Natural history of eosinophilic esophagitis: a systematic review of epidemiology and disease course

SUMMARY Eosinophilic esophagitis is a chronic immune-mediated esophageal disorder. For its timely diagnosis, clinicians must recognize common symptoms, and understand differences in symptoms across patient groups. The aim of this study is to systematically review the epidemiology and natural history of eosinophilic esophagitis. The MEDLINE, Embase, and Cochrane databases were searched from 1974 to February 2017 for studies describing the epidemiology and natural history of eosinophilic esophagitis. Congress abstracts from 2014 to 2016 were also searched. Search results were screened against predetermined inclusion/exclusion criteria by two independent reviewers, and data extraction was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Of 1376 articles identified, 47 met the inclusion criteria: 20 on epidemiology and 27 on natural history. Incidence and prevalence of eosinophilic esophagitis varied widely across North America and Europe, and increased over time. Incidence increased 131-fold in the Netherlands (1996–2010), 20-fold in Denmark (1997–2006), and 5.1-fold in Calgary, Canada (2004–2008). The most commonly reported symptoms were emesis and abdominal pain in children, and dysphagia and food impaction in adults. Age at diagnosis was 5.9–12.0 years in children, and approximately 30 years in adults. Time between symptom onset and diagnosis was 1.2–3.5 years in children and 3.0–8.0 years in adults. Diagnostic delay was associated with an increased risk of endoscopic features of fibrostenosis. Symptoms of eosinophilic esophagitis differed significantly by age and race. In conclusion, there is an increasing incidence and prevalence of eosinophilic esophagitis. The considerable delay between symptom onset and diagnosis suggests that clinicians do not readily recognize the disease, which may have important clinical ramifications.


INTRODUCTION
Eosinophilic esophagitis (EoE) is a chronic, immunemediated disorder of the esophagus, which is associated with a large number of eosinophils infiltrating the esophageal mucosa. 1 The predominant symptoms in EoE are dysphagia, feeding dysfunction, food impaction, chest pain, gastroesophageal reflux disease increasing awareness of the disease, a thorough understanding of the evolving epidemiology of EoE is essential. Furthermore, common symptom patterns and disease progression of EoE are incompletely described.
This systematic review addresses two objectives. First, it evaluates the incidence and prevalence of EoE in children and adults, including trends over time. Second, it elucidates the symptom patterns and the natural history of EoE in all age groups.

Identification of studies
The systematic review was designed to identify literature that included data on the epidemiology and natural history of EoE in children and adults.
Searches were carried out using search strategies including a combination of free text and Medical Subject Headings (MeSH) terms (Supplementary Table  S1). Epidemiology searches queried EoE and search terms related to incidence, prevalence, risk factors, and trend analysis. Natural history searches queried EoE and search terms related to disease features, functional abilities (e.g. physical, cognitive, and psychiatric), severity and frequency of signs and symptoms, time course and predictors of manifestation progression, and survival.
Searches were conducted in Ovid R on February 27, 2017 using: MEDLINE R In-process & Other Nonindexed Citations and Ovid MEDLINE (covering publications from 1946 to present), Embase R (covering publications from 1974 to present), and Evidence-Based Medicine Reviews, comprising the Cochrane Database of Systematic Reviews, the American College of Physicians Journal Club archives, the Database of Abstracts of Reviews of Effects, the Cochrane Central Register of Controlled Trials, the Cochrane Methodology Register, the Health Technology Assessment Database and the National Health Service Economic Evaluation Database. Searches were limited to studies in humans and English language publications. Supplementary searching included manual review of abstracts from congresses held from 2014 to 2016 by the following organizations: the American College of Gastroenterology; the North American Society for Pediatric Gastroenterology, Hepatology and Nutrition; the American Academy of Allergy, Asthma & Immunology; the American College of Allergy, Asthma & Immunology; the International Society for Pharmacoeconomics and Outcomes Research Annual International Meeting; and Digestive Diseases Week. Further publications were identified on the basis of assessment of the reference titles listed in articles selected for inclusion in this study and in systematic reviews identified in the searches described above.

Study selection
Once publications had been identified, they were independently screened and evaluated by two independent reviewers based on their title and abstract in accordance with the 2009 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. 4 Disagreements between reviewers regarding study selection were resolved by consensus.
To be included in the full-text review, publications had to meet specific criteria defined in a predetermined protocol (Supplementary Table S2). Searches included studies of adults and/or children with EoE. In agreement with the most recent European consensus guidelines, 3 studies were included regardless of whether they excluded individuals with symptoms of GERD or patients who were responsive to proton pump inhibitor (PPI) treatment. Natural history searches excluded interventional studies. Studies were restricted to English language publications.

Data extraction
Data extraction was carried out by one researcher using a standardized data extraction sheet, and these extractions were reviewed and confirmed by a second independent researcher. A third researcher performed a quality check of all extracted data. Data were presented as reported in the identified publications without applying any meta-analytic techniques.
The extracted data from epidemiology studies included the name of the first author, year of publication, study population, time period of estimate, sample size, diagnostic and exclusion criteria, and annual prevalence and incidence values per 100 000 people. Data ranges indicating the lowest and highest annual prevalence or incidence values reported in a given time period of estimate were also extracted.
Data extracted from natural history studies included the name of the first author, year of publication, study region, ethnicity, gender and age of study participants, diagnostic criteria, and information on whether patients with GERD or proton pump inhibitor-responsive esophageal eosinophilia (PPI-REE) were excluded in the study. Moreover, the four most common symptoms reported in each study were summarized in a table and reported both as absolute numbers and as the proportion of patients in a study population (in percentages). We further assessed the mean age and mean duration of symptoms at diagnosis (in years; including standard deviations, if available) and analyzed symptoms by age category as proportion of patients (in percentages).

Assessment of quality of evidence
The identified studies were evaluated for quality of evidence using the Newcastle-Ottawa scale for quality assessment of cohort, case-control and crosssectional studies (Supplementary Table S3).

Identified papers
The searches identified 1674 articles, of which 1376 were screened after removal of duplicates. In total, 47 references met the inclusion criteria (Fig. 1); of these, 20 articles reported on the epidemiology and 27 on the natural history of EoE.

Incidence and prevalence
Incidence and prevalence estimates varied widely across the USA, Canada, Europe, and Australia (Fig. 2 and Supplementary Table S4). Studies also varied widely regarding methods of identifying cases of EoE; three relied on ICD-9 codes, 5,6,8 one used the SNOMED-CT diagnosis, 13 one used a combination of SNOMED and ICD-10 codes, 14 and 15 used histology alone for the diagnosis of EoE. 7,[9][10][11][12][15][16][17][18][19][20][21][22][23][24] Of the studies using histology for the diagnosis of EoE, the majority used a cut-off of 15 or more eosinophils per high-power field. 7,[10][11][12][14][15][16][17][18][19][20]23 In the USA, prevalence estimates in children ranged from 7.3 per 100 000 per year in 1995-2004 in West Virginia, 7 to 50.5 per 100 000 across the USA for 2009-2011. 6 In adults, prevalence estimates across the USA ranged from 9.5 per 100 000 in 2008-2009 5 to 58.9 per 100 000 in 2009-2011. 6 The incidence of EoE in adults and children in North America ranged from 0.35 cases per 100 000 person-years in 1991-1995 in Minnesota, USA, 10 to 10.7 cases per year in 2008 in Calgary, Canada. 11,12 In European studies, no Representative data are presented for each country as incidence per 100 000/year and prevalence per 100 000. For each country where data are available, rates were taken either from the only study available (even if it relates only to a region) or from the study with the largest sample size. † Data relate to a region of this country. ‡ Age-and sex-adjusted values are reported. § Study population included children only. ¶ Study population included adults only. † † For the USA, we represented data from Dellon et al. 6 21,22 The largest study analyzed the prevalence of EoE in the USA in 2010-2015 using the Explorys database, covering a source population of 30 301 440 patients. 13 The overall period prevalence was 25.9 per 100 000. The prevalence of EoE was significantly higher in male patients than in female patients (odds ratio [OR] 2.00, 95% confidence interval [CI] 1.92-2.10). It was also significantly higher in Caucasians than in Asians and African-Americans (OR 2.00, 95% CI 1.86-2.14), and in the adult population (18-65 years of age) compared with children or the elderly (OR 1.63, 95% CI 1.54-1.71).

Trends over time
Seven studies reported temporal trends in the incidence and prevalence of EoE in adults and children, all of which reported an increase over time (Fig. 3). 11,12,14,17,[20][21][22] Three studies examined the incidence and prevalence of the disease in two regions of Switzerland. [20][21][22] In one of these regions (Olten County), the prevalence increased from 2 per 100 000 in 1989 to 23 per 100 000 in 2004, and the incidence increased from 2 per 100 000 in 1989 to 7.4 per 100 000 in 2007-2009. 21,22 In the second region (the Canton of Vaud), no inhabitants were diagnosed with EoE between 1993 and 2003. 20 The first patients were diagnosed in 2004, and the incidence of disease increased from 0.16 per 100 000 in 2004 to 6.32 per 100 000 in 2013. 20 The prevalence increased from 0.16 per 100 000 to 24.08 per 100 000 in the same time frame. Similarly, the incidence of the disease increased 131fold in the Netherlands from 1996 to 2010, 17 20

Disease progression
A number of studies examined disease progression, including changes in symptoms over time, recurrence and remission of disease. These are summarized below. However, they are difficult to compare due to heterogeneity in study design, in particular whether treatment was given and, if so, whether the impact of this on disease course was assessed.

Changes in EoE symptoms
Three studies evaluated changes in symptoms postdiagnosis. 31,35,36 One study of 30 adults with EoE found that after a mean follow-up of 7.2 years during which they had not been receiving medical therapy, 11 (36.7%) experienced a decrease in intensity of dysphagia, 7 (23.3%) reported a worsening in intensity of dysphagia, 11 (36.7%) described dysphagia as being stable-persistent, and one patient (3.3%) reported that symptoms of dysphagia had disappeared by the end of the follow-up period. 31 Eleven of the patients received dilation during the study; however, it is not clear from the report how this related to the severity of their dysphagia. A second study of 562 children used retrospective and prospective medical chart review to assess changes in a subpopulation of 24 children who had not received medical or dietary treatment. 35 After a mean follow-up period of 5.2 years, dysphagia and food impaction became worse. The third study used the Endoscopic Reference Score to classify patients into those with fibrostenotic versus inflammatory EoE. 36 Comparison of these groups showed that a similar proportion of patients who received steroid, diet, and/or PPI therapy reported an improvement in symptoms such as dysphagia and food impaction regardless of whether they presented with fibrostenotic or inflammatory disease (69% vs. 60%, P = 0.256). 36 More than half of 256 pediatric and  of patients received swallowed topical steroids for 4-8 weeks, 40% received PPIs for 6-12 weeks, and 20% were treated with medication and dilation, but did not analyze recurrence by treatment group.

Remission of EoE
Six studies reported on the proportion of patients achieving 'remission' of EoE. 18,32,33,35,36,44 However, only three of these examined remission off treatment 18,33,35 and only one reported the proportion of patients with both histological and symptomatic resolution off treatment. 18 This study of 25 children with EoE in Slovenia, all of whom were treated primarily with dietary therapy, found that 20% of patients had complete remission despite food reintroduction and cessation of steroid or PPI treatment. However, it should be noted that the definitions of histological and symptomatic remission are not given in the report, and that the duration of time between treatment cessation and biopsy may have been too short to accurately assess disease remission.

Disease progression from childhood to adulthood
Three studies explored the progression of EoE from childhood to adulthood. 39,40,44 In all of these studies, the current status of adults who had had EoE since childhood was assessed via a telephone survey.
In the first study, 40 53 adults with well-characterized EoE that had been diagnosed in childhood responded to the cross-sectional survey. At the time of the survey, their mean age was 20.5 (SD: 2.47) years and their mean age at diagnosis was 13.5 years (SD: 3.52). The survey found that two patients (4%) had a positive dysphagia score (defined as scores ≥40) on the Mayo Dysphagia Questionnaire-30 Day (MDQ-30), 18 (34%) reported some difficulty swallowing in their MDQ-30 responses, and 33 patients (62%) reported regurgitation or heartburn over the past month. During this time, 49% were receiving ongoing PPI therapy, and 76% were following an allergy-directed diet. Results from this study indicate that the majority of young adults diagnosed with EoE during childhood continue to require treatment or dietary modification for EoE during early adulthood. In a second study, adults (mean age 22 years) who had histological evidence of EoE as children were contacted a mean of 15 years after their initial childhood biopsy. 39 At this followup, 49% of patients had dysphagia, 73% had recent upper gastrointestinal tract symptoms, 40% had a history of food impaction, and 41% needed ongoing care from a gastroenterologist. 39 Again, this study suggests that symptomatology associated with EoE diagnosed in childhood commonly persists to adulthood. However, results from a third study (N = 58; mean duration between diagnosis and follow-up: 8.3 years) revealed that most young adults (81%) who were diagnosed with EoE as children reported resolution (47%) or improvement of their symptoms (34%). 44 Given that within the previous 12 months, only 12% and 24% of patients were treated with steroids or PPI, respectively, the results from this study, in contrast to the previous two, suggest that a large proportion of patients had symptom resolution or improvement and did not require ongoing healthcare.

Association between age and endoscopic features
Four studies showed an association between older age and the presence of endoscopic features of fibrostenosis, 25,34,36,41 while two studies found no evidence for such an association. 18,30 In children, the prevalence of strictures significantly increased with age (<3 years vs. 3-10 years vs. >10 years, 0% vs. 1.8% vs. 6.9%, P = 0.04). 25 In populations that included children and adults, the number of rings (peaked in the 18-35 years age group and plateaued thereafter, P < 0.001), strictures (peaked in the ≥51 years age group, P < 0.001), and Schatzki rings (peak in the ≥51 years age group, P < 0.001) increased with age, whereas white plaques decreased with age (lowest value in the ≥51 years age group, P < 0.001). 34 Moreover, inflammatory endoscopic features such as plaques presented at a younger age than fibrostenotic features such as rings and strictures (24 years vs. 39 years, P = 0.001; 24 years vs. 34 years, P = 0.02). 36,41 Association between diagnostic delay and endoscopic features Data from four studies indicated an association between longer diagnostic delay and increasing presence of some endoscopic features of EoE. 30,34,36,41 The duration of symptoms before diagnosis was significantly longer in patients with rings and strictures than in those without (rings: 24 months vs. 13 months, P < 0.001; strictures: 36 months vs. 15 months, P < 0.001). 34 Similarly, the duration of symptoms before diagnosis was longer in patients with fibrostenotic features than in those without (8.1 years vs. 5.3 years, P = 0.002; 12.9 years vs. 5.1 years, P < 0.0001). 36,41 Patients with a small esophageal diameter of 6.0-9.9 mm or a medium diameter of 10.0-16.9 mm had a significantly longer duration of symptoms prior to diagnosis than individuals with larger diameters of ≥17 mm (P < 0.0001 and P = 0.003, respectively). 41

DISCUSSION
This systematic review shows that the incidence and prevalence of EoE vary widely across North America and Europe, and have increased dramatically over time. The most common symptoms of EoE are emesis, abdominal pain, and dysphagia in children, and dysphagia, food impaction, and heartburn in adults. These symptoms differ significantly by age and race.
Our review further reveals the common occurrence of delays of several years from the onset of symptoms until the diagnosis of the disease. Eosinophilic esophagitis appears to be a progressive disease that persists from childhood to adulthood. Data on the disease course are very limited. The studies that are available suggest a relatively high recurrence rate and that few patients achieve resolution without treatment.
Differences and inconsistencies in diagnostic criteria and methodology may partially explain the varying estimates of the prevalence and incidence of EoE across countries and studies. Such variance can be seen even within a single country, as was noted in the data from the Canton of Vaud compared with Olten County in Switzerland. 21,22 One hypothesis could be that studies including patients with PPI-REE are likely to find a higher estimated prevalence and incidence than those in which such patients are excluded. However, much variation in incidence and prevalence was still seen between studies that excluded patients with PPI-REE, suggesting that factors such as differences in diagnostic criteria, diagnosis codes used to define disease, or other factors may impact the reported epidemiology of EoE.
The data also suggest that the incidence and prevalence of EoE are rising among both children and adults in developed countries. Awareness of the disease may play a crucial role in these variations, given that the lowest incidence and prevalence data are from earlier publications when less was known about the disease. However, it is unclear how much of this increase is attributable to true changes in incidence of disease. In pediatrics, variation may also be due to the increasing availability and accessibility of pediatric endoscopy equipment and expertise. Conversely, causative environmental exposures may have changed over time. The observed variations in the incidence and prevalence of EoE, and its increase over time, are consistent with observations from previous work. 47 Moreover, it is unclear whether rates are similar in less developed countries, and whether the lack of epidemiological data in these countries is due to lack of access to technology/healthcare, due to symptoms being reported only as a nuisance and considered of low priority, or if there is a real difference in presentation and prevalence outside of the developed world.
The most commonly reported symptoms of EoE are emesis, abdominal pain, dysphagia, and food impaction in children, and dysphagia, food impaction, heartburn, and acid regurgitation in adults. It is interesting to note that dysphagia is considerably more prominent in adults than in children (46.2-94.5% vs. 4.8-60.9%), suggesting age-dependent differences in disease manifestation. However, it is important to consider the impact of language development in reported symptoms. For instance, since very young children may struggle to express complicated symptoms such as dysphagia, these symptoms may manifest themselves in food refusals for solid textures. Capturing such outcomes is difficult and may not be reflected in the data above.
The prevalence of symptoms varies significantly across studies and throughout the course of the disease, and differs significantly by age and race, while the impact of sex remains controversial. 34,42 For instance, occurrence of dysphagia and food impaction increases considerably from early childhood, and both of these symptoms are more common in Caucasians than in African-Americans and other races. 34 These findings suggest that heterogeneity of patient populations, particularly differences in symptom presentation depending on the age and race of the patient, may explain some of the variation in symptoms seen across studies. However, it should be noted that conclusions related to epidemiologic differences in symptomatology by race and age are limited by the lack of worldwide data. This is the first systematic review to assess the disease course of EoE. The mean time from symptom onset to diagnosis of EoE was up to 3.5 years in children and 8.0 years in adults, 26,32,[36][37][38] suggesting a need for a better understanding of the common symptoms and early indicators of the disease, which would enable clinicians to provide earlier diagnosis and therefore more effective treatment as well as making patients aware of a potential underlying disease.
Early diagnosis may be particularly important because evidence from this systematic review indicates that EoE is a persistent disorder that continues from childhood to adulthood. More than a third of adult patients who were diagnosed with EoE as a child have lasting difficulties in swallowing and are in need of continuing healthcare. 39,40 In addition, fibrostenotic features, such as rings and strictures, which are considered to be hallmarks of progression of the disease, are more prominent in older patients, suggesting that they had the disease for a number of years. 36,41 Therefore, if fibrostenotic disease can be averted by timely treatment, delay in diagnosis may have long-lasting ramifications for patients.
This systematic review identified very limited data on the recurrence and remission rates in EoE. Furthermore, the studies that have been published in this area are very heterogeneous and difficult to interpret. More work is urgently needed in this area to guide the approach to maintenance therapy for patients with EoE.
Several strengths and limitations of this work deserve mention. Interpretation of both epidemiology and natural history data is challenged by the fact that some studies excluded patients with PPI-REE, whereas others did not specifically address this issue. Moreover, the need to account for changing disease definitions, confounders, and effect modifiers is a legitimate concern, and further analysis of epidemiologic incidence and prevalence data, particularly from population-based cohorts, is necessary. Confounders may include the availability of routine access to gastrointestinal endoscopy, as well as disease recognition, changing endoscopy practices and the work force of pediatric gastroenterologists in a given country or region. Several limitations are related to natural history data. For instance, studies concerning remission are difficult to interpret, because the definitions of remission and of the disease itself differed between studies, and very few confirmed whether resolution was histological as well as symptomatic, whether the recorded patients were undergoing treatment, and whether they were compliant with therapy. Moreover, all studies examining the effect of race and sex were undertaken in the USA, affecting the generalizability of results.
This study highlights a lack of epidemiological data for many parts of the world, including South America, Africa, and Asia. EoE is largely unstudied in these areas: only one non-English language study was identified in our searches, and inclusion of this abstract would not have impacted the conclusions presented in this systematic review. In the future, it will also be important to assess more thoroughly the incidence and prevalence estimates of EoE in specific patient subgroups to allow a better understanding of the influence of age, sex, and race on the disease. Moreover, while the prevalence and incidence of EoE appear to be increasing, the reasons behind this remain unclear. Further long-term data are needed to clarify whether the increasing incidence and prevalence of EoE result from an increasing recognition and awareness of symptoms, or instead reflect a true increase in this disease driven by some as yet unknown external factor.
Although considerable data are available on symptom presentation and disease progression of EoE, future work is needed to understand variations in symptom patterns in general and after treatment, as well as the duration and the progression of the disease. In particular, increased knowledge about whether symptoms become worse if untreated, and whether complications can be altered or prevented by maintenance therapy, is crucial for optimizing the care of patients with EoE.
In conclusion, this systematic review found that EoE is increasing in incidence and prevalence, which may be partly due to increasing recognition and awareness of symptoms. Delay in diagnosis appears to be associated with fibrostenotic disease manifestations, suggesting that timely recognition of the disease may impact its clinical course. A better understanding of the progressive nature of EoE, its symptom burden over time, and the impact of current therapies on symptom resolution are crucial in directing current clinical practice.

SUPPLEMENTARY DATA
Supplementary data are available at DOTESO online.