Gastrointestinal Infection and Risk of Microscopic Colitis ...

Page 1

Gastroenterology 2021;160:1599–1607

Gastrointestinal Infection and Risk of Microscopic Colitis: ANationwide Case-Control Study in Sweden

CLINICAL

AT

Hamed Khalili,1,2,3,4 Jordan E. Axelrad,5 Bjorn Roelstraete,6 Ola Olén,3,7 Mauro D’Amato,3,7

and Jonas F. Ludvigsson8,9,10,11

1Massachusetts General Hospital, Clinical and Translational Epidemiology Unit, Mongan Institute, Boston, Massachusetts;2Massachusetts General Hospital, Gastroenterology Unit, Harvard Medical School, Boston, Massachusetts; 3Division of ClinicalEpidemiology, Department of Medicine Solna, Karolinska Institutet, Stockholm, Sweden; 4Broad Institute of MIT and Harvard,Cambridge Massachusetts; 5Inflammatory Bowel Disease Center at NYU Langone Health, Division of Gastroenterology,Department of Medicine, NYU Grossman School of Medicine, New York, New York; 6Sachs’ Children and Youth Hospital,Stockholm South General Hospital, Stockholm, Sweden; 7School of Biological Sciences, Monash University, Clayton, Victoria,Australia; 8Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Stockholm, Sweden; 9Department ofPediatrics, Orebro University Hospital, Orebro, Sweden; 10Division of Epidemiology and Public Health, School of Medicine,University of Nottingham, UK; and 11Department of Medicine, Columbia University College of Physicians and Surgeons, NewYork, New York

Abbreviations used in this article: aOR, adjusted odds ratio; CC, collage-nous colitis; CI, confidence interval; ESPRESSO, Epidemiology Strength-ened by histoPathology Reports in Sweden; HR, hazard ratio; IBD,inflammatory bowel disease; ICD, International Classification of Disease;LC, lymphocytic colitis; MC, microscopic colitis; MHC, major histocom-patibility complex; NSAID, nonsteroidal anti-inflammatory drug; OR, oddsratio; PPI, proton pump inhibitor; SNOMED, Systematized Nomenclature

BACKGROUND AND AIMS: Gastrointestinal infections havebeen linked to changes in the composition and function of gutmicrobiome and development of inflammatory bowel diseases.We therefore sought to examine the relationship betweengastroenteritis and risk of microscopic colitis (MC). METHODS:We conducted a case-control study of all adult patients with MCdiagnosed between 1990 and 2016 in Sweden matched to up to5 general population controls according to age, sex, calendaryear, and county. Cases of MC were identified using Systema-tized Nomenclature of Medicine codes from the ESPRESSO(Epidemiology Strengthened by histoPathology Reports inSweden) study, a cohort of gastrointestinal pathology reportsfrom all 28 pathology centers in Sweden. We used logisticregression modeling to estimate adjusted odds ratios (aORs)and 95% confidence intervals (CIs). RESULTS: ThroughDecember of 2016, we matched 13,468 MC cases to 64,479controls. The prevalence of previous diagnosed gastrointestinalinfection was 7.5% among patients with MC, which wassignificantly higher than in controls (3.0%, Pcomparison < .001).After adjustment, gastroenteritis was associated with anincreased risk of MC (aOR 2.63; 95% CI 2.42–2.85). Amongspecific pathogens, Clostridioides difficile (aOR 4.39; 95% CI3.42–5.63), Norovirus (aOR 2.87; 95% CI 1.66–4.87), andEscherichia species (aOR 3.82; 95% CI 1.22–11.58), but notSalmonella species, were associated with an increased risk ofMC. The association between gastrointestinal infections andrisk of MC was stronger for collagenous subtype (aOR 3.23;95% CI 2.81–3.70) as compared with lymphocytic colitis (aOR2.51; 95% CI 2.28–2.76; Pheterogeneity ¼ .005). The associationsremained significant after adjustment for immune-mediatedconditions and polypharmacy and when compared with unaf-fected siblings. CONCLUSION: In a nationwide study, we foundthat gastrointestinal infection, particularly Clostridioides diffi-cile, is associated with an increased risk of subsequent MC. Thisstudy was approved by the Regional Ethics Committee, Stock-holm, Sweden (Protocol no. 2014/1287-31/4).

of Medicine.

Most current article

© 2021 by the AGA Institute0016-5085/$36.00

https://doi.org/10.1053/j.gastro.2021.01.004

Keywords: Pathogens; Epidemiology; Collagenous Colitis; Lym-phocytic Colitis.

icroscopic colitis (MC) is an inflammatory bowel

Mdisease (IBD) of the large intestine that primarilyaffects older adults. Histologically, the disease is furthercategorized into lymphocytic (LC; dense lymphocytic infil-tration, >20 lymphocytes per 100 epithelial cells) andcollagenous colitis (CC; expansion of collagen fiber, >15 mmwith lymphocytic infiltration) subtypes. The pathophysi-ology of MC is largely unknown but is thought to be relatedto loss of immune tolerance to environmental-inducedperturbances in the aging microbiota.1 Genetic associa-tions with the HLA region have been detected for MC inearly small candidate studies,2 and later larger Immunochipanalyses,3,4 which were consistently replicated in indepen-dent cohorts and biobanks.5,6 Increased risk of CC, but notLC,3,4 was observed in relation to an extended haplotype(8.1) that includes several genes coding for major histo-compatibility complex (MHC) molecules with a critical rolein immune response to microbial pathogens. This observa-tion supports the hypothesis that enteric infection throughits effect on innate immune system and the communitystructure and function of the gut microbiome may increaserisk of MC. This is further supported by previous observa-tions that gastrointestinal infections have been linked todevelopment of classical IBD (reviewed in Axelrad et al7). Inaddition, a number of studies have demonstrated new-onsetMC in patients with recurrent Clostridioides difficile orfollowing fecal microbiota transplantation.8–12 Last, a recent

Page 2

WHAT YOU NEED TO KNOW

BACKGROUND AND CONTEXT:

Gastrointestinal infections have been linked to changes inthe composition and function of gut microbiome anddevelopment of inflammatory bowel diseases. However,the relationship between gastrointestinal infections andrisk of microscopic colitis has not been fully investigated.

NEW FINDINGS:

In a nationwide study, we found that gastrointestinalinfection, particularly Clostridioides difficile is associatedwith an increased risk of microscopic colitis. Theassociations were stronger with collagenous colitissubtype as compared to lymphocytic colitis.

LIMITATIONS:

Information on gastrointestinal infections were extractedfrom diagnostic codes and stool culture information wasnot available. Additionally, the vast majority ofparticipants in the study were white, possibly limitingthe generalizability of the findings.

IMPACT:

Understanding the underlying mechanism of theseassociations may provide significant insights into thepathogenesis of microscopic colitis.

1600 Khalili et al Gastroenterology Vol. 160, No. 5

CLINICALAT

study from Denmark demonstrated an association betweenCampylobacter infections and risk of MC.13 Despite thesecompelling lines of evidence, no prior study has systemati-cally examined the association between other bacterial,parasitic, and viral gastrointestinal infections and risk of MC.

We therefore sought to examine the association betweenall gastrointestinal infections and risk of MC in a nationwidehistopathology database in Sweden.

MethodsThe study was approved by the Stockholm Ethics Review

Board. Informed consent was waived because the study wasstrictly a register-based analysis.14

Study PopulationAscertainment of MC cases. We have previously out-

lined our method for identifying MC cases in Sweden.15–17

Briefly, we used the ESPRESSO (Epidemiology Strengthenedby histoPathology Reports in Sweden) cohort,18 a cohort ofgastrointestinal histology from all 28 pathology departments inSweden from January of 1965 until April of 2017. Each pa-thology report in ESPRESSO includes data on personal identitynumber, date of biopsy, anatomic location, and morphologyaccording to the Systematized Nomenclature of Medicine(SNOMED) system (CC: M40600 and LC: M47170).Supplementary Table 1 includes a list of definitions for allvariables used to define the study population. For this study,we included only MC cases diagnosed from January 1, 1990,until December 31, 2016. This time window was selected basedon our prior study demonstrating a significant increase inincidence starting in the early 1990s.16 A similar time periodhas been used in other epidemiologic studies of MC in theUnited States.19–21 We defined date of diagnosis of MC as the

date of first colorectal biopsy demonstrating MC. In a priorstudy, when comparing to a clinical diagnosis of MC obtainedfrom review of medical records, our method of identifying caseshad a positive predictive value of 95% demonstrating the val-idity of our outcome definition.22

Selection of population controls. We matched eachMC case to up to 5 general population controls without a priordiagnosis of IBD or MC according to age, sex, calendar year, andcounty of residence at date of diagnosis. We used the TotalPopulation Register in Sweden to identify controls using per-sonal identity number, a unique number assigned to all resi-dents in Sweden.14

Selection of sibling controls. We identified siblingsthrough the Multigeneration Register, a component of the TotalPopulation Register that has collected information on familyrelationships from all registered residents in Sweden bornsince 1932 and alive on January 1, 1961.14

Ascertainment of Exposure InformationThe Swedish Patient Register includes individual-level data

on hospital discharges on a nationwide level since 1987 (andcountywide since 1964).14 The register was expanded in 2001to also include specialized outpatient care. Information onenteric infections (bacterial, parasitic, viral, and not otherwisedefined) before study entry (among cases or matched controls)were collected from the register using International Classifi-cation of Diseases (ICD) codes (Supplementary Table 2). Theaccuracy of ICD coding for ascertainment of diagnoses for theinpatient component of Swedish Patient Register has beenpreviously validated with a positive predictive value of 85% to95%.23

Other CovariatesDemographic information on age, sex, death, or emigration

were collected from the Swedish Total Population Register.14 Inaddition, we obtained data on education level as a measure ofsocioeconomic status from the LISA (longitudinal integrateddatabase for health insurance and labor market studies). Thisdatabase compiles and annually updates administrative infor-mation from the labor market and educational and social sec-tors from 1990 onward on all individuals 16 years or olderregistered as residents in Sweden.24

We used the Swedish Prescribed Drug Register to obtaindata on medication use.25 Briefly, this Register was started onJuly 1, 2005, when information on all dispended medicationsincluding date of redemption, amount of drug, and dosage weredirectly transferred from pharmacy to the registry. Consistentwith our prior analyses, we defined active treatment for MC asuse of budesonide (Anatomic Therapeutic Chemical Classifica-tion code: A07EA06). This was based on prior published ran-domized controlled trials26–28 and AmericanGastroenterological Association and European guidelines29,30

showing that budesonide is effective in treatment of MC andis the recommended first-line treatment. In this study, wedefined active treatment as any use of budesonide within thefirst year of diagnosis of MC.

Last, information on other medical conditions, includingtype 1 diabetes and thyroid disease, were collected from theSwedish Patient Register, and data on diagnosis of celiac dis-ease were collected from the ESPRESSO study (Supplementary

Page 3

Figure 1. Flow chart of eligible participants in the study. GI,gastrointestinal.

April 2021 Gastrointestinal Infections and Risk of MC 1601

CLINICAL

AT

Table 1). We defined history of immune-mediated conditions ashaving at least 1 encounter for celiac disease, type 1 diabetes,psoriasis, systemic lupus, sarcoidosis, rheumatoid arthritis,thyroid disease, primary biliary cirrhosis, autoimmune hepati-tis, and primary sclerosing cholangitis (SupplementaryTable 3).

Statistical AnalysisWe conducted a case-control study of all patients with

incident MC diagnosed in Sweden from January 1, 1990, untilDecember 1, 2016, matched to up to 5 population controlsaccording to age, sex, calendar year, and county of residence atdate of diagnosis. We excluded patients with history of IBDbefore study entry or those with a gastrointestinal infectionwithin 6 months of diagnosis for cases and index date forcontrols to minimize the bias related to miscoding of gastro-intestinal infections in MC cases with delayed diagnosis(Figure 1). We modeled gastrointestinal infections as anyinfection, bacterial, parasitic, viral, and others. We also exam-ined the following specific bacterial and viral infections: Sal-monella, Escherichia, C difficile, and Norovirus. We used logisticregression modeling to estimate odds ratio (OR) and 95%confidence interval (CI) while adjusting for age, sex, county,calendar year, education (� 9, 10–12, > 12 years), and otherimmune-mediated conditions. Because MC is a rare outcome,we assumed that ORs are good approximations of relative riskestimates. In addition, we examined the associations accordingto histologic subtypes of CC and LC and used log likelihoodratio test to compare model fit between models assumingsimilar effect and models allowing for differential associationaccording to histologic subtypes.31

A number of exploratory and sensitivity analyses weredone. First, we conducted latency analyses examining the as-sociation between gastrointestinal infections and risk of MCaccording to different exposure time window (0.5 to <1, 1 to<3, 3 to <5, �5 years). Second, we additionally adjusted ourmodels for number of encounters (3 years to 6 months beforeindex date) and use of medications commonly associated withMC including nonsteroidal anti-inflammatory drugs (NSAIDs),proton pump inhibitors (PPIs), selective serotonin reuptakeinhibitors, and statins to account for health-seeking behaviorsamong those diagnosed with MC as well as residual con-founding related to polypharmacy. Third, we explored the as-sociations according to strata defined by sex, age of diagnosis(< 50 vs � 50 years), calendar period (1990–2006, 2007–2011, and 2012–2016), and history of an immune-mediateddisease. Fourth, we limited our cases of MC to those receivingtreatment with budesonide. Last, we limited our definition ofenteric infections to those that resulted in dispensed antibiotics(Anatomic Therapeutic Chemical Classification codes are listedin Supplementary Table 1) within 2 weeks of diagnosis andexamined their associations with risk of MC. All statistical an-alyses were carried out using R statistical software (version3.5.2; R Foundation for Statistical Computing, Vienna, Austria).A P value <.05 was considered statistically significant.

ResultsWe matched 13,468 MC cases (4295 CC and 9173 LC) to

64,479 controls (Table 1). The mean age at MC diagnosiswas 60 years (SD 17) with women accounting for more than

70% of participants. Compared with population controls,patients with MC were more likely to have been born in aNordic country (94% vs 85%, respectively) and have anearlier diagnosis of celiac disease (3.4% vs 0.1%, respec-tively) and an immune-mediated condition (23.7% vs11.6%, respectively). There was no difference in socioeco-nomic status as defined by number of years of educationbetween MC and population controls (Table 1).

The prevalence of previously diagnosed gastrointestinalinfection was 7.5% among patients with MC, which wassignificantly higher than in population controls (3.0%)(Pcomparison < .001). Specifically, compared with populationcontrols, the OR of MC after adjusting for matching factorsfor any gastrointestinal infection was 2.72 (95% CI 2.51–2.94) (Figure 2). The estimate was similar when addi-tionally adjusting for education and immune-mediateddiseases (adjusted OR [aOR] 2.63; 95% CI 2.42–2.85).The increased risk was consistent when the analyses werelimited to infections from bacteria (aOR 2.85; 95% CI2.45–3.30), parasites (aOR 2.46; 95% CI 1.32–4.41), andvirus (aOR 2.58; 95% CI 1.95–3.38). Among specificpathogens, we saw an increased risk of MC with C difficile(aOR 4.39; 95% CI 3.42–5.63), Escherichia species (aOR3.82; 95% CI 1.22–11.58), and Norovirus (aOR 2.87; 95%CI 1.66–4.87), but not Salmonella. We observed an asso-ciation with both histologic subtypes of CC and LC, albeitthe magnitude of association was stronger with CC subtype(Supplementary Table 4). Specifically, compared withpopulation controls, any gastrointestinal infection wasassociated with an aOR of 3.12 (95% CI 2.70–3.59) for CCand 2.42 (95% CI 2.19–2.68) for LC. Formal statisticalcomparison of these estimates did reach significance(Pheterogeneity ¼ .005). Similar to our primary analysis, weobserved an association between C difficile infection andrisk of both CC (aOR 4.07; 95% CI 2.70–6.12) and LC (aOR4.56; 95% CI 3.33–6.25).

In exploratory analyses, we examined the associationbetween gastrointestinal infection and MC according todifferent exposure time window. Although the relative riskof MC appeared to be highest within 6 to 12 months (aOR3.42; 95% CI 2.35–4.94), it remained elevated even 3 yearsafter a gastrointestinal infection (aOR 2.63; 95% CI

Page 4

Table 1.Characteristics of Cases and Controls at Index Data

Population controlsn ¼ 64,479

Microscopic colitisn ¼ 13,468

Collagenous colitisn ¼ 4295

Lymphocytic colitisn ¼ 9173

Sex, female 46,525 [72.2] 9697 [72.0] 3310 [77.1] 6387 [69.6]

Age at start of follow-up, y

Mean [SD] 59.6 [16.7] 60.2 [16.8] 63.13 [15.0] 58.79 [17.4]

Median [IQR] 63.0 [50.0–72.0] 63.0 [51.0–73.0] 65.0 [55.0–75.0] 62.0 [49.0–72.0]

< 50 15,945 [24.7] 3230 [24.0] 750 [17.5] 2480 [27.0]

� 50 48,534 [75.3] 10,238 [76.0] 3545 [82.5] 6693 [73.0]

Country of birth

Nordic 57,170 [88.7] 12,642 [93.9] 4114 [95.8] 8528 [93.0]

Other 7309 [11.3] 826 [6.1] 181 [4.2] 645 [7.0]

Age at diagnosis, y

Mean [SD] N.A. 60.2 [16.8] 63.13 [15.0] 58.79 [17.4]

Years of diagnosis

1990–2006 21,420 [33.2] 4442 [33.0] 1444 [33.6] 2998 [32.7]

2007–2011 21,645 [33.6] 4528 [33.6] 1468 [34.2] 3060 [33.4]

2012–2016 21,414 [33.2] 4498 [33.4] 1383 [32.2] 3115 [34.0]

History of gastroenteritis 1920 [3.0] 1030 [7.6] 358 [8.3] 672 [7.3]

Time from gastroenteritisto index date, y

Mean [SD] 16.2 [12.4] 14.3 [11.7] 13.7 [11.9] 14.6 [11.6]

Median [IQR] 13.4 [4.7–26.3] 11.0 [4.0–23.1] 9.7 [3.5–22.0] 11.7 [4.4–23.9]

0.5 < 1 79 [0.1] 56 [0.4] 22 [0.5] 34 [0.4]

1 < 3 223 [0.4] 138 [1.0] 52 [1.2] 86 [0.9]

3 < 5 199 [0.3] 111 [0.8] 44 [1.0] 67 [0.7]

�5 1419 [2.2] 725 [5.4] 240 [5.6] 485 [5.3]

Education, ya

� 9 17,957 [27.9] 3526 [26.2] 1290 [30.0] 2236 [24.4]

10–12 25,785 [40.0] 5486 [40.7] 1771 [41.2] 3715 [40.5]

� 13 18,490 [28.7] 4080 [30.3] 1122 [26.1] 2958 [32.3]

Missing 2247 [3.5] 376 [2.8] 112 [2.6] 264 [2.9]

Comorbidities

Celiac disease 45 [0.1] 462 [3.4] 156 [3.6] 306 [3.3]

Immune-mediated diseasesb 7,507 [11.6] 3196 [23.7] 1223 [28.5] 1973 [21.5]

IQR, interquartile range; N.A., not applicable; SD, standard deviation.aCategories based on compulsory school, high school, and college.bIncludes psoriasis, thyroid disease, rheumatoid arthritis, systemic lupus erythematosus, sarcoidosis, vasculitis, autoimmunehepatitis, primary sclerosing cholangitis, primary biliary cirrhosis, and celiac disease.

1602 Khalili et al Gastroenterology Vol. 160, No. 5

CLINICALAT

2.04–3.36). We also explored the association betweengastrointestinal infection and MC according to strata definedby sex, calendar year, and concomitant autoimmune disease,and observed no evidence for effect modification (all

Pintereaction > .08) (Table 2). We did observe a stronger as-sociation between gastrointestinal infection and risk of MCwith older age of diagnosis �50 years (Pinteraction ¼ .038).Specifically, among participants with age at index date �50

Page 5

Figure 2. Association between gastrointestinal infection and risk of MC. *MV, multivariable. DModels adjusted for matchingfactors: age, sex, county, and calendar year. †Models were additionally adjusted for education, and comorbidities according tocategories.

April 2021 Gastrointestinal Infections and Risk of MC 1603

CLINICAL

AT

years, the aOR of MC was 2.76 (95% CI 2.50–3.04). Thecorresponding aOR participants with age index date <50years was 2.31 (95% CI 1.98–2.68).

In sensitivity analyses, we considered the possibilitythat the observed associations may in part be related todifferential health-seeking behaviors between MC casesand controls and therefore additionally adjusted ourmodels for the number of health-related encounters withinthe last 3 years of index date and observed similar asso-ciations. Specifically, compared with population controls,gastrointestinal infection was associated with aORs of 1.92(95% CI 1.76–2.09) for MC, 2.16 (95% CI 1.85–2.52) forCC and 1.81 (95% CI 1.63–2.01) for LC. We also consid-ered the possibility that differential health-seekingbehavior (ie, patients with MC are more likely to havebeen prescribed medications) may have explained ourfindings. Therefore, we limited our study population to2007, 1.5 years after introduction of the Swedish Pre-scribed Register and further adjusted our models formedications known to be associated with risk of MC,including NSAIDs, PPI, selective serotonin reuptake in-hibitors, and statins as well as number of health-relatedencounters within 6 months to 3 years of index date.Compared with population controls, gastrointestinal infec-tion was associated with an aOR of 1.52 (95% CI 1.28–1.81) for MC. Similar to our primary analysis, C difficileinfection was particularly associated with an increased riskof MC (aOR 2.15; 95% CI 1.38–3.33). We also consideredthe possibility that our finding may be related to antibioticexposure and therefore excluded participants who receivedantibiotics with diagnosis of gastrointestinal infection andobserved an aOR for MC of 1.52 (95% CI 1.26–1.83) withany infection. Similarly, when we limited cases of gastro-intestinal infections to those followed by antibiotic ther-apy, the association was slightly strengthened, aORs 1.74(95% CI 1.15–2.61) with any infection.

Last, we explored the possibility that shared genetics orearly life factors may have accounted for our observed

associations and therefore conducted an additional sensi-tivity analysis examining the relationship between gastro-intestinal infections and risk of MC using unaffected siblingsas comparators (Table 3, Supplementary Tables 5 and 6).The aORs of MC were 1.79 (95% CI 1.56–2.07) with anyinfection, 1.75 (95% CI 1.34–2.28) with bacterial infection,3.04 (95% CI 1.11–8.29) with parasitic infection, and 1.84(95% CI 1.12–3.03) with viral infection, when comparedwith their unaffected siblings.

DiscussionIn a nationwide case-control study of more than 13,000

patients with MC, we found that gastrointestinal infectionwas associated with a nearly 3-fold increased risk of MC,and this excess risk persisted beyond 3 years after theinfection. The increased risk of MC appeared to be strongestfollowing infection with C difficile and with older age ofdiagnosis and CC histologic subtype. Our primary finding ofan increased association between gastrointestinal infectionand risk of MC remained consistent across multiple sensi-tivity and exploratory analyses.

One recent study by Nielsen and colleagues13 fromDenmark examined the association between positive stoolculture for Campylobacter concisus, Campylobacter jejuni,nontyphoidal Salmonella, or negative stool culture and riskof MC. The study found an association with C concisus(hazard ratio [HR] 9.3; 95% CI 4.1–20.1) and negative stoolculture (HR 5.6; 95% CI 4.6–7.2), but not Salmonella species(HR 1.2; 95% CI 0.2–11.1). These findings are at least in partconsistent with our observation that no or negative bacterialculture but not Salmonella species are associated withincreased risk of MC. Nevertheless, our study, whichincluded nationwide coverage of MC cases and compre-hensively assessed the relationship with all gastrointestinalinfections (including viral and parasites), significantly ex-pands on these prior findings. Our finding that C difficileinfection is particularly associated with an increased risk of

Page 6

Table 2.Association Between Any Gastrointestinal Infectionand Risk of MC According to Selected Strata

Strata

ControlsNumberexposed

MCNumberexposed

MV-aOR[95% CI] Pinteraction

Sex

Male 462 250 2.61 [2.21–3.08] .940

Female 1458 780 2.63 [2.39–2.89]

Age at indexdata, y

< 50 659 308 2.31 [1.98–2.68] .038

� 50 1261 722 2.76 [2.50–3.04]

Years sinceexposed

< 1 79 55 3.42 [2.35–4.94] .222

1 � 3 223 137 2.99 [2.38–3.74]

3 � 5 199 111 2.63 [2.04–3.36]

� 5 1419 725 2.53 [2.30–2.78]

Calendar year

1990–2006 498 285 2.80 [2.39–3.27] .086

2007–2011 629 369 2.85 [2.48–3.27]

2012–2016 793 376 2.35 [2.05–2.67]

Immune-mediateddiseases

No 1609 722 2.53 [2.31–2.78] .599

Yes 311 308 2.68 [2.24–3.21]

NOTE. Models were adjusted for age, sex, calendar year,county, education, and comorbidities according to categorieslisted in Table 1.MV, multivariable.

Table 3.Association Between Gastrointestinal Infection andRisk of MC Using Unaffected Sibling asComparators

Exposures

ControlsNumberexposed

MCNumberexposed

Multivariable-aOR, 95% CI

Any 538 543 1.79 [1.56–2.07]

Bacteria 138 141 1.75 [1.34–2.28]

Salmonella sp 15 9 1.24 [0.51–3.00]

Escherichia sp 5 2 0.59 [0.11–3.25]

Clostridioidesdifficile

35 53 2.55 [1.58–4.10]

Parasites 8 13 3.04 [1.11–8.29]

Virus 43 44 1.84 [1.12–3.03]

Norovirus 11 9 2.32 [0.83–6.47]

Others 379 400 1.88 [1.59–2.23]

1604 Khalili et al Gastroenterology Vol. 160, No. 5

CLINICALAT

MC is consistent with prior epidemiologic studies of Cdifficile infection with classical IBD (reviewed in Axelradet al7). Importantly, we have previously confirmed thisrelationship using a similar study population and studydesign in the Swedish Patient Register.32 Last, a number ofcase series have reported new-onset MC following recurrentC difficile infection or fecal microbiota transplantation.8–12

The pathophysiology of MC remains largely unknown.Emerging evidence suggests that similar to classical IBD, thedisease occurs as a result of an aberrant immune responseto the luminal microenvironment in a genetically suscepti-ble host. First, genetic association studies have shownthat variants within HLA genes are associated with risk ofMC,2,4–6 similar to other chronic diseases with aberrantimmune response to the gut microenvironment (ie, celiacdisease33 and classical IBD34). Second, human translationalstudies have demonstrated complete histologic remissionwith fecal diversion (ie, creation of an ileostomy bag) inpatients with refractory disease, and recurrence of

histologic inflammation after reestablishing the fecalstream,35–37 underscoring the critical role of gut microbiotain the pathogenesis of MC. Moreover, recent work from ourgroup has demonstrated that compared with patients withfunctional diarrhea (ie, chronic diarrhea without underlyinginflammation), gut microbiota in active MC is characterizedby significant dysbiosis.38 Last, medications such as PPIsand NSAIDs previously shown to be associated with risk ofMC are known to have significant impact on the compositionof the gut microbiota.39–43 Therefore, it is plausible thatgastrointestinal infections increase risk of MC broadlythrough their effect on the gut microbiome composition andfunction. In addition, our findings that C difficile infection,concomitant antibiotics therapy, and older age are morestrongly associated with risk of MC is likely related to thecharacteristically profound and persistent proinflammatorychanges in the gut microbiota observed in these subgroups.However, we note that our observed associations were in-dependent of antibiotic use.

Although it is unclear how microbiome changes cantrigger an immune response, genetic association studieshave suggested a possible link between genes coding forMHC molecules,3–6 which are involved in recognition ofextracellular pathogens by innate immunity, and risk of MC.This raises the possibility that gastrointestinal infectionseither directly or through alteration of the gut microenvi-ronment may lead to T-cell activation through MHC mole-cules. This proposed mechanism is in part supported by ourobservation of a stronger association between entericinfection and CC, which appears to be the only subtypepossibly linked to genes encoding MHC molecules.3,4

Our study has several strengths. This is the firstnationwide cohort study that comprehensively examines theassociation between gastrointestinal infections and risk ofMC in cohorts that collected data prospectively in routinemedical care. Therefore, our study is less likely to be

Page 7

April 2021 Gastrointestinal Infections and Risk of MC 1605

CLINICAL

AT

susceptible to selection and recall biases that are inherent tomany observational studies. We identified cases of MC usinga validated method demonstrating a positive predictivevalue of 95%. Last, we used multiple comparator groups (ie,siblings) and definitions (ie, active treatment) and per-formed latency analysis to demonstrate that our resultswere robust and not due to misclassification of exposurerelated to delayed diagnosis of MC.

We highlight several limitations. We used ICD coding toascertain information on gastrointestinal infection, whichmay be susceptible to underreporting or inaccurate repre-sentation of the type of infection, introducing exposuremisclassification. However, because such misclassification islikely nondifferential, it would have likely biased our resultstoward null. In addition, although the sensitivity ofcapturing all gastrointestinal infection in patient registermay be only modest, in a recent validation study inDenmark, the positive predictive value was 82% (95% CI73%–88%) for definitive gastroenteritis and 92% (95% CI85%–96%) for definitive or probable gastroenteritis.44 Wealso did not have specific information on the results of stoolculture and therefore were limited in examining the asso-ciations according to more specific pathogens. However, wenote that most stool studies in routine clinical practice arenegative due to difficulties in culturing pathogens.45 Last,we did not have information on lifestyle factors that mayhave confounded our observed associations and could notassess whether the associations were consistent accordingto other racial and ethnic groups.

In conclusion, in a nationwide case-control study wedemonstrate that previously diagnosed gastrointestinalinfection is associated with an increased risk of MC. Theassociations were particularly stronger for C difficile infec-tion and older age of diagnosis and CC histologic subtypebut appeared to be independent of shared genetics andearly life factors. Our findings further highlight the role ofperturbances in the gut microenvironment in the patho-genesis of MC. Future investigations on the potentialmechanism by which gastrointestinal infection increasesrisk of MC, particularly in genetically susceptible hosts, arewarranted.

Supplementary MaterialNote: To access the supplementary material accompanyingthis article, visit the online, version of Gastroenterology atwww.gastrojournal.org, and at, https://doi.org/10.1053/j.gastro.2021.01.004.

References

1. Miehlke S, Verhaegh B, Tontini GE, et al. Microscopic

colitis: pathophysiology and clinical management. Lan-cet Gastroenterol Hepatol 2019;4:305–314.

2. Fernandez-Banares F, Esteve M, Farre C, et al. Pre-disposing HLA-DQ2 and HLA-DQ8 haplotypes ofcoeliac disease and associated enteropathy in micro-scopic colitis. Eur J Gastroenterol Hepatol 2005;17:1333–1338.

3. Westerlind H, Bonfiglio F, Mellander MR, et al. HLA as-sociations distinguish collagenous from lymphocyticcolitis. Am J Gastroenterol 2016;111:1211–1213.

4. Westerlind H, Mellander MR, Bresso F, et al. Densegenotyping of immune-related loci identifies HLA vari-ants associated with increased risk of collagenous coli-tis. Gut 2017;66:421–428.

5. Green HD, Beaumont RN, Thomas A, et al. Genome-wide association study of microscopic colitis in the UKBiobank confirms immune-related pathogenesis.J Crohns Colitis 2019;13:1578–1582.

6. Stahl E, Roda G, Dobbyn A, et al. Collagenous colitis isassociated with hla signature and shares genetic riskswith other immune-mediated diseases. Gastroenterology2020;159:549–561.e8.

7. Axelrad JE, Cadwell KH, Colombel JF, et al. Systematicreview: gastrointestinal infection and incident inflamma-tory bowel disease. Aliment Pharmacol Ther 2020;51:1222–1232.

8. Agrawal M, Aroniadis OC, Brandt LJ, et al. The long-termefficacy and safety of fecal microbiota transplant forrecurrent, severe, and complicated Clostridium difficileinfection in 146 elderly individuals. J Clin Gastroenterol2016;50:403–407.

9. Fasullo MJ, Al-Azzawi Y, Abergel J. Microscopic colitisafter fecal microbiota transplant. ACG Case Rep J 2017;4:e87.

10. Erim T, Alazmi WM, O’Loughlin CJ, et al. Collagenouscolitis associated with Clostridium difficile: a cause ef-fect? Dig Dis Sci 2003;48:1374–1375.

11. Byrne MF, McVey G, Royston D, et al. Association ofClostridium difficile infection with collagenous colitis.J Clin Gastroenterol 2003;36:285.

12. Khan MA, Brunt EM, Longo WE, et al. Persistent Clos-tridium difficile colitis: a possible etiology for the devel-opment of collagenous colitis. Dig Dis Sci 2000;45:998–1001.

13. Nielsen HL, Dalager-Pedersen M, Nielsen H. High risk ofmicroscopic colitis after Campylobacter concisus infec-tion: population-based cohort study. Gut 2020;69:1952–1958.

14. Ludvigsson JF, Almqvist C, Bonamy AK, et al. Registersof the Swedish total population and their use in medicalresearch. Eur J Epidemiol 2016;31:125–136.

15. Khalili H, Burke KE, Roelstraete B, et al. Microscopiccolitis and risk of inflammatory bowel disease in anationwide cohort study. Gastroenterology 2020;158:1574–1583.e2.

16. Bergman D, Clements MS, Khalili H, et al.A nationwide cohort study of the incidence of micro-scopic colitis in Sweden. Aliment Pharmacol Ther2019;49:1395–1400.

17. Khalili H, Bergman D, Roelstraete B, et al. Mortality ofpatients with microscopic colitis in Sweden. Clin Gas-troenterol Hepatol 2020;18:2491–2499.e3.

18. Ludvigsson JF, Lashkariani M. Cohort profile:ESPRESSO (Epidemiology Strengthened by histoPa-thology Reports in Sweden). Clin Epidemiol 2019;11:101–114.

Page 8

1606 Khalili et al Gastroenterology Vol. 160, No. 5

CLINICALAT

19. Burke KE, Ananthakrishnan AN, Lochhead P, et al.Identification of menopausal and reproductive risk fac-tors for microscopic colitis-results from the Nurses’Health Study. Gastroenterology 2018;155:1764–1775.e2.

20. Liu PH, Burke KE, Ananthakrishnan AN, et al. Obesityand weight gain since early adulthood are associatedwith a lower risk of microscopic colitis. Clin Gastro-enterol Hepatol 2019;17:2523–2532.e1.

21. Liu PH, Lebwohl B, Burke KE, et al. Dietary gluten intakeand risk of microscopic colitis among US women withoutceliac disease: a prospective cohort study. Am J Gas-troenterol 2019;114:127–134.

22. Svensson M, Bergman D, Olen O, et al. Validatingmicroscopic colitis (MC) in Swedish pathology registers.Scand J Gastroenterol 2018;53:1469–1475.

23. Ludvigsson JF, Andersson E, Ekbom A, et al. Externalreview and validation of the Swedish national inpatientregister. BMC Public Health 2011;11:450.

24. Ludvigsson JF, Svedberg P, Olen O, et al. The longitu-dinal integrated database for health insurance and labourmarket studies (LISA) and its use in medical research. EurJ Epidemiol 2019;34:423–437.

25. Wettermark B, Hammar N, Fored CM, et al. The newSwedish Prescribed Drug Register—opportunities forpharmacoepidemiological research and experience fromthe first six months. Pharmacoepidemiol Drug Saf 2007;16:726–735.

26. Miehlke S, Aust D, Mihaly E, et al. Efficacy and safety ofbudesonide, vs mesalazine or placebo, as inductiontherapy for lymphocytic colitis. Gastroenterology 2018;155:1795–1804.e3.

27. Miehlke S, Madisch A, Kupcinskas L, et al. Budesonide ismore effective than mesalamine or placebo in short-termtreatment of collagenous colitis. Gastroenterology 2014;146:1222–1230.e1–2.

28. Miehlke S, Madisch A, Karimi D, et al. Budesonide iseffective in treating lymphocytic colitis: a randomizeddouble-blind placebo-controlled study. Gastroenter-ology 2009;136:2092–2100.

29. Nguyen GC, Smalley WE, Vege SS, et al. AmericanGastroenterological Association Institute guideline onthe medical management of microscopic colitis.Gastroenterology 2016;150:242–246; quiz e17–e18.

30. Fernandez-Banares F, Casanova MJ, Arguedas Y, et al.Current concepts on microscopic colitis: evidence-based statements and recommendations of the Span-ish Microscopic Colitis Group. Aliment Pharmacol Ther2016;43:400–426.

31. Wang M, Spiegelman D, Kuchiba A, et al. Statisticalmethods for studying disease subtype heterogeneity.Stat Med 2016;35:782–800.

32. Axelrad JE, Olen O, Askling J, et al. Gastrointestinalinfection increases odds of inflammatory bowel diseasein a nationwide case-control study. Clin GastroenterolHepatol 2019;17:1311–1322.e7.

33. van Heel DA, Franke L, Hunt KA, et al. A genome-wideassociation study for celiac disease identifies risk vari-ants in the region harboring IL2 and IL21. Nat Genet2007;39:827–829.

34. Goyette P, Boucher G, Mallon D, et al. High-densitymapping of the MHC identifies a shared role for HLA-DRB1*01:03 in inflammatory bowel diseases and het-erozygous advantage in ulcerative colitis. Nat Genet2015;47:172–179.

35. Daferera N, Kumawat AK, Hultgren-Hornquist E, et al.Fecal stream diversion and mucosal cytokine levels incollagenous colitis: a case report. World J Gastroenterol2015;21:6065–6071.

36. Veress B, Lofberg R, Bergman L. Microscopic colitissyndrome. Gut 1995;36:880–886.

37. Jarnerot G, Bohr J, Tysk C, et al. Faecal stream diversionin patients with collagenous colitis. Gut 1996;38:154–155.

38. Morgan DM, Cao Y, Miller K, et al. Microscopic colitis ischaracterized by intestinal dysbiosis. Clin GastroenterolHepatol 2020;18:984–986.

39. Imhann F, Vich Vila A, Bonder MJ, et al. The influence ofproton pump inhibitors and other commonly usedmedication on the gut microbiota. Gut Microbes 2017;8:351–358.

40. Zhernakova A, Kurilshikov A, Bonder MJ, et al. Popula-tion-based metagenomics analysis reveals markers forgut microbiome composition and diversity. Science2016;352:565–569.

41. Imhann F, Bonder MJ, Vich Vila A, et al. Proton pumpinhibitors affect the gut microbiome. Gut 2016;65:740–748.

42. Freedberg DE, Toussaint NC, Chen SP, et al. Protonpump inhibitors alter specific taxa in the human gastro-intestinal microbiome: a crossover trial. Gastroenter-ology 2015;149:883–885.e9.

43. Masclee GM, Coloma PM, Kuipers EJ, et al. Increasedrisk of microscopic colitis with use of proton pump in-hibitors and non-steroidal anti-inflammatory drugs. Am JGastroenterol 2015;110:749–759.

44. Skyum F, Andersen V, Chen M, et al. Infectious gastro-enteritis and the need for strict contact precaution pro-cedures in adults presenting to the emergencydepartment: a Danish register-based study. J HospInfect 2018;98:391–397.

45. Axelrad JE, Freedberg DE, Whittier S, et al. Impact ofgastrointestinal panel implementation on health careutilization and outcomes. J Clin Microbiol 2019;57:e01775-18.

Received August 18, 2020. Accepted January 2, 2021.

CorrespondenceAddress correspondence to: Hamed Khalili, MD, MPH, Clinical andTranslational Epidemiology Unit, Massachusetts General Hospital, HarvardMedical School, Boston Massachusetts. e-mail: [email protected];fax: (978)-882-6710.

CRediT Authorship ContributionsHamed Khalili, MD (Conceptualization: Lead; Methodology: Equal; Writing –

original draft: Lead)Jordan E. Axelrad, MD (Writing – review & editing: Supporting)Bjorn Roelstraete, PhD (Formal analysis: Lead). Ola Olén, MD, PhD (Writing –

review & editing: Supporting)Mauro D’Amato, PhD (Conceptualization: Equal; Writing – review & editing:

Supporting)

Page 9

April 2021 Gastrointestinal Infections and Risk of MC 1607

Jonas F Ludvigsson, MD, PhD (Data curation: Lead; Methodology: Equal;Writing – original draft: Supporting; Writing – review & editing: Supporting)

Conflict of interestThese authors disclose the following: Hamed Khalili receives grant fundingfrom Takeda and Pfizer. He has received consulting fees from AbbVie andTakeda. Jordan E. Axelrad receives grant funding from BioFire Diagnostics.He has received consulting fees from Janssen and BioFire Diagnostics.Jonas F. Ludvigsson coordinates a study on behalf of the Swedish IBDquality register (SWIBREG). That study has received funding from Janssencorporation. Karolinska Institutet has received investigator-initiated studygrants and fees for lectures and consulting performed by Ola Olén forJanssen, Pfizer, Ferring, and Takeda. The remaining authors disclose noconflicts.

Funding This project is funded by the Karolinska Institutet (Jonas F. Ludvigsson)and by Stockholm county council (Jonas F. Ludvigsson). Hamed Khalili issupported by the American College of Gastroenterology, a Senior ResearchAward from the Crohn’s and Colitis Foundation, and the Beker Foundation.Ola Olén is supported by grants from the Swedish Medical Society, KarolinskaInstitutet Foundations, and the Strategic Research Area Epidemiologyprogram at Karolinska Institutet while working on this project. Financialsupport is also provided through the regional agreement on medical trainingand clinical research between Stockholm County Council and KarolinskaInstitutet (ALF), None of the funding organizations has had any role in thedesign and conduct of the study; in the collection, management, and analysisof the data; or in the preparation, review, and approval of the manuscript. Thisproject is also supported by National Institute of Aging AG068390 to HamedKhalili and Jonas Ludvigsson.

CLINICAL

AT

Page 10

Supplementary Table 1.ATC, ICD, SNOMED, and Topography Codes Used for Identification of Study Populations

TOPOGRAPHY SNOMED ICD-9a ICD-10 ATC

Antibiotics J01

Crohn’s disease 555 K50

Celiac disease T64x, T65x (excludingT652, T65200, T653,T65300, T654, T656,T65600, T65620,T6592, T6593)

M58, M5800, M58000,M58001, M58005,M58006, M58007,M58000X, D6218,D62180, D6218X,D6218Y, D62188

Inflammatory bowel disease 556, 555 K50, K51, K52.3

Microscopic colitis T67x, T68x M40600x, M47170x

Non-steroidal anti-inflammatorydrugs

M01A

Proton pump inhibitors A02BC

Serotonin Reuptake Inhibitors N06AB

Statins C10AA

Steroids (budesonide) A07EA06

Ulcerative colitis 556 K51

ATC, Anatomic Therapeutic Chemical Classification.aICD-9 started in 1987. Since the first patients with MC were included in 1990, all study participants had �3 years of exposurein the Patient Register before identification of comorbidities.

1607.e1 Khalili et al Gastroenterology Vol. 160, No. 5

Page 11

Supplementary Table 2.Definition of Types of Gastrointestinal (GI) Infection Using ICD Coding

Type of gastroenteritis Infectious agent ICD-7 ICD-8 ICD-9 ICD-10

Infectious diseasesof the GI tract

Any 040–049 000–009 001–009 A00-A09

Bacterial infections Typhoid/Paratyphoid 040–041 001–002 002 A01

Salmonella 042 003 003 A02

Shigella 045 004 004 A03

Escherichia coli - 008,0 008A A04.0-A04.4

Campylobacter - - 008E A04.5

Yersinia enterocolitica - - 008E A04.6

Clostridioides difficile - - 008E A04.7

Other bacterial infection - 008,2–008,3 008F A04 (exceptA04.0-A04.7)

Bacterial food poisoning 049 005 005 A05

Parasitic infections Amoeba 046 006 006 A06

Giardia 047,02 007,10 007B A07.1

Cryptosporidium - - - A07.2

Isospora - - - A07.3

Other protozoic GI infections 047 (except047,02)

007 (except007,.10)

007 (exceptthan 007B)

A07 (exceptA07.1-A07.3)

Viral infections Rotavirus - - 008L A08.0

Norovirus - - - A08.1

Adenovirus - - 008H A08.2

Other viral GI infection - 008,8–008,9 008M (except008L, 008H)

A08 (exceptA08.0-A08.2)

Infectious disease,not otherwise defined

Not defined if viral orbacterial

048, 571,764, 785,6

008 (except 008,0;008,2; 008,3), 009

008, 008W, 009 A09

April 2021 Gastrointestinal Infections and Risk of MC 1607.e2

Page 12

Supplementary Table 3.Definition of Immune-Mediated Conditions Using ICD Coding and Histopathology

Disease ICD-7 ICD-8 ICD-9 ICD-10

Diabetes 260 250 250 E10–14

Psoriasis 706,09 696 (969,30excluded)

696A-C þ E-W L40

SLE 456,20 734,1 710A M32.1; M32.8; M32.9

Rheumatoid arthritis 722,00 þ 722,01 712,3; 714,93 714 M05; M06; M08;M09; M12.3

Thyroiditis 254 245 245A-X E06

Hyperthyroidism 252.00, 252.01,252.02, 254.00;

242 242 E05

Sarcoidosis 138,00 to 138,04 135 135 D86

Primary biliary cirrhosis 571G K74.3

ANCA vasculitis andother vasculitis

446 446 M31

Spondylopathy M45.9

Autoimmune hepatitis 573,0x (chronichepatitis)571,9x

573D þ 573E K75.4; K75.5; K75.9(for exclusion criteria,if AIH had been the

outcome, we would haveused only K75.4)

Primary sclerosingcholangitis

K83.0A

Celiac disease Defined according to histopathology

ANCA, anti-neutrophilic cytoplasmic autoantibodies; SLE, systemic lupus erythematosus.

Supplementary Table 4.Association Between Gastrointestinal Infection and Risk of MC According to Histologic Subtypes

Exposures Controls No. exposed CC No. exposed CC OR, 95% CI LC No. exposed LC OR, 95% CI Pheterogeneity

Any 1920 358 3.12 [2.70–3.59] 672 2.42 [2.19–2.68] .005

Bacteria 502 94 2.84 [2.17–3.70] 201 2.85 [2.38–3.41] .992

Salmonella sp 52 5 1.43 [0.47–3.62] 13 1.66 [0.80–3.18] .813

Escherichia sp 8 2 8.66 [0.81–189.22] 4 3.01 [0.77–10.60] .450

Clostridioides difficile 142 47 4.07 [2.70–6.12] 79 4.56 [3.33–6.25] .664

Parasites 34 6 3.28 [1.10–9.13] 12 2.14 [0.98–4.35] .509

Virus 167 41 3.23 [2.10–4.91] 49 2.19 [1.50–3.13] .172

Norovirus 40 8 2.66 [1.04–6.36] 14 2.96 [1.48–5.73] .849

Others 1327 272 3.49 [2.95–4.13] 486 2.43 [2.16–2.73] <.001

NOTE. Models were adjusted for age, sex, county, calendar year, education, and comorbidities according to categories listedin Table 1.

1607.e3 Khalili et al Gastroenterology Vol. 160, No. 5

Page 13

Supplementary Table 5.Characteristics of Cases and Their Unaffected Siblings

Population controls n ¼ 13,699 MC n ¼ 7091 CC n ¼ 2216 LC n ¼ 4875

Sex, female 6882 [50.2] 5115 [72.1] 1705 [76.9] 3410 [70.0]

Age at start of follow-up, y

Mean [SD] 54.2 [14.6] 54.1 [15.2] 56.9 [13.1] 52.8 [15.9]

Median [IQR] 57.0 [46.0–65.0] 57.0 [45.0–65.0] 59.0 [50.0–66.0] 56.0 [42.0–65.0]

< 50 4370 [31.9] 2317 [32.7] 553 [25.0] 1764 [36.2]

� 50 9329 [68.1] 4774 [67.3] 1663 [75.1] 3111 [63.8]

Country of birth

Nordic 13515 [98.7] 7005 [98.8] 2205 [99.5] 4800 [98.5]

Other 184 [1.3] 86 [1.2] 11 [0.5] 75 [1.5]

Years of diagnosis

1990–2006 3890 [28.4] 1951 [27.5] 629 [28.4] 1322 [27.1]

2007–2011 4740 [34.6] 2425 [34.2] 751 [33.9] 1674 [34.3]

2012–2016 5069 [37.0] 2715 [38.3] 836 [37.7] 1879 [38.5]

History of gastroenteritis 538 [3.9] 543 [7.7] 158 [7.1] 385 [7.9]

Time from gastroenteritisto index date, y

Mean [SD] 17.1 [12.0] 15.4 [12.0] 15.25 [12.3] 15.43 [11.9]

Median [IQR] 14.8 [6.3–26.4] 12.3 [4.7–25.4] 11.42 [5.4–23.1] 12.47 [4.6–25.9]

0.5 < 1 7 [0.1] 24 [0.3] 8 [0.4] 16 [0.3]

1 < 3 48 [0.4] 63 [0.9] 17 [0.8] 46 [0.9]

3 < 5 57 [0.4] 51 [0.7] 14 [0.6] 37 [0.8]

> ¼ 5 426 [3.1] 405 [5.7] 119 [5.4] 286 [5.9]

Education,a y

� 9 3371 [24.6] 1333 [18.8] 495 [22.3] 838 [17.2]

10–12 5956 [43.5] 3125 [44.1] 1024 [46.2] 2101 [43.1]

� 13 4065 [29.7] 2513 [35.4] 674 [30.4] 1839 [37.7]

Missing 307 [2.2] 120 [1.7] 23 [1.0] 97 [2.0]

IQR, interquartile range; SD, standard deviation.aCategories based on compulsory school, high school, and college.

April 2021 Gastrointestinal Infections and Risk of MC 1607.e4

Page 14

Supplementary Table 6.Association Between Gastrointestinal Infection and Risk of MC According to Histologic SubtypesUsing Unaffected Sibling as Comparators

Exposures Controls No. exposed CC No. exposed CC OR, 95% CI LC No. exposed LC OR, 95% CI Pheterogeneity

Any 538 158 1.79 [1.38–2.33] 385 1.81 [1.53–2.15] .997

Bacteria 138 39 2.29 [1.38–3.82] 102 1.63 [1.20–2.22] .904

Salmonella sp 15 2 1.00 [0.16–6.19] 7 1.36 [0.50–3.74] .854

Escherichia sp 5 0 2 0.81 [0.14–4.78]

Clostridioides difficile 35 15 4.57 [1.73–12.11] 38 2.22 [1.29–3.82] .887

Parasites 8 5 1.94 [0.47–8.03] 8 4.73 [1.09–0.61] .862

Virus 43 13 2.03 [0.81–5.03] 31 1.77 [0.98–3.21] .960

Norovirus 11 2 1.92 [0.29–12.92] 7 2.55 [0.73–8.92] .929

Others 379 121 1.73 [1.28–2.33] 279 1.97 [1.61–2.42] .960

NOTE. Models were adjusted for age, sex, county, calendar year, education, and comorbidities according to categories listedin Table 1.

1607.e5 Khalili et al Gastroenterology Vol. 160, No. 5