Article Text

Download PDFPDF

Eight year prognosis of postinfectious irritable bowel syndrome following waterborne bacterial dysentery
  1. John K Marshall1,
  2. Marroon Thabane1,
  3. Amit X Garg2,
  4. William F Clark2,
  5. Paul Moayyedi1,
  6. Stephen M Collins1,
  7. Walkerton Health Study Investigators
  1. 1Department of Medicine (Division of Gastroenterology) and Farncombe Family Digestive Health Research Institute,McMaster University, Hamilton, Ontario, Canada
  2. 2Department of Medicine (Division of Nephrology), University of Western Ontario, London, Ontario, Canada
  1. Correspondence to Dr John K Marshall, Division of Gastroenterology (2F59), McMaster University Medical Centre, 1200 Main Street West, Hamilton, Ontario L8N 3Z5, Canada; marshllj{at}


Background Although postinfectious irritable bowel syndrome (PI-IBS) is a well-recognised complication of acute gastroenteritis, its prognosis remains poorly defined. The natural history of PI-IBS was assessed among participants in the Walkerton Health Study (WHS), which has followed the long-term effects of a large outbreak of acute gastroenteritis related to municipal water contamination in May 2000.

Methods WHS participants were invited to return for annual assessment at a research clinic. Adult residents of Walkerton at the time of the outbreak who enrolled in 2002/2003 and returned for assessment in 2008 were eligible for a PI-IBS study cohort if they had no prior history of IBS or inflammatory bowel disease. A modified Bowel Disease Questionnaire was used to diagnose IBS by Rome I criteria and to identify IBS subtypes.

Results Of 4561 WHS participants, 2451 returned for their 8 year assessment and 1166 were eligible for the PI-IBS study cohort (688 females, mean age 46.2 years). The prevalence of IBS among 742 eligible subjects who suffered acute gastroenteritis during the outbreak declined from 28.3% after 2–3 years to 15.4% after 8 years, but remained significantly increased compared with controls who did not have acute gastroenteritis (OR 3.12; 95% CI 1.99 to 5.04). Independent risk factors for PI-IBS at 8 years included female gender, younger age, prior anxiety/depression, and fever or weight loss during the acute enteric illness. IBS subtypes were not stable over time.

Conclusions Acute gastroenteritis can trigger IBS symptoms that persist for at least 8 years. Characteristics of the host and the acute enteric illness can predict the long-term risk of PI-IBS.

  • Epidemiology
  • infectious diarrhoea
  • irritable bowel syndrome

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Postinfectious irritable bowel syndrome (PI-IBS) is a well-recognised complication of acute gastroenteritis. Between 5% and 30% of people experience new gastrointestinal symptoms that persist after bacterial dysentery despite clearance of the inciting pathogen.1 The mechanisms that underlie this chronic disturbance of gut function remain poorly understood, but are thought to involve chronic inflammation, increased epithelial permeability and altered neuromuscular function.2 3 The cumulative economic impact of PI-IBS may be substantial.4

There are few data that characterise the long-term natural history of PI-IBS, as most studies of its epidemiology have assessed small cohorts for relatively short intervals.5–19 Only four studies have followed participants for >2 years.14 20–22 In each of these cohorts, up to half of the subjects with PI-IBS after acute gastroenteritis recovered over 3,14 520 21 or 622 years. However, these studies followed only 12, 11, 41 and 14 subjects with PI-IBS, respectively. Larger studies are clearly needed to provide more precise estimates of prognosis.

The Walkerton Health Study (WHS) was initiated to assess long-term health outcomes following a large waterborne outbreak of acute gastroenteritis in May 2000.23 The WHS cohort has provided a unique opportunity to study the phenomenon of PI-IBS following a large, simultaneous and well-characterised exposure. Herein, we report observations on the epidemiology and natural history of PI-IBS made from biennial assessment of the WHS cohort over 8 years.


Study population

Walkerton is a small rural town in Ontario, Canada with a population of ∼5000. In May 2000, heavy rainfall washed livestock fecal residue from nearby farms into inadequately chlorinated drinking water supplied from a shallow well.24 Subsequent contamination of the municipal water supply with Escherichia coli 0157:H7, Campylobacter jejuni and other pathogens led to a large outbreak of acute bacterial gastroenteritis that affected at least 2300 local residents, with 27 recognised cases of the haemolytic uraemic syndrome and six attributable deaths.24 25 The WHS was initiated 2 years after the outbreak to study the epidemiology and long-term health outcomes of the outbreak, and to facilitate local residents' access to specialised healthcare. Details of its methodology have been reported elsewhere.23 26 27

The WHS established a special multidisciplinary ambulatory clinic at a local hospital to conduct annual structured assessments of affected and unaffected residents of Walkerton and its surrounding area. Since its inception, the WHS has enrolled a total of 4561 subjects. All participants were invited to return for annual reassessment at the WHS clinic.23 Participants were eligible for the PI-IBS study cohort if they were adults (age ≥16 years), lived in the Walkerton postal code at the time of the outbreak, and had no prior history of IBS, coeliac disease or inflammatory bowel disease based on self-report and review of their medical records. To minimise recall bias and misclassification, only subjects who enrolled in the first 2 years of the WHS (2002 or 2003) were included in the PI-IBS study cohort.

Data collection

Each year, all returning WHS participants underwent a standardised interview that included a 20 min structured questionnaire about various domains of general health. In alternate years, this assessment included a modified Bowel Disease Questionnaire (BDQ),28 that allowed symptom classification by Rome I criteria.29

Subgroup comparison

As in our initial study of the short-term epidemiology of PI-IBS,23 we divided all participants into three exposure strata to characterise acute gastroenteritis experienced at the time of the outbreak: (1) those who reported no acute illness during the outbreak (‘controls’); (2) those who reported an acute illness that could not be substantiated by prior health records (‘self-reported’ gastroenteritis); and (3) those whose acute enteric illness was substantiated by review of health records (‘clinically suspected’ gastroenteritis). Gastroenteritis could be substantiated by any of the following: (1) documented healthcare contact for acute gastrointestinal symptoms during the outbreak; (2) bloody diarrhoea in May 2000 reported in health records or in responses to a local public health survey conducted during the outbreak; (3) diarrhoea lasting at least 3 days with more than three stools per day reported on the public health survey; or (4) a documented positive stool culture. Such confirmation of gastroenteritis has been found to correlate with severity of the acute illness.27

IBS subtypes were identified as follows. IBS with diarrhoea (IBS-D) was diagnosed if subjects reported more than three bowel movements per day, watery stools or urgency at least 25% of the time, and denied having fewer than three bowel movements per week, hard or lumpy stools or straining to evacuate stools >25% of the time. IBS with constipation (IBS-C) was diagnosed if subjects reported fewer than three bowel movements per week, hard or lumpy stools or straining to evacuate stools at least 25% of the time, and denied having three bowel movements per day, watery stools or urgency more than 25% of the time. Subjects who were neither IBS-D nor IBS-C were considered to have unclassified IBS (IBS-U).

Statistical analysis

At each assessment, the prevalence of PI-IBS within each cohort exposure stratum was estimated as a proportion with 95% CI. The association between acute gastroenteritis (clinically suspected and/or self-reported) and a diagnosis of IBS was tested using χ2, with Bonferroni correction for multiple comparisons. ORs were calculated using unexposed controls as the referent group.

Among subjects who experienced acute gastroenteritis, risk factors for PI-IBS after 4, 6 and 8 years were tested at each time point using univariable and multivariable logistic regression to adjust for confounding. Variables associated with an increased risk of PI-IBS (p≤0.10) in univariate logistic regression were included in multivariable backward stepwise logistic regression models. Risk factors considered in this analysis included age, gender, exposure status, specific features of illness during outbreak, self-reported premorbid anxiety or depression, antibiotic use and direct contact with livestock. All analyses were conducted using SPSS software (Version 17.0).


Study participants

From 2002 to 2008, a total of 4561 subjects participated at least once in the WHS (table 1). The 246 participants who enrolled after 2003 were excluded from the PI-IBS study cohort to minimise recall bias and case misclassification (figure 1). The cumulative number of terminates in 2008 was 2110, with 130 subjects deaths, 1032 withdrawals and 948 losses to follow-up (participants who missed one assessment but returned in a subsequent year were not counted as terminates) (figure 1). The demographic characteristics of participants who returned in 2004, 2006 and 2008 were similar (table 1). Returning participants in 2008 were older than terminates (37.9 vs 32.6 years, p<0.0001), and more likely to be female (59.7% vs 51.8%, p<0.0001). However, participants and terminates did not differ in their exposure to gastroenteritis (table 2).

Table 1

Characteristics of returning Walkerton Health Study (WHS) participants

Figure 1

Summary of study participation and withdrawal.

Table 2

Assessment of response bias among returning participants who enrolled in 2002 or 2003

Of the 4315 participants who enrolled in the WHS in 2002 or 2003, 3280 returned in 2004, 2572 returned in 2006 and 2309 returned in 2008 (table 1). The latter included 1166 of the 2069 original participants eligible for the PI-IBS study cohort. Of the 1166 eligible subjects, 742 subjects had been exposed to acute gastroenteritis (either self-reported or clinically confirmed) and 424 were unexposed controls. A total of 1026 members of the PI-IBS study cohort returned for reassessment at all time points (2004, 2006 and 2008).

Prevalence and natural history of PI-IBS

The overall prevalence of IBS among the 742 returning members of the PI-IBS study cohort who had suffered acute gastroenteritis in 2000 declined from 210/742 (28.3%) at 2–3 years, to 159/742 (21.4%) at 4 years, 106/742 (14.3%) at 6 years and 114/742 (15.4%) at 8 years (figure 2). The prevalence of IBS at 8 years remained significantly lower among controls who did not have acute gastroenteritis during the outbreak (23/424 (5.4%), p<0.0001). Among the 210 participants with acute gastroenteritis and subsequent IBS at enrolment in 2002 or 2003, Rome I criteria were still fulfilled at 4 years in 105 (50.0%), at 6 years in 72 (34.3%) and at 8 years in 73 (34.8%) (figure 3). In comparison, ‘sporadic’ IBS (in subjects with no exposure to acute gastroenteritis during the outbreak who fulfilled Rome I criteria in 2002 or 2003) persisted over the same interval in 10 of 45 subjects (22.2%, p=0.103). At the 2008 assessment, 13 of 742 subjects with a history of acute gastroenteritis (1.8%) fulfilled Rome I criteria for the first time (incident IBS), versus 8 of 424 controls with no such history (1.9%) (p=0.869).

Figure 2

Prevalence of irritable bowel syndrome (IBS) by Rome I criteria.

Figure 3

Prognosis of postinfectious irritable bowel syndrome (PI-IBS) among subjects who returned for assessment in 2008.

An additional posthoc analysis was conducted under the conservative assumption that subjects who failed to return for assessment did not have IBS. In this analysis, the prevalence of IBS by Rome I criteria fell from 30.4% in 2002/2003 to 8.3% in 2008 among subjects exposed to acute gastroenteritis during the outbreak, and from 10.1% to 3.3% among those who remained well during the outbreak. The difference in prevalence between exposed and non-exposed subjects was significant at each assessment (p<0.0001).

Risk factors for PI-IBS

Risk factors for PI-IBS were assessed at each time point by multivariable logistic regression (table 3). Female gender and weight loss during the acute enteric illness emerged as an independent predictor at all time points. Younger age was significant at all assessments except 4 years, while prolonged diarrhoea was significant at all assessments except 8 years. Abdominal cramps during the acute illness were an independent predictor only in 2002/2003 and 2004. Premorbid psychological disorders, peak stool, frequency, fever and blood per rectum during the acute enteric illness emerged as independent predictors at only one time point. Characteristics of eligible subjects who were exposed to gastroenteritis and returned for assessment in 2008 are summarised in table 4, stratified by IBS status in 2008.

Table 3

Multiple logistic regression of predictors of postinfectious irritable bowel syndrome at each time point

Table 4

Characteristics of subjects exposed to gastroenteritis with and without postinfectious irritable bowel syndrome (PI-IBS) at the year 8 follow-up

Stability of IBS subtype

Of the 2069 subjects enrolled in 2002/2003 who were eligible for the original PI-IBS study cohort, 488 fulfilled Rome I criteria for IBS at enrolment (23.6%). Of these, 204 were subclassified as IBS-D (41.8%), 52 were subclassified as IBS-C (10.7%) and 232 were subclassified as IBS-U (47.5%). IBS subtypes were also assessed in accordance with Rome III criteria, which consider only stool consistency. Of the 488 subjects with Rome I IBS at enrolment, 206 (42.2%) were thus subclassified as IBS-D, 98 (20.1%) as IBS-C, 75 (15.4%) as mixed IBS (IBS-M) and 109 (22.3%) as unclassified IBS (IBS-U).

Among the 204 subjects with IBS-D in 2002/2003, a significant number were reclassified as IBS-C or IBS-U in subsequent years (figure 4A). Persistence of the IBS-D phenotype did not differ significantly between patients who developed gastroenteritis during the outbreak (PI-IBS) and those who had remained well (sporadic IBS). Similar phenotypic instability was seen among the 52 and 232 subjects initially classified as IBS-C and IBS-U (figure 4B,C).

Figure 4

Stability of irritable bowel syndrome (IBS) phenotype among subjects with IBS with diarrhoea (IBS-D; n=204) (A), IBS with constipation (IBS-C; n=52) (B) and unclassified IBS (IBS-U; n=232) (C) when enrolled in 2002 or 2003.


The WHS is the largest and longest prospective study of PI-IBS ever published. Our results confirm that PI-IBS resolves over time in most patients. However, one in seven affected individuals suffers symptoms that persist for at least 8 years after infection. Among study participants who returned for assessment in 2008, the prevalence of IBS was 15.4% among subjects who had suffered acute gastroenteritis during the outbreak of May 2000 versus 5.4% among those without such a history (p<0.0001). While these results are broadly consistent with those of previously published long-term cohort studies,14 20–22 none of these studies was of comparable size or duration.

We assessed predictors of PI-IBS at each study visit and identified a number of important risk factors. Risk factors for persistent symptoms 8 years after the initial infection included female gender, younger age, psychological illness (anxiety or depression) and both fever and weight loss during the acute gastroenteritis. At various earlier time points, other risk factors for PI-IBS included various features of the acute enteric illness such as the duration and peak frequency of acute diarrhoea, cramps and rectal bleeding. These findings are similar to those reported in other cohorts,6 7 21 22 30 and demonstrate that the risk of PI-IBS is determined by features of both the host and the enteric illness itself.

Several weaknesses of this study warrant recognition. First, only a minority of its participants had a culture-proven infection, as local health authorities were overwhelmed during the outbreak. However, an extensive public health investigation did confirm E coli 0157:H7 and Campylobacter species as the pathogens. Secondly, only 2309 of the original 4315 WHS participants (53.5%) returned for reassessment in 2008. Such attrition is not uncommon among cohort studies of this duration, particularly where a voluntary in-person assessment is required. While returning participants were older than terminates and more often female, they did not differ in their exposure to gastroenteritis during the outbreak. Thirdly, because the study could only be planned once the outbreak was established, we were unable to assess rigorously some important aspects of premorbid health, such as previous gastrointestinal symptoms and psychological profile. To supplement patient recall, all primary care and hospital records of participants were also reviewed for related diagnoses. However, we cannot exclude the possibility that some subjects were misclassified.

PI-IBS is the only reproducible human model of IBS pathogenesis, and has now been well characterised. It has been estimated that 1 in 10 patients with IBS can recall an infectious onset to their symptoms.2 However, it is likely that this recall underestimates the role that infection plays in the pathogenesis of IBS and other chronic gastrointestinal disorders. While only a minority of patients with IBS recall a clear infectious onset, it is plausible that many more such patients develop their symptoms after an infection that it either unrecognised or forgotten. Indeed, enteric infection and related perturbations of gut flora are likely to play an important role in the pathogenesis of many gastrointestinal disorders.

The results of this study demonstrate considerable instability of the IBS phenotype, with few patients remaining as either IBS-D or IBS-C over continued follow-up. This observation emphasises the limitations of symptom-based systems for classifying functional bowel disorders and the need to develop classifications based on underlying pathophysiology. It is also problematic for treatments that target dominant symptoms, as clinicians must constantly reassess the appropriateness of treatment. Previous longitudinal studies have also shown that symptoms of functional gastrointestinal disorders fluctuate and vary over time, but these did not use Rome criteria or consider IBS subtypes.31–33 Our analysis diagnosed IBS using Rome I criteria, and a predefined algorithm to classify subjects as IBS-C, IBS-D or IBS-U. However, Rome I criteria were not developed to assess IBS subtypes, and this approach has not been validated. Further studies are needed to characterise the stability of IBS subtypes using the updated Rome III criteria.

The Walkerton outbreak of waterborne gastroenteritis was an awful human tragedy that took at least six lives and caused considerable human suffering. However, this public health disaster has also provided a unique opportunity to learn about the long-term health consequences of acute enteric infection. The results of this study confirm a strong association between acute bacterial dysentery and development of IBS. While the prognosis of PI-IBS after waterborne bacterial dysentery is largely favourable, some patients develop chronic symptoms. Female gender, younger age, psychological co-morbidity and more severe infection all appear to increase the risk of PI-IBS. Future studies should evaluate strategies to prevent or attenuate PI-IBS in high-risk individuals.

Significance of this study

What is already known about this subject?

  • Acute gastroenteritis is an established risk factor for new-onset of irritable bowel syndrome.

  • Features of both the patient and the acute enteric illness can predict the short-term risk of developing postinfectious irritable bowel syndrome.

  • The pathogenesis of postinfectious irritable bowel syndrome is poorly understood.

What are the new findings?

  • Over prolonged follow-up, most patients with postinfectious irritable bowel will improve. However, a significant minority experience symptoms that persist for at least 8 years.

  • Features of the patient and the acute illness can predict the risk of sustained postinfectious irritable bowel syndrome.

  • Subtypes of irritable bowel syndrome are not stable over time.

How might it impact on clinical practice in the foreseeable future?

  • This information will help clinicians to counsel patients on their long-term risk of irritable bowel syndrome following acute gastroenteritis.

  • As treatments for postinfectious irritable bowel syndrome emerge, these data will help to identify high-risk subgroups who might benefit from early intervention.


The authors are grateful to the people of Walkerton, participants in the Walkerton Health Study and the Walkerton Health Study investigators for their contribution to this research.



  • Walkerton Health Study Investigators: William Clark, Stephen M. Collins, Amit Garg, R. Brian Haynes, John Howard, Jennifer MacNab, Jeff Mahon, John K. Marshall, Douglas Matsell, Louise Moist, Janet Pope, Joel Ray, Patricia Rosas-Arellano, Marina Salvadori, Rita Suri.

  • Funding Ontario Ministry of Health and Long-Term Care and the Crohn's and Colitis Foundation of Canada (CCFC).

  • Competing interests None.

  • Ethics approval The study protocol received full approval from both the Hamilton Health Sciences/McMaster University Research Ethics Board (Hamilton, Ontario) and the University of Western Ontario's Office of Research Ethics (London, Ontario).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • Digest
    Emad El-Omar Severine Vermeire Alexander Gerbes