ORIGINAL ARTICLE

Cyclooxygenase-2 (COX-2) Polymorphisms and Risk ofInflammatory Bowel Disease in a Scottish and DanishCase–Control Study

Vibeke Andersen, PhD,* Elaine Nimmo, PhD,† Henrik B. Krarup, PhD,‡ Hazel Drummond, BSc,†

Jane Christensen, MSc,§ Gwo-tzer Ho, PhD,† Mette Østergaard, PhD,*,k Anja Ernst, PhD,‡

Charlie Lees, Prof,† Bent A. Jacobsen, PhD,¶ Jack Satsangi, Prof,† and Ulla Vogel, Prof**,††,‡‡

Background: Inflammatory bowel diseases (IBDs) are a result

of interactions between luminal pathogens and the intestinal

immune response. Cyclooxygenase-2 (COX-2) plays a key role in

the regulation of the inflammatory response upon stimulation by

luminal pathogens via Toll-like receptors.

Methods: Genotypes of the COX-2/PTGS2/PGHS2 A-1195G

(rs689466), G-765C (rs20417), and T8473C (rs5275) polymor-

phisms were assessed in a Scottish and Danish case–control study

including 732 Crohn’s disease (CD) cases, 973 ulcerative colitis

(UC) cases, and 1157 healthy controls using logistic regression.

Results: Carriers of the COX-2 A-1195G variant allele had

increased risk of UC (odds ratio [OR], 95% confidence interval

[CI] ¼ 1.25 [1.02–1.54], P ¼ 0.03) and of both UC and IBD

among never smokers (OR [95% CI] ¼ 1.47 [1.11–1.96], P ¼0.01 and OR [95% CI] ¼ 1.37 [1.06–1.77], P ¼ 0.02, respec-

tively). Furthermore, this variant genotype was associated with

increased risk of diagnosis of UC before age 40 years and with

extensive UC (OR [95% CI] ¼ 1.34 [1.11–1.62], P ¼ 0.002 and

OR [95% CI] ¼ 1.32 [1.03–1.69], P ¼ 0.03, respectively).

Conclusions: COX-2 A-1195G polymorphism was associated

with the risk of UC, especially among never-smokers, suggesting

that low activity of COX-2 may predispose to UC. Our results

suggest that inclusion of smoking status may be essential for the

evaluation of the role of genetic predisposition to IBD.

(Inflamm Bowel Dis 2011;17:937–946)

Key Words: Crohn’s disease, single nucleotide polymorphisms,smoking status, ulcerative colitis

T he inflammatory bowel diseases (IBDs), ulcerative colitis

(UC) and Crohn’s disease (CD), are complex diseases

whose etiology comprises both genetic and environmental fac-

tors. Although the use of candidate gene and genome-wide

association studies have added significantly to the number of

replicated IBD-associated genes and loci,1–3 still many more

genes are expected to contribute with small effects to IBD eti-

ology.4,5 Specifically, the populations in the northern part of

Europe are characterized by a high incidence of IBD6,7 and

low frequencies of NOD2 polymorphisms, which significantly

contribute to the IBD etiology in Central Europe.8,9

IBD occurs as a consequence of a dysregulated immune

response to normal constituents of the intestinal flora.10–12 Cy-

clooxygenase-2 (COX-2) plays a key role in the regulation of

the intestinal immune response. Upon the recognition of, e.g.,

bacterial products in the intestinal lumen by the Toll-like

receptors (TLRs), in particular TLR4,13 COX-2 expression is

induced by transcription factors such as nuclear factor j-B(NF-jB).14 Next, COX-2-activation may impact inflammatory

processes via the inhibition of NF-jB and activation of peroxi-

some proliferator-activated receptor c (PPAR-c)15 and by the

modification of the mucosal barrier function.16 Although tradi-

tionally considered a proinflammatory cytokine,17–20 it has

long been recognized that COX-2 also possesses antiinflamma-

tory properties.21,22 Thus, COX-2-derived prostaglandins have

Additional Supporting Information may be found in the online version

of this article.

Received for publication June 21, 2010; Accepted June 29, 2010.

From the *Medical Department, Viborg Regional Hospital, Viborg,

Denmark; †Gastrointestinal Unit, Molecular Medicine Centre, Western

General Hospital, University of Edinburgh, UK, ‡Department of Clinical

Biochemistry, Aarhus University Hospital, Aalborg, Denmark, §Danish

Cancer Society, Institute of Cancer Epidemiology, Copenhagen, Denmark,¶Department of Medical Gastroenterology, Aarhus University Hospital,

Aalborg, Denmark, kBiochemical Department, Viborg Regional Hospital,

Viborg, Denmark, **National Food Institute, Technical University of

Denmark, DK-2860 Søborg, Denmark, ††Institute for Science, Systems

and Models, University of Roskilde, Denmark, ‡‡National Research Centre

for the Working Environment, Copenhagen, Denmark.

Reprints: Vibeke Andersen, PhD, Medical Department, Regional

Hospital Viborg, Heibergs Alle 4, DK-8800 Viborg, Denmark (e-mail:

va9791@gmail.com)

Supported by the Lundbeck Foundation, the Danish Research Council,

the Western Danish Research Forum for Health Science, the County of

Viborg, the Danish Colitis-Crohn Association, the Harboe Foundation, the

‘‘Familien Erichsen Mindefond,’’ ‘‘John M Klein og hustrus mindelegat,’’

and ‘‘Fonden til Lægevidenskabens fremme.’’

Copyright VC 2010 Crohn’s & Colitis Foundation of America, Inc.

DOI 10.1002/ibd.21440

Published online 27 August 2010 in Wiley Online Library

(wileyonlinelibrary.com).

Inflamm Bowel Dis � Volume 17, Number 4, April 2011 937

been shown to be critical for intestinal healing.19,23 Indeed,

COX-2 is upregulated in inflamed tissue in IBD patients.24,25

The functional COX-2 polymorphisms A-1195G and

G-765C in the promoter region and T8473C in the 30 untrans-lated (UTR) region have been the most widely studied.26 The

COX-2 G-765C and T8473C variant alleles are in tight linkage,

whereas the COX-2 A-1195G variant allele cosegregate with

the wildtype alleles of the two other polymorphisms in Cauca-

sians (Fig. 1).27,28 The variant G allele of the -1195 polymor-

phism causes the removal of a c-Myb binding site resulting in

lower COX-2 mRNA levels compared to the A allele.26 The G

to C substitution at nucleotide -765 eliminates an Sp1 binding

site, but meanwhile creates an E2F binding site.26 In vitro stud-

ies have revealed a lower COX-2 expression from a haplotype

containing the C-allele of G-765C allele in combination with

the G-allele of A-1195G,26 whereas the COX-2 haplotype con-

taining the variant G-765C allele together with the variant allele

of the T8473C single nucleotide polymorphism (SNP) has been

associated with high COX-2 activity,29 indicating that the

30UTR variant allele may stabilize the mRNA level consider-

ably. Using a candidate gene approach, COX-2 polymorphisms

have been associated with IBD risk in Dutch and Danish study

populations.30,31 In the small Dutch study encompassing 291

cases and two control groups of 175 and 192 subjects, respec-

tively, two polymorphisms in the 30UTR region were associated

with IBD.30 The T5209G variant allele was weakly associated

with CD and the T8473C T-allele with UC.30 In the Danish

study, homozygous carriers of the COX-2 G-765C variant were

at slightly higher risk of CD and UC than wildtype carriers.31

Smoking has been shown to induce inflammation32

and COX-2 expression.33,34 Also, smoking affects the risk

of IBD differentially—conferring increased risk of CD and

reduced risk of UC.35–37 Therefore, smoking may interact

with COX-2 polymorphisms in relation to IBD risk.

In the present study we wanted to assess the role of the

functional COX-2 A-1195G, G-765C, and T8473C polymor-

phisms in relation to risk of developing IBD combining a

Scottish and a Danish case–control study to achieve a total

of 732 CD cases, 973 UC cases, and 1157 healthy controls.

The functional polymorphisms were chosen in order to allow

a biological interpretation of the results.

MATERIALS AND METHODS

Patients and ControlsThe Scottish patients were recruited from Edinburg hos-

pitals from January 2001 to July 2007. The Danish patients

were recruited from Viborg, Aalborg, and Herning regional

hospitals from January 2004 to March 2005. The diagnosis of

CD or UC was based on clinical, radiological, endoscopic, and

histological examinations (infectious and other cases of IBDs

were excluded).38 The Scottish control group consists of two

samples, one was healthy controls (n ¼ 154) consisting of the

laboratory personnel where data on sex, years of age at inclu-

sion, and smoking status were collected. The other was samples

from the Blood Transfusion Centre (n ¼ 250), where only data

on sex and age were available. The Danish control group con-

sists of healthy blood donors who were recruited from Viborg

Hospital. Smoking habits at study entry was collected. All

Danish subjects were Caucasian and older than 18 years of age.

Characteristics of Scottish and Danish IBD patients and

controls are shown in Table 1. A total of 933 (423 CD and 510

UC) Scottish patients and 404 controls and 772 (309 CD and

463 UC) Danish patients and 753 controls were included. No

differences regarding genotype frequencies were found

between the two Scottish groups of healthy controls; they were

therefore grouped together as one large control group. Smok-

ing was more common among the CD patients than among UC

patients, 41% and 51%, for the Scottish and Danish CD

patients, respectively, versus 19% and 17% for the Scottish

and Danish UC patients, respectively. The Scottish UC patients

had more propagated disease; 45% had extensive colitis com-

pared to 19% of the Danish patients. The Danish UC patients

and healthy controls were, in general, older than the Scottish.

GenotypingDNA was extracted from EDTA-stabilized peripheral

blood samples from all patients and healthy controls using

FIGURE 1. The COX-2 gene. A: The position of the COX-2 A-1195G, G-765C, and T8473C polymorphisms. B: Frequencyand linkage of haplotypes (from Ref. 28). [Color figure canbe viewed in the online issue, which is available atwileyonlinelibrary.com.]

Inflamm Bowel Dis � Volume 17, Number 4, April 2011Andersen et al

938

TABLE

1.Basic

DescriptionsoftheScottishan

dDan

ishPatients

withCrohn’sDisease(CD),Ulcerative

Colitis(UC),an

dHealthyControls

ScottishPatients

andControls

DanishPatients

andControls

CrohnsDisease

a

(n¼

423)

UlcerativeColitisa

(n¼

510)

Controls

(n¼

404)

CrohnsDisease

(n¼

309)

UlcerativeColitis

(n¼

463)

Controls

(n¼

753)

Gender

M/F:n(%

)164/259(39/61)

253/257(50/50)

183/221(45/55)

122/187(39/61)

223/240(48/52)

388/365(52/48)

MedianAge:

(5%–95%)

43(21–73)

41(21–71)

39(21–64)

43(23–76)

48(24–77)

43(23–61)

MedianAgeat

diagnosis:

(5%–95%)

27(14–64)

46(22–76)

30(15–65)

35(17–66)

Fam

ilyhistory

ofIBD:n(%

)

Yes/No/NAb

89/321/13(21/76/3)

86/415/9

(17/81/2)

Smokinghabitsat

diagnosisc:n(%

)

Smokers

175(41)

99(19)

157(51)

77(17)

199(26)

Never

smokers

189(45)

248(49)

109(35)

215(46)

378(50)

Form

ersm

okers

46(11)

148(29)

43(14)

171(37)

176(23)

Datanotavailable

13(3)

15(3)

LocationUCd:

Proctitis

(E1)

81(16)

199(43)

Leftside(E2)

189(37)

167(36)

Extensive(E3)

227(45)

87(19)

Datanotavailable

13(3)

10(2)

LocationCDd:n(%

)

Ileal(L1,L1þL

4)

147(35)

145(47)

Colonic

(L2,L2þL

4)

149(35)

71(23)

Ileocolonic

(L3,L3þL

4)

99(23)

83(27)

Other

13(3)

Datanotavailable

2(1)

10(3)

Behaviorat

5yrafterdiagnosis

Inflam

matory

þ/�

perianal

(B1,B1p)

259(60)

Stricturingþ/

�perianal

(B2,B2p)

56(15)

Penetratingþ/

�perianal

(B3,B3p)

70(16)

Other

36(9)

Datanotavailable

2(1)

Advancedmedicatione:n(%

)

Yes/No/NAb

133/174/2

(43/56/1)

103/360/0

(22/78/0)

Operation:n(%

)

Yes/No/NA

36/387/0

(9/91/0)

119/387/4

(23/76/1)

140/168/1

(45/54/0)

13/449/1

(3/97/0)

ThelocationofCD

andUCin

accordance

withtheViennaclassificationareshown.

a413(98%)oftheCD

patientsand488(96%)ofUCwereofwhiteEuropeanethnicity.

bNA;Datanotavailable.

cForcontrols:sm

okinghabitsat

thetimeofthesampling.

dMaxim

um

disease

extent.

eAdvancedmedication:azathioprine,

6-m

ercaptopurine,

TNF-a

inhibitors.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011 COX-2 Polymorphisms and IBD

939

either a PureGene (Gentra Systems, Minneapolis, MN) or

Wizard Genomic (Promega, Madison, WI) DNA purifica-

tion kit, according to the manufacturer’s recommendations.

Genotyping of COX-2 G-765C (rs20417)andCOX2 A-

1195G (rs689466) was performed by TaqMan real-time po-

lymerase chain reaction (PCR) on an ABI7900HT (Applied

Biosystems, Nærum, Denmark), using allelic discrimination

as previously described.27 COX2 C8473T (rs5275) was gen-

otyped as previously described.39 Data on COX-2 G-765C

genotypes in the Danish cohort have previously been pub-

lished.31 Twenty ng of DNA was genotyped in 5 lL contain-

ing 1x Mastermix (Applied Biosystems), 100 nM probes,

and 900 nM primers. Controls of known genotypes were

included in each run and repeated genotyping of a random

10% subset yielded 99%–100% identical genotypes.

The genotype distributions among the controls did

not deviate from Hardy–Weinberg equilibrium.

Statistical AnalysesWe pooled the genotype data from Scotland and Den-

mark and logistic regression was used for the statistical

analyses. The analyses included only subjects where all in-

formation was available. Age was entered linearly in the

model after checking for linearity using a linear spline.40

Subgroup analyses were done for the three COX-2polymorphisms in relation to location of the disease (CD:

L1 and L1þL4, L2 and L2þL4, L3 and L3þL4, UC: E1,

E2, E3), operation (Y/N) and age at diagnosis (�40/>40).

For the Scottish patients, subgroup analyses for behavior

5 years after diagnosis (CD: B1 and B1p, B2 and B2p, B3

and B3p) and family history (Y/N) were also made. For

the Danish patients, subgroup analyses for advanced medi-

cal treatment (Y/N) were also made. The haplotypes were

inferred manually as described previously.27

Ethical ConsiderationsAll subjects received written and oral information

and gave written informed consent. The study was done in

accordance with the Declaration of Helsinki and approved

by the Lothian Research and Ethics Committee (LREC-

2000/4/192 and 2004/S1103/22) and the Danish Regional

Ethics Committee (VN 2003/124).

RESULTS

Associations Between Polymorphisms andDisease Phenotypes

There was no heterogeneity between Scottish and

Danish CD and UC patient and control allele frequencies

for the COX-2 polymorphisms and, therefore, the two study

groups were combined. The odds ratios (ORs) for the asso-

ciation between COX-2 genotypes and risk of CD, UC, and

IBD, respectively, in the combined study groups are shown

in Table 2. Carriers of COX-2 A-1195G variant genotype

were at higher risk of UC (OR, [95% confidence interval,

CI] ¼ 1.25 [1.02–1.54], P ¼ 0.03) than homozygous wild-

type carriers in the combined group (Table 2). No statisti-

cally significant associations with CD or all IBD were

TABLE 2. Odds Ratios for the COX-2 Gene Polymorphisms in the Combined Scottish and Danish Study Groups; Crohn’sDisease (CD), Ulcerative Colitis (UC), and Inflammatory Bowel Disease (IBD)

NCD NUC Ncontrol ORCD (95%CI)a P-valueb ORUC (95%CI)a P-valueb ORIBD (95%CI)a P-valueb

G-765C

GG 510 676 659 1.00 — 1.00 — 1.00 —

CG 192 254 229 0.93 (0.72–1.20) 0.57 0.90 (0.72–1.14) 0.39 0.92 (0.75–1.13) 0.43

CC 17 28 19 1.34 (0.62–2.89) 0.45 1.28 (0.67–2.46) 0.46 1.28 (0.71–2.30) 0.41

CG and CC 209 282 248 0.95 (0.74–1.23) 0.72 0.93 (0.74–1.16) 0.53 0.95 (0.78–1.15) 0.59

A-1195G

AA 456 578 589 1.00 — 1.00 — 1.00 —

AG 236 347 288 1.10 (0.86–1.40) 0.46 1.25 (1.01–1.55) 0.04 1.17 (0.97–1.42) 0.10

GG 27 33 30 1.37 (0.75–2.49) 0.31 1.31 (0.76–2.26) 0.33 1.34 (0.83–2.16) 0.24

AG and GG 263 380 318 1.12 (0.89–1.42) 0.34 1.25 (1.02–1.54) 0.03 1.19 (0.99–1.43) 0.06

T8473C

TT 292 400 393 1.00 — 1.00 — 1.00 —

CT 328 441 403 1.08 (0.85–1.37) 0.54 1.04 (0.84–1.29) 0.69 1.07 (0.88–1.29) 0.50

CC 99 117 111 1.23 (0.86–1.76) 0.25 0.89 (0.64–1.23) 0.48 1.00 (0.75–1.33) 0.99

CT and CC 427 558 514 1.11 (0.89–1.39) 0.36 1.01 (0.83–1.24) 0.92 1.05 (0.88–1.26) 0.58

Statistical analyses included subjects where all information was available.aAdjusted for country, age, gender, and smoking status.bP-value for estimate.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011Andersen et al

940

found and no statistically significant associations with the

COX-2 G-765C or T8473C polymorphisms were found.

Suppl. Table 2A,B shows the ORs for the individual Scot-

tish and Danish CD and UC groups, respectively. A statisti-

cally significant association between the COX-2 A-1195G

variant genotype and risk of UC (OR [95% CI] ¼ 1.53

[1.03–2.27], P ¼ 0.03) was found in the Scottish UC

patients, whereas a nonsignificant association was found in

the Danish UC patients (Suppl. Table 2B). The effect of

smoking status at the time of diagnosis on the ORs is

shown in Table 3 and Suppl. Table 3A,B. The risk modifi-

cation conferred by smoking was not dependent on the ge-

notypes. Among never-smokers, variant allele carriers of

COX-2 A-1195G were at higher risk of UC (OR [95% CI]

¼ 1.47 [1.11–1.96], P ¼ 0.01), and IBD (OR [95% CI] ¼1.37 [1.06–1.77], P ¼ 0.02) than the homozygous wildtype

carriers (Table 3). After correction for multiple testing,

only the association between COX-2 A-1195G and risk of

UC among never-smokers was statistically significant. No

other genotype associations were found. Smoking at diagno-

sis was associated with increased risk of CD for all geno-

types compared to never-smokers in both the Scottish and

Danish CD (Suppl. Table 3A). Former smoking at the time

of diagnosis was associated with increased risk of UC com-

pared to never smokers in both Scottish and Danish UC

(Suppl. Table 3B). There was no interaction between any of

the polymorphisms and smoking status.

Subgroup analyses showed that COX-2 A-1195G var-

iant allele carriers were at statistically significantly higher

risk of extensive UC with OR [95% CI] of 1.32 [1.03–

1.69], P ¼ 0.03 than the homozygous wildtype carriers

(Table 4A). The individual values for the Scottish and Dan-

ish UC groups were 1.54 [1.14–2.10] and 1.14 [0.73–1.78],

respectively. Furthermore, the COX-2 A-1195G variant al-

lele carriers were at statistically significantly higher risk of

UC before the age of 40 years with OR [95% CI] of 1.34

[1.11–1.62], P ¼ 0.002 than the homozygous wildtype car-

riers (Table 4B). The individual values for the Scottish and

Danish UC groups were 1.49 [1.13–1.98] and 1.35 [1.03–

1.77], respectively. No associations between the genotypes

and operation were found.

Haplotype analysis of the COX-2 polymorphisms

revealed that the four most frequently occurring haplotypes

described more than 75% of the genetic variation in the

control group (Table 5). The most common haplotype con-

sisted of the wildtype alleles of the three polymorphisms

(Table 5). Carriers of two different rare haplotypes encom-

passing the COX-2 A-1195G and G-765C variant geno-

types were at higher risk of IBD (GCC; OR [95% CI] ¼1.48 [1.05–2.09], P ¼ 0.03 and GCT; OR [95% CI] ¼ 4.08

[1.22–13.65], P ¼ 0.02). No other haplotypes were associ-

ated with risk of IBD.

Minor allele frequencies of COX-2 A-1195G,

G-765C, and T8473C in the Scottish and Danish control

group and in other published studies on Caucasian study

populations are shown in Table 6. The allele frequencies of

the present study did not deviate from those found in previ-

ous studies.28,41–51

DISCUSSIONIn the present case–control study of 1705 IBD

patients (732 CD and 973 UC) and 1157 healthy controls,

variant allele carriers of COX-2 A-1195G were at increased

risk of UC compared to homozygous wildtype carriers, pri-

marily among patients diagnosed with UC at young age,

and, furthermore, among patients with extensive UC. Two

rare haplotypes encompassing the COX-2 A-1195G and

G-765C variant alleles were associated with the risk of

TABLE 3. Odds Ratios for the COX-2 Gene Polymorphisms in the Combined Scottish and Danish Never-smoker Patientswith Crohn’s Disease (CD), Ulcerative Colitis (UC), and Inflammatory Bowel Disease (IBD)

NCD NUC Ncontrol ORCD (95%CI)a P-value ORUC (95%CI)a P-value ORIBD (95%CI)a P-valueb

G-765C

GG 217 320 343 1.00 — 1.00 — 1.00 —

CG and CC 81 143 129 0.89 (0.62–1.28) 0.53 1.02 (0.75–1.38) 0.91 0.97 (0.73–1.27) 0.81

A-1195G

AA 186 267 310 1.00 — 1.00 — 1.00 —

AG and GG 112 196 162 1.25 (0.89–1.75) 0.19 1.47 (1.11–1.96) 0.01 1.37 (1.06–1.77) 0.02

T8473C

TT 129 194 203 1.00 — 1.00 — 1.00 —

CT and CC 169 269 269 0.95 (0.69–1.32) 0.76 0.99 (0.75–1.31) 0.93 0.98 (0.76–1.25) 0.85

Statistical analyses included subjects where all information was available.aAdjusted for country, age, and gender.bP-value for estimate.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011 COX-2 Polymorphisms and IBD

941

UC. The COX-2 A-1195G variant genotype was associated

with the risk of UC and of IBD among never smokers,

whereas no gene effects were detected among past or cur-

rent smokers.

Our study suggests that COX-2 expression is related

to the risk of UC. Furthermore, the risk seems to be related

to low expression levels of COX-2 conferred primarily by

the COX-2 A-1195G variant allele.26 Carriers of the COX-2T8473C T-allele were at risk of UC in a small Dutch

study.30 The COX-2 T8473C T-allele has been associated

with a low COX-2 function.26 Our findings are thus in

accordance with the Dutch study as both studies suggest

that the ‘‘low-level’’ alleles are associated with UC. Fur-

thermore, our study suggests that two rare haplotypes

encompassing the COX-2 A-1195G and G-765C variant al-

leles were at risk of UC. This haplotype resulted in

low expression levels in in vitro promoter studies.26 A

weak association between IBD and the homozygous COX-2G-765C variant genotype has previously been found in the

Danish study group, where the COX-2 A-1195G polymor-

phism was not assessed.31

All in all, our study suggests that a low level of

COX-2 may dispose to UC. Upon stimulation of the TLRs

by luminal pathogens a lower COX-2 response in the

variant allele carriers may lead to a lower production of

COX-2 dependent prostaglandins.15,16 Prostaglandins E2,

D2, and 15-deoxy-J2 mediate antiinflammatory effects via the

NF-jB and PPAR-c pathways.15 Thereby, a lower COX-2

activity may negatively affect the regulation of mucosal

inflammation and mucosal integrity, leading to chronic

TABLE 4A. Odds Ratios (OR) for COX-2 Genotypes in Scottish and Danish Patients with Crohn’s Disease (CD) andUlcerative Colitis (UC) Dependent on Disease Location

Crohn’s Disease

L1 L2 L3

Ncases Ncontrols OR (95%CI)a P-valueb Ncases Ncontrols OR (95%CI)a P-valueb Ncases Ncontrols OR (95%CI)a P-valueb

COX2 G765C

GG 199 1646 1.00 — 155 1690 1.00 — 129 1716 1.00 —

CG and CC 87 652 1.07 (0.82–1.40) 0.62 62 677 0.90 (0.66–1.23) 0.52 49 690 0.90 (0.64–1.27) 0.55

COX2 A1195G

AA 194 1429 1.00 — 134 1489 1.00 — 102 1521 1.00 —

AG and GG 92 869 0.78 (0.60–1.01) 0.06 83 878 1.04 (0.78–1.39) 0.79 76 885 1.28 (0.94–1.74) 0.12

COX2 T8473C

TT 105 980 1.00 — 91 994 1.00 — 81 1004 1.00 —

CT and CC 181 1318 1.27 (0.99–1.64) 0.06 126 1373 0.97 (0.73–1.30) 0.86 97 1402 0.85 (0.62–1.15) 0.28

Ulcerative Colitis

E1 E2 E3

Ncases Ncontrols OR (95%CI)a P-valueb Ncases Ncontrols OR (95%CI)a P-valueb Ncases Ncontrols OR (95%CI)a P-valueb

COX2 G765C

GG 199 1646 1.00 — 250 1595 1.00 — 211 1634 1.00 —

CG and CC 77 662 1.01 (0.77–1.34) 0.93 103 636 0.99 (0.77–1.26) 0.91 96 643 1.03 (0.79–1.34) 0.83

COX2 A1195G

AA 175 1448 1.00 — 214 1409 1.00 — 175 1448 1.00 —

AG and GG 101 860 0.98 (0.75–1.27) 0.86 139 822 1.11 (0.88–1.40) 0.38 132 829 1.32 (1.03–1.69) 0.03

COX2 T8473C

TT 126 959 1.00 — 147 938 1.00 — 118 967 1.00 —

CT and CC 150 1349 0.86 (0.67–1.10) 0.23 206 1293 1.00 (0.80–1.26) 0.98 189 1310 1.15 (0.90–1.48) 0.27

Statistical analyses included subjects where all information was available.aAdjusted for country.bP-value for estimate.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011Andersen et al

942

inflammation.13 The association found between COX-2 A-

1195G and UC seems to be carried by an association

among never smokers, as we did not detect an association

among smokers or past smokers due to a very strong risk-

modifying effect of smoking. Smoking induces COX-2expression33,34 and, moreover, reduces the risk of UC.35

One may therefore speculate that induction of COX-2 by

one or more of the more than 3000 different substances in

cigarette smoke may contribute to the healing process in

UC. Nicotine has previously been reported to improve

UC.52 Differences in exposure to IBD-modifying environ-

mental factors may contribute to differential effects of

known IBD risk genes in the northern and central parts of

Europe.9 Therefore, including smoking status may be

essential in order to study IBD etiology.

It is important to stress the strengths and limitations

of the present study. Hence, our results could be due to

population heterogeneity. A priori, we did not expect

TABLE 4B. Odds Ratios (OR) for COX-2 Genotypes in Scottish and Danish Patients with Crohn’s Disease (CD) andUlcerative Colitis (UC) Dependent on Age at Diagnosis

�40 >40

Ncases Ncontrols OR (95%CI)a P-valueb Ncases Ncontrols OR (95%CI)a P-valueb

Crohn’s Disease

COX2 G765C

GG 374 1471 1.00 — 136 1709 1.00 —

CG and CC 162 577 1.03 (0.83–1.27) 0.79 47 692 0.81 (0.57–1.14) 0.22

COX2 A1195G

AA 336 1287 1.00 — 120 1503 1.00 —

AG and GG 200 761 1.00 (0.82–1.22) 0.99 63 898 0.87 (0.64–1.20) 0.40

COX2 T8473C

TT 214 871 1.00 — 78 1007 1.00 —

CT and CC 322 1177 1.09 (0.90–1.33) 0.37 105 1394 0.96 (0.71–1.30) 0.78

Ulcerative Colitis

COX2 G765C

GG 432 1413 1.00 — 244 1601 1.00 —

CG and CC 170 569 0.93 (0.76–1.14) 0.48 112 627 1.13 (0.88–1.44) 0.34

COX2 A1195G

AA 346 1277 1.00 — 1391 232 1.00 —

AG and GG 256 705 1.34 (1.11–1.62) 0.002 837 124 0.88 (0.70–1.12) 0.30

COX2 T8473C

TT 266 819 1.00 — 134 951 1.00 —

CT and CC 336 1163 0.88 (0.73–1.05) 0.16 222 1277 1.22 (0.97–1.54) 0.09

Statistical analyses included subjects where all information was available.aAdjusted for country.bP-value for estimate.

TABLE 5. Odds Ratios for the COX-2 Gene Haplotypes inthe Scottish and Danish Patients with InflammatoryBowel Disease

Ncases Ncontrols OR 95% CI P-valueb

AGTa 1711 951 1.00 —

ACT 30 15 1.11 (0.60–2.08) 0.74

GGT 390 220 0.99 (0.82–1.18) 0.87

GCT 22 3 4.08 (1.22–13.65) 0.02

GCC 125 47 1.48 (1.05–2.09) 0.03

GGC 166 78 1.18 (0.89–1.57) 0.24

AGC 551 298 1.03 (0.87–1.21) 0.74

ACC 359 202 0.99 (0.82–1.19) 0.90

Statistical analyses included subjects where all information was available.aThe polymorphisms are in the following order: COX-2 A-1195G, COX-2G-765C, COX-2 T8473C.bP-value for estimate.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011 COX-2 Polymorphisms and IBD

943

heterogeneity between the Scottish and Danish population

as the contribution of NOD2 variants to CD susceptibility

in the Scottish and Danish population is markedly smaller

than in the central part of Europe.8,9,53 Moreover, the

COX-2 genotype and haplotype distributions in the Scottish

and the Danish study groups were similar and in agreement

with what has previously been found in the Danish popu-

lation (Table 6).27,28,42,45,47,48,50,51 Therefore, differences

in study group stratification are an unlikely reason for

the result. Next, in light of the number of statistical tests

performed, the results may be due to chance findings,

i.e., false-positive findings. However, the study group

consisted of more than 1700 IBD patients and 1100 con-

trols, corresponding to a power of more than 90% to

detect an OR of 1.3 for the combined group and of 1.4

for CD or UC alone.54 Moreover, the association was

stronger among younger patients and among patients

with more extensive disease, a pattern that is expected

for genetically susceptible groups. Still, the risk may be

modified by environmental factors and gene–environmen-

tal interactions, which stresses the need for large, well-

characterized study groups.

TABLE 6. Minimal Allelic Frequencies (MAF) for the COX-2 Gene Polymorphisms in the Scottish and Danish StudyGroups; Crohn’s Disease (CD), Ulcerative Colitis (UC), Inflammatory Bowel Disease (IBD), and Controls

N G-765C P-value A-1195G P-value T8473C P-valueb References

CD

Cases Scottish 410 0.16 0.69 0.21 0.25 0.35 0.24

Cases Danish 309 0.15 0.19 0.38

Controls Scottish 154 0.17 0.16 0.36

Control Danish 753 0.14 0.20 0.34

UC

Cases Scottish 495 0.19 0.61 0.21 0.42 0.38 0.26

Cases Danish 463 0.13 0.22 0.32

Controls Scottish 154 0.17 0.16 0.36

Control Danish 753 0.14 0.20 0.34

IBD

Cases Scottish 905 0.18 0.92 0.21 0.30 0.37 0.63

Cases Danish 772 0.14 0.21 0.34

Controls Scottish 154 0.17 0.16 0.36

Control Danish 753 0.14 0.20 0.34

Caucasian control groups

Danish (colorectal cancer) 765 0.202 0.139 0.356 Andersen (28)

Danish (breast cancer) 354 0.165 0.143 0.341 (unpubl V. Andersen)

Danish (basal cell carcinoma) 304 0.173 0.150 0.366 Vogel (41)

Danish (lung cancer) 744 0.202 0.354 Vogel (42)

Swedish (Parkinson) 616 0.30 Hakansson (43)

Swedish (prostata c) 782 0.381 Hedelin (44)

HapMap European 60 0.142 0.175 0.367 HapMap (45)

Mostly Caucasian (breast c) 8135 0.163 0.208 0.338 BPC3 (46)

Mostly Caucasian (prostata c) 8604 0.166 0.201 0.348 BPC3 (46)

Dutch (CRC) 399/391a 0.215 0.300 Siezen (47)

373/744a 0.287 Siezen(47)

Dutch colorectal adenomas 373/689a 0.224 0.301 Siezen (48)

Minnesota (colorectaladenomas) 584 0.170 Ulrich (49)

Chinese control groups

Chinese (MC) 643 0.488 0.064 0.179 Gao (50)

Chinese (CRC) 1300 0.500 0 Tan (51)

MAF for control groups/populations from other studies are added for comparison. Statistical analyses included subjects where all information wasavailable.aDisagreement between the number stated and the sum of the genotypes.bP-value for heterogeneity between countries.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011Andersen et al

944

CONCLUSIONIn conclusion, COX-2 A-1195G was associated with

the risk of UC, especially among never smokers, with risk

of diagnosis of UC before age 40 years, and with extensive

UC, suggesting that genetically determined low expression

levels of COX-2 may predispose to UC. Inclusion of smok-

ing status may be essential for the evaluation of the role of

genetic predisposition in IBD.

ACKNOWLEDGMENTSWe thank Lars Bentsen and Anne-Karin Jensen for

expert technical assistance. We thank the patients and

donors for participation. We thank the staff at the Library

and Information Department, Viborg Regional Hospital.

REFERENCES1. Andersen V, Ernst A, Christensen J, et al. The polymorphism

rs3024505 proximal to IL-10 is associated with risk of ulcerative coli-tis and Crohns disease in a Danish case-control study. BMC MedGenet.2010;11:82.

2. Lees CW, Satsangi J. Genetics of inflammatory bowel disease: impli-cations for disease pathogenesis and natural history. Expert Rev Gas-troenterol Hepatol. 2009;3:513–534.

3. Ernst A, Andersen V, Ostergaard M, et al. Genetic variants of gluta-thione S-transferases mu, theta, and pi display no susceptibility toinflammatory bowel disease in the Danish population. Scand J Gastro-enterol. 2010 [Epub ahead of print].

4. Duerr RH. Update on the genetics of inflammatory bowel disease. JClin Gastroenterol. 2003;37:358–367.

5. Brant SR, Shugart YY. Inflammatory bowel disease gene hunting bylinkage analysis: rationale, methodology, and present status of thefield. Inflamm Bowel Dis. 2004;10:300–311.

6. Economou M, Pappas G. New global map of Crohn’s disease:Genetic, environmental, and socioeconomic correlations. InflammBowel Dis. 2008;14:709–720.

7. Jacobsen BA, Fallingborg J, Rasmussen HH, et al. Increase in inci-dence and prevalence of inflammatory bowel disease in northern Den-mark: a population-based study, 1978–2002. Eur J GastroenterolHepatol. 2006;18:601–606.

8. Ernst A, Jacobsen B, Ostergaard M, et al. Mutations in CARD15 andsmoking confer susceptibility to Crohn’s disease in the Danish popula-tion. Scand J Gastroenterol. 2007;42:1445–1451.

9. Hugot JP, Zaccaria I, Cavanaugh J, et al. Prevalence of CARD15/NOD2 mutations in Caucasian healthy people. Am J Gastroenterol.2007;102:1259–1267.

10. Baumgart DC, Carding SR. Inflammatory bowel disease: cause andimmunobiology. Lancet. 2007;369:1627–1640.

11. Strober W, Fuss I, Mannon P. The fundamental basis of inflammatorybowel disease. J Clin Invest. 2007;117:514–521.

12. Xavier RJ, Podolsky DK. Unravelling the pathogenesis of inflamma-tory bowel disease. Nature. 2007;448:427–434.

13. Fukata M, Chen A, Klepper A, et al. Cox-2 is regulated by Toll-likereceptor-4 (TLR4) signaling: role in proliferation and apoptosis in theintestine. Gastroenterology. 2006;131:862–877.

14. Tsatsanis C, Androulidaki A, Venihaki M, et al. Signalling networks reg-ulating cyclooxygenase-2. Int J Biochem Cell Biol. 2006;38:1654–1661.

15. Scher JU, Pillinger MH. The anti-inflammatory effects of prostaglan-dins. J Investig Med. 2009;57:703–708.

16. Wallace JL. COX-2: a pivotal enzyme in mucosal protection and reso-lution of inflammation. Sci World J. 2006;6:577–588.

17. Sheibanie AF, Yen JH, Khayrullina T, et al. The proinflammatoryeffect of prostaglandin E2 in experimental inflammatory bowel diseaseis mediated through the IL-23—>IL-17 axis. J Immunol. 2007;178:8138–8147.

18. Karmeli F, Cohen P, Rachmilewitz D. Cyclo-oxygenase-2 inhibitorsameliorate the severity of experimental colitis in rats. Eur J Gastroen-terol Hepatol. 2000;12:223–231.

19. Reuter BK, Asfaha S, Buret A, et al. Exacerbation of inflammation-associated colonic injury in rat through inhibition of cyclooxygenase-2. J Clin Invest. 1996;98:2076–2085.

20. Okayama M, Hayashi S, Aoi Y, et al. Aggravation by selective COX-1 and COX-2 inhibitors of dextran sulfate sodium (DSS)-induced co-lon lesions in rats. Dig Dis Sci. 2007;52:2095–2103.

21. Kapoor M, Kojima F, Yang L, et al. Sequential induction of pro- andanti-inflammatory prostaglandins and peroxisome proliferators-acti-vated receptor-gamma during normal wound healing: a time coursestudy. Prostaglandins Leukot Essent Fatty Acids. 2007;76:103–112.

22. Rajakariar R, Yaqoob MM, Gilroy DW. COX-2 in inflammation andresolution. Mol Interv. 2006;6:199–207.

23. Mizuno H, Sakamoto C, Matsuda K, et al. Induction of cyclooxygen-ase 2 in gastric mucosal lesions and its inhibition by the specificantagonist delays healing in mice. Gastroenterology. 1997;112:387–397.

24. Hendel J, Nielsen OH. Expression of cyclooxygenase-2 mRNA inactive inflammatory bowel disease. Am J Gastroenterol. 1997;92:1170–1173.

25. Singer II, Kawka DW, Schloemann S, et al. Cyclooxygenase 2 isinduced in colonic epithelial cells in inflammatory bowel disease. Gas-troenterology. 1998;115:297–306.

26. Zhang X, Miao X, Tan W, et al. Identification of functional geneticvariants in cyclooxygenase-2 and their association with risk of esopha-geal cancer. Gastroenterology. 2005;129:565–576.

27. Vogel U, Christensen J, Wallin H, et al. Polymorphisms in COX-2,NSAID use and risk of basal cell carcinoma in a prospective study ofDanes. Mutat Res. 2007;617:138–146.

28. Andersen V, Ostergaard M, Christensen J, et al. Polymorphisms in thexenobiotic transporter multidrug resistance 1 (MDR1) gene and inter-action with meat intake in relation to risk of colorectal cancer in aDanish prospective case-cohort study. BMC Cancer. 2009;9:407.

29. Sanak M, Szczeklik W, Szczeklik A. Association of COX-2 gene hap-lotypes with prostaglandins production in bronchial asthma. J AllergyClin Immunol. 2005;116:221–223.

30. Cox DG, Crusius JB, Peeters PH, et al. Haplotype of prostaglandinsynthase 2/cyclooxygenase 2 is involved in the susceptibility toinflammatory bowel disease. World J Gastroenterol. 2005;11:6003–6008.

31. Ostergaard M, Ernst A, Labouriau R, et al. Cyclooxygenase-2, multi-drug resistance 1, and breast cancer resistance protein gene polymor-phisms and inflammatory bowel disease in the Danish population.Scand J Gastroenterol. 2009;44:65–73.

32. Hermann M, Krum H, Ruschitzka F. To the heart of the matter: cox-ibs, smoking, and cardiovascular risk. Circulation. 2005;112:941–945.

33. Moraitis D, Du B, De Lorenzo MS, et al. Levels of cyclooxygenase-2are increased in the oral mucosa of smokers: evidence for the role ofepidermal growth factor receptor and its ligands. Cancer Res. 2005;65:664–670.

34. Badawi AF, Habib SL, Mohammed MA, et al. Influence of cigarettesmoking on prostaglandin synthesis and cyclooxygenase-2 gene expres-sion in human urinary bladder cancer. Cancer Invest. 2002;20:651–656.

35. Mahid SS, Minor KS, Soto RE, et al. Smoking and inflammatorybowel disease: a meta-analysis. Mayo Clin Proc. 2006;81:1462–1471.

36. Aldhous MC, Drummond HE, Anderson N, et al. Does cigarettesmoking influence the phenotype of Crohn’s disease? Analysis usingthe Montreal classification. Am J Gastroenterol. 2007;102:577–588.

37. Aldhous MC, Drummond HE, Anderson N, et al. Smoking habit andload influence age at diagnosis and disease extent in ulcerative colitis.Am J Gastroenterol. 2007;102:589–597.

38. Podolsky DK. Inflammatory bowel disease. N Engl J Med. 2002;347:417–429.

39. Campa D, Zienolddiny S, Maggini V, et al. Association of a commonpolymorphism in the cyclooxygenase 2 gene with risk of non-smallcell lung cancer. Carcinogenesis. 2004;25:229–235.

40. Greenland S. Dose-response and trend analysis in epidemiology: alter-natives to categorical analysis. Epidemiology. 1995;6:356–365.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011 COX-2 Polymorphisms and IBD

945

41. Vogel U, Christensen J, Wallin H, et al. Polymorphisms in COX-2,NSAID use and risk of basal cell carcinoma in a prospective study ofDanes. Mutat Res. 2007;617:138–146.

42. Vogel U, Christensen J, Wallin H, et al. Polymorphisms in genesinvolved in the inflammatory response and interaction with NSAIDuse or smoking in relation to lung cancer risk in a prospective study.Mutat Res. 2008;639:89–100.

43. Hakansson A, Bergman O, Chrapkowska C, et al. Cyclooxygenase-2polymorphisms in Parkinson’s disease. Am J Med Genet B Neuropsy-chiatr Genet. 2007;144B:367–369.

44. Hedelin M, Chang ET, Wiklund F, et al. Association of frequent con-sumption of fatty fish with prostate cancer risk is modified by COX-2polymorphism. Int J Cancer. 2007;120:398–405.

45. HapMap: Rs689466:http://www.ncbi.nlm.nih.gov/projects/SNP/snp_ref.cgi?rs¼689466. 2010.

46. Dossus L, Kaaks R, Canzian F, et al. PTGS2 and IL6 Genetic Varia-tion and Risk of Breast and Prostate Cancer: results from the Breastand Prostate Cancer Cohort Consortium (BPC3). Carcinogenesis.2010;31:455–461.

47. Siezen CL, van Leeuwen AI, Kram NR, et al. Colorectal adenomarisk is modified by the interplay between polymorphisms in arachi-donic acid pathway genes and fish consumption. Carcinogenesis.2005;26:449–457.

48. Siezen CL, Bueno-de-Mesquita HB, Peeters PH, et al. Polymorphismsin the genes involved in the arachidonic acid-pathway, fish consumptionand the risk of colorectal cancer. Int J Cancer. 2006;119:297–303.

49. Ulrich CM, Whitton J, Yu JH, et al. PTGS2 (COX-2)-765G > C pro-moter variant reduces risk of colorectal adenoma among nonusers ofnonsteroidal anti-inflammatory drugs. Cancer Epidemiol BiomarkersPrev. 2005;14:616–619.

50. Gao J, Ke Q, Ma HX, et al. Functional polymorphisms in the cycloox-ygenase 2 (COX-2) gene and risk of breast cancer in a Chinese popu-lation. J Toxicol Environ Health A. 2007;70:908–915.

51. Tan W, Wu J, Zhang X, et al. Associations of functional polymor-phisms in cyclooxygenase-2 and platelet 12-lipoxygenase with risk ofoccurrence and advanced disease status of colorectal cancer. Carcino-genesis. 2007;28:1197–1201.

52. McGrath J, McDonald JW, Macdonald JK. Transdermal nicotine forinduction of remission in ulcerative colitis. Cochrane Database SystRev. 2004;CD004722.

53. Arnott ID, Nimmo ER, Drummond HE, et al. NOD2/CARD15, TLR4and CD14 mutations in Scottish and Irish Crohn’s disease patients:evidence for genetic heterogeneity within Europe?Genes Immun.2004;5:417–425.

54. Genetic Power Calculator; http://pngu.mgh.harvard.edu/�purcell/gpc/cc2.html. 2009.

Inflamm Bowel Dis � Volume 17, Number 4, April 2011Andersen et al

946