Objective Eosinophilic oesophagitis (EoO) and IBD are immune-mediated diseases of the gastrointestinal tract with possible overlapping pathogenic mechanisms. Our aim was to define the epidemiology and clinical implications of concurrent EoO and IBD diagnoses.
Design We conducted a prospective cohort analysis using the Truven MarketScan database (2009–2016) to estimate the incidence and prevalence of EoO in patients with Crohn’s disease (CD) or UC and vice versa. Cox proportional hazards and Kaplan-Meier methods were used to estimate the risk of EoO-related or IBD-related complications among patients with concurrent diagnoses.
Results Among 134 013 536 individuals, the incidence of EoO, CD and UC were 23.1, 51.2 and 55.2 per 100 000 person-years, respectively. The risk of EoO was higher among patients with CD (incidence rate ratio [IRR] 5.4, p<0.01; prevalence ratio (PR) 7.8, p<0.01) or UC (IRR 3.5, p<0.01; PR 5.0, p<0.01), while the risk of IBD was higher among patients with EoO (CD: IRR 5.7, p<0.01; PR 7.6, p<0.01; UC: IRR 3.4, p<0.01; PR 4.9, p<0.01) versus individuals without either diagnosis. Concurrent diagnosis of EoO and IBD was associated with greater composite risk of IBD-related complications (CD: adjusted HR (aHR) 1.09, p=0.01; UC: aHR 1.10, p=0.04) but lower composite risk of EoO-related complications (aHR 0.59; p<0.01).
Conclusion Based on a population-based prospective cohort analysis, the risk of EoO is significantly higher among patients with IBD and vice versa. Concurrent diagnoses might modify the risk of IBD-related and EoO-related complications. Studies defining the mechanisms underlying these observations are needed.
- inflammatory bowel disease
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Eosinophilic oesophagitis (EoO) and IBD are immune-mediated diseases of the gastrointestinal tract.
EoO and IBD share similar patterns of rising incidence and prevalence worldwide.
The epidemiology and clinical implications of EoO and IBD codiagnosis are unknown.
What are the new findings?
The incidence and prevalence of concurrent IBD and EoO are rising.
The incidence of EoO is threefold to fivefold greater in patients with IBD than in the general population, while the incidence of IBD is threefold to sixfold greater in patients with EoO than in the general population.
EoO is fivefold to eightfold more prevalent in patients with IBD, and IBD is fivefold to eightfold more prevalent in patients with EoO, compared with the general population.
Compared with patients without concurrent EoO, the codiagnosis of EoO among patients with IBD might be associated with an increased need for systemic corticosteroids for IBD treatment, but either a decreased or unchanged risk of biological therapy initiation or IBD-related hospitalisation.
Compared with patients without concurrent IBD, the codiagnosis of IBD among patients with EoO might be associated with a decreased risk of EoO-related complications, including food bolus impaction and oesophageal stricture requiring endoscopic dilatation.
How might it impact on clinical practice in the foreseeable future?
These observations suggest there may be some overlap between IBD and EoO disease pathogeneses and phenotypic manifestations.
Further investigation focused on defining the underlying reasons for this overlapping epidemiology and altered disease course may help elucidate approaches to disease prevention, novel targets for treatment and reduction of disease-related complications.
Eosinophilic oesophagitis (EoO) and the IBD are both chronic, incurable, immune-mediated diseases of complex pathogenesis that most often present in adolescence or early adulthood and are associated with high morbidity. EoO is characterised by eosinophilic infiltration of the oesophagus with consequent clinical manifestations of oesophageal dysfunction that are sometimes progressive.1 2 The prevalence of EoO appears to be greater in Western industrialised countries and lower elsewhere with rising incidence rates that exceed detection bias from increased recognition and surveillance.3 These observations notably resemble the global geographic prevalence and rising incidence of IBD, inclusive of Crohn’s disease (CD) and UC.4 As chronic immune-mediated diseases of the gastrointestinal tract, EoO and IBD share cytokine and helper T cell (Th)-mediated mechanisms of pathogenesis.5 6 In fact, CD sometimes manifests increased mucosal expression of interleukin 5 and eosinophilic infiltration,7 while UC is associated with increased expression of eotaxin, a protein that promotes eosinophil recruitment and inflammation.8 9 Moreover, epithelial barrier dysfunction and mucosal inflammation are common to both EoO and CD and phenotypically might manifest over time as fibrostenotic disease.10 11 Few case reports have described an overlap between Eoo and IBD.12 13 However, the epidemiology and clinical implications of this relationship are otherwise unknown; defining this relationship and identifying shared aetiopathogenic mechanisms could have important therapeutic, prognostic and even preventative implications. Our primary objectives were to define the epidemiology of concurrent EoO and IBD, as well as the clinical implications of concurrent diagnosis. To achieve these objectives, we constructed a population-based cohort using the Truven MarketScan database, which includes longitudinal clinical and procedural data for approximately half of the commercially insured US population.
The Truven Health MarketScan Commercial Database is the largest multipayer health claims database in the USA and has been extensively used for population health studies.14 15 The database comprises over 450 million patient-specific entries per calendar year and represents approximately 50% of the US population covered by employer-sponsored insurance plans. Enrollees include active employees, early retirees, Consolidated Omnibus Budget Reconciliation Act participants and their dependents. Deidentified data entries capture claims and clinical encounters for outpatient, inpatient and pharmaceutical services across the continuum of care as well as enrolment data. Enrollee data are linked even if employment or insurance coverage changes within a large selection of employers, public organisations and health plans. Data quality control includes internal auditing by each private payer system and across the entire database.
All enrollees in the database from 2009 through 2016 were included in this study. This timeframe represents the interval from the year after the EoO diagnosis code was first introduced in the USA through the latest year of available data. The International Classification of Diseases (ICD), 9th Revision (ICD-9) and 10th Revision (ICD-10), codes in all outpatient and inpatient encounters were used to identify patients with EoO (ICD-9: 530.13; ICD-10: K20.0), CD (ICD-9: 555; ICD-10: K50) and UC (ICD-9; 556; ICD-10: K51). A prior study demonstrated that the ICD code for EoO had a 99% specificity.16 To reduce the risk of misclassification for either EoO or IBD, identification of these diagnoses required at least two recurrent codes in any diagnosis position separated by at least 30 days.17 18 The definition of EoO did not specifically distinguish between proton-pump inhibitor responsive oesophageal eosinophilia (PPIREE) and the classic definition of EoO (clinical symptoms of oesophageal dysfunction and at least 15 eosinophils per high power field on oesophageal biopsies and exclusion of other potential causes of oesophageal eosinophilia including oesophageal reflux disease), as separate ICD codes do not exist for the two entities. Notably, recent international consensus guidelines have incorporated PPIREE into the diagnosis of EoO.19
We also analysed and compared the risk of certain Th1 and Th2 predominant immune-mediated conditions among patients with EoO, IBD or EoO-IBD codiagnosis. Predominant Th1-mediated conditions included coeliac disease (ICD-9: 579.0; ICD-10: K90.0) and IBD-associated rheumatological conditions, such as inflammatory arthropathy (ICD-9: 714; ICD-10: M05-M08), inflammatory spondylopathy (ICD-9: 720; ICD-10: M45-M49) and psoriasis (ICD-9: 696; ICD-10: L40). These rheumatological conditions, which are common extraintestinal manifestations of IBD, were included in the primary epidemiological analyses, but excluded from the analysis of disease-related complications to minimise confounding. Th2-mediated conditions included eczema (ICD-9: 691.8; ICD-10: L20) and asthma (ICD-9: 493; ICD-10: J45).
IBD-related complications were defined as the use of systemic corticosteroids for IBD indication, initiation of IBD-specific biologic therapy, IBD-related hospitalisation or intestinal resection. Medication data and admission dates were abstracted. An IBD-related hospitalisation was defined as an admission with CD or UC within the first two diagnosis positions. ICD and Common Procedural Terminology (CPT) codes in any procedure position were used to identify small bowel resections (ICD: 45.6, 45.73, 45.79, 0DB8, 0DBA, 0DBB, 0DT8, 0DTA and 0DTB; CPT: 44120, 44121, 44160, 44202, 44203 and 44205) and colectomies (ICD: 17.3, 45.7, 45.8, 48.4, 48.5, 0DBE-0DBH, 0DBK-0DBQ, 0DTE-0DTH and 0DTK-0DTQ; CPT: 44 140–44160, 44 204–44212, 45 110–45123 and 45 395–45397). Complications related to EoO included the need for oesophageal stricture dilatation and the development of food bolus obstruction. ICD and CPT codes in any procedure position were used to identify oesophageal dilatations (ICD: 42.92 and 0DT71-0D75; CPT: 43195, 43196, 43 212–43214, 43220, 43226, 43229, 43233, 43249, 43450 and 43453). Multiple procedures were considered independent episodes if they occurred at least 7 days apart. ICD codes in any diagnosis position were used to identify food bolus obstructions (935.1 and T18.12).
Descriptive analyses were used to estimate the incidence and prevalence and to compare the demographics between patients with IBD with versus without incident EoO, as well as for patients with EoO with versus without incident CD or UC. Categorical and continuous variables were compared using the χ2 statistic and Student’s t-test, respectively. The incidence rates of EoO and IBD were calculated as the number of new cases over the total follow-up time in person-years for the population of interest. Direction standardisation based on the entire study population was used to calculate age-adjusted and sex-adjusted rates. Patients who had less than 6 months of follow-up time prior to the diagnosis of EoO or IBD were considered prevalent cases and were excluded from the analysis of incidence rates. We evaluated secular trends in epidemiological patterns, given that EoO is an emerging diagnosis over the past decade. The annual prevalence of EoO and IBD was calculated as the number of cases over the total number of individuals in the population of interest for the given year.
Cox proportional hazards were used to compare the longitudinal risk of major disease-related complications between IBD patients with versus without EoO as well as EoO patients with versus without IBD. The time of entry was the date of IBD or EoO diagnosis, respectively. Follow-up times in the survival analyses were limited to the time of event, loss to follow-up, insurance disenrolment or predefined administrative censoring at 5 years (due to anticipated dearth of data beyond then). When evaluating individual outcomes, the development of another outcome did not lead to censoring of the evaluated outcome. Multivariable regression models were adjusted for age, sex, geographic region and baseline markers of IBD severity (systemic corticosteroid use, biological therapy, IBD-related hospitalisation and intestinal resection); each variable was defined at study entry and baseline variables included all relevant events within the 1-year time interval prior to study entry. Kaplan-Meier methods were used to compare the time with disease-related complications according to the presence of concurrent EoO diagnosis or IBD as appropriate. The analyses were stratified by IBD type (CD or UC). Analyses of the incident risk of developing EoO or IBD were further subgrouped according to age group (paediatric, age <18 years; adult >18 years) or sex (male or female). To reduce the risk of missing data, data were limited to the times of continuous enrolment, as all medical claims in or out of network would have been captured during this timeframe. We were unable to account for encounters that used self or non-commercial payment mechanisms, although these are expected to be minimal. Missing data on individual patient characteristics were few (<3.0%) and were thus handled using listwise deletion in the regression models. Statistical significance was defined as a two-sided α of less than 0.05. All statistical analyses were performed using SAS V.9.4 and Stata SE V.14.2.
Use of the primary data was performed in compliance with contractual agreements between the Stanford Center for Population Health Sciences and Truven Health Analytics, an IBM Company.
A total of 134 013 536 individuals were included for the primary analysis. The mean age of the cohort was 37.0 years (SD 18.6; range: 0–72), with a similar distribution of men and women (48.9% vs 51.1%), and a mean follow-up time of 2.2 years (SD 2.2). The mean follow-up times for patients with EoO or IBD were 4.4 years (SD 2.3) and 3.7 years (SD 2.3), respectively. The overall incidence rate of EoO was 23.1 per 100 000 person-years, and the overall incidence rates for CD and UC were 51.2 and 55.2 per 100 000 person-years, respectively. The overall prevalence of EoO was 78.8 per 100 000 individuals and the prevalence of CD and UC were 281.2 and 245.0 per 100 000 individuals, respectively (table 1).
Epidemiology of EoO in patients with IBD
Among the 376 822 patients with CD, the incidence and prevalence of EoO were 127.0 per 100 000 person-years and 594.2 per 100 000 persons, respectively (table 1). CD patients with concurrent EoO were more often male (62.0% vs 38.0%) and younger (mean 33.4 vs 45.7 years; p<0.01) compared with patients with CD without EoO. Among the 328 375 patients with UC, the incidence and prevalence of EoO were 82.7 per 100 000 person-years and 385.8 per 100 000 persons. UC patients with EoO were also predominantly male (56.3% vs 43.7%) and younger (mean 41.3 vs 49.4 years; p<0.01).
The overall risk of incident EoO among patients with CD (incidence rate ratio [IRR] 5.4, 95% CI 5.1 to 5.6) or UC (IRR 3.5, 95% CI 3.3 to 3.7) was significantly higher compared with individuals without IBD. As expected, the prevalence of EoO was also significantly higher among patients with CD (prevalence ratio [PR] 7.8, 95% CI 7.7 to 7.8) of UC (PR 5.0, 95% CI 5.0 to 5.1) compared with individuals without IBD. The mean times to diagnosis of EoO were 2.1 years (SD 1.7) for patients with CD and 2.0 years (SD 1.6) for patients with UC.
Stratified by sex, the age-adjusted incidence of EoO was greater among men with IBD (CD: 329.7 [95% CI 305.8 to 353.7] per 100 000 person-years; UC: 234.0 [95% CI 209.1 to 259.0]) versus women with IBD (CD: 172.4 [95% CI 155.8 to 189.1] per 100 000 person-years; UC: 130.2 [95% CI 114.5 to 145.8]). Similar to the demography of EoO among the general US population, EoO prevalence among IBD patients was greater in adult versus paediatric patients (CD: 481.4 vs 112.5, p<0.01 per 100,000; UC: 347.2 vs 38.1, p<0.01) and male versus female patients (CD: 368.1 vs 225.8 per 100,000, p<0.01; UC: 216.8 vs 168.4, p<0.01).
Epidemiology of IBD in patients with EoO
Among 105 554 patients with EoO, the incidence of CD and UC were 283.4 and 185.7 per 100 000 person-years, respectively. The prevalence of CD and UC were 2121.2 and 1200.3 per 100 000 persons, respectively. There was 5.7-fold higher risk of incident CD (95% CI 5.4 to 6.0) and 3.4-fold higher risk of incident UC (95% CI 3.2 to 3.6) among patients with versus without EoO. There was 7.6-fold and 4.9-fold higher prevalence of CD (95% CI 6.8 to 8.4) and UC (95% CI 4.4 to 5.5), respectively, among patients with versus without EoO. The mean times to diagnosis of IBD among patients with EoO were 1.6 years (SD 1.4) for CD and 1.9 years (SD 1.6) for UC.
Men and women with EoO had similar age-adjusted incidence rates of CD (310.0 [95% CI 267.4 to 352.5] vs 253.6 [95% CI 226.3 to 280.9] per 100 000 person-years) and UC, although there was a trend towards higher rates of UC among men with EoO compared with women with EoO (210.4 [95% CI 175.8 to 245.0] vs 156.4 [95% CI 134.8 to 178.0] per 100 000 person-years). Compared with adult patients, paediatric patients with EoO had a higher age-adjusted and sex-adjusted incidence of CD (383.7 [95% CI 319.0 to 448.4] vs 259.3 [95% CI 234.0 to 284.8] per 100 000 person-years) but a similar incidence of UC (161.5 [95% CI 117.6 to 205.4] vs 189.2 [95% CI 167.4 to 210.9] per 100 000 person-years) (table 2). As expected, the prevalence of IBD among patients with EoO was also greater in adult versus paediatric patients (CD: 1718.6 vs 401.7, p<0.01; UC: 1080.0 vs 118.4 per 100 000, p<0.01) and men versus women (CD: 1314.0 vs 806.2, p<0.01; UC: 674.5 vs 523.9 per 100 000, p<0.01). However, the prevalence of IBD among patients without EoO was more common among women (CD: 157.3 vs 122.2, p<0.01; UC: 134.7 vs 109.5 per 100 000, p<0.01).
Compared with the general population, the annual incidence of EoO rose more rapidly among patients with CD (+12.2 more cases per 100 000 persons per year; 95% CI 6.1 to 18.3; p<0.01) or UC (+10.7; 95% CI 6.3 to 15.2; p<0.01) (figure 1A). Similarly, compared with the general population, the annual prevalence of EoO also rose more rapidly among patients with CD (+129.9 more cases per 100 000 persons per year; 95% CI 111.9 to 147.9; p<0.01) or UC (+79.6; 95% CI 70.0 to 89.6; p<0.01) (figure 1B). During this same time interval, the annual rate of upper endoscopies was either stable or declined slightly for CD (−4.2 cases per 1000 persons per year; 95% CI −6.1 to −2.4; p<0.01), UC (−1.7; 95% CI −4.3 to 0.8; p=0.15) and the general population (−0.4; 95% CI −0.1 to −0.6; p<0.01). In addition, the rate of EoO diagnoses per initial upper endoscopy increased fourfold (from 5.5 per 1000 endoscopies) for CD and fivefold (from 3.8 per 1000 endoscopies) for UC between 2009 and 2016. The greater increase in EoO diagnoses among patients with IBD relative to the stable increase in upper endoscopies suggests that the observed epidemiological trends are not wholly attributable to detection bias.
The annual incidence (per 100 000 persons) of CD declined by a greater magnitude among patients with EoO (−13.9 fewer cases per year; 95% CI −24.0 to −3.7; p<0.01) versus without EoO (−3.6; 95% CI −5.3 to −1.8; p<0.01), although not statistically significant, while the annual incidence of UC remained stable for patients with EoO (6.8; 95% CI −8.2 to 21.9; p=0.29) versus without EoO (−1.6; 95% CI −4.3 to 1.2; o=0.20) (figure 2A). Nonetheless, the annual incidence of CD and UC remained between 3 and 7 times greater in patients with versus without EoO for all years during the study period. The annual prevalence of IBD (per 100 000 persons) rose more rapidly among patients with EoO (CD:+191.7 more cases per year, 95% CI 178.0 to 205.4, p<0.01; UC: +119.4 more cases per year, 95% CI 102.9 to 135.9, p<0.01) versus without EoO (CD: +20.6 more cases per year, 95% CI 12.5 to 28.7, p<0.01; UC: +23.2 more cases per year, 95% CI 19.0 to 27.4, p<0.01) (figure 2B).
Risk of other immune-mediated comorbidities
When comparing patients with versus without EoO, the relative risk (RR) of predominantly Th1-mediated conditions was greater for coeliac disease (RR 19.6; 95% CI 18.6 to 20.8; p<0.01) and IBD-associated rheumatological conditions (RR 2.6; 95% CI 2.6 to 2.7; p<0.01) (figure 3). The concurrent diagnosis of EoO and IBD increased the risk of these immune-mediated conditions by an additional 1.4-fold to 2.4-fold. For Th2-mediated conditions, the risk of eczema (RR 5.1; 95% CI 4.9 to 5.3; p<0.01) and asthma (RR 5.5; 95% CI 5.4 to 5.6; p<0.01) was greater in patients with EoO, although the concurrent diagnosis of EoO and IBD did not markedly increase the risk (1.0-fold to 1.3-fold) of these diagnoses.
Among patients with CD, EoO was associated with an increased composite risk of CD-related complications (adjusted HR [aHR] 1.09; 95% CI 1.02 to 1.16) (table 3, figure 4A). When analysed by complication type, diagnosis of EoO concurrent with CD was associated with 14% higher risk of needing systemic corticosteroids for CD treatment (aHR 1.14; 95% CI 1.06 to 1.22), but a 23% lower risk of needing to initiate biological therapy (aHR 0.77; 95% CI 0.66 to 0.89) and 20% decreased risk of intestinal resection (aHR 0.80; 95% CI 0.66 to 0.96) when compared with patients with CD without concurrent EoO. Concurrent EoO was not associated with an increased risk of hospitalisation (aHR 0.95; 95% CI 0.85 to 1.06).
Among patients with UC, EoO was similarly associated with an increased composite risk of UC-related complications (aHR 1.10; 95% CI 1.00 to 1.21) (figure 4B). Concurrent diagnosis of EoO was associated with a 10% higher risk of needing systemic corticosteroids for UC treatment (aHR 1.10; 95% CI 1.00 to 1.22). There was a statistically non-significant trend towards decreased risk of initiation of biological therapy (aHR 0.70; 95% CI 0.45 to 1.11), decreased risk of colonic resection (aHR 0.69; 95% CI 0.45 to 1.06) and increased risk of hospitalisation (aHR 1.18; 95% CI 0.97 to 1.44) compared with patients with UC without EoO.
Among patients with EoO, having concurrent IBD was associated with a 53% lower risk of food bolus impaction (aHR 0.47; 95% CI 0.34 to 0.66) and 35% lower risk of needing endoscopic oesophageal stricture dilatation (aHR 0.65; 95% CI 0.54 to 0.79) (table 4, figure 5). Due to data use agreements that restricted the ability to evaluate small cell sizes for privacy reasons, we were unable to further stratify the analyses or report more detailed results.
Based on a prospective US population-based cohort of over 134 million individuals, we demonstrated for the first time that the incidence and prevalence of EoO among patients with IBD, and vice versa, is significantly higher than the general population. Specifically, the risk of EoO among patients with IBD was 3–5 fold higher, while the incidence of IBD among patients with EoO was 3–6 fold higher. Concurrent EoO and IBD was also associated with greater risk of some IBD-related complications but lower risk of EoO-related complications
EoO and IBD are both chronic, often progressive, immune-mediated diseases of the luminal GI tract, with complex pathogeneses that implicate environmental factors, aberrant host immune responses and likely microbial factors, interacting on a background of genetic proclivity. Some overlap in pathogenesis and phenotype (eg, fibrostenotic CD and EoO oesophageal stricture), thus seems plausible, although not yet well defined. EoO and IBD were previously rare diagnoses outside of Western industrialised countries, but each is steadily rising in both incidence and prevalence and both are now considered global diseases.3 4 However, apart from case reports, there are sparse data describing the epidemiology of concurrent EoO and IBD. Despite this knowledge gap, current gastroenterology and allergy society guidelines recommend exclusion of IBD in the diagnosis of EoO due to concerns that oesophageal eosinophilia may reflect oesophageal involvement of CD.2 19 Oesophageal disease, however, is an extremely uncommon manifestation of CD with endoscopic features (erosion, ulceration and pseudopolyps) that are distinct from EoO (rings, furrows and exudate).20–22 Moreover, oesophageal eosinophilia as a manifestation of CD would be even more rare. As support, in a study of 24 patients with CD oesophagitis (out of 12 367 patients with CD) seen at the Mayo Clinic between 1998 and 2012, histological findings were characterised by the presence of lymphocytes and plasma cells; none were described to have eosinophils.20 That we observed an increased incidence and prevalence of EoO in patients with UC, whose inflammation would have theoretically been limited to the colon, provides further biological credence to our findings.
Validating these epidemiological trends and identifying shared disease determinants will have important clinical implications for both EoO and IBD. The rate of increase in both diseases over a short period of time strongly implicates environmental factors and gene-environment interactions, as opposed to genetic factors alone. Early life risk factors common to both diseases include caesarean delivery, intestinal infections and antibiotic use.23 24 Dietary intake has also been implicated in the pathogenesis of both conditions, either through introduction of food antigens or alterations in the microbiome.25–27 Aberrant responses to alterations in the gastrointestinal microbiota is another potential shared mechanism between EoO and IBD.24 28 Both EoO and IBD pathogenesis involve upregulation of toll-like receptors (TLRs) in the oesophageal29 and intestinal mucosa,30 respectively. TLRs activate the innate immune response to pathogen-associated molecular patterns of gastrointestinal tract microbes. Additional overlaps between EoO and IBD pathogenesis include the increased expression of proeosinophilic cytokines in both diseases,5 6 the involvement of Th2-mediated pathways of inflammation in EoO and UC (although CD primarily invokes the Th1 and Th17 pathways)7–9 and disruption of the mucosal barrier.10 While IBD and EoO were each independently associated with other immune-mediated conditions, codiagnosis of IBD and EoO was associated with an even greater risk of Th1-mediated conditions (coeliac disease, IBD-related rheumatological disorders), but this was minimal for the Th2-mediated conditions (eczema and asthma). The risk of asthma and eczema among patents with EoO without concurrent IBD was stronger than the risk among patients with CD or UC alone. Given that the risk was not markedly different between EoO patients with versus without IBD, these findings suggest that EoO might be the stronger driver in in the pathogenesis of Th2-mediated conditions when there are codiagnoses.
The risk of disease-related complications associated with concurrent EoO and IBD diagnosis has not been previously reported. Compared with patients with EoO and no IBD, a concurrent diagnosis of IBD (irrespective of CD vs UC) was associated with a more benign clinical course of EoO, evidenced by a lower risk of food bolus impaction and need for oesophageal stricture dilatation. Given that virtually all patients who received systemic corticosteroids had IBD, it is possible that corticosteroids served as the mediator for the observed reduction in EoO-related complications; systemic corticosteroids are known to improve clinical and histological outcomes in EoO.31 Exploratory analyses revealed that systemic corticosteroid use for IBD was associated with 53% (odds ratio (OR) 0.47, 95% CI 0.28 to 0.78) lower risk of food bolus impaction and 23% (OR 0.77, 95% CI 0.59 to 1.01) statistically non-significant lower risk of needing oesophageal stricture dilations among patients with concurrent IBD and EoO. Another possibility is that IBD patients (who are habituated to long-term medication use) with concurrent EoO might be more compliant with their EoO treatment regimen and thus have fewer EoO-related complications.32 33
Patients with concurrent EoO-IBD had a significantly higher risk of needing systemic corticosteroids for an IBD-related indication, although there was a lower risk of biological initiation and subsequent intestinal resection among patients with CD (but not UC). The inherent nature of administrative claims databases unfortunately limits the ability to confidently identify underlying aetiologies for certain observations. We can offer some hypotheses, but these are speculative. It is possible that earlier endoscopic evaluation in patients with EoO led to earlier diagnosis of CD, earlier treatment and improved longitudinal outcomes.34 35 Additionally, there may have been an increased reliance on corticosteroids instead of biological therapy for potentially milder disease. Due to data use restrictions, we were not able to adjust for some patient-related and disease-related factors; how these factors modify the observed association between EoO or IBD and respective disease course is not known. Additional studies confirming or refuting our findings related to IBD-related and EoO-related complications with codiagnosis are needed, as are experimental studies to define mechanisms underlying these observations.
Our study has several strengths. The relative rarity of EoO in the general population, further amplified when limited to patients with a diagnosis of IBD, necessitates a large prospective cohort to adequately power an investigation aimed at accurately defining epidemiological patterns of concurrent EoO and IBD. As the largest multipayer health claims database in the USA with longitudinal data for over half the US population with commercial coverage, the MarketScan database (which has been used extensively for other population-based investigations) provides adequate power for achieving our predefined objectives. Our access to comprehensive healthcare utilisation and medication data allowed us to identify specific diagnoses, medication prescriptions, procedures and hospitalisations (along with their respective indications), as well as the temporal relationship of these occurrences. Because of the nature of the health claims data, enrollee data are linked even if insurance coverage changes, thus maximising the opportunity for complete data and event capture as well as continuous patient follow-up. As evidence, missing data were minimal (<3%) in our study. The MarketScan database also undergoes routine quality checks; indeed, that the non-concurrent EoO and IBD incidence and prevalence estimates corresponded to other published population-based data supports both the reliability and validity of the data as well as our diagnostic algorithms for case identification. We also showed that the observed trends were not explained by detection bias or coding patterns. First, the rising incidence persisted over time, several years after the introduction of the ICD-9 code for EoO in 2008. Second, the rise in incidence of EoO was also more pronounced among patients with IBD relative to those without IBD, and we would not expect differential ICD-9 coding for EoO between patients with versus without IBD. Finally, we demonstrated that the rise in EoO incidence was not attributable to a rise in the number of endoscopic procedures performed over time, as the annual rates of upper endoscopies either remained stable or declined throughout the study period.
Our study has also some limitations, which mostly reflect inherent limitations of large administrative databases. Our population was limited to individuals with commercial insurance coverage, while excluding the uninsured and those covered by government-funded plans, such as Medicare. Thus, our results might not be generalisable to this population. Due to data protection agreements for the MarketScan database, we did not have access to more granular data (eg, IBD location and behaviour, disease activity, anthropometrics, endoscopic and histological reports, date of IBD diagnosis if prior to enrolment, smoking history and family history) that might further characterise the epidemiology and natural course of concurrent diagnoses of EoO and IBD. Data protection of small cell sizes also limited our analyses of disease-related complications, particularly for EoO-related complications, so these findings should be interpreted with caution pending additional confirmatory studies. A related limitation is that we could not confirm that IBD-related hospitalisations were for IBD flares per se; that said, it is likely that the vast majority of these hospitalisations were for flares since we used the same methodological approach as other studies to increase specificity and limited the criteria for inclusion to two diagnosis positions.36–38
In conclusion, in this first large population-based study of EoO and IBD, we provide strong evidence that EoO is more common in patients with IBD and that IBD is more common in patients with EoO. Concurrent diagnosis of EoO and IBD also have clinical implications with respect to disease-related outcomes for both diseases. As chronic immune-mediated diseases of the gastrointestinal tract, EoO and IBD have overlapping disease pathogeneses, but neither the extent of overlap nor specific mechanisms are well defined. Our findings clearly highlight the need for additional clinical and translational investigations to further unpack the relationship between these two diseases and to also externally validate our observations.
BNL and SCS contributed equally.
Contributors BNL participated in the study concept and design, data analysis, data interpretation, drafting of the manuscript, critical revision and final approval of the manuscript. SCS and IH participated in the study design, data interpretation, critical revision and final approval of the manuscript. EB participated in the study concept and design, data interpretation, critical revision and final approval of the manuscript. J-FC participated in the study concept and design, data interpretation, critical revision and final approval of the manuscript. J-FC provided study supervision.
Funding Access to the Truven Health MarketScan Commercial Database was provided by the Stanford Center for Population Health Sciences (PHS) Data Core. The PHS Data Core is supported by a National Institutes of Health National Center for Advancing Translational Science Clinical and Translational Science Award (UL1 TR001085) and internal Stanford funding.
Competing interests IH received consulting fees and research support from Adare, Allakos, Celgene, Regeneron and Shire. J-FC received research grants from Abbvie, Janssen and Takeda; received payments for lectures from Abbvie, Amgen, Ferring, Shire and Takeda; received consulting fees from Abbvie, Amgen, Boehringer Ingelheim, Celgene, Celltrion, Enterome, Ferring, Genentech, Janssen, Eli Lilly, Mediummune, Merck, Novartis, Pfizer, Protagonist, Sandoz, Second Genome, Seres, Shire, Takeda, Theradiag and Theravance; and holds stock options in Intestinal Biotech Development and Genfit.
Ethics approval The study protocol was reviewed and deemed exempt by the Institutional Review Board of Stanford University.
Provenance and peer review Not commissioned; externally peer reviewed.
Patient consent for publication Not required.
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.