Promoter Methylation in APC, RUNX3, and GSTP1 and
Mortality in Prostate Cancer Patients
Lorenzo Richiardi, Valentina Fiano, Loredana Vizzini, Laura De Marco, Luisa Delsedime, Olof Akre,
Anna Gillio Tos, and Franco Merletti
From the Cancer Epidemiology Unit,
Center for Experimental Research and
Medical Sciences and Center for Onco-
logic Prevention Piemonte, University
of Turin; Department of Pathology,
Molinette Hospital, Turin, Italy; and
Clinical Epidemiology Unit, Karolinska
Institutet, Stockholm, Sweden.
Submitted May 19, 2008; accepted
November 14, 2008; published online
ahead of print at www.jco.org on May
26, 2009.
Supported in part by the Italian Associa-
tion for Cancer Research; the Compag-
nia San Paolo/FIRMS; the Piedmont
Region; the Italian Ministry of Univer-
sity; and by the Master in Epidemiol-
ogy, University of Turin, and San Paolo
Foundation (L.V.).
Presented in part at the 98th Annual
Meeting of the American Association
for Cancer Research, April 14-18, 2007,
Los Angeles, CA; and the XXXI Annual
Meeting of the Italian Association for
Epidemiology, Marina di Ostuni (Brin-
disi), Italy, October 17-19, 2007.
Terms in blue are defined in the glos-
sary, found at the end of this article
and online at www.jco.org.
Authors’ disclosures of potential con-
flicts of interest and author contribu-
tions are found at the end of this
article.
Corresponding author: Lorenzo
Richiardi, MD, PhD, Cancer Epidemiology
Unit, University of Turin, Via Santena 7,
10126 Turin, Italy; e-mail: lorenzo
.richiardi@unito.it.
The Acknowledgment is included in
the full-text version of this article,
available online at www.jco.org.
It is not included in the PDF version
(via Adobe® Reader®).
© 2009 by American Society of Clinical
Oncology
0732-183X/09/2719-3161/$20.00
DOI: 10.1200/JCO.2008.18.2485
ABSTRACT
Purpose
There is a need to better understand prostate cancer progression and identify new prognostic
markers for this tumor. We investigated the association between promoter methylation in a priori
selected genes and survival in two independent large series of prostate cancer patients.
Methods
We followed up with two cohorts of patients (216 patients diagnosed in 1982 to 1988 and 243
patients diagnosed in 1993 to 1996) diagnosed at one hospital pathology ward in Turin, Italy. DNA
was obtained from paraffin-embedded tumor tissues and evaluated for promoter methylation
status in glutathione S-transferase (GSTP1), adenomatous polyposis coli (APC), and runt-related
transcription factor 3 (RUNX3).
Results
The two cohorts had different prevalences of methylation in APC (P ⫽ .047), GSTP1 (P ⫽ .002),
and RUNX3 (P ⬍ .001). Methylation in APC was associated with an increased risk of prostate
cancer–specific mortality (hazard ratio [HR] ⫽ 1.42; 95% CI, 0.98 to 2.07 in the 1980s cohort;
HR ⫽ 1.57; 95% CI, 0.95 to 2.62 in the 1990s cohort; HR ⫽ 1.49; 95% CI, 1.11 to 2.00 in the two
cohorts combined). In subgroup analyses, the HRs were higher among patients with a Gleason
score less than 8 (HR ⫽ 1.52; 95% CI, 0.85 to 2.73 in the 1980s cohort; HR ⫽ 2.09; 95% CI, 1.02
to 4.28 in the 1990s cohort). Methylation in RUNX3 was associated with prostate cancer mortality
only in the 1990s cohort, and methylation in GSTP1 did not predict mortality in either cohort.
Conclusion
The pattern of hypermethylation may have changed after the introduction of prostate-specific
antigen testing in the beginning of the 1990s. Promoter methylation in APC was identified as a
marker for prostate cancer progression.
J Clin Oncol 27:3161-3168. © 2009 by American Society of Clinical Oncology
INTRODUCTION
Prostate cancer is the most frequent cancer in the
United States and Western Europe.
1
The incidence
has been increasing by approximately 3% per year
during several decades.
2
The trends shifted in the
late 1980s and early 1990s when prostate-specific
antigen (PSA) testing became widespread.
3
In the
United States, for example, the annual percent
change in incidence was 2.4% before 1989 and
18.4% between 1989 and 1992.
4
The use of PSA testing remains under debate.
First, PSA testing has low sensitivity and positive
predictive value, implying high proportions of both
false-positive and false-negative tests.
5,6
Second,
PSA screening identifies indolent cancers, resulting
in overdiagnosis.
7
Patients diagnosed with localized
prostate cancers, notably those detected through
PSA, may have an excellent prognosis when left
untreated. For these patients, it would be impor-
tant to distinguish between indolent and aggres-
sive tumors. Several preoperative nomograms
developed in the last 10 years have been based on
clinicopathologic variables, including PSA, Gleason
score, clinical stage, and number of positive and
negative biopsy cores.
8,9
The collective prognostic
value of these factors is unsatisfactory, and better
understanding of the biology of prostate cancer pro-
gression is needed to identify new markers.
9,10
Emerging evidence indicates that epigenetic
alterations, particularly DNA hypermethylation,
play a role in human carcinogenesis and tumor
progression.
11,12
Several studies found that pres-
ence of CpG island (clusters of dinucleotides of a
cytosine and a guanosine) methylation in the
promoter of some cancer-related genes, such as
glutathione S-transferase (GSTP1), adenoma-
tous polyposis coli (APC), and PTGS2, may be
JOURNAL OF CLINICAL ONCOLOGY
ORIGINAL REPORT
VOLUME 27 䡠 NUMBER 19 䡠 JULY 1 2009
© 2009 by American Society of Clinical Oncology
3161
Downloaded from jco.ascopubs.org on April 3, 2012. For personal use only. No other uses without permission.
Copyright © 2009 American Society of Clinical Oncology. All rights reserved.
used as a diagnostic test to distinguish between normal and
prostate cancer tissue.
13,14
In addition, methylation in single
genes and methylation indices have been found to be associated
with clinicopathologic indicators of poor prognosis, although
there is inconsistency between studies.
15-23
The association be-
tween promoter hypermethylation in prostate cancer and clin-
ical outcome or mortality has been less investigated, and
investigation was performed mostly in relatively small patient
series with short follow-up using heterogeneous or intermedi-
ate outcomes.
15,18,24-27
We studied the prostate cancer survival in association with
promoter methylation in GSTP1, APC, and runt-related transcrip-
tion factor 3 (RUNX3). GSTP1 is the most frequently investigated
gene in prostate cancer epigenetics and has been found to be
frequently methylated in prostate tumor tissues in several stud-
ies. APC and RUNX3 were selected among genes (EDNRB,
COX2, PTGS2, APC, RASSF1, and RUNX3) for which aberrant
methylation status was reported to be associated with clinical fea-
tures of poor prognosis in prostate cancer patients at the time when
our study was designed.
28
Because they are involved in signaling
and transcription pathways, their inhibition by promoter methyl-
ation may plausibly have a role in prostate cancer progression.
28,29
In addition, their polymerase chain reaction (PCR) target se-
quences were short enough to be successfully investigated by
published primers in paraffin-embedded tissues preserved for sev-
eral years.
Two independent cohorts of, in total, 459 prostate cancer
patients were observed for prostate cancer mortality. The first
cohort included patients diagnosed in the 1980s, whereas members
of the second cohort were diagnosed in the 1990s. The second
cohort was used to validate the findings in the first cohort and to
study any possible changes in the methylation patterns between the
1980s, before the introduction of PSA testing, and the 1990s, dur-
ing the PSA era.
Table 1. Characteristics of the Two Cohorts of Prostate Cancer Patients After 14 Years of Follow-Up
Characteristic
1980s Cohort (1982-1988) 1990s Cohort (1993-1996)
P
ⴱ
No. of Patients % No. of Patients %
No. of patients 216 243
No. of person-years 1,040 1,591
Mortality
Overall 195 177
As a result of prostate cancer 121 76
As a result of other causes 74 101
Missing cause of death 8 0
Survival time, years
Median 3.1 6.3
Range 0-14 0-14
Age, years .003
Mean 72.3 70.0
Standard deviation 7.5 8.7
Residence .692
Turin 153 70.8 168 69.1
Turin hinterland 63 29.2 75 30.9
Source of tumor tissue ⬍ .001
Biopsy 182 84.3 164 67.5
TURP 11 5.1 45 18.5
Radical prostatectomy 23 10.6 34 14.0
Gleason score ⬍ .001
⬍ 7 32 14.8 136 56.0
7 85 39.4 34 14.0
8⫹ 99 45.8 73 30.0
Methylation in GSTP1 .002
Yes 159 76.1 150 62.8
No 50 23.9 89 37.2
Missing 7 — 4 —
Methylation in APC .047
Yes 76 35.9 106 45.1
No 136 64.1 129 54.9
Missing 4 — 8 —
Methylation in RUNX3 ⬍ .001
Yes 170 84.6 103 48.1
No 31 15.4 111 51.9
Missing 15 — 29 —
Abbreviation: TURP, transurethral resection of the prostate.
ⴱ
P value for difference between 1980s cohort and 1990s cohort.
Richiardi et al
3162 © 2009 by American Society of Clinical Oncology
J
OURNAL OF CLINICAL ONCOLOGY
Downloaded from jco.ascopubs.org on April 3, 2012. For personal use only. No other uses without permission.
Copyright © 2009 American Society of Clinical Oncology. All rights reserved.
METHODS
Cohorts
The cohorts consist of consecutive prostate cancer patients of any age
identified at a single pathology ward of the San Giovanni Battista Hospital,
the main hospital in the city of Turin (900,000 inhabitants), Italy. The first
cohort, hereafter referred to as the 1980s cohort, includes patients who
received a biopsy of the prostate, transurethral resection of the prostate, or
radical prostatectomy between 1982 and 1988. Patients in the second
cohort, the 1990s cohort, were diagnosed at the hospital between 1993
and 1996. The 1980s and 1990s cohorts included 298 and 280 eligible
patients, respectively. The study was approved by the local ethical commit-
tee.
Slices of formalin-fixed and paraffin-embedded tumor tissue (PETs)
were obtained from each patient. DNA extraction from the PETs was success-
ful in 77% (n ⫽ 228) of the patients in the 1980s cohort and 90% (n ⫽ 253) of
patients in the 1990s cohort. Patients with successful extraction remained for
further analysis.
From each patient’s pathology report, we obtained information on name
and surname, age, tumor grade, place of residence, and, limited to the 1990s
cohort, Gleason score. Three patients in the 1980s cohort and two patients in
the 1990s cohort with incorrect demographic informationwere excluded from
the study. Diagnostic slides for patients in the 1980s cohort were traced and
re-evaluated by a pathologist (L.D.), who assigned the Gleason score. We could
not trace the slides of eight tumors. In those cases, we used the information on
tumor grade that was available in the pathology report; well-differentiated
tumors were translated to a Gleason score of 6 or less, moderately differenti-
ated tumors corresponded to a score of 7, and poorly differentiated tumors
had a score of 8⫹.
Follow-Up
We observed the patients from the date of the pathology report to
February 13, 2006 for the 1980s cohort and to January 15, 2007 for the 1990s
cohort. Dead patients were censored on their date of death. Information on
vital status and copies of the death certificates came from the demographic
offices of Turin and the towns of the hinterland, and we ascertained migration
at the Migration Office. Follow-up information was 95% complete (nine
patients lost) for the 1980s cohort and 96% complete (eight patients lost) for
the 1990s cohort. Patients with no follow-up information were excluded from
the study. The death certificates for eight patients in the 1980s cohort were not
retrievable. These patients were excluded from the analyses focusing on pros-
tate cancer mortality.
Molecular Analyses
We extracted genomic DNA from 3 to 5 (10-
m thick) sequential
sections of PETs and checked for adequacy by PCR amplification of the

-globin gene.
30
If a patient had multiple blocks of PET, a block embedding
tissue with tumor cells was chosen after histologic review of the corresponding
slide. If a patient had both biopsy and prostatectomy blocks, we analyzed the
biopsy. For all patients retained in the study,

-globin PCR analysis resulted in
clearly sharp detectable amplicons after gel electrophoresis, allowing adequacy
for methylation-specific analyses.
The genomic DNA samples, including positive controls for methyl-
ated and unmethylated status, underwent bisulfite modification using
CpGenome DNA Modification Kit (Intergen Co, Purchase, NY) following
the manufacturer’s instructions.
31
Bisulfite-modified DNA was used as a
template for PCR amplification using primers specific for either the meth-
ylated or the modified unmethylated DNA. The sets of specific primers and
their annealing temperatures for methylated and unmethylated forms of
GSTP1, APC, and RUNX3 gene promoters were selected from published
sequences.
20,32
For PCR amplification, 4
L of bisulfite-modified DNA was
added in a final volume of 25
L PCR mix containing 1X PCR buffer (15
mmol/L Tris, pH 8.0; 50 mmol/L KCl; and 6.7 mmol/L MgCl
2
), deoxynucle-
otide triphosphates (2 mmol/L each), primers (0.4
mol/L each per reaction),
and 1.25 U of AmpliTaq Gold DNA polymerase (Applied Biosystems, Foster
City, CA). PCR conditions were as follows: 10 minutes at 95°C, 30 seconds at
95°C, 1 minute at primer-specific annealing temperature, 1 minute at 72°C for
45 cycles, and 7 minutes at 72°C.
20,32
All PCR amplifications were performed
in a Gene Amp PCR System 9700 Thermal Cycler (Applied Biosystems).
Bisulfite-modified CpGenome universal methylated DNA (Intergen Co) was
used as positive control for methylated alleles, and bisulfite-modified DNA
from normal human lymphocytes was used as a positive control for unmeth-
ylated alleles. Negative PCR controls without DNA were included in each PCR
run. Ten microliters of each PCR amplification were loaded onto 2% agarose
gel stained with ethidium bromide and visualized by ultraviolet transillumina-
tion. As shown in previous studies, this method has high sensitivity, detecting
one methylated nucleotide in 1,000 unmethylated nucleotides.
33
Statistical Analyses
We compared clinical and pathologic characteristics of the patients in the
1980s cohort and the 1990s cohort using univariate
2
and t tests.
34
Variables
were classified as reported in Table 1. No information on PSA variables, tumor
stage, and number of positive and negative biopsies was available in the
pathology reports.
Through logistic regression, we estimated, separately for the 1980s and
1990s cohorts, the prevalence odds ratios for the association between methyl-
ation in the genes and clinical and pathologic characteristics.
35
Patients with
missing methylation status in one gene were excluded from the correspond-
ing analyses.
For the older cohort, we ended follow-up after 14 years to apply the same
maximum follow-up for both cohorts. The effect of methylation status in each
of the three genes on cumulative mortality from prostate cancer was investi-
gated taking into account competing risks,
36
and differences in overall prostate
Table 2. Association Between Gleason Score and Prevalence of Methylation in GSTP1, APC, and RUNX3 in 1980s and 1990s Cohorts
Gleason
Score
GSTP1 APC RUNX3
Prevalence of
Methylation (%) POR
ⴱ
95% CI
Prevalence of
Methylation (%) POR
ⴱ
95% CI
Prevalence of
Methylation (%) POR
ⴱ
95% CI
1980s cohort
⬍ 7 65.5 1 — 25.8 1 — 89.3 1 —
7 82.1 2.27 0.81 to 6.34 40.5 1.66 0.64 to 4.31 81.0 0.40 0.10 to 1.70
8⫹ 74.0 1.18 0.45 to 3.13 35.1 1.25 0.48 to 3.25 86.2 0.58 0.13 to 2.53
1990s cohort
⬍ 7 57.0 1 — 38.5 1 — 44.7 1 —
7 62.5 1.21 0.53 to 2.77 48.5 1.39 0.63 to 3.08 48.4 1.21 0.53 to 2.76
8⫹ 73.6 2.02 1.06 to 3.84 55.6 2.07 1.13 to 3.81 53.6 1.36 0.73 to 2.52
Abbreviation: POR, prevalence odds ratio.
ⴱ
POR adjusted for age, source of tumor tissue, year of tissue collection, and residence.
Gene Methylation and Prostate Cancer Mortality
www.jco.org © 2009 by American Society of Clinical Oncology 3163
Downloaded from jco.ascopubs.org on April 3, 2012. For personal use only. No other uses without permission.
Copyright © 2009 American Society of Clinical Oncology. All rights reserved.
cancer mortality were evaluated with the Gray’s test.
37
Using age as the tem-
poral axis, we used Cox proportional hazards regression models to estimate
hazard ratios (HRs) with 95% CIs of prostate cancer mortality by methylation
status and Gleason score (two categories: ⬍ 8 and ⱖ 8). Patients were censored
at death from causes other than prostate cancer. Both a graphical check and
formal tests based on Schoenfeld residuals (P ⬎ .15) indicated that the pro-
portional hazard assumption was met. We included the following covariates in
the models: source of tumor tissue, Gleason score, and follow-up duration
(time-dependent variable in three categories: ⬍ 1 year, between 1 and 3 years,
and 3⫹ years). Inclusion of place of residence in the models did not affect HR
estimates. HRs specific for categories of follow-up duration were estimated
introducing terms for the interaction between this variable and methyl-
ation status.
We also investigated the interaction between Gleason score and methyl-
ation in APC and RUNX3. In exploratory analyses, we further studied the
association between survival and number of methylated genes. Patients with
missing methylation status in at least one of the genes were excluded from this
analysis (n ⫽ 53, 12% of the patients).
To understand whether a lack of cancer cells in some tissue slices biased
our estimates, we provisionally restricted survival analyses to patients positive
for methylation in GSTP1, for whom tumor cells were most likely sufficiently
represented, and patients who underwent biopsy, where all available tissue was
paraffin embedded in one single block.
RESULTS
Patients in the 1990s cohorts were younger and had twice the
median survival time than those in the 1980s cohort (Table 1). In
the 1990s cohort, tumor tissue was obtained from transurethral
resections of the prostate and radical prostatectomies more often
than in the 1980s cohort.
Prevalence of Promoter Methylation
In the 1980s cohort, the prevalence of methylation in APC was
lower (P ⫽ .047) but methylation in GSTP1 (P ⫽ .002) and RUNX3
(P ⬍ .001) was more frequent compared with the 1990s cohort. These
differences remained after adjustment for Gleason score, age, and
source of the tumor tissue.
Methylation in GSTP1 and APC was positively associated with
Gleason score only in the 1990s cohort (Table 2). Age, source of tumor
tissue, and year of tissue collection were not associated with methyl-
ation (data not shown). In both cohorts, pair-wise comparisons re-
vealed that methylation in each gene was independent from
methylation in the other two genes (P ⬎ .25).
A
246 1210814
Cumulative Prostate Cancer Mortality
Follow-Up Duration (years)
0
0.2
0.4
0.6
0.8
1.0
APC in 1980s cohort
Gray’s test P = .11
246 1210814
Cumulative Prostate Cancer Mortality
Follow-Up Duration (years)
0
0.2
0.4
0.6
0.8
1.0
APC in 1990s cohort
Gray’s test P = .02
Unmethylation
Methylation
B
246 1210814
Cumulative Prostate Cancer Mortality
Follow-Up Duration (years)
0
0.2
0.4
0.6
0.8
1.0
RUNX3 in 1980s cohort
Gray’s test P = .32
246 1210814
Cumulative Prostate Cancer Mortality
Follow-Up Duration (years)
0
0.2
0.4
0.6
0.8
1.0
RUNX3 1990s cohort
Gray’s test P = .05
Unmethylation
Methylation
Unmethylation
Methylation
Unmethylation
Methylation
Fig 1. Cumulative prostate cancer mor-
tality by methylation status in (A) APC
and (B) RUNX3 in the 1980s cohort and
1990s cohort.
Richiardi et al
3164
© 2009 by American Society of Clinical Oncology
J
OURNAL OF CLINICAL ONCOLOGY
Downloaded from jco.ascopubs.org on April 3, 2012. For personal use only. No other uses without permission.
Copyright © 2009 American Society of Clinical Oncology. All rights reserved.
Promoter Methylation and Tumor Progression
Patients with methylation in APC had a higher prostate
cancer mortality than patients with an unmethylated cancer
(Fig 1A). This association was statistically significant only in the
1990s cohort (P ⫽ .02). Methylation in RUNX3 was associated
with survival in the 1990s cohort (P ⫽ .05) but not in the 1980s cohort
(Fig 1B). Methylation in GSTP1 was not associated with survival (data
not shown).
The HR of prostate cancer mortality for methylation in APC
was 1.42 (95% CI, 0.98 to 2.07) in the 1980s cohort and 1.57 (95%
CI, 0.95 to 2.62) in the 1990s cohort (HR ⫽ 1.49; 95% CI, 1.11 to
2.00 for the two cohorts combined; Table 3). In the 1990s cohort,
the adjusted HR estimate was lower than the crude one, mainly
because of the introduction of Gleason score into the model. In the
1980s cohort, the association between methylation in APC and
mortality was stronger and statistically significant in the first year
of follow-up, whereas in the 1990s cohort, the HR estimate increased
with duration of follow-up.
Results were not consistent between the two cohorts for methyl-
ation in RUNX3, which was associated with prostate cancer mortality
in the 1990s cohort (HR ⫽ 1.56; 95% CI, 0.95 to 2.56) but not in the
1980s cohort (Table 3). Methylation in GSTP1 was not associated with
prostate cancer mortality. Restriction of the analyses to patients who
underwent biopsy or were positive for GSTP1 methylation did not
substantially change increased HR estimates for methylation in APC
and RUNX3 (data not shown).
The effect of methylation in APC or RUNX3 at different levels
of Gleason score is summarized in Table 4. In the 1980s cohort,
Gleason score had a small impact on the HR estimates, whereas in
the 1990s cohort, we found a doubled HR from prostate cancer
mortality among patients with a Gleason score less than 8 both for
methylation in APC (HR ⫽ 2.09; 95% CI, 1.02 to 4.28) and in
RUNX3 (HR ⫽ 2.40; 95% CI, 1.18 to 4.91). In the two cohorts
combined, the HR of prostate cancer mortality increased with
increasing number of methylated genes (P ⫽ .002 for linear trend;
Table 5).
DISCUSSION
We found that methylation in APC is associated with prostate cancer
mortality, particularly among those with a highly to moderately dif-
ferentiated tumor. A similar association was found for methylation in
RUNX3 in the 1990s cohort, whereas methylation in GSTP1 was not
associated with risk. The results also indicated a shift in the methyl-
ation patterns from the 1980s to the 1990s.
Table 3. Prostate Cancer Mortality for Methylation in GSTP1, APC, and RUNX3 by Duration of Follow-Up in 1980s and 1990s Cohorts
Methylation and Gleason
Score No. of Deaths
Overall 14-Year Follow-Up
Follow-Up Period
⬍ 1 Year 1-3 Years 3⫹ Years
Crude HR HR
ⴱ
95% CI HR
ⴱ
95% CI HR
ⴱ
95% CI HR
ⴱ
95% CI
1980s cohort
Methylation in GSTP1
No 2611111
Yes 89 1.08 1.00 0.64 to 1.58 0.42 0.17 to 1.03 0.94 0.41 to 2.12 1.64 0.79 to 3.41
Methylation in APC
No 7411111
Yes 46 1.46† 1.42 0.98 to 2.07 2.66 1.12 to 6.31 1.41 0.72 to 2.75 1.10 0.63 to 1.93
Methylation in RUNX3
No 1511111
Yes 97 1.32 1.22 0.70 to 2.14 0.97 0.27 to 3.43 2.39 0.72 to 7.89 0.93 0.45 to 1.94
Gleason score
⬍ 8 5011111
8⫹ 71 2.53† 2.17 1.48 to 3.18 3.39 1.29 to 8.91 1.87 0.97 to 3.60 2.07 1.21 to 3.52
1990s cohort
Methylation in GSTP1
No 1711111
Yes 58 2.02† 1.44 0.82 to 2.54 1.62 0.52 to 5.04 0.94 0.34 to 2.62 1.74 0.75 to 4.04
Methylation in APC
No 3111111
Yes 41 1.99† 1.57 0.95 to 2.62 1.48 0.55 to 3.96 1.15 0.42 to 3.16 1.86 0.93 to 3.72
Methylation in RUNX3
No 3011111
Yes 40 1.74 1.56 0.95 to 2.56 1.58 0.60 to 4.14 1.33 0.48 to 3.68 1.66 0.85 to 3.27
Gleason score
⬍ 8 3611111
8⫹ 36 3.50† 3.27 2.00 to 5.37 6.50 2.08 to 20.3 3.61 1.31 to 9.94 2.34 1.16 to 4.73
Abbreviation: HR, hazard ratio.
ⴱ
HR was adjusted for follow-up duration, Gleason score, and source of tumor tissue; Gleason score was also adjusted for methylation in APC.
†P ⬍ .05.
Gene Methylation and Prostate Cancer Mortality
www.jco.org © 2009 by American Society of Clinical Oncology 3165
Downloaded from jco.ascopubs.org on April 3, 2012. For personal use only. No other uses without permission.
Copyright © 2009 American Society of Clinical Oncology. All rights reserved.
