AI

L*d

BbiloR2esmpmiy1r3ct4CtyCwet

TtaiWItapmalt

GASTROENTEROLOGY 2006;130:1588–1594

cute Gastroenteritis Is Followed by an Increased Risk ofnflammatory Bowel Disease

UIS ALBERTO GARCÍA RODRÍGUEZ,* ANA RUIGÓMEZ,* and JULIÁN PANÉS‡

Centro Español de Investigación Farmacoepidemiológica, Madrid; and ‡Servei de Gastroenterologia, Hospital Clínic de Barcelona, Institut

’Investigacions Biomèdiques August Pi i Sunyer, Barcelona, Spain

tafitii

bSvpteoi

stioiItuq

aufc2mP

e

ackground & Aims: Bacterial intestinal infections haveeen implicated as a possible cause of exacerbation of

nflammatory bowel disease (IBD). We explored the re-ationship between infectious gastroenteritis and theccurrence of IBD using data from the General Practiceesearch Database. Methods: A cohort of patients aged0–74 years with an episode of acute infectious gastro-nteritis (n � 43,013) was identified. From the sameource population, an age-, sex-, and calendar time-atched control group free of gastroenteritis was sam-

led (n � 50,000). Both cohorts were followed up for aean duration of 3.5 years. Results: The estimated

ncidence rate of IBD was 68.4 per 100,000 person-ears after an episode of gastroenteritis and 29.7 per00,000 person-years in the control cohort. The hazardatio of IBD was 2.4 (95% confidence interval [CI], 1.7–.3) in the gastroenteritis cohort compared with theontrol cohort, and the excess risk was greater duringhe first year after the infective episode (hazard ratio,.1; 95% CI, 2.2–7.4). The relative risk of developingrohn’s disease in the gastroenteritis cohort was greaterhan that of ulcerative colitis, especially during the firstear after the infective episode (hazard ratio, 6.6; 95%I, 1.9–22.4). Conclusions: Our results are compatibleith the hypothesis that infectious agents causing anpisode of infectious gastroenteritis could play a role inhe initiation and/or exacerbation of IBD.

he prevailing theory of the pathogenesis of inflam-matory bowel disease (IBD) suggests that the intes-

inal immune system is inappropriately activated due toconfluence of genetic and environmental factors, lead-

ng to the generation of inflammatory tissue damage.hile critical to the understanding and treatment of

BD, little is known about the proximal events that sethe process in motion. In addition to the use of tobaccond nonsteroidal, anti-inflammatory drugs, infections byathogens such as Mycobacterium paratuberculosis, Listeriaonocytogenes, and paramyxoviruses have been suggesteds etiologic agents in IBD. However, a recently pub-ished critical appraisal of the literature concluded that

he current evidence does not support a causal role for

hese infectious agents in the etiology of IBD.1 Anlternative explanation compatible with the failure tond specific pathogenic agents in IBD is the hypothesishat enteric pathogens may trigger an initial overshoot-ng response or a defect in down-regulating the mucosalmmune response, leading to chronic inflammation.

Several clinical observations support this notion. It haseen observed that following epidemics of Salmonella,higella, or Yersinia, a small percentage of patients de-elop typical IBD,2,3 but these analyses lacked a generalopulation comparison cohort. A study performing ahorough microbiologic study could detect concurrentnteral infections at the time of diagnosis of IBD in 21%f cases.4 Bacterial intestinal infections have also beenmplicated as a possible cause of exacerbation of IBD.5,6

In the past, we reported that the risk of irritable bowelyndrome was greatly increased after an episode of infec-ious gastroenteritis (GE).7 The questions of whethernfectious GE represents a risk factor for the developmentf IBD, the magnitude of this risk, the time span ofncreased risk after the episode of bacterial infection, or ifBD following infectious GE has some specific charac-eristics remain unanswered. In the current study, wesed a large population-based study to address theseuestions.

Materials and Methods

Study Population

We explored the relationship between infectious GEnd the occurrence of IBD, including Crohn’s disease (CD),lcerative colitis (UC), and indeterminate colitis, using datarom the General Practice Research Database. The database isomposed of computerized medical records of approximately000 general practitioners from the United Kingdom and isanaged by the United Kingdom’s Medicines and Healthcareroducts Regulatory Agency.8 The age and sex distribution of

Abbreviations used in this paper: CI, confidence interval; GE, gastro-nteritis; HR, hazard ratio; OR, odds ratio.© 2006 by the American Gastroenterological Association Institute

0016-5085/06/$32.00

doi:10.1053/j.gastro.2006.02.004

tGgsbqtabarmnftrIPPvtmar

d2dcpeawtppImPcetiaus2�

mrcc

tG

tcdc

(osrisssprtvDsa

cpIfitddr

fccrddcw

blsccpcc

May 2006 RISK OF IBD AFTER GASTROENTERITIS 1589

he population served by physicians collaborating with theeneral Practice Research Database is similar to that of theeneral UK population.9 Each medical practice must demon-trate competency at entering data into the electronic databaseefore their data are considered “up to standard.” Subse-uently, each practice is subject to monthly audits to ensurehat the quality of the data remains up to standard. Theseudits include examining the electronic data for weekly num-ers of consultations, completeness of drug indication data,nd completeness of data on births and cause of death. Dataecorded in the electronic record include demographic infor-ation, prescription information, clinical events and diag-

oses, preventive care, hospital admissions, cause of death, andree text. In addition, significant diagnoses occurring beforehe initiation of the electronic medical record are recordedetrospectively. Diagnoses are recorded using Oxford Medicalndexing System codes and more recently using Read codes.rescribed medications are recorded using codes issued by therescription Pricing Authority of the National Health Ser-ice.8,10 Several studies have shown that the clinical informa-ion in the electronic record is sufficiently accurate for use inost epidemiologic studies, including studies of IBD.11–13 In

ddition, prior research has documented very high rates ofecording diagnoses resulting from specialist consultations.14

Study Cohorts

We identified all registered patients aged 20–74 yearsuring the study period (January 1, 1992, to December 31,001) with a code suggesting an episode of infectious GEuring the study period. We applied the following eligibilityriteria. First, patients had to be free of cancer, alcohol abuse,rior GE, IBD, related gastrointestinal infectious disease, ornteritis/colitis at any time before the episode of GE. Inddition, patients with a code of diarrhea or rectal bleeding orith recorded use of specific IBD treatment in the year before

he episode of GE entry date were also excluded. For allatients identified with a code suggesting GE, computerizedatient profiles were produced and reviewed (n � 47,852).nformation included demographic data and all clinical infor-ation. Patient profiles were free of any personal identifiers.atients in whom an enteric pathogen was identified in a stoolulture performed for reasons other than an episode of GE werexcluded (n � 4820). Finally, 6414 patients were consideredo have “documented” bacterial GE with a specific bacteriasolated (Salmonella, Campylobacter, Shigella, or other bacteria),nd 36,599 patients were classified as having “bacteriologicallyndocumented” GE (clinical diagnosis of GE with negativetool culture or no mention of a stool culture recorded). Thesesubgroups formed our final cohort of acute infectious GE (N43,013).A comparison cohort of 50,000 individuals frequencyatched by age, sex, and calendar year to the GE cohort was

andomly sampled from the same source population (where theohort of GE was ascertained), applying the same eligibility

riteria as used in the ascertainment of the cohort of GE with f

he additional criteria of not having a recorded diagnosis ofE.

Follow-up to Ascertain Incident Cases ofIBD

We followed up all individuals in the 2 cohorts fromhe date of GE diagnosis or the random date in the comparisonontrol cohort. Follow-up ended at the first occurrence of IBDiagnosis, cancer, death, date of last data collection, or De-ember 31, 2001, whichever came first.

All patients with a coded diagnosis of UC (5631), CD5630CR; including regional enteritis [5630ER]), or “IBD nottherwise specified” (92N) were potentially eligible for inclu-ion in the study. Because previous studies have shown a lowereliability of the diagnostic code for “IBD not otherwise spec-fied” (92N),11 these patients were only considered cases ifpecific medication for the treatment of IBD, mainly me-alamine-containing drugs and prednisone, had been pre-cribed after the diagnosis was established. In this group ofatients, if a diagnosis of UC or CD appeared later in theecord, the latter was considered the final diagnosis; otherwise,he patients were categorized as IBD type unclassified. Thealidity and completeness of the General Practice Researchatabase for the diagnoses of IBD have been specifically as-

essed, and 92% of diagnoses recorded were found to beccurate.11

For all patients identified with one of the aforementionedodes of IBD (n � 335), we reviewed computerized patientrofiles and classified them into “definite” IBD, “possible”BD, and noncases. Definite cases were those patients with arst-ever recorded diagnosis of IBD together with specificreatment and/or confirmation by consultant letter or hospitalischarge (n � 152). Possible cases were those patients with aiagnosis of IBD without specialist confirmation (n � 15). Allemaining individuals were considered noncases (n � 168).

We sent a questionnaire to the general practitioners askingor confirmation of diagnosis among all definite and possibleases registered with collaborating practices (n � 83). Patientonfidentiality was always preserved. Based on all informationeceived from the general practitioners, we confirmed all 76efinite cases as IBD and 6 of 7 possible cases. As a result, weecided to include as confirmed IBD cases all other definiteases of IBD (n � 76) but not the other 8 possible cases forhich we could not get access to the general practitioner.

Analysis

Incidence rates of IBD in the 2 cohorts were calculatedy dividing the number of incident cases by the total fol-ow-up experience in each study cohort. Incidence rates ofpecific IBD diagnoses stratified by age and sex were alsoalculated. The risk of developing IBD in the GE cohortompared with the control cohort was estimated using Coxroportional hazards regression, adjusting by age, sex, andalendar year. Estimates of hazard ratios (HR) and their 95%onfidence intervals (CIs) were computed for the first year of

ollow-up and for the whole period of follow-up.

cFadgretTttwIapsp

(pcacwtrtbGcm

ti(SG

2dTaytbs1Iwpww

F

T

S

A

I

I

1590 GARCÍA RODRÍGUEZ ET AL GASTROENTEROLOGY Vol. 130, No. 6

To adjust for additional risk factors for IBD, a nestedase-control study within our 2 study cohorts was performed.or this analysis, we used all 158 incident cases of IBDscertained in the 2 cohorts and considered the date of firstiagnosis as the index date. A date during the study period wasenerated at random for every member of the 2 cohorts. If theandom date of a study member was included in his or herligible person-time of follow-up, we used this random date ashe index date and marked that patient as an eligible control.his mechanism (ie, incidence density sampling) allows that

he likelihood of being selected as a control is proportional tohe person-time at risk. Two thousand controls (free of IBD)ere frequency matched by sex, age, and calendar year to the

BD cases. Estimates of odds ratios (ORs) and 95% CIs of IBD,ssumed to be valid estimates of the relative risk, were com-uted using unconditional logistic regression adjusted for age,ex, calendar year, visits to the general practitioner in the yearrior, and other risk factors for IBD listed in Table 3.

Results

During a mean follow-up period of 3.5 yearsrange, 0–10 years; SD, 2.3 years), we ascertained 108atients with a first-ever diagnosis of IBD in the GEohort and 50 in the comparison cohort. Table 1 presentssummary description of IBD cases occurring in the GE

ohort and comparison cohort. Forty were diagnosedith CD, 64 with UC, and 4 as IBD type unclassified in

he GE cohort. The distribution by type of IBD occur-ing in the comparison cohort free of GE was similar tohat in the GE cohort; no differences were observedetween the documented (bacterial) and undocumentedE subgroups (data not shown). Among the bacterial GE

ohort, Campylobacter was the microorganism most com-

able 1. Summary of Incident Cases of IBD in the GE and Co

exMaleFemale

ge (y)Younger than 4040–5960�

BD diagnosisUCCDUnclassified IBD

nterval in months between IBD occurrence and the episodeof GE in the GE cohort or the random date in thecomparison cohort (mo)12–1213–24�24

only detected (4124 patients), followed by 1885 pa- a

ients with Salmonella infection, and Shigella was isolatedn 312 patients. The infection was due to other bacteriamost commonly Escherichia coli, Clostridium difficile, ortaphylococcus) in the remaining 93 patients with bacterialE.Figure 1 shows the incidence rate of IBD by age in thecohorts, and Figure 2 shows the corresponding inci-

ence rate of UC and CD in the 2 cohorts. As shown inable 2, the overall incidence rate of IBD in patientsfter an episode of GE was 68.4 per 100,000 person-ears: 77 (95% CI, 48–123.2) in the documented bac-erial GE cohort and 67 (95% CI, 54.5–82.2) in theacteriologically undocumented GE cohort. The corre-ponding estimate in the control cohort was 29.7 per00,000 person-years. The incidence rate of developingBD in Salmonella- and Campylobacter-infected patientsas 104 (95% CI, 50–214) and 59 (95% CI, 30–115)er 100,000 person-years, respectively. The HR of IBDas 2.4 (95% CI, 1.7–3.3) in the GE cohort comparedith the control cohort, and the excess risk was greater

igure 1. Incidence rate of IBD in the GE and control cohorts by

rison Cohorts

E cohort (n � 108) (%) Comparison cohort (n � 50) (%)

39 (36) 27 (54)69 (64) 23 (46)

37 (34) 16 (32)44 (41) 24 (48)27 (25) 10 (20)

64 (59) 31 (62)40 (37) 14 (28)4 (4) 5 (10)

4 (4) 4 (8)44 (41) 10 (20)23 (21) 8 (16)37 (34) 28 (56)

mpa

G

ge.

d(d((ucfiC

afI(dwU1ipOw

b5G

rcecptgsCotir

ppceetgdjtwoiim

Fb

T

P

CIHH

CIHH

CIHH

a

May 2006 RISK OF IBD AFTER GASTROENTERITIS 1591

uring the first year after the infective intestinal episodeHR, 4.1; 95% CI, 2.2–7.4). Patients with an episode ofocumented bacterial GE presented an HR of IBD of 2.695% CI, 1.9–10.8), and the corresponding HR was 2.395% CI, 1.6–3.3) in the patients with bacteriologicallyndocumented GE. The risk of developing CD in the GEohort was greater than that of UC, especially during therst year after the infective episode (CD HR, 6.6; 95%I, 1.9–22.4; UC HR, 3.7; 95% CI, 1.8–7.5).Table 3 presents the estimates of ORs of UC and CD

ssociated with GE after adjusting for a number of riskactors in the nested case-control analysis. The OR ofBD was 2.2 (95% CI, 1.5–3.2) in patients with GE: 2.295% CI, 1.2–4.0) in patients with bacteriologicallyocumented GE and 2.2 (95% CI, 1.5–3.3) in patientsith bacteriologically undocumented GE. The OR ofC was 2.1 (95% CI, 1.4–3.4) in the patients with GE:.7 (95% CI, 0.8–3.8) in the patients with bacteriolog-cally documented GE and 2.2 (95% CI, 1.4–3.6) in theatients with bacteriologically undocumented GE. TheR of CD was 2.9 (95% CI, 1.5–5.6) in the patientsith GE: 3.7 (95% CI, 1.4–9.9) in the patients with

igure 2. Incidence rate of CD and UC in the GE and control cohortsy age.

able 2. Incidence Rate and Relative Risk of IBD in the GE C

erson-years

ases of IBDncidence rate per 100,000 person-years (95% CI)R (95% CI)a

R in first year of follow-up (95% CI)a

ases of UCncidence rate per 100,000 person-years (95% CI)R (95% CI)a

R in first year of follow-up (95% CI)a

ases of CDncidence rate per 100,000 person-years (95% CI)R (95% CI)a

R in first year of follow-up (95% CI)a

HR estimated by Cox proportional hazards regression analysis including a

acteriologically documented GE and 2.8 (95% CI, 1.4–.4) in the patients with bacteriologically undocumentedE.No clear association was found between osteoarthritis,

heumatoid arthritis, depression, anxiety, asthma,hronic pulmonary disease, gastroesophageal reflux dis-ase or appendicitis, and UC or CD in the nested case-ontrol analysis. The increased OR for peptic ulcer inatients with CD, but not in those with UC, suggestshat some of these diagnoses may correspond to upperastrointestinal CD. As expected, smokers and formermokers both presented an increased risk of developingD, whereas a small increased risk of UC was observednly among former smokers. Among patients with GE,reatment with antibiotics within the 2 weeks after thenitial episode of GE was associated with an increasedisk of IBD (OR, 2.7; 95% CI, 1.6–4.6).

Discussion

The incidence rate of IBD in our cohort of 43,013atients with an episode of GE was 68 per 100,000erson-years, a doubling of the risk observed in theontrol cohort of individuals free of GE. To our knowl-dge, this is the first large epidemiologic study providingvidence on the association between an episode of GE andhe occurrence of IBD. The nested case-control analysisave results similar to the cohort analysis, with the soleifference of slightly smaller estimates of risk after ad-usting for a number of additional risk factors. Of note,he incidence rate of IBD in the control cohort fallsithin the range of highest estimates reported.15–17 Twof the factors that may contribute to the relatively highncidence of IBD observed in the studied populationnclude the geographic location and the source of infor-ation. Previous studies have shown that incidence rates

t Compared With the Control Cohort

trol cohort (n � 50,000) GE cohort (n � 43,013)

167,942 157,801

50 10829.7 (22.6–39.2) 68.4 (56.7–82.6)

1 2.4 (1.7–3.3)1 4.1 (2.2–7.4)

31 6418.5 (13.0–26.2) 40.6 (31.8–51.8)

1 2.3 (1.5–3.6)1 3.7 (1.8–7.5)

14 408.3 (5–14.0) 25.3 (18.6–34.5)

1 3.1 (1.7–5.7)1 6.6 (1.9–22.4)

ohor

Con

ge, sex, and calendar year.

afFCmcKrrauaovib

oattacmpt

tatCwws

tpaggtitdasatsmbs

T

C

ORDAACGPAS

B

U

a

b 7 con

1592 GARCÍA RODRÍGUEZ ET AL GASTROENTEROLOGY Vol. 130, No. 6

re considerably higher in northern Europe, where dataor this study were extracted, than in southern Europe.or example, an incidence of UC as high as 24.3 and ofD of 8.2 per 100,000 person-years have been docu-ented in Iceland,17 although in the same study the

orresponding figures in a population from the Unitedingdom were 9.2 and 3.2 per 100,000 person-years,

espectively. These incidence rates are closer to thoseeported in southern Europe for UC (8.7 per 100,000)18

nd for CD (3.9 per 100,000).17 Another factor contrib-ting to the high estimate of IBD incidence is the use ofprimary care database to identify incident cases insteadf hospital records. This results in the ascertainment ofirtually all cases occurring in the general population,ncluding those with relatively mild disease who may note admitted to hospital care.The current study also indicates that the excess risk

f IBD occurrence was higher during the first yearfter the infective intestinal episode, with an HR closeo 4 compared with an overall HR of about 2 duringhe entire period of follow-up. The HRs of both UCnd CD in the GE cohort were increased relative to theontrol cohort, suggesting that the intestinal infectionay trigger the initiation of a chronic inflammatory

rocess and that other factors, including genetic de-

able 3. Relative Risk of UC and CD Associated With GE, OtCase-Control Analysis

UC cases(n � 95) (%)

CD ca(n � 5

ohort statusCohort free of GE 31 (35.2) 14 (2GE cohort 64 (67.4) 40 (7

steoarthritis 21 (22.1) 9 (1heumatoid arthritis 3 (3.2) 3 (5epression 19 (20.0) 12 (2nxiety 12 (12.6) 10 (1sthma 13 (13.7) 10 (1hronic pulmonary disease 8 (8.4) 3 (5astroesophageal reflux 15 (15.8) 9 (1eptic ulcer 4 (42) 8 (1ppendicitis 4 (4.2) 3 (5mokingNonsmoker 50 (52.6) 19 (3Smoker 20 (21.1) 26 (4Former smoker 12 (12.6) 5 (9Unknown 13 (13.7) 4 (7

ody mass index (kg/m2)�20 2 (2.1) 7 (120–25 32 (33.7) 22 (425–30 28 (29.5) 11 (2�30 12 (12.6) 2 (3Unknown 21 (22.1) 12 (2

se of antibiotics after GE episodeb 18 (28.1) 9 (2

Estimates adjusted for age, sex, calendar year, visits to the generaOnly among patients belonging to the GE cohort: 108 cases and 87

erminants and other environmental factors such as t

obacco use, will determine the type of IBD that mayppear thereafter. In keeping with previous observa-ions,19,20 the nested case-control study confirmed thatD was significantly associated with smoking,hereas an association with UC was only observedith former smoker status, although not statistically

ignificant.In the group of patients with a positive stool cul-

ure, the incidence rate of IBD was increased in bothatients with Campylobacter and Salmonella infection,lthough the rate was slightly higher in the Salmonellaroup. Although the number of cases in each of theseroups of bacterial GE was small, the data suggesthat initiation of IBD is not exclusively linked tonfection by specific enteric bacteria. The observationhat the incidence rate of IBD after an episode ofocumented bacterial GE is similar to the incidencefter episodes of GE with negative or undocumentedtool culture, in which a significant proportion of casesre probably of viral origin, also supports the notionhat it is the presence of an acute inflammatory re-ponse in the gut, rather than a specific pathogen, thatay lead to initiation of IBD in genetically suscepti-

le individuals. One of the limitations of the currenttudy is that we may not determine if certain charac-

isk Factors for IBD, and Unrelated Diseases in a Nested

)Controls

(n � 2000) (%) UC ORa (95% CI) CD ORa (95% CI)

1123 (56.1) 1 1877 (43.9) 2.1 (1.4–3.4) 2.9 (1.5–5.6)339 (17.0) 1.3 (0.7–2.2) 0.7 (0.3–1.7)27 (1.4) 2.1 (0.6–7.9) 4.5 (1.0–20.0)

399 (20.0) 0.7 (0.4–1.3) 0.6 (0.3–1.4)294 (14.7) 0.7 (0.4–1.4) 1.1 (0.5–2.6)226 (11.3) 0.7 (0.4–1.4) 1.1 (0.5–2.5)60 (3.0) 2.2 (0.9–5.5) 0.8 (0.2–3.3)

219 (10.9) 1.3 (0.7–2.3) 1.2 (0.5–2.6)64 (3.2) 1.1 (0.4–3.2) 3.8 (1.5–9.7)

104 (5.2) 0.8 (0.3–2.4) 1.0 (0.3–3.6)

1108 (55.4) 1 1493 (24.6) 0.9 (0.5–1.5) 3.3 (1.7–6.2)139 (7.0) 1.6 (0.8–3.1) 1.7 (0.6–5.1)260 (13.0) 1.4 (07–3.2) 1.0 (0.3–3.5)

103 (5.1) 0.5 (0.1–2.0) 2.4 (0.9–6.4)662 (33.1) 1 1524 (26.2) 1.0 (0.6–1.8) 0.7 (0.3–1.5)232 (11.6) 0.9 (0.4–1.8) 0.2 (0.1–1.0)479 (23.9) 0.8 (0.4–1.6) 1.1 (0.5–2.7)129 (14.7) 2.9 (1.5–5.7) 2.0 (0.8–4.8)

titioner in the year prior, and all variables from the table.trols.

her R

ses4) (%

8.6)4.1)6.7).6)2.2)8.5)8.5).6)6.7)4.8).6)

5.2)8.1).3).4)

3.0)0.7)0.4).7)2.2)2.5)

l prac

eristics of the episode of GE, such as severity, may

mewdioc

trsmbtppKotctttitwpdceoic

bIolcthtCrsybadtim

It

1

1

1

1

1

1

1

1

1

1

2

May 2006 RISK OF IBD AFTER GASTROENTERITIS 1593

odify the risk of initiation of IBD after the acutepisode. We found that treatment of the GE episodeith antibiotics was associated with a greater risk ofeveloping IBD. Whether treatment with antibioticss a marker of severity of the episode of GE or is partf the mechanism underlying a greater risk of IBDannot be answered in our data set.

Previous studies have suggested a relationship be-ween infections with enteropathogenic organisms andelapses in patients with known IBD, although this istill a controversial issue. In a prospective study usingultiple biopsy specimens for histologic and micro-

iologic investigation in parallel to classic stool cul-ure, Weber et al found enteropathogenic bacteria in 6atients (9%) and Clostridium difficile toxin B in 5atients (8%) with a relapse of IBD.6 In another study,allinowski et al5 examined the prevalence of enter-

pathogenic bacteria in patients with surgicallyreated chronic IBD in whom a classic infectiousolitis had been ruled out by negative stool and in-estinal tissue cultures. Using polymerase chain reac-ion amplification, pathogenic E coli was identified inhe stool samples of 3 of 14 patients with UC, and thenvestigators concluded that E coli may contribute tohe disease progression in a small subgroup of patientsith UC. Although in the current study we excludedrevalent cases of IBD, patients receiving specificrugs to treat IBD, and individuals with a recordedode of diarrhea or rectal bleeding, the possibility stillxists that the episode of GE induced an exacerbationf a previously clinically silent IBD rather than thenitiation of the chronic intestinal inflammatory pro-ess in some of our cases.

This study provides no insight into the mechanismy which an episode of GE may lead to initiation ofBD. Recently, genetic variants of various componentsf the innate immune response, including the Toll-ike receptor 4 and the caspase recruitment domain-ontaining protein 15 gene (CARD15), which affecthe response of these receptors to lipopolysaccharide,ave shown to be associated with UC (Toll-like recep-or 4) and CD (Toll-like receptor 4 andARD15).21–23 Interestingly, variants of the Toll-like

eceptor 4 have been associated with late-onset IBD,pecifically with disease beginning after the age of 40ears,22 which is an unusual pattern of presentationut coincides with the observed pattern in patientsfter an episode of GE. It will be interesting toetermine in the future the relationship between cer-ain genetic determinants of disease susceptibility andntestinal infections as a possible trigger of IBD. This

ay provide new insights into the pathogenesis of

BD and offer the opportunity for a more rationalreatment.

References1. Loftus EV Jr. Clinical epidemiology of inflammatory bowel dis-

ease: incidence, prevalence, and environmental influences. Gas-troenterology 2004;126:1504–1517.

2. Treacher DF, Jewell DP. Yersinia colitis associated with Crohn’sdisease. Postgrad Med J 1985;61:173–174.

3. Powell SJ, Wilmot AJ. Ulcerative post-dysenteric colitis. Gut 1966;7:438–443.

4. Schumacher G, Kollberg B, Sandstedt B, Jorup C, Grillner L,Ljungh A, Mollby R. A prospective study of first attacks of inflam-matory bowel disease and non-relapsing colitis. Microbiologicfindings. Scand J Gastroenterol 1993;28:1077–1085.

5. Kallinowski F, Wassmer A, Hofmann MA, Harmsen D, Heese-mann J, Karch H, Herfarth C, Buhr HJ. Prevalence of enteropatho-genic bacteria in surgically treated chronic inflammatory boweldisease. Hepatogastroenterology 1998;45:1552–1558.

6. Weber P, Koch M, Heizmann WR, Scheurlen M, Jenss H, Hart-mann F. Microbic superinfection in relapse of inflammatory boweldisease. J Clin Gastroenterol 1992;14:302–308.

7. Rodriguez LA, Ruigomez A. Increased risk of irritable bowel syn-drome after bacterial gastroenteritis: cohort study. BMJ 1999;318:565–566.

8. Lis Y, Mann RD. The VAMP Research multi-purpose database inthe U.K. J Clin Epidemiol 1995;48:431–443.

9. The General Practice Research Database. Information for Re-searchers. London, UK: Office for National Statistics, 1994.

0. Garcia Rodriguez LA, Perez GS. Use of the UK General PracticeResearch Database for pharmacoepidemiology. Br J Clin Pharma-col 1998;45:419–425.

1. Lewis JD, Brensinger C, Bilker WB, Strom BL. Validity and com-pleteness of the General Practice Research Database for studiesof inflammatory bowel disease. Pharmacoepidemiol Drug Saf2002;11:211–218.

2. Gupta G, Gelfand JM, Lewis JD. Increased risk for demyelinatingdiseases in patients with inflammatory bowel disease. Gastroen-terology 2005;129:819–826.

3. Lewis JD, Aberra FN, Lichtenstein GR, Bilker WB, Brensinger C,Strom BL. Seasonal variation in flares of inflammatory boweldisease. Gastroenterology 2004;126:665–673.

4. Jick H, Jick SS, Derby LE. Validation of information recorded ongeneral practitioner based computerised data resource in theUnited Kingdom. BMJ 1991;302:766–768.

5. Bjornsson S, Johannsson JH. Inflammatory bowel disease inIceland, 1990–1994: a prospective, nationwide, epidemiologicalstudy. Eur J Gastroenterol Hepatol 2000;12:31–38.

6. Ekbom A, Helmick C, Zack M, Adami HO. The epidemiology ofinflammatory bowel disease: a large, population-based study inSweden. Gastroenterology 1991;100:350–358.

7. Shivananda S, Lennard-Jones J, Logan R, Fear N, Price A, Car-penter L, van Blankenstein M. Incidence of inflammatory boweldisease across Europe: is there a difference between north andsouth? Results of the European Collaborative Study on Inflam-matory Bowel Disease (EC-IBD). Gut 1996;39:690–697.

8. Lanfranchi GA, Michelini M, Brignola C, Campieri M, Cortini C,Marzio L. Epidemiological study on intestinal inflammatory dis-eases in the Province of Bologna [in Italian]. G Clin Med 1976;57:235–245.

9. Calkins BM. A meta-analysis of the role of smoking in inflamma-tory bowel disease. Dig Dis Sci 1989;34:1841–1854.

0. Birrenbach T, Bocker U. Inflammatory bowel disease and smok-ing: a review of epidemiology, pathophysiology, and therapeutic

implications. Inflamm Bowel Dis 2004;10:848–859.

2

2

2

C2

SJ

ba

1594 GARCÍA RODRÍGUEZ ET AL GASTROENTEROLOGY Vol. 130, No. 6

1. Franchimont D, Vermeire S, El Housni H, Pierik M, Van Steen K,Gustot T, Quertinmont E, Abramowicz M, Van Gossum A, DeviereJ, Rutgeerts P. Deficient host-bacteria interactions in inflamma-tory bowel disease? The toll-like receptor (TLR)-4 Asp299gly poly-morphism is associated with Crohn’s disease and ulcerativecolitis. Gut 2004;53:987–992.

2. Oostenbrug LE, Drenth JP, de Jong DJ, Nolte IM, Oosterom E, vanDullemen HM, van der Linde K, te Meerman GJ, van der Steege G,Kleibeuker JH, Jansen PL. Association between Toll-like receptor 4 andinflammatory bowel disease. Inflamm Bowel Dis 2005;11:567–575.

3. Eckmann L, Karin M. NOD2 and Crohn’s disease: loss or gain of

function? Immunity 2005;22:661–667. a

Received October 2, 2005. Accepted January 18, 2006.Address requests for reprints to: Luis Alberto García Rodríguez, MD,

entro Español de Investigación Farmacoepidemiológica, Almirante8, 2° 28004 Madrid, Spain. e-mail: lagarcia@ceife.es.Supported in part by a research grant from AstraZeneca R&D,

weden, and grant C03/02 from Instituto de Salud Carlos III (to.P.).

The authors thank the staff at the General Practice Research Data-ase, the participating general practitioners for their collaboration,nd the Boston Collaborative Drug Surveillance Program for providing

ccess to the database.