Short- and Long-term Effects of Bacterial Gastrointestinal Infections

Bacterial gastrointestinal infections are associated with short- and long-term complications from several organ systems.

B acterial gastrointestinal infections continue to cause illness and death and contribute to economic loss in most parts of the world, including high-income countries that have developed surveillance and control programs. The symptoms of acute bacterial intestinal infection are usually mild to moderate, and spontaneous remission occurs (1), but in some cases, the disease can cause rapid deterioration of a patient's condition.
There may be other, perhaps unusual and less documented, late effects of acute enteric infections, such as infl ammatory bowel disease (6).
In Sweden, there is no active follow-up on reported cases of bacterial enteric infection in terms of disease outcome or long-term complications. During the 8-year period 1997-2004, >100,000 persons with acute gastrointestinal infection were reported within the national surveillance program for communicable diseases. We present a retrospective cohort study of these patients to investigate the association between exposure to a bacterial pathogen and the risk for autoimmune illness, gastrointestinal complications, and extraintestinal infectious disease.

Materials and Methods
Participants comprised persons with intestinal infection (nontyphoidal Salmonella spp., Campylobacter spp., Yersinia enterocolitica, Shigella spp., or enterohemorrhagic Escherichia coli [EHEC]) reported to the Swedish Institute for Infectious Disease Control during 1997-2004. We collected data on age, sex, date reported, and country of infection and used social security numbers for identifi cation. This identifi cation number was used to link our cohort of cases (those with short-term complications occurring within 3 months or long-term effects within 1 year after infection) to the Swedish Hospital Discharge (covers all hospital in Sweden) and Causes of Death registers. Ethics permission was obtained from the Ethical Committee, Karolinska Institute. Discharge diagnoses must be reported to the register; therefore, any study using this register is, in practice, population based. The Hospital Discharge Register was validated by using a diagnosis of acute myocardial infarction; underreporting was <1%, main diagnosis was missing for <1% of cases, and correct diagnosis was made for 86% (7).
We calculated the follow-up time for each case as person-time from reported date of infection to an event, death, or study termination. Person-years were then compared with a Swedish standard population of 5-year age groups to calculate the expected number of cases for each disease. Standardized incidence ratios (SIRs) were constructed by dividing the observed number of cases with the expected number of cases. Ninety-fi ve percent exact confi dence intervals (CIs) were calculated under the assumption that the number of observed cases were Poisson distributed. CIs that do not overlap 1 indicate that the number of observed cases is signifi cantly different from the number of cases expected in a population cohort of similar age and sex distribution. The described method is called indirect standardization, and interpretation of results is similar to relative risk interpretation, i.e., comparing the risk for disease in an exposed cohort to the risk for disease in an unexposed cohort.
We previously estimated standardized mortality ratios (SMRs) for Salmonella (8) and Campylobacter infections (9) and showed that country of infection (domestic or abroad) was an effect modifi er; i.e., the SMR differed substantially between these 2 strata and no pooled SMR could be calculated. The underlying factor for this interaction was probably that the term abroad served as a proxy for healthiness or a healthy traveler effect. For our present analysis, we divided the cohort into 2 strata on the basis of country of infection (Sweden or abroad), but no statistical signifi cant interaction was evident. We concluded that crude SIRs irrespective of country of infection could be estimated. All analyses were conducted by using SAS statistical software, version 8.2 (SAS Institute, Inc., Cary, NC, USA).

Results
Demographic data on the 101,855 study participants and frequency counts for infectious agents are summarized in Table 1. Campylobacter spp. caused the most cases, 57,425 (56%). The second most frequent pathogen was Salmonella spp., the causative agent in 34,664 cases (34%); distribution of serovars is shown in Table 2. Of all cases of gastroenteritis, Yersinia spp. accounted for 5,133 (5%) cases; Shigella spp. 3,813 (4%); and EHEC 820 (<1%). Table 3 shows the number of reported case-patients with specifi c diseases within 3 months of an episode of bacterial gastrointestinal infection, along with expected num-ber of cases and SIRs. Not surprisingly, the highest risks were found for HUS after EHEC infection and GBS following campylobacter infection. Although SIRs were quite elevated, absolute risks were more moderate; among 820 cases of EHEC infection, we found 13 episodes of HUS (1.6%), 57,425 cases of campylobacteriosis, 13 cases of GBS (0.02%), 5,133 cases of Yersinia infection, and 9 cases of reactive arthritis (0.2%). The risk for aortic aneurysm among patients with salmonellosis was signifi cantly higher than expected (SIR 6.4, 95% CI 3.1-11.8). The absolute risk for bacteremia/sepsis was 0.02% for case-patients with Campylobacter infection and 0.03% for those with salmonellosis. For many complications, we did not fi nd any statistically signifi cant elevated risks. Other complications that we had hypothesized to be associated with gastrointestinal infections could not be shown. Only a few cases were found within 3 months, contributing to imprecise estimates of SIRs.
Within 1 year of acute bacterial gastrointestinal infection, case-patients with Yersinia enteritis were at increased risk for reactive arthritis (SIR 47.0, 95% CI 21.5-89.2), Salmonella infection (SIR 18.2, 95% CI 12.0-26.5), and Campylobacter infection (SIR 6.3, 95% CI 3.5-10.4) ( Table 4). The risk for ulcerative colitis was elevated among patients with salmonellosis (SIR 3.2, 95% CI 2.2-4.6) and, to a lesser extent, among patients with campylobacteriosis (SIR 2.8, 95% CI 2.0-3.8). Of the 29 patients in our salmonellosis cohort who had ulcerative colitis, 13 (44%) had fi rst experienced ulcerative colitis during the 10-year period before the acute infection. Among patients with campylobacteriosis, we found 42 with ulcerative colitis, of whom 18 (43%) had received a diagnosis of ulcerative colitis in the 10-year period before the infection. We did not fi nd any increased risk for Crohn's disease in the same group of patients. We did not fi nd any elevated risk for many of the rheumatologic diseases included in the present study in any of the participants. The distribution of Salmonella serotypes among patients with aortic aneurysm, reactive arthritis, and ulcerative colitis in our cohort did not differ in any substantial way from the whole salmonellosis cohort (Table 5), although the number of patients was rather small.

Discussion
Our data confi rm the elevated risk for complications and long-term sequelae after an episode of acute bacterial gastroenteritis. We have presented new estimates of the absolute and relative risk for well-described complications such as HUS after EHEC infection, GBS after an episode of Campylobacter enteritis, and reactive arthritis after Yersinia enteritis. Another complication that we have been able to verify is aortic aneurysm after an episode of salmo-nellosis. Perhaps more unexpected, the risk for ulcerative colitis was elevated in the cohort of patients with salmonellosis and campylobacteriosis. The distribution of Salmonella serovars was the same among patients with and without complications. The fi nding of no major difference in the distribution of Salmonella serovars between the group of patients with and without complications indicates that factors other than Salmonella serovar alone determine the risk for complications.
Compared with other studies, our new estimate of the risk for HUS after EHEC infection is lower than previously reported (10,11). An explanation of our lower estimates could be that we used only International Classifi cation of Diseases (ICD) codes specifi c for HUS. Several of these cases may in fact be classifi ed under nonspecifi c ICD codes that also include a large proportion of cases unrelated to HUS. However, had we included them in the analysis, any association with the infections would have been diluted. Our estimate of risk for GBS and campylobacter are in line with a study in England that showed a risk of <2/10,000 that GBS will develop in a patient with campylobacteriosis (12). These results are also in line with a previous study in Sweden (13). All estimates of complications in this study are based on discharge data from the Hospital Discharge Register; this means that minor complications that either were not presented to any doctor or were handled only by general practitioners were not available for this analysis. At the population level, re-active rheumatologic symptoms associated with infection are typically mild and transient (14). This is probably the reason why our estimate of reactive arthritis after Yersinia infection is quite low, although similar low risks have been reported elsewhere (15).
In patients with atherosclerotic disease, or in those with preexisting aneurysms, transient bacteremia with nontyphoidal Salmonella infection can result in vascular infections (16)(17)(18). Most of these aneurysms described previously have been localized in the subrenal segment of the abdominal aorta (17). Salmonella spp. in these patients can invade the arterial intima and cause a localized endothelial infection that results in an aneurysm or the enlargement of a previously existing aneurysm. This may explain the association between Salmonella infection and aortic aneurysm in this study.
Our fi ndings of an elevated risk for ulcerative colitis in the cohort of patients with salmonellosis and campylobacteriosis need further study. In another large cohort study, an association between acute gastroenteritis and infl ammatory bowel disease was identifi ed (n = 43,013), where the incidence rate for ulcerative colitis was 40 per 100,000 person-years, a doubling of the risk for those unexposed to infection (19). We do not know why an episode of infectious gastroenteritis could contribute to the initiation or exacerbation of ulcerative colitis. Seasonal variation in the onset of ulcerative colitis, and reports that excessive childhood infections are associated with higher risk for ulcerative colitis, may support the hypothesis that infections could be triggers of disease (20). From this study, we cannot say whether there is a causal relationship between Salmonella and Campylobacter infections and relapse of disease in patients with known ulcerative colitis, or whether the infection could trigger ulcerative colitis in susceptible persons. We cannot entirely rule out that the fi ndings are an artifact, resulting from an increased number of medical examinations and stool cultures in a group of patients with diarrhea because of a known or unknown infl ammatory bowel disease. More study is needed to confi rm or refute our fi ndings.
Because irritable bowel syndrome is diagnosed and treated at hospital in only a minority of patients, our estimates are probably too low. Many studies have not used a control group but reported only the numbers and percentages of patients who had irritable bowel syndrome after gastroenteritis (21); 1 study with controls estimated a relative risk of 11.9 (CI 6.7-21) after 1 year of follow-up (22).
Our study has some limitations. Perhaps the most serious one is the selection bias of patients entering the gastroenteritis cohort. Only a small fraction of all patients with Salmonella infection, for example, seek medical care, have a stool sample taken, and are eventually reported to national surveillance (23). This could have an effect on the results, especially if we are collecting data on those with the most severe disease; disease severity itself affects complications and sequelae. Another limitation is the lack of information on confounding factors among study participants, especially coexisting illnesses such as malignant disease or immunodefi ciencies of any cause. Such coexisting illnesses could perhaps increase to some extent the risk for complications (6), but our results on the effect of disease from gastrointestinal infections would not have changed. Although the quality of the Swedish Hospital Discharge Register is quite good, there is always a general problem of reliability in registry-based epidemiologic research.
In conclusion, we studied the risk for complications 3 months and 1 year after acute bacterial gastroenteritis and found disease manifestations from several organ systems that required hospitalization of patients. These fi ndings are a reminder of, and could be an argument for, the usefulness of existing control programs targeted to control bacterial enteric disease. This study was approved by the Regional Ethical Committee, Karolinska Institute, Stockholm, Sweden. Dr Ternhag is a resident physician at Karolinska University Hospital and a PhD student at Karolinska Institute, Department of Medical Epidemiology and Biostatistics, Sweden. His research interests are in infectious disease epidemiology, registry-based research, and long-term prognosis of infectious diseases.