Carcinogenesis vol.29 no.3 pp.568–572, 2008
Advance Access publication November 13, 2007
Polymorphic variants in PTGS2 and prostate cancer risk: results from two large
nested case–control studies
Kim N.Danforth1,?, Richard B.Hayes1, Carmen
Rodriguez2, Kai Yu1, Lori C.Sakoda3, Wen-Yi Huang1,
Bingshu E.Chen4, Jinbo Chen5, Gerald L.Andriole6,
Eugenia E.Calle2, Eric J.Jacobs2, Lisa W.Chu7, Jonine
D.Figueroa1,7, Meredith Yeager8, Elizabeth A.Platz9,
Dominique S.Michaud10, Stephen J.Chanock1,8,11,
Michael J.Thun2and Ann W.Hsing1
1Division of Cancer Epidemiology and Genetics, National Cancer Institute,
National Institutes of Health, Department of Health and Human Services,
Bethesda, MD 20852,2Department of Epidemiology and Surveillance
Research, American Cancer Society, Atlanta, GA 30303,3Department of
Epidemiology, School of Public Health and Community Medicine, University
of Washington, Seattle, WA 98195,4Department of Mathematics and
Statistics, Concordia University, Montreal, Quebec H3G 1M8, Canada,
5Department of Biostatistics and Epidemiology, University of Pennsylvania
School of Medicine, Philadelphia, PA 19104,6Division of Urologic Surgery,
Washington University School of Medicine, St Louis, MO 63110,7Cancer
Prevention Fellowship Program, Office of Preventive Oncology, Division of
Cancer Prevention, National Cancer Institute, National Institutes of Health,
Department of Health and Human Services, Bethesda, MD 20852,8Core
Genotyping Facility, Division of Cancer Epidemiology and Genetics,
Advanced Technology Program, SAIC Frederick, Inc., NCI-Frederick,
Frederick, MD 20877,9Department of Epidemiology, Johns Hopkins
Bloomberg School of Public Health, Baltimore, MD 21205,10Department of
Epidemiology, Harvard School of Public Health, Boston, MA 02115 and
11Center for Cancer Research, National Cancer Institute, National Institutes of
Health, Department of Health and Human Services, Bethesda, MD 20892
?To whom correspondence should be addressed. Tel: þ1 301 594 5631;
Fax: þ1 301 402 0916;
Chronic inflammation has been hypothesized to increase prostate
cancer risk. Prostaglandin-endoperoxide synthase 2 (PTGS2) en-
codes the proinflammatory cyclooxygenase 2 enzyme believed to be
the rate-limiting step in the synthesis of prostaglandins, important
mediators of inflammation. We investigated associations between
PTGS2 polymorphisms and prostate cancer risk among 2321
prostate cancer cases and 2560 controls in two large case–control
studies nested within the Prostate, Lung, Colorectal and Ovarian
(PLCO) Cancer Screening Trial and the Cancer Prevention Study
II Nutrition Cohort. Five single nucleotide polymorphisms
(SNPs) (rs5277, rs20432, rs4648276, rs5275 and rs689470) were
examined in SNP and haplotype analyses (five SNPs in PLCO and
four SNPs in the Nutrition Cohort). In PLCO, the Ex10 1837
T>C marker (rs5275) was initially associatedwith prostate cancer
risk (P-trend 5 0.02) but became non-significant after adjust-
ment for multiple comparisons (P 5 0.08); this SNP showed no
association with prostate cancer risk in the Nutrition Cohort
(P-trend 5 0.54) or in an analysis pooling the two cohorts
(P-trend 5 0.20). No other SNP was associated with prostate can-
cer risk in PLCO or the Nutrition Cohort individually or com-
bined. Haplotype analyses suggested an association between
PTGS2 variants in PLCO alone (global P 5 0.007), but not in
the Nutrition Cohort (global P 5 0.78) or pooled analysis (global
P 5 0.18). In conclusion, despite the potential importance of in-
flammation in prostate carcinogenesis, results from our large
study of five PTGS2 SNPs does not support a strong association
between PTGS2 variants and prostate cancer risk in non-Hispanic
The prostaglandin-endoperoxide synthase 2 (PTGS2) gene encodes
the proinflammatory cyclooxygenase 2 enzyme believed to be the
rate-limiting step in the synthesis of prostaglandins, important medi-
ators of inflammation (1). Chronic inflammation has been implicated
in the development of several cancers, including prostate cancer (2,3),
and proliferative inflammatory atrophy, an inflammatory condition
in the prostate, has been hypothesized to be a precursor lesion for
prostate cancer (4). Recently, a large Swedish study examined 9275
single nucleotide polymorphisms (SNPs) in 1086 inflammatory genes
and reported significant (a 5 0.01) associations between 106 SNPs
and prostate cancer (5), suggesting that variation in inflammation
genes may play a role in prostate cancer risk. Research on non-genetic
factors, including non-steroidal anti-inflammatory drug (NSAID) use,
obesity and prostatitis, also supports a possible etiologic role for in-
flammation in prostate cancer risk (6–9).
Three previous studies have examined the relationships betweenvar-
ious PTGS2 polymorphisms and prostate cancer risk (10–12) with
mixed results. To further clarify the role of PTGS2 variants in prostate
cancer development, we investigated associations between five PTGS2
polymorphisms and prostate cancer risk among 2321 prostate cancer
cases and 2560 controls in two large case–control studies nested within
the Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening
Trial and the Cancer Prevention Study II (CPS-II) Nutrition Cohort.
Materials and methods
PLCO Cancer Screening Trial. The PLCO Cancer Screening Trial (hereafter
referred to as PLCO) is an ongoing, randomized controlled trial designed to
evaluate the impact of screening tests on cancer-specific mortality. Details on
the trial have been published previously (13,14). From 1993–2001, 154 000
men and women, aged 55–74 years, were enrolled at 10 screening centers
throughout the country (Washington, DC; Detroit, MI; Salt Lake City, UT;
Denver, CO; Honolulu, HI; Minneapolis, MN; Marshfield, WI; Pittsburgh,
PA; St Louis, MO; Birmingham, AL) and randomized to the trial’s screening
arm or usual care. During screening visits, blood samples were collected. This
analysis uses non-Hispanic white men from the screening arm of the trial.
Institutional review boards at the National Cancer Institute (NCI) and each
of the participating institutions approved the PLCO protocol, and each partic-
ipant provided written informed consent.
Prostate tumors were identified by screening exams (prostate-specific anti-
gen test, digital rectal exam), reports from participants, physicians or relatives,
linkage with the National Death Index or linkage with state cancer registries.
All cases were pathologically confirmed.Cases were classified as ‘advanced’ if
score of ?7 (using the highest available Gleason score and the best available
information from pathology and/or clinical data for staging). Controls (n 5
1399) were matched to cases (n 5 1162) identified between October 1993 and
September 2001 on age, time since initial screening and year of blood draw
using incidence density sampling.
CPS-II NutritionCohort. The CPS-IINutritionCohort (hereafter referredto as
the Nutrition Cohort) is a prospective cohort study designed to examine asso-
ciations between a wide range of exposures and cancer incidence, as described
previously (15). It was established by the American Cancer Society in 1992
and 1993 among 86 000 men and 97 000 women in 21 USA states (California,
Connecticut, Florida, Georgia, Illinois, Iowa, Louisiana, Maryland, Massachu-
setts, Michigan, Minnesota, Missouri, New Jersey, New Mexico, New York,
North Carolina, Pennsylvania, Utah, Virginia, Washington, Wisconsin). Blood
samples were collected from a subset of Nutrition Cohort participants between
June 1998 and June 2001 (17 411 men and 21 965 women). This analysis uses
non-Hispanic white men who provided a blood sample. The Emory University
Abbreviations: CGEMS, Cancer Genetic Markers of Susceptibility; CPS-II,
Cancer Prevention Study II; CI, confidence interval; NCI, National Cancer
Institute; NSAID, non-steroidal anti-inflammatory drug; OR, odds ratio;
PTGS2, prostaglandin-endoperoxide synthase 2; PLCO, Prostate, Lung,
Colorectal and Ovarian; SNP, single nucleotide polymorphism.
? The Author 2007. Published by Oxford University Press. All rights reserved. For Permissions, please email: firstname.lastname@example.org
by guest on June 1, 2013
Institutional Review Board approved the study, and participants provided writ-
Prostate cancers diagnosed between enrollment in the Nutrition Cohort and
August 2001 were identified through self-report or National Death Index link-
age and were subsequently confirmed by medical record review or registry
linkage, except for two prostate cancer deaths for which information was
available only from death certificates. Prostate cancer was classified as ad-
vanced if the Gleason score was ?7, the tumor was classified as stage III or
IV, or it was a fatal case of unknown stage at diagnosis. Cases (n 5 1159) and
controls (n 5 1161) were matched on age and date of blood collection using
incidence density sampling.
SNPs were selected based on putative function, reported minor allele fre-
quency (?5%) and the availability of a validated assay at the NCI’s Core
Genotyping Facility (http://snp500cancer.nci.nih.gov). Five SNPs were geno-
typed in PLCO and four in the Nutrition Cohort (all also in PLCO) using the
TaqMan assay. The genotyping completion rate was ?98% in PLCO and
.92% in the Nutrition Cohort. The interassay concordance from blinded qual-
ity control samples was .99% in both studies.
Many participants (?75%) in our original PLCO analysis had one of their
SNPs (rs5275) genotyped again, using an Illumina assay, when they were later
included in a genome-wide scan as part of the Cancer Genetic Markers of
Susceptibility (CGEMS) Study (16,17). The CGEMS study, which over-
sampled advanced cancers, also genotyped several hundred additional PLCO
participants who were identified during further follow-up and were not in-
cluded in our original sample. For men assayed by both platforms (n 5 1817),
genotype concordance was .99.9% after excluding missing data (,1.5% for
For SNP analyses, odds ratios (ORs) and 95% confidence intervals (CIs) were
calculated using unconditional logistic regression models. In each study
ing factors) models were similar; therefore, only unadjusted results are pre-
sented.(Thematchingfactorsconsideredascovariates in adjustedmodelswere
age, time since initial screening and year of blood draw in PLCO and age and
date of blood collection in the Nutrition Cohort.)
Associations were calculated for each genotype (heterozygote and homozy-
gote for the variant allele) separately and combined, compared with the refer-
ent genotype (homozygote for the most common allele). Tests for trend were
based on the number of copies of the minor allele. For pooled analyses (PLCO
and Nutrition Cohort), heterogeneity was assessed using a two degrees of
freedom Wald test for gene-by-study interaction terms. Pooled analyses for
rs5275 included PLCO participants with and without data from CGEMS and
Nutrition Cohort participants.
Haplotype frequencies and ORs were estimated using the expectation–
maximization algorithm in haplo.stats (R) (18) based on the four SNPs in both
studies. Haplotype frequencies among controls were compared for PLCO and
the Nutrition Cohort before pooling data; in the pooled analysis, heterogeneity
by study was assessed using a global Wald test (19).
SNP analyses were adjusted for multiple testing using the Simes test, which
is based on the test for linear trend using the number of copies of the minor
allele (0, 1 and 2) (20). A global score test by Schaid et al. (21) was used to test
for overall differences in the frequency of the haplotypes between cases and
In each study, ?65% of study participants were between 60 and 70
years old (Table I). Men in both studies were highly educated,
with .40% receiving at least a college degree. A little over half of
participants reported use of NSAIDs. Men in the Nutrition Cohort
were less likely to report a history of diabetes, but more likely to
report a family history of prostate cancer, than men in PLCO.
Among controls, all polymorphisms were in Hardy–Weinberg equi-
librium (P . 0.05). In PLCO, PTGS2 variants were not associated
overall with prostate cancer risk (Table II). Although the CC genotype
of the Ex10 þ837T.C marker (rs5275) was initially associated with
a 37% increased risk compared with the TT genotype, this association
was not significant after adjustment for multiple testing (P 5 0.08).
Results for this SNP also became borderline significant (P-
trend 5 0.05) when .350 men genotyped through CGEMS were
included (CC versus TT genotype, OR 5 1.26, 95% CI: 0.98–1.60).
No other SNP was significantly associated with prostate cancer risk in
PLCO, although point estimates for three other SNPs (rs5277,
rs20432 and rs4648276) were similar to those observed for rs5275.
Table I. Characteristicsaof non-Hispanic white prostate cancer cases and controls in the PLCO Cancer Screening Trial and the CPS-II Nutrition Cohort
PLCO Nutrition Cohort
Cases (n 5 1162) Controls (n 5 1399)Cases (n 5 1159) Controls (n 5 1161)
N (%)N (%)N (%)N (%)
Age at enrollment (years)
Less than high school
High school graduate
Vocational or some college
College graduate or higher
Body mass index (kg/m2)
History of diabetes
Family history of prostate cancer
Missing categories not shown.
aAll characteristics presented at baseline (1993–2001 in PLCO and 1992–1993 in the Nutrition Cohort), except for family history of prostate cancer in the Nutrition
Cohort, which was updated in 1997, and percent of advanced tumors.
bRegular NSAID use in PLCO and ever NSAID use in the Nutrition Cohort; ‘regular use’ was not specifically defined on the PLCO questionnaire.
cAdvanced cancer: PLCO, Gleason score ?7 or stage III/IV cancer; Nutrition Cohort, Gleason score ?7, stage III/IV cancer or fatal case of unknown stage at
PTGS2 variants and prostate cancer risk
by guest on June 1, 2013
Among Nutrition Cohort participants, no SNP was associated with
prostate cancer risk, with all point estimates close to one. The Ex10
þ837T.C marker (rs5275) that appeared suggestive in PLCO
showed no relationship to prostate cancer risk in the Nutrition Cohort
(TC versus TT genotype, OR 5 0.95, 95% CI: 0.80–1.13; CC versus
TT genotype, OR 5 0.94, 95% CI: 0.70–1.25).
There was no statistical evidence of heterogeneity between results
of these two cohorts (P-heterogeneity . 0.10); thus, data from the
two studies were pooled. In the pooled analyses, no SNP was associ-
ated with prostate cancer risk (including rs5275, P-trend 5 0.20).
Haplotype frequency was significantly different between cases and
controls in PLCO (global P 5 0.007), but not in the Nutrition Cohort
(global P 5 0.78) (Table II). Using the four SNPs in both PLCO
and the Nutrition Cohort, haplotype frequencies were similar among
controls (P 5 0.32). As was done for the SNP analyses, data were
pooled from the two studies in a combined haplotype analysis
(P-heterogeneity 5 0.05). In the pooled analysis, PTGS2 haplotypes
were not associated with prostate cancer risk (global P 5 0.18).
In PLCO, risk patterns were generally similar among non-NSAID
and NSAID users (data not shown). However, in the Nutrition Cohort,
risk estimates varied by NSAID use, although no PTGS2 SNP was
significantly associated with prostate cancer risk among either non-
NSAID or NSAID users. For example, for the Ex10 þ837T.C
marker (rs5275), among non-NSAID users point estimates suggested
an increased risk of prostate cancer (CC versus TT genotype, OR 5
1.32, 95% CI: 0.84–2.08; n 5 50 cases and 42 controls), whereas
among NSAID users point estimates suggested a decreased risk of
prostate cancer (CC versus TT genotype, OR 5 0.75, 95% CI: 0.51–
1.09; n 5 62 cases and 74 controls) with thevariant allele. When non-
advanced/advanced tumor status was accounted for in this analysis,
results remained non-significant but point estimates were slightly
stronger for the advanced tumors compared with non-advanced
Table II. ORs and 95% CIs for prostate cancer risk and PTGS2 polymorphisms among non-Hispanic white men in the PLCO Cancer Screening Trial and the
CPS-II Nutrition Cohort
PLCONutrition Cohort PLCO þ Nutrition Cohort combined
(n 5 2321)(n 5 2560)
(n 5 1162)
(n 5 1399)
OR (95% CI) Cases
(n 5 1159)
(n 5 1161)
OR (95% CI) Controls OR (95% CI)
Ex3 ?8G.C (rs5277)
P-trend 5 0.84
P-trend 5 0.16
P-trend 5 0.08
1.15 (0.97–1.36)GC or CC
IVS5 ?275T.G (rs20432)
353386 344 330 697716
P-trend 5 0.43
P-trend 5 0.15
P-trend 5 0.22
1.09 (0.92–1.30) TT or GG
IVS7 þ111T.C (rs4648276)
327370325 308652 678
P-trend 5 0.14
1.14 (0.95–1.37)TC or CC
Ex10 þ837T.C (rs5275)
P-trend 5 0.54
P-trend 5 0.20
P-trend 5 0.02a
1.17 (1.00–1.36)TC or CC 658
With additional CGEMSbparticipants
With additional CGEMSbparticipants
P-trend 5 0.30
P-trend 5 0.05
1.13 (0.97—1.31)TC or CC
Ex10 ?90C.T (rs689470)
CT or TT
47 6337 89 100
Global P 5 0.78
Global P 5 0.18
Global P 5 0.007
ORs and 95% CIs not shown for SNP analyses with fewer than five men in a cell or haplotype analyses where the haplotype frequency is ,5%.
aAfter adjustment for multiple comparisons, P-value 5 0.08.
bGenotyping data were available for additional PLCO participants (not included in the original analysis) through CGEMS.
cBold letters indicate SNP changes from the referent haplotype. Haplotype: rs5277, rs20432, rs5275 and rs689470; rare haplotypes not shown. In PLCO, the
haplotype analysis with all five SNPs (rs5277, rs20432, rs4648276, rs5275 and rs689470) was nearly identical to the haplotype analysis using only the four
K.N.Danforth et al.
by guest on June 1, 2013
tumors in both non-NSAID and NSAID users. In analyses stratified
only by non-advanced/advanced tumor status, results were similar for
advanced and non-advanced cases (data not shown).
In this nested case–control study of .2300 prostate cancer cases, we
did not observe significant associations between PTGS2 variants and
prostate cancer risk overall. Although SNP and haplotype findings
were suggestive in PLCO, they did not persist after adjustment for
multiple comparisons or the inclusion of additional PLCO partici-
pants available through CGEMS. SNP and haplotype results for four
SNPs were null in the Nutrition Cohort and pooled data set, and they
did not support an association between PTGS2 variants and prostate
Overall, we did not find a strong association between prostate can-
cer risk and the Ex10 þ837T.C marker (rs5275), although SNP and
haplotype results were somewhat suggestive in PLCO. The role of
rs5275 has been investigated in two previous studies, one in Sweden
(n 5 2160) (11) and one in the USA (n 5 834 whites) (12), and in
both studies, this SNP was not associated with prostate cancer risk.
Other SNPs (rs689470, rs5277, rs4648276 and rs20432) in our
analysis were also examined by previous studies. The Ex10
?90C.T marker (rs689470) showed no association in two USA
studies (12), including ours, but the variant (T) allele of this marker
was associated with a decreased risk of prostate cancer in a Swedish
case–control study (11). In addition, consistent with our results, null
associations were observed for the Ex3 ?8G.C (rs5277) marker in
a USA study (12) and the IVS7 þ111T.C marker (rs4648276) in
a Swedish study (11). However, in contrast to our results, the Swedish
study found an association for the IVS5 ?275T.G marker (rs20432)
with prostate cancer risk, although point estimates did not suggest
a dose–response trend across copies of the variant allele (11).
A SNP not included in our study, rs2745557, was included in two
other studies (11,12). This polymorphism had the strongest associa-
tion among SNPs examined in a USA case–control study of advanced
prostate cancer (12), but no dose–response trend was observed across
copies of the variant allele, and no association was found in the
Swedish case–control study for this SNP with prostate cancer risk
(11). One other epidemiologic study reported associations for variants
in the PTGS2 promoter and prostate cancer risk, but risk could not
be reliably evaluated among whites due to their small sample size
(n 5 92 cases and 92 controls) and low minor allele frequencies
(most ?1%) (10).
No consistent association emerged for various PTGS2 SNPs across
multiple study populations. Thus, despite some suggestive findings in
each of three studies [our study, the USA case–control study (12) and
the Swedish case–control study (11)], associations between PTGS2
variants and prostate cancer risk do not appear robust. It is unclear
why associations vary across studies. Within our own study, differ-
ences were observed across study populations (PLCO and Nutrition
Cohort) despite relatively large sample sizes, similar allele frequen-
cies and similar demographic characteristics (e.g. both highly edu-
cated, white, older men from multiple states in the USA). It is possible
that the varying associations resulted from chance alone, as supported
by the attenuation of results within PLCO when additional PLCO
participants were added to our original analysis (who had genotyping
data available through CGEMS for rs5275, the SNP whose associa-
tion was initially significant). However, it is also possible that the
varied findings reflect differences in unidentified characteristics
(genes or lifestyle factors) that interact with PTGS2 in affecting pros-
tate cancer risk.
There was some suggestion that associations between the PTGS2
SNPs and prostate cancer risk were different among non-NSAID and
NSAID users in the Nutrition Cohort, although no SNP was signifi-
cantly associated with prostate cancer risk. Similarly, another USA
case–control study of advanced prostate cancer (12) suggested possi-
ble differences in associations between PTGS2 variants and prostate
cancer risk by NSAID use. Although the interaction between
rs2745557 and NSAID use was not statistically significant, point
estimates for NSAID use were stronger (more inverse) among men
with the homozygote genotype for the most common allele than
among men with the variant allele (OR=0.62 and 0.86, respectively)
(12). Thus, interactions between PTGS2 variants and NSAID use
might be further examined in studies large enough to detect these
A notable strength of our study is its relatively large size with
.2300 cases, providing sufficient power to detect modest main ef-
fects. Genotyping error in the study is low, as evidenced by the high
genotyping success rates and high interassay concordance. However,
one major weakness of our study is that, with five candidate SNPs, we
had limited gene coverage of PTGS2. Our analysis included two (rs
5277 and rs5275) of the four (rs5277, rs5275, rs2066826 and
rs2206593) PTGS2 SNPs needed tocapture common genetic variation
(minor allele frequency . 0.05) according to HapMap data, based on
Caucasian Utah residents of northern and western Europe ancestry
(22). Furthermore, results from a recent genome-wide scan (CGEMS)
in one of our study populations (PLCO) provided information on four
additional PTGS2 SNPs (rs2206593, rs10911905, rs2143417 and
rs2383529) not included in our original PLCO analysis, one of which
was identified in HapMap (rs2206593) as a tagging SNP. No signif-
icant associations with prostate cancer risk were observed for any of
these PTGS2 SNPs (all P-values ? 0.36) (17), indicating that PTGS2
is not associated with prostate cancer.
Another limitation is that we did not examine other types of genetic
that these types of genetic variation in the PTGS2 gene are important.
Furthermore, variations in othergenes in the inflammation pathway may
interact with variants in PTGS2 to affect the risk of prostate cancer.
Thus, future studies should seek to examine the combined effects of
multiple genes in the inflammation pathway on prostate cancer risk.
In conclusion, despite the potential importance of inflammation in
prostate carcinogenesis, results from our large study of five PTGS2
SNPs does not support a strong association between PTGS2 variants
and prostate cancer risk in non-Hispanic white men. However, it is
possible that PTGS2 may interact with other genes or lifestyle factors,
such as NSAID use, to influence prostate cancer risk. Larger studies
will be needed to evaluate such interactions.
Intramural Research Program of the National Cancer Institute,
National Institutes of Health, Department of Health and Human Serv-
ices; this project funded in part with federal funds from the National
Cancer Institute, National Institutes of Health (under contract N01-
in Cancer Research (to K.N.D.), National Cancer Institute, National
Institutes of Health, Department of Health and Human Services.
We thankDrsChristineBergandPhilip Prorok(Divisionof CancerPrevention,
NCI); the ScreeningCenter investigators and staff of the PLCO Cancer Screen-
ing Trial; Tom Riley and staff (Information Management Services, Silver
Spring, MD); Barbara O’Brien, Shelley Niwa and staff (Westat, Rockville,
MD); Drs Bill Kopp, Wen Shao and staff (Science Applications International
Corporation-Frederick) and Kimberly Walker–Thurmond and Cari Lichtman
(American Cancer Society) for their contributions to making this study possi-
ble. The content of this publication does not necessarily reflect the views or
policies of the Department of Health and Human Services, nor does mention of
trade names, commercial products or organizations imply endorsement by the
Conflict of Interest Statement: None declared.
1.Hussain,T. et al. (2003) Cyclooxygenase-2 and prostate carcinogenesis.
Cancer Lett., 191, 125–135.
PTGS2 variants and prostate cancer risk
by guest on June 1, 2013
2.De Marzo,A.M. et al. (2007) Inflammation in prostate carcinogenesis. Nat. Download full-text
Rev. Cancer, 7, 256–269.
3.Moran,E.M. (2002) Epidemiological and clinical aspects of nonsteroidal
anti-inflammatory drugs and cancer risks. J. Environ. Pathol. Toxicol. On-
col., 21, 193–201.
4.De Marzo,A.M. et al. (2007) Inflammation, atrophy, and prostate carcino-
genesis. Urol Oncol, 25, 398–400.
5.Zheng,S.L. et al. (2006) A comprehensive association study for genes in
inflammation pathway provides support for their roles in prostate cancer
risk in the CAPS study. Prostate, 66, 1556–1564.
6.Mahmud,S. et al. (2004) Prostate cancer and use of nonsteroidal anti-
inflammatory drugs: systematic review and meta-analysis. Br. J. Cancer,
7.MacInnis,R.J. et al. (2006) Body size and composition and prostate cancer
risk: systematic review and meta-regression analysis. Cancer Causes Con-
trol, 17, 989–1003.
8.Hsing,A.W. et al. (2006) Prostate cancer epidemiology. Front. Biosci., 11,
9.Hsing,A.W. et al. (2007) Obesity, metabolic syndrome, and prostate cancer.
Am. J. Clin. Nutr., 86 (suppl), 843S–857S.
10.Panguluri,R.C. et al. (2004) COX-2 gene promoter haplotypes and prostate
cancer risk. Carcinogenesis, 25, 961–966.
11.Shahedi,K. et al. (2006) Genetic variation in the COX-2 gene and the
association with prostate cancer risk. Int. J. Cancer, 119, 668–672.
12.Cheng,I. et al. (2007) COX2 genetic variation, NSAIDs, and advanced
prostate cancer risk. Br. J. Cancer, 97, 557–561.
13.Gohagan,J.K. et al. (2000) The Prostate, Lung, Colorectal and Ovarian
(PLCO) Cancer Screening Trial of the National Cancer Institute: history,
organization and status. Control. Clin. Trials, 21, 251S–272S.
14.Hayes,R.B. et al. (2005) Methods for etiologic and early marker investiga-
tions in the PLCO trial. Mutat. Res., 592, 147–154.
15.Calle,E.E. et al. (2002) The American Cancer Society Cancer Prevention
Study II Nutrition Cohort: rationale, study design, and baseline character-
istics. Cancer, 94, 2490–2501.
16.Yeager,M. et al. (2007) Genome-wide association study of prostate cancer
identifies a second risk locus at 8q24. Nat. Genet., 39, 645–649.
17.National Cancer Institute Cancer Genetic Markers of Susceptibility Study.
http://cgems.cancer.gov/data. (7 September 2007 date last accessed).
18.Excoffier,L. et al. (1995) Maximum-likelihood estimation of molecular
19.Lake,S.L. et al. (2003) Estimation and tests of haplotype-environment
interaction when linkage phase is ambiguous. Hum. Hered., 55, 56–65.
20.Simes,R.J. (1986) An improved Bonferroni procedure for multiple tests of
significance. Biometrika, 73, 751–754.
21.Schaid,D.J. et al. (2002) Score tests for association between traits and
haplotypes when linkage phase is ambiguous. Am. J. Hum. Genet., 70,
22.International HapMap Consortium. (2005) A haplotype map of the human
genome. Nature, 437, 1299–1320.
Received September 7, 2007; revised November 2, 2007;
accepted November 4, 2007
K.N.Danforth et al.
by guest on June 1, 2013