Bacterial gastroenteritis is usually self-limited, but sequelae such as reactive arthritis (ReA) may follow acute infection. The epidemiology and clinical spectrum of ReA in the US is poorly defined, and is based primarily on information obtained through outbreak investigations.

We present results of a study that involved follow-up of culture-confirmed cases of Campylobacter, Escerichia coli O157, Salmonella, Shigella and Yersinia infections reported among residents of Minnesota and Oregon. Our main objective was to use the population-based framework of the Foodborne Disease Active Surveillance Network (FoodNet) to define more clearly the epidemiology and clinical spectrum of ReA following foodborne illness in the US.

METHODS

The study was reviewed and approved by the investigational review boards at each participating medical institution, state health department and CDC. Informed consent was obtained from all participating subjects. A description of the FoodNet program and surveillance methods is available at http://www.cdc.gov/foodnet/index.htm.

Beginning March 15, 2002, we attempted to interview all subjects age >1 year whose incident culture-confirmed Campylobacter, E coli O157, Salmonella, Shigella, or Yersinia infection had been reported to FoodNet in Minnesota and Oregon. Telephone interviews were conducted 4 to 8 weeks after the specimen collection date using a standardised screening questionnaire developed for this study (see Appendix). Interviews were conducted in Oregon until 14 March 2004, and in Minnesota until 14 August 2004. Children (age <18 years) were not interviewed directly; parents or legal guardians provided proxy interviews. Exclusion criteria are shown in fig 1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Total cases reported and screened for possible reactive arthritis (ReA), Minnesota and Oregon, 2002–2004.

We defined a possible ReA case as a subject with new onset of any of the rheumatological symptoms listed in the screening questionnaire. Subjects with possible ReA were invited to complete a more detailed medical history and physical examination. We defined a case of definite ReA as a subject with possible ReA and a history or physical examination findings consistent with monoarthritis, oligoarthritis, dactylitis, enthesitis, or inflammatory back pain without other rheumatological explanation. Enthesitis was measured by the Mander Enthesis Index (MEI),1 which scores the degree of tenderness at 30 individual and grouped enthesis sites on a scale of 0 to 3 (0 = no tenderness, 1 = mild tenderness, 2 = moderate tenderness, 3 = wince or withdraw tenderness) with a maximum score of 90. One of three rheumatologists (MT, AD, AB) in Oregon and a single rheumatologist (HK) in Minnesota conducted the physical examinations. In Minnesota, no examinations were performed on subjects screened during the first year of surveillance; children were excluded from the examination portion of the study at that site. A follow-up questionnaire that addressed ongoing rheumatological symptoms was mailed to subjects with definite ReA 6 months later.

Subjects with culture-confirmed infections but without new rheumatological symptoms (controls) were also asked to submit a blood specimen. No physical examination of control subjects was performed. Blood specimens collected from possible cases and controls were screened for human leukocyte antigen (HLA)-B27 using a standard lymphocyte microcytotoxicity test.2 Erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) were reported for confirmed ReA cases. Repeat stool specimens were not collected from subjects at the follow-up interview.

Data from the two sites were pooled and analysed together using SPSS V 14.0 (SPSS Inc., Chicago, Illinois, USA). The χ² test was used to compare differences in proportions. Non-parametric tests were used to test locations of continuous variables. p Values of <0.05 were considered significant. Incidence calculations were performed using US Census Bureau population estimates for year 2002 rounded to the nearest 100 000 (Minnesota population 5.1 million, Oregon population 3.5 million).

RESULTS

A total of 6379 culture-confirmed infections were reported to FoodNet in Minnesota and Oregon during the study. Figure 1 shows the proportion of potential subjects in each state that were screened for new rheumatological symptoms, and the number of subjects meeting exclusion criteria. We interviewed 4468 (70%) subjects within 8 weeks of specimen collection. Although a higher proportion of interviews were completed in Minnesota (80%) than in Oregon (51%), few subjects in either state refused the screening interview. Interviews were completed a median of 38 days (mean 40 days) after the date of specimen collection, and a median of 43 days (mean 46 days) after diarrhoea onset. The demographic and clinical characteristics of interviewed subjects were similar between the two states. Table 1 displays features of the combined cohort by aetiology.

A total of 575 (13%) subjects reported having new rheumatological symptoms lasting ⩾3 days after enteric illness onset. The time to onset, duration and type of rheumatological symptoms did not differ by pathogen. Among those with new rheumatological symptoms, 46% reported more than one symptom; 56% had new joint pain or discomfort, 44% had new low back pain, 32% had new morning joint stiffness lasting longer than an hour; and 23% had new heel pain. Only 14% reported joint swelling or redness. The median interval from onset of diarrhoea to onset of the first rheumatological symptom was 16 days (mean 18 days, range 0–180 days). At least one of these symptoms was still present at the time of the interview for 389 (68%) subjects. Among the 575 subjects with rheumatological symptoms, 85 (15%) had visited a health care provider for evaluation or treatment of these symptoms; of these, five (6%) reported they had been diagnosed with ReA, none of whom presented for examination by a rheumatologist involved in this study. Risk of developing new joint symptoms was not associated with month or year of specimen collection, or specimen source.

Table 2 shows the frequency and proportion of subjects with new onset of rheumatological symptoms by pathogen. For all pathogen types, adults (age⩾18) and females accounted for the largest proportion of subjects with new onset of rheumatological symptoms.

Table 3 shows the proportion of subjects with new rheumatological symptoms by serotype or species. Risk of developing new symptoms did not differ significantly by Salmonella serotype (p = 0.75) or by Campylobacter (p = 0.39) or Shigella (p = 0.50) species.

Risk of new rheumatological symptoms associated with demographic and clinical features, with odds ratios adjusted for age group (adult vs child) and sex, are shown in table 4.

Subjects who reported clinical features suggesting an increased severity of acute illness (eg, fever, chills, headache, bloody stools, hospitalisation and persistent diarrhoea) were more likely to report subsequent new onset of rheumatological symptoms. A total of 39 subjects with culture-confirmed infection were asymptomatic; 1 of these (3%) reported developing rheumatological symptoms. Antibiotic treatment of the acute infection was common among those with and without new rheumatological symptoms; the mean number of days from onset of diarrhoea to start of antibiotics was 7 days for both groups. After controlling for age group and sex, antibiotic use was not associated with risk of new rheumatological symptoms.

Rheumatology history and examination findings

Rheumatology history and physical examinations were completed on 68 (38%) of the 178 Oregon subjects with rheumatological symptoms. Most had been diagnosed with Campylobacter or Salmonella infections (57% and 29% respectively); evaluations were completed on only four subjects each with a history of E coli O157 or Shigella infection, and on one person with Yersinia infection. In Minnesota, 124 adults had rheumatological symptoms with culture dates after August 1, 2003, when the examination phase was approved; 14 (11%) completed rheumatology evaluations, 10 had a history of Campylobacter and four had Salmonella infection. Evaluations were completed a median of 17 days after the screening interviews (range 3–79 days) in Oregon and 33 days after interviews in Minnesota (range 3–64 days). There were no significant differences in the demographic and clinical characteristics of subjects who completed rheumatology evaluations and those who did not, except for the over-representation of Campylobacter cases in the former group.

A total of 54 subjects met the case definition for confirmed ReA (62% of subjects with rheumatological symptoms evaluated in Oregon and 86% of subjects evaluated in Minnesota). The median age of confirmed cases was 41 years (range 10–66). A total of 52 (96%) confirmed ReA cases were adults; 36 (67%) were female. Among the 28 subjects with new rheumatological symptoms who did not meet the case definition, 9 had no rheumatological diagnosis; 10 had a single diagnosis and 9 had more than one. Diagnoses other than ReA included osteoarthritis (n = 10), mechanical low back pain (n = 12), bursitis (n = 3), myofascial pain syndrome (n = 3), patellofemoral pain syndrome (n = 3), fibromyalgia (n = 3), hypermobility syndrome (n = 1) and rheumatoid arthritis (n = 1).

Table 5 shows the clinical and laboratory features of the confirmed ReA cases.

Although most subjects rated their pain on the day of the examination as absent or moderate, a substantial proportion had previously seen a health care provider and had taken medication for pain relief. Mild enthesitis was the most common examination finding. The most frequently involved entheses were the calcaneal insertions of the plantar fascia and the Achilles tendon; 54% of confirmed ReA cases were tender at one of these sites. Tenderness in spinous processes was also common, present in 44% of confirmed cases. The knee was the joint most frequently affected by arthritis; 7 of 10 subjects with arthritis had knee involvement. Only one subject had a joint effusion that was amenable to arthrocentesis. The synovial fluid was inflammatory (total white blood cell (WBC) count of 24 000, 86% neutrophils) and the gram stain, culture and microscopy for crystals were negative. Four subjects had dysuria; none had concomitant arthritis, conjunctivitis and dysuria.

Blood specimens were collected from 52 subjects with confirmed ReA, and from 56 subjects who had infection but did report any new rheumatological symptoms (controls). The frequency of HLA-B27 among subjects with confirmed ReA was not significantly higher than among control subjects; HLA-B27 was positive for 6/52 (12%) ReA cases compared with 6/56 (11%) controls.

Follow-up questionnaire at 6 months

Of 54 subjects with ReA confirmed by history and examination, 34 (63%) returned the 6-month follow-up questionnaire. Of these, 23 (68%) had persistent symptoms at the time of the questionnaire; 17 (50%) had joint pain; 4 (12%) had joint swelling or redness; 9 (27%) had morning joint stiffness lasting longer than an hour; 7 (21%) had heel pain; 16 (47%) had low back pain; 15 (44%) had neck pain. Further, 12 (35%) visited a health care provider for these symptoms and 4 (12%) went to an emergency room or urgent care centre; none were hospitalised. Only two subjects with ongoing symptoms at 6 months were HLA-B27 positive.

Incidence of ReA

Based on our screening questionnaire and follow-up examinations, we estimate a range of incidence of confirmed ReA following culture-confirmed infections due to enteric pathogens in Oregon in the following manner: assuming none of the subjects who did not complete interviews or examinations were confirmed as having ReA, then a minimum of the 42 (1.9%) cases that we confirmed occurred among 2256 culture-confirmed reported cases during the 2 years of surveillance, resulting in a minimum annual incidence estimate of 0.6/100 000. Alternatively, assuming that persons who did not completed interviews or examinations were equally likely to have ReA as persons who did, then a maximum of 215 ReA cases may have occurred (since 15.6% of 2256 culture-confirmed cases in Oregon had rheumatological symptoms, and 62% of these were confirmed as ReA), resulting in a maximum annual incidence of 3.1 cases/100 000. A similar calculation was not made for Minnesota given the small proportion of cases examined there.

DISCUSSION

In this large-scale population-based study of rheumatological sequelae of infection due to bacterial enteric pathogens in the US, we documented the frequency of new rheumatological symptoms developing after culture-confirmed infection with Campylobacter, Salmonella, Shigella, E coli O157 and Yersinia in Minnesota and Oregon, examined risk factors for developing these symptoms, and described the clinical presentation and physical examination findings in a subset of patients with confirmed ReA. Our study demonstrated that Campylobacter and Salmonella infections are the most important contributors to the incidence of ReA related to foodborne illness in the Minnesota and Oregon. Repeating the study in other FoodNet sites with greater ethnic and racial diversity would provide a better understanding of the generalisabilty of our conclusions for the US population as a whole.

Case reports and outbreak investigations have demonstrated an association between ReA and infection with Salmonella,3^–6 Shigella,7 8 Campylobacter,9^–13 and Yersinia species.14^–17 The frequency of ReA found in these studies has ranged from 1–21%.18 Only a few population-based studies of ReA have been published,19^–21 primarily in Scandinavian countries where the distribution of enteric pathogens and population genetics may differ significantly from those in the US. None of these studies has compared rates of reactive musculoskeletal symptoms and ReA after infection with multiple different organisms in a single cohort study, as we have done. To our knowledge, no population-based studies other than outbreak investigations have specifically addressed risk of rheumatological illness associated with Salmonella infections.

E coli O157 has not been previously linked to ReA in published outbreak investigations. However, we included this pathogen in our study because of the genetic similarities between E coli and Shigella species.22 In 1999, Laasila et al described a single case of ReA in an HLA-B27 positive woman with recurrent E coli urinary tract infections.23 In 2002, a study comparing rates of rheumatological symptoms following Campylobacter and enterotoxigenic E coli (ETEC) in Denmark found 10 (6%) cases of ReA in the ETEC group.19 We similarly found a surprisingly high rate of rheumatological symptoms among adults with E coli O157 infections, approximately the same as the rate among adult Campylobacter cases. However, because few of these cases were available for follow-up rheumatological examinations, the question of whether there is an association between E coli infections and confirmed ReA remains open.

Our findings are consistent with a Danish population-based study that found that 16.1% of subjects with Campylobacter infection had new onset of musculoskeletal symptoms within 4 weeks of infection,19 and a population-based study in Finland that confirmed ReA in 7% of Campylobacter cases.20 A pilot FoodNet survey in California 24 found a substantially lower proportion (2%) of subjects with musculoskeletal symptoms, perhaps owing to the low response rate among Campylobacter and Salmonella cases in that study, and our use of a telephone interview as opposed to a mailed questionnaire.

As in other studies, we found that risk of developing new rheumatological symptoms was greater among subjects with more severe acute illness6 and prolonged diarrhoea,3 25 26 and few asymptomatic carriers developed reactive symptoms. It is not known whether this finding relates to virulence properties of the infecting pathogen, innoculum size, variations in host immune response, or some combination of these or other factors. Although genetic factors may play a role, we did not find an association between risk of acquiring ReA and HLA-B27 positivity. Hannu et al also found no association of ReA with HLA-B27 after Campylobacter infection, and noted that studies that report a high frequency of HLA-B27 often involve hospital-based series with more severe and complicated disease.20 We did not have sufficient numbers of subjects with HLA-B27 to determine whether there is an association between HLA-B27 and ReA disease severity.

We did not identify any modifiable risk factors for ReA. Although antibiotics were commonly used for these infections, we found that their use did not prevent development of rheumatological symptoms regardless of pathogen or type of antibiotic used. Some previous studies have suggested that effective antibiotics might indeed prevent the development of ReA,27 28 while others showed no protective effect.19 29 One outbreak investigation found an increased risk of ReA in patients with S enteritidis treated with antibiotics.6 Antibiotics including ciprofloxacin have been shown to increase rates of faecal carriage of Salmonella30 31 and could theoretically increase risk of ReA by causing antigen persistence in the gut.

This study has several limitations. First, it is difficult to prove that the rheumatological symptoms described by our subjects are truly attributable to the antecedent infections. However, by examining a subset of those with subjective symptoms, we were able to provide objective confirmation that true illness was present, and was not related to alternate rheumatological diagnoses. Use of imaging studies such as ultrasound 32^–34 to confirm enthesitis could have provided additional objective evidence of disease. Including a control group without diarrhoeal illness in our study would also have helped to clarify the amount of risk attributable to infection. Future studies are planned to ascertain rates of onset of these rheumatological symptoms in the general populations of FoodNet sites including Oregon and Minnesota.

Second, although all persons with possible ReA were offered a follow-up examination, only a small subset agreed to participate in this part of the study. Long travel distance to the examination sites, work schedule conflicts and resolution of symptoms were among reasons given for refusal to participate. It is possible that the group of subjects who agreed to be examined were more likely to do so because of greater severity or duration of ReA. Examination of a larger subset of possible cases would have given more a more precise assessment of the frequency and severity of confirmed ReA. Additionally, because of the low proportion of subjects from whom we were able to obtain blood specimens, small differences in risk associated with HLA-B27 might have been missed.

Finally, a major difficulty in the interpretation of epidemiological studies of ReA such as this one is the absence of a universally accepted definition of the term. There is general agreement that ReA shares clinical characteristics with other members of the spondyloarthitis group of disorders such as ankylosing spondylitis and psoriatic arthritis, 36 37 but there is difference of opinion as to what clinical findings are required for the diagnosis.38^–40 We elected to include enthesitis and inflammatory back pain in our case definition. Including only those with frank arthritis would obviously have resulted in a substantially lower estimate of the incidence of ReA.

In summary, we have described the epidemiology and clinical spectrum of rheumatological sequelae of common bacterial enteric pathogens in two population-based study sites. Symptoms were generally mild, but often persistent, and resulted in a substantial use of medical resources and days missed from work or daily activities. Our findings will provide more accurate estimates of the total burden of illness in the US due to these enteric bacterial pathogens.