The epidemiology of infectious gastroenteritis related reactive arthritis in U.S. military personnel: a case-control study

Although infectious gastroenteritis (IGE) is typically a short-lived, acute illness, disease morbidity is significant. Incidence rates in travelers have been reported at 29 episodes per 100 person-months, based on self-reported symptoms and cohort studies[1]. Total IGE morbidity is likely underestimated due to unmeasured chronic sequelae, including reactive arthritis (ReA)[2–8]. ReA is a systemic inflammatory disorder characterized by an asymmetric, additive, aseptic arthritis triggered by an infection at a distant site. Previously considered a self-limited disease, the potential of ReA to cause prolonged, recurrent, or erosive disease has increasingly been recognized[2, 3, 5, 7, 9–11].

IGE-associated ReA is most commonly described after infection with Campylobacter [11–15], Salmonella[16–19], Shigella[4, 20], and Yersinia[21, 22]. Other ReA case reports have implicated asymptomatic enteric infections[23], enterotoxigenic E. coli (ETEC)[10, 24], Cryptosporidium spp[25, 26], Giardia lamblia[27], Strongyloides stercoralis[28], and possibly Schistosoma mansoni[23], and Clostridium difficile[29–33].

Two pseudonyms for ReA exist. Reiter's disease (RD) refers to the classic triad of arthritis, urethritis, and conjunctivitis, while postdysenteric arthritis (PA) refers specifically to IGE-related disease. In recent years, both subsets of disease have been referred to as reactive arthritis, although the naming convention has not been standardized[2, 34]. Current ICD-9 coding includes PA and RD, but not ReA.

Prior studies based on culture-confirmed bacterial disease have estimated incidence rates of 3.1-5.0 per 100,000[35, 36]. However, these estimates do not account for non-bacterial etiologies or uncultured IGE-associated ReA. A Scandinavian study showed a higher incidence (18 per 100,000) based on serologic and/or culture-based evidence of a prior infection[37].

The US military is at high risk of exposure to IGE-causing pathogens during deployment [1, 38, 39] and may subsequently be at an increased risk of ReA. The considerable health-care associated costs and decreased quality of life associated with ReA and other IGE sequelae could represent an important consideration to further justify efforts aimed at decreasing the incidence of IGE among US military.

This was a matched case-control study to evaluate ReA associated with IGE in a US military population. Data were obtained from the Armed Forces Health Surveillance Center (AFHSC) which operates the Defense Medical Surveillance System (DMSS), a database containing medical data on all active duty US military personnel[40]. All records meeting outcome case definitions from 1999 to 2007 were included. These data were linked via social security number to demographic data from the Defense Manpower Data Center (DMDC), and deployment data. Data were compiled into a single dataset, de-identified, and provided to the investigators by AFHSC personnel. The number of active military personnel, stratified by service, year, and age, was obtained through the Defense Medical Epidemiology Database (DMED)[41]. The study protocol was approved by the Naval Medical Research Center and the Uniformed Services University of the Health Sciences Institutional Review Boards in compliance with all applicable Federal regulations governing the protection of human subjects.

Between 1999 and 2007 a total of 506 subjects met the specific ReA case definition, including 482 with Reiter's disease (RD) and 32 with postdysenteric arthritis (PA). Eight cases met both the RD and PA case definitions. An additional 16,365 cases met the nonspecific arthritis/arthalgia (N.A.A) case definition. Incidence rates for specific ReA and N.A.A were 4.1 (95% CI: 3.7, 4.5) and 132.0 (95% CI: 130.0, 134.0) per 100,000, respectively. A rising incidence over time of N.A.A was observed (p < 0.0001), while incidence rates for specific ReA remained stable (Figure 1).

For both specific ReA and N.A.A, incidence rose with increasing age (Beta = 1.04, p = 0.002 and Beta = 55.68, p = 0.006, respectively). Distribution of cases by age category is shown in Figure 2. Nonspecific arthritis/arthralgia rates increased nearly 10-fold over the age spectrum, from 39.8 per 100,000 (95%CI 36.1-44.0) in the < 20 age group, to 349.7 per 100,000 (95%CI 339.5-360.2) in the > 40 age group. Similar increases were seen for specific ReA diagnoses, from 0.99 per 100,000 (95%CI 0.52-1.82) in the < 20 age group, to 7.09 per 100,000 (95% CI 5.70 - 8.72) in the > 40 age group.

Deployment to a high risk area was associated with an increased risk of specific ReA (OR: 2.07; 95% CI: 1.36, 3.15). However, it was protective against N.A.A (OR: 0.47; 95% CI: 0.41, 0.51). Army service (compared to other branches of service) was associated with an increased risk of N.A.A (OR 1.54; 95%CI: 1.49, 1.60). This association was reversed for specific ReA (OR: 0.77; 95% CI: 0.62, 0.94). Odds ratios for marriage and enlisted status in the N.A.A group were 1.21 (95% CI: 1.16, 1.27) and 1.66 (95% CI: 1.58, 1.74), respectively. Neither marital status nor rank were associated with a change in specific ReA risk.

Duration of ReA symptoms was prolonged after initial diagnosis. Among specific ReA cases that remained on active duty, 35.5% (89/251) were still receiving ReA-related medical care for a minimum of 2 years after initial presentation (Figure 3). This proportion was lower (p < 0.001) for the N.A.A cases (1,019/6,380; 16.0%). Increased duration of care was associated with increasing age for both the specific ReA (Cochran Armitage Trend p < 0.001) and non-specific (p < 0.001) arthropathy/arthralgia outcomes with the following proportion receiving care at least 2 years after initial diagnosis (specific ReA: ≤ 25: 13.5%, > 25 to < 35: 29.2%, ≥35: 57.3%; N.A.A: ≤ 25: 12.7%, > 25 to < 35: 28.5%, ≥35: 58.9%). Males had significantly higher proportion receiving care for specific ReA at 2 years than did females (38.7% vs 7.7%, respectively) (Pearson Chi-Square p = 0.002). This gender effect was not observed for the N.A.A outcome.

An analysis of the affected site showed that the majority (47%) of cases had no indicated joint. Although categories are not mutually exclusive, RD (limited to ICD-9 code 711.1) predominately affected the knee (9.8%) or multiple joints (17.1%). The hand (15.6%), knee (15.6%), and multiple joints (28.1%) were the most commonly affected in PA. N.A.A most commonly affected the shoulder (22.6%), knee (15.3%), and ankle/foot (13.0%), with only 2.4% coded as multiple joints.

We found an incidence of 4.1 per 100,000 persons for specific ReA in an active duty population. Prior publications have reported similar population incidence rates of 3-5 per 100,000, for ReA with culture confirmed antecedent bacterial IGE[35, 36]. However, as clinical recognition of antecedent infection is required for the specific diagnoses, both the incident rate we report and prior published rates for specific ReA likely underestimate the true incidence of IGE-associated ReA.

Prior studies have reported undifferentiated arthritis in an attempt to describe cases that do not meet the full diagnostic criteria for RD or PA. Schillerup et al. reported an incidence of 41 per 100,000 for undifferentiated inflammatory arthritis[44], and Fendler et al. described a ReA case series in which 59% of patients presented with a undifferentiated oligoarthritis[45]. Similarly, we used a N.A.A case definition and found an incidence of 132 per 100,000. Our N.A.A. case definition is overly sensitive due to limitations associated with the use of ICD-9 codes. It is very possible that a proportion of our N.A.A cases are somehow different than the specific ReA cases. However, our description of these cases illustrate an upper bound to the incidence of disease to be used in conjunction with the lower bound obtained from the highly restrictive postinfectious gastroenteritis and Reiter's disease case definitions. We suspect the large difference in incidence rates indicate there is a lack of clinical recognition of the reactive nature of at least some arthritis/arthropathy cases. Of note, we found an increasing incidence of N.A.A cases from 2002 to 2007 with an unknown etiology. The impact of deployment tempo during this time frame on N.A.A incidence could not be determined, as only data on deployment to certain high-risk areas was obtained.

The incidence of both specific ReA and N.A.A increased with age. Published descriptions of age in ReA indicate most cases occur in a young population, generally less than 40 years of age[2, 23]. This description likely stems from the earliest case reports of ReA, which largely consisted of STD-related cases. We found no prior publications reporting an increasing incidence with age. Further evaluation of the incidence of ReA in an older population would be worthwhile; however our use of an active duty study population limits our ability to do so.

As expected, we found increased odds of prior IGE in ReA cases compared to matched controls, despite the low rate of documented antecedent gastroenteritis. After controlling for covariates, the OR for specific ReA was 4.42 (95% CI: 2.24, 8.73). The effect was smaller in the N.A.A group (OR: 1.76, 95% CI: 1.49, 2.07), though it remained statistically significant. Our actual number of documented cases of IGE was very low; however relying on medical records for documentation of IGE severely underestimates the true incidence[42]. A prior population-based survey indicated less than a quarter of those with acute diarrheal illness seek medical care[46]. This is clearly evident in the paucity of documented IGE exposures in our cases diagnosed with RD and PA, both of which require a clinical recognition of antecedent GI disease.

Our cases are not strictly limited to IGE-related ReA. While we excluded ICD-9 documented Chlamydia and gonococcal infections, our methods potentially misses both asymptomatic chlamydial infections and STI cases diagnosed outside of military data catchment (such as through a public health department). However, IGE related disease likely comprises a large proportion of all reactive arthritis cases, as has been reported previously[37].

We assessed the effect of deployment to high traveler's diarrhea risk regions as a potential surrogate for IGE, due to incomplete data capture in a deployed population. We found an increased odds of prior deployment in those with specific ReA diagnoses. Conversely, there was a protective effect of deployment for the N.A.A diagnoses. The etiology of this relative difference is unclear, but the unique difficulties in capture of deployment related illness and injury data may account for the qualitative shift in the effect estimate. More severe injuries may be preferentially captured as those patients reach higher levels of care where integrated electronic medical systems are more prevalent. Additionally, stress injuries such as tendonitis, trauma, etc are commonly expected when deployed. Patients may be given one of these more common diagnoses, rather than one of the nonspecific codes used in our N.A.A. case definition. Finally, military members with underlying medical illness may not be deployed resulting in a type of "healthy worker" effect which might also explain the lower rate of diagnoses in deployed personnel.

Inherent difference between the specific ReA and N.A.A group could account for directional differences in some effect estimates noted above. This is supported by the difference in concurrent diagnoses for the two case definitions. Specifically, approximately 13% of the N.A.A cases were seen for physical therapy, 4% for lumbago, and 3% for aftercare following surgery. None of these co-morbid diagnoses were among the top 20 for the specific ReA cases. Interestingly, functional gastrointestinal diagnoses were prevalent in both groups (data not shown).

Although the original reports of ReA indicated a self-limited disease, recent publications have described chronic disease in up to half of cases, especially in the presence of HLA-B27[9, 34, 47]. Our data supports the idea of a more chronic disease, especially among patients over the age of 35. While we were unable to assess the symptoms associated with a ReA-related medical encounter, or HLA-B27 status, ReA continued to be commonly coded as a primary diagnosis up to 7 years after the initial medical encounter. The chronicity of this disease has not been well recognized by clinicians. However, like the functional gastrointestinal postinfectious disorders, the morbidity associated with ReA needs to be considered by providers and policy makers when assessing the impact associated with deployment of countermeasures, in order to mitigate the effects of acute enteric infections during deployment.

The U.S. active duty military population was used as the study population. There are several unique qualities of this population that should be considered when comparing our findings to the general population. The military population is younger (the average age is roughly 28) and generally physically fit, yet extensive physical stressors likely increase the number of musculoskeletal injuries in this population. Military members may travel extensively to underdeveloped areas of the world, placing them at increased risk for IGE exposures. In addition, as there is a direct link between health and duty status, deployment eligibility, and disability claim eligibility, members may have additional ICD-9 coded visits to address the administrative aspects of health issues.

Utilization of a medical encounter database for epidemiologic studies carries an intrinsic risk of misclassification or systemic biases, in addition to those described specifically above. Multiple studies have demonstrated inaccuracies in ICD-9 coding[48–50]. As described in previous publications, we required multiple medical encounters for the same diagnosis in an attempt to minimize case misclassification[43, 51].

In summary, we found that ReA was more prevalent than previously recognized in the U.S. active duty military population, especially when non-specific ReA cases were considered. This represents a significant burden on the military healthcare system and may be an important medical condition in returning veterans. Due to the strong association between IGE and ReA as well as other post-infectious sequelae, and the high risk of travelers diarrhea among deployed US service members, efforts aimed at minimizing IGE incidence in DoD personnel are warranted and support enhancement of current primary prevention strategies and development of novel solutions.