investigators have hypothesized that reduced intraprostatic DHT suppresses growth of androgen-dependent cancer clones, which allows the preferential growth of androgen-independent, aggressive cancers (19). In contrast, several studies have suggested that high-grade cancers are more easily detected in finasteride-treated men, because their prostates are reduced in size and a larger proportion of the gland is sampled during biopsy (20). If low intraprostatic DHT causes increased risk of high-grade disease, we would expect that men with the lowest posttreatment 3α-dG concentration would have a greater risk of high-grade disease. In contrast, in this study there was a nonsignificant but large 81% (−9% to 260%, P_trend = 0.04) increased risk of Gleason Score 8 to 10 cancer among men in the highest quartile of 3α-dG. Thus, although this study does not address directly whether or not the association of 5-α reductase inhibitors with high-grade cancer is causal, it does not support the hypothesis that the reduction in DHT after treatment allows the preferential growth of high-grade disease.

The association of high pretreatment T with decreased risk is somewhat inconsistent with study findings overall and difficult to interpret. This association was limited to men in the highest quartile and not consistent with the lack of an association for free T. Furthermore, because neither posttreatment T nor free T were associated with risk, we judge this likely to be a chance finding. The association of high pretreatment E₁ with increased cancer risk is also difficult to interpret. Pretreatment, estrogen associations were limited to E₁ with low-grade disease, compared with significant associations of posttreatment E₁ and E₂ with both low- and high-grade disease. Also, the pretreatment estrogen findings given here on treatment-compliant men differed somewhat from those previously published from a larger sample of PCPT participants that did not exclude noncompliant men (21), in which both E₁ and E₂ were associated with increased risk of low-grade disease. Further research will be needed to clarify these findings.

The associations of high posttreatment estrogens with increased cancer risk are noteworthy because trends were significant and the ORs were similar for low- and high-grade disease. The mechanism underlying these associations is unclear. It is possible that estrogens influence prostate cancer risk when intraprostatic DHT is pharmacologically reduced, or the association could be indirect; for example, genetic and/or environmental characteristics that increase estrogen levels (e.g., through increased aromatase activity) may in some way modify the response to finasteride. Unfortunately, this nested case–control study does not allow us to directly measure the effect of finasteride in men with low and high posttreatment estrogen concentrations, because controls were matched on treatment arm. Overall, these findings on posttreatment estrogens support a new area of research to investigate the effects of estrogens on prostate tissues in a low androgen environment.

Strengths of this study include the use of highly sensitive and specific assays for serum steroids, the large sample size, the use of prostate biopsy to verify absence or presence of cancer, and the exclusion of men not compliant with finasteride treatment. One important weakness of this study is our assumption that the reduction in 3α-dG after finasteride treatment accurately reflects the reduction in intraprostatic DHT. It is also uncertain whether, when measured 3 years after the initiation of finasteride treatment, the associations of steroids with cancer risk are the same as those that would be observed if steroids were measured at other times after treatment initiation. The reduction in serum DHT after finasteride treatment is roughly 90% 1 year posttreatment and does not change thereafter; however, PSA falls by approximately 45% at 1 year to a maximum of 60% at 3 years posttreatment (22). Thus, it is not entirely clear whether the posttreatment steroid concentrations used in this study precisely reflect changes that occurred soon after treatment initiation. We also relied upon self-report and/or a single blood finasteride concentration to determine treatment adherence. Nevertheless, when nonadherent men were included in the analysis there was a significant positive association of posttreatment 3α-dG with cancer risk, suggesting that this approach was valid. Additional weaknesses include the small number of men with high-grade disease, and the inability to directly calculate the conditional effects of finasteride on posttreatment estrogen concentrations.

In conclusion, we found no support for the hypothesis that lower 3α-dG after finasteride treatment, reflecting a larger reduction in intraprostatic DHT, is associated with lower risks of total or high-grade prostate cancer. There was some evidence that high pretreatment T and E₁ concentrations predict reduced and increased prostate cancer risk, respectively, but we consider this weak. There was stronger and more consistent evidence that high posttreatment concentrations of E₁, E₂, and free E₂ are associated with increased low- and high-grade prostate cancer risk. It is possible that estrogens play a significant role in prostate cancer risk only when intraprostatic DHT is lowered pharmacologically. Further research is needed to evaluate whether low posttreatment serum estrogens could be used to identify men most likely to benefit from finasteride for prostate cancer prevention.

No potential conflicts of interest were disclosed.

Conception and design: A. R. Kristal, C. M. Tangen, P. J. Goodman, M. L. Neuhouser, I. M. Thompson, Jr., J. Reichardt, A. M. Hoque, W. D. Figg, A. W. Hsing

Development of methodology: A. R. Kristal, P. J. Goodman, A. W. Hsing

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A. R. Kristal, P. J. Goodman, F. Z. Stanczyk, S. K. Patel, I. M. Thompson, Jr., A. van Bokhoven, A. W. Hsing

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A. R. Kristal, C. A. Till, C. M. Tangen, P. J. Goodman, M. L. Neuhouser, F. Z. Stanczyk, L. W. Chu, A. M. Hoque, S. M. Lippman, A. W. Hsing

Writing, review, and/or revision of the manuscript: A. R. Kristal, C. M. Tangen, P. J. Goodman, M. L. Neuhouser, F. Z. Stanczyk, L. W. Chu, I. M. Thompson, Jr., J. Reichardt, A. M. Hoque, E. A. Platz, W. D. Figg, A. van Bokhoven, S. M. Lippman, A. W. Hsing

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A. R. Kristal, P. J. Goodman, L. W. Chu, A. van Bokhoven, A. W. Hsing

Study supervision: A. R. Kristal, P. J. Goodman, I. M. Thompson, Jr., A. W. Hsing

The authors thank Dr. Ronald K. Ross, University of Southern California (deceased), who participated in the design of this study.

This work was funded by the following: P01-CA108964 from the National Cancer Institute; Intramural Research Program of the U.S. NIH, National Cancer Institute, Division of Cancer Epidemiology and Genetics; P30-CA054174, the Cancer Center Support Grant for the Cancer Therapy and Research Center at the University of Texas Health Science Center at San Antonio; and P30 CA015704-36, the Cancer Center Support Grant for the Seattle Cancer Consortium, Seattle, WA.

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.