Abstract: 236 words
Manuscript text: 2810 words
Endogenous and exogenous testosterone and the risk of prostate cancer and
increased prostate specific antigen (PSA): a meta-analysis
Boyle P.
1,2,7
, Koechlin A.
1,2
, Bota M.
1,2
, d’Onofrio A.
2
, Zaridze D.G.
3
, Perrin P.
4
, Fitzpatrick
J.
5,*
, Burnett A.L.
6
and Boniol M.
1,2
1. Strathclyde Institute of Global Public Health at iPRI, Espace Européen d’Ecully, Bâtiment G,
Allée Claude Debussy, 69130 Ecully Lyon ouest, France
2. International Prevention Research Institute, 95 cours Lafayette, 69006 Lyon, France
3. Director, Institute of Carcinogenesis, Kashiroe Sh. 24, Moscow 115478, Russian Federation
4. Urologie, Centre Hospitalier Lyon Sud, 69495 Pierre-Bénite, France
5. Irish Cancer Society, 43/45 Northumberland Rd, Dublin 4, Ireland
6. Patrick C. Walsh Distinguished Professor of Urology, Cellular and Molecular Medicine, Johns
Hopkins Medicine, 600 North Wolfe Street, Marburg 407, Baltimore, Maryland 21287, United
States of America
7. Address for correspondence: Professor Peter Boyle, Strathclyde Institute of Public Health at
iPRI, Espace Européen d’Ecully, Bâtiment G, Allée Claude Debussy, 69130 Ecully Lyon
ouest, France. Telephone +33 (0)4 72 17 11 84; email peter.boyle@strath.ac.uk
*Tragically, Professor John Fitzpatrick passed away between the completion of the full
Technical Report and the completion of the manuscript.
Abstract
Objective: To review and quantify the association between endogenous and exogenous
testosterone and prostate specific antigen (PSA) and prostate cancer.
Methods: Literature searches were performed following the PRISMA guidelines. Prospective
cohort studies that reported data on the associations between endogenous testosterone and
prostate cancer, and placebo controlled randomised trials of testosterone replacement
therapy (TRT) that reported data on PSA and/or prostate cancer cases were retained. Meta-
analyses were performed using random-effects models with tests for publication bias and
heterogeneity.
Results: Twenty estimates were included in a meta-analysis which produced a summary
relative risk of prostate cancer for an increase of 5 nmol/L of testosterone of 0.99 (95% CI
(0.96, 1.02)) without heterogeneity (I² = 0%). Based on 26 trials, the overall difference in PSA
levels following onset of use of TRT was 0.10 ng/mL (-0.28, 0.48). Results were similar when
conducting heterogeneity analyses by mode of administration, region, age at baseline,
baseline testosterone, trial duration, type of patients and type of testosterone replacement
therapy. The summary relative risk of prostate cancer as an adverse effect from 11 TRT trials
was 0.87 (0.30; 2.50). Results were consistent across studies.
Conclusions: Prostate cancer appears to be unrelated to endogenous testosterone levels.
Testosterone replacement therapy for symptomatic hypogonadism does not appear to
increase PSA levels nor the risk of prostate cancer development. The current data are
reassuring although some care is essential until multiple studies with longer follow-up are
available.
Keywords: prostate cancer, testosterone, PSA, meta-analysis
Introduction
Testosterone is important for normal growth, development and maintenance of the prostate
gland. Testosterone deficiency in aging men has become a topic of increasing interest and
debate worldwide. Cross-sectional and longitudinal data indicate that testosterone levels are
reduced progressively with age and that a significant percentage of men aged over 60 years
have serum testosterone levels that are below the lower limits of young adult men aged 20–
30 years [1-3]. Late onset hypogonadism (LOH) is a clinical and biochemical syndrome
associated with advancing age and characterized by a deficiency in serum testosterone
levels, among other signs and symptoms [4, 5]. Late onset hypogonadism may result in
significant detriment to quality of life and adversely affect the function of multiple organ
systems. Therefore, there has been a growing awareness of the potential health benefits of
testosterone therapy for men with testosterone deficiency, including improved sexual desire
and performance, improved mood, increased muscle mass and strength, decreased fat mass
and improved bone mineral density [6].
More than 60 years ago, Huggins demonstrated that suppression of testosterone levels
caused regression of prostate cancer, and it is now commonplace for men with metastatic
prostate cancer to undergo treatment designed to lower testosterone levels [7]. Many
urologists are concerned that testosterone replacement therapy may accelerate prostate
growth not only in benign disease but also in cancer. If lowering testosterone causes prostate
cancer to regress in men, does elevating testosterone cause prostate cancer to appear?
Prostate cancer is the second most common cancer in men and a high proportion of men
harbour microscopic foci of prostate cancer [8]. It has been hypothesized that sex hormones
and androgens in particular might be involved in prostate cancer carcinogenesis. Zaridze and
co-workers concluded from qualitative analysis that there was little, if any, association
between serum levels of testosterone and the risk of prostate cancer [9]. A collaborative
analysis of 18 prospective studies on endogenous sex hormones and prostate cancer
conducted in 2008 found no associations between the risk of prostate cancer and serum
concentrations of testosterone, or other sex hormones [10].
Additionally, an increase in
exogenous testosterone replacement therapies has been encouraged by media attention and
marketing of new transdermal formulations (patches or gels) for the treatment of LOH.
Several systematic reviews discussed about the benefits and risks of testosterone treatment
[11-13], and several meta-analyses tried to quantify these risks [14-16].
This review aims at evaluating the potential relationship between testosterone and prostate
cancer, and to synthesise all available data regarding the impact of testosterone replacement
therapy (TRT) on changes in Prostate Specific Antigen (PSA) levels and on the risk of
development of prostate cancer. Two approaches have been employed, separately for
endogenous and exogenous testosterone. A meta-analysis of long-term observational data
has been retained for quantifying the relationship between endogenous levels of testosterone
and prostate cancer. A meta-analysis of short-term placebo-controlled randomised trials of
testosterone supplementation and their adverse effects was done in order to test the
hypothesis of a rapid acceleration of prostatic cancer development with the use of
exogenous testosterone.
Methods
A systematic literature search and quantitative analysis was planned, conducted and
reported following PRISMA guidelines [17]. Published reports were obtained from the
Cochrane library and PUBMED. Other sources were found in the reference lists of the
retrieved articles and preceding reviews on the topic. There was no restriction on
geographical location of studies, but only articles published in English were considered.
Articles were first screened by title and abstract. Then, full copies of the potentially relevant
articles were retrieved and read by at least two co-authors. Data were then extracted in a
predefined database by one author, and double-checked by the statistician who performed
the analyses.
Endogenous testosterone
The Endogenous Hormones and Prostate Cancer Collaborative Group (EHPCCG) published
a pooled analysis of 18 prospective studies [10]. Original articles from each study identified
by the EHPCCG, including those not included in the pooled analysis, were retrieved. Then, a
literature search was performed in May 2015 for the period 2007-2015 in order to collect
articles published after the EHPCCG pooled analysis.
Only prospective studies including cohort studies, nested case-control studies, case-cohort
studies and control arms of randomized controlled trials were eligible. Articles had to report
risk estimates for prostate cancer according to baseline serum testosterone levels. The list of
keywords for literature search is reported in supplementary figure (Figure S1). A dose-
response meta-analysis was conducted for the calculation of the summary relative risk (SRR)
corresponding to an increase of 5 nmol/L of baseline testosterone [18]. A change of 5nmol/L
corresponds broadly to the average difference between tertiles of population in the studies
included in the present meta-analysis. Heterogeneity analyses were performed in order to
investigate potential sources of variation between studies: by region (USA, Europe), age at
baseline (cut-off 65 years) and follow-up duration (cut-off 10 years).
The following variables were extracted: first author name and year of publication, country,
baseline age, baseline testosterone (nmol/L), follow-up duration and for each quantile of
testosterone, mean or median testosterone (nmol/L), number of subjects, person-years,
relative risk estimate and 95% confidence interval (CI).
Exogenous testosterone
The meta-analyses focused on the potential associations between use of TRT and 1)
changes in PSA level and 2) incidence of prostate cancer. Placebo-controlled, randomized
trials reporting statistics on these two outcomes were eligible for the meta-analysis. The
included outcome for change in PSA level could have been PSA level at baseline and end of
study or change in PSA level during the study. For incidence of prostate cancer, it could have
been number of prostate cancer occurring in each treatment arm, incidence rate, or odds-
ratio comparing treatment arm to control group.
The following variables were extracted: first author name and year of publication, country,
baseline age, baseline testosterone (nmol/L), follow-up duration, type of subjects (healthy or
other), drug name and dose, mode of administration, number of participants by group, PSA
level related data and number of prostate cancer cases by group.
Increase in PSA level
Data were abstracted on change in PSA level and its corresponding standard deviation. For
studies that reported only baseline and end of study PSA with corresponding standard
deviation, the variance of the average difference was estimated with the variance of PSA
baseline and PSA end of study using a correlation factor to account for auto-correlation. This
correction factor was calculated using data from the Prostate, Lung, Colorectal and Ovarian
(PLCO) cancer screening trial [19]. Detailed information about this correction factor are
available in the Supplementary methods.
Heterogeneity analyses were undertaken by region (USA, Europe), mode of administration
(transdermal, intramuscular, oral), type of patients (healthy subjects, subjects with pre-
existing comorbidities), age at baseline (cut-off 65 years), trial duration (cut-off one year) and
baseline testosterone (cut-off 11 nmol/L). This cut-off was chosen because it was the 5
th
percentile of the serum testosterone distribution of healthy men aged 35-44 [20]. An analysis
by dose of testosterone could not be undertaken because of differences in modes and
frequencies of administrations.
Prostate cancer
Following Cochrane review guidelines [21], studies reporting zero cases in both intervention
and control arms were excluded. Since these studies deal with rare events, odds ratios and
corresponding 95% CI were calculated applying a correction of 0.5 to all entries. The same
heterogeneity analyses as for the TRT and PSA were conducted.
Meta-analysis
Meta-analyses were undertaken with restricted maximum likelihood (REML) random effects
models [22]. The 95% confidence intervals of summary estimates were calculated using a t-
distribution.
To summarize incidence of prostate cancer from TRT trials, because of very low event rates,
the Mantel-Haenszel method was used [23, 24].
Heterogeneity across studies was evaluated by the Cochran’s Q statistic and with Higgins’ I²
[25]. Publication bias was investigated using the funnel plot and Macaskill test [26], Begg test
[27],
and Egger test [28]. The meta-analyses were carried out in programming language R
(version 3.0.0, GNU General Public License, 2013) and package metafor [29]. In addition to
the heterogeneity analyses described above, sensitivity analyses, excluding one study at a
time, were performed in order to evaluate the influence of individual studies on the overall
result (leave-one-out). In sensitivity analysis, the association between endogenous
testosterone and prostate cancer was also computed as a high versus low meta-analysis
comparing the risk in the highest quantile of PSA to the lowest quantile of PSA.
Results
