Systematic Review With Meta-analysis

The Growing Incidence and Prevalence of Eosinophilic Oesophagitis in Children and Adults in Population-based Studies

Pilar Navarro; Ángel Arias; Laura Arias-González; Emilio J. Laserna-Mendieta; Miriam Ruiz-Ponce; Alfredo J. Lucendo


Aliment Pharmacol Ther. 2019;49(9):1116-1125. 

In This Article


Literature Search

The search strategy yielded 2386 references; 2324 were excluded from the search mainly due to (a) no measure of prevalence or incidence of EoE being given, (b) not being population-based studies, (c) being review articles or (d) the focus not being on EoE. In all, we identified 29 studies that reported on the population-based incidence and prevalence of EoE. Figure 1 summarises the results of the search strategy. Most of the studies of the prevalence of EoE were conducted in United States (US),[8,9,19–29] Canada[30,31] and Europe,[32–41,7,42] but there were also studies from Western Australia,[43] and South America.[44] Key differences in prevalence rates depended on whether the study population included only children, only adults or individuals of all ages, as well as on the time the study was undertaken and the definition of prevalence such as a point (16 studies[21–32,36,37,42] or a period (11 studies);[9,19,33–35,38,40,41,7,43,44] and methodology used, such as hospital-based case series (14 studies),[8,19,21,32–35,38–41,7,43] administrative database (7 studies),[9,20,30,31,36,37,42] or insurance database (8 studies)[22–29] (Table S1). Further variation was related to the definition of EoE considered by the various authors, which differed significantly over the period with regard to the role of pH-monitoring and consideration of PPIs in the diagnostic and therapeutic algorithm of the disease.[1,2,18]

Figure 1.

Flow chart for the process of identifying studies included in and excluded from the systematic review

Overall prevalence rates and changes according to regional distribution and diagnostic criteria for EoE

The overall prevalence of EoE in the 24 retrieved studies was 34.2 cases per 100 000 inhabitants (95% CI, 23.1-47.5; I 2 = 99.9%; Figure 2A). Differences in the overall prevalence rates were also documented according to study region, being higher for North America (41; 95% CI, 25.7-59.9; I 2 = 99.7%) than for Europe (29; 95% CI, 19.9-39.8; I 2 = 99.6%), although these differences were not statistically significant (P = 0.571).

Figure 2.

Summary estimates for population-based prevalence of EoE, including overall (A) and subgroup analysis of studies conducted in children (B) or adults (C). Summary estimates are expressed as the number of EoE patients/100 000 inhabitants. An I 2 value (statistical heterogeneity) over 75% indicates a high variability in intra-study differences in the overall effect size

Subgroup analysis according to risk of bias of source documents (Table S2) did not provide significant changes in the prevalence of EoE (35.5; 95% CI, 25.7-46.8; I 2 = 99.8% vs 32; 95% CI, 16.6-52.4; I 2 = 100%, for studies with low and high risk of bias, respectively; P = 0.710).

When the studies were classified according to the criteria used by their authors to define EoE and its variations over time, the overall prevalence reported for the disease increased progressively, with a fourfold increase from the oldest studies (that considered criteria before the 2007 consensus[18]) and those that used the most up-to-date evidence-based diagnostic criteria provided from 2017 onwards[1,6] (Figure 3). This change in prevalence reached a statistically significant difference (15.4; 95% CI, 10.4-21.2 vs 63.2; 95% CI, 34.6-100.3, respectively; P = 0.011) (Table 1).

Figure 3.

Population-based studies that have assessed the prevalence of EoE over time. Graphic representation of the prevalence rates (value per 100 000 inhabitants with 95% CI) of each individual study (identified by first author's name), distributed (A) throughout the years of publication and (B) according to the diagnostic criteria for EoE used in each study. Boxes and whisker plots in B represent summary of prevalence with 95% CIs after meta-analysis of individual studies

Prevalence of EoE in Children

Fourteen studies reported the prevalence of EoE in children (defined as those aged < 16 years) (Figure 2B). In general, the overall prevalence of EoE in children was up to 34.4 cases/100 000 inhabitants (95% CI, 22.3-49.2; I2 = 99.7%), with no significant differences between the US and Europe (38.3; 95% CI 23.7-56.4 and 41; 95% CI, 3.2-121.1, respectively). Studies based on insurance and administrative databases, confined to those aged under 16 years, reported the same prevalence provided by hospital-based case series (Table 1). According the three most recently published studies[40,7,44] following the most up-to-date diagnostic criteria for EoE, the current prevalence for EoE in children is 53.4 cases/100 000 inhabitants (95% CI, 27.1-88.5; I 2 = 95.6%). There was a nonsignificant trend towards a higher prevalence of EoE among the studies with a lower risk of bias (42.7; 95% CI, 14.1-86.8 vs 28.2; 95% CI, 16.4-43.3)

Prevalence of EoE in Adults

The nine studies focused on prevalence of EoE in adults (Figure 2C) yielded higher estimates than studies focused on children, with an overall prevalence of 42.2 (95% CI, 31.1-55; I2 = 99.9%). However, there appeared to be little consistency between countries, with European-based studies providing significantly higher prevalence figures for EoE than American ones (95.8; 95% CI 68.4-127.8 vs 31.9; 95% CI, 21.5-44.3; P = 0.006). In Europe, Spain had the highest consistent estimates reported, with two recent studies on different populations carried out with hospital-based, prospectively maintained databases.[41,7] In contrast, prevalence estimates for EoE in adults in the US and Canada were obtained from insurance and administrative databases, some of them specifically excluding patients with codes related to GORD,[8,27,29,37] which could have resulted in an underestimation of the true magnitude of prevalence (Table 1). No population-based epidemiological study on adults has been published by the US since the release of the AGREE conference paper[6] that supports the elimination of PPI in the diagnostic algorithm of EoE.[1,45] Again source studies with lower risk of bias tended to provide higher prevalence rates than those with some methodological weakness (64.2; 95% CI, 14.9-148.2 vs 34.2; 95% CI, 21.4-49.9; P = 0.167).

Overall Incidence of EoE and Changes According to Regional Distribution and Diagnostic Criteria

Eighteen studies examined the incidence of EoE in the general population. These studies were conducted in North America (US,[20,22–24] Canada[30,31]) and Europe (The Netherlands,[36,42] Denmark,[37] Switzerland[38] and Spain[7]) and looked at different groups of the population (children or adults or all ages) and different time periods. The studies were pooled to give an overall incidence rate estimate of 4.4 (95% CI, 2.8-6.4) new cases of EoE per 100 000 inhabitants/year at risk of EoE based on a random-effects model (I 2 = 99.9%; Table 2; Figure 4A). No significant differences were noted for pooled incidence rates when studies were grouped by geographical origin, despite incidence figures being higher for those conducted in North American compared to European countries (7; 95% CI, 2.6-13.6 vs 2.7; 95% CI, 2-3.6, P = 0.108). No significant differences in incidence of EoE based on the origin of source data (i.e., hospital-based case series regarding insurance and administrative databases) were noted, although summaries for incidence tended to be slightly higher in the former (5.5; 95% CI, 2.2-10.3 vs 3.3; 95% CI, 1.4-5.9, respectively).

Figure 4.

Summary estimates for population-based incidence of EoE overall (A), as well as subgroup analysis of studies conducted in children (B) or adults (C). Summary estimates are expressed as the number of EoE patients/100 000 persons-year. An I 2 value (statistical heterogeneity) over 75% indicates a high heterogeneity

Subgroup analyses by grouping source studies according to risk of bias, did not show significant differences in overall incidence rate of EoE, being 4.3 (95% CI, 2.9-5.9) and 4.7 (95% CI, 1.4-9.8) new cases per 100 000 persons-year for studies with low and high risk of bias, respectively.

Subgroup analysis according to diagnostic criteria for EoE demonstrated changes through time in incidence rates when comparing studies carried out before 2007 (2.6; 95% CI, 1.6-3.9) and after 2017, with current pooled incidence rates for EoE being 6.2/100 000 inhabitants/year; 95% CI, 3.5-9.5; P = 0.059).

Incidence of EoE in Children

There were 10 studies of the incidence of EoE in children; these were conducted in the US,[8,9,19,21,39] Denmark,[32] Spain[40,7] and Brazil[44] over a 14-year period, with data from 1991 to the end of 2017 (Table S1). It was found that the incidence of EoE in children for the whole period was 6.6/100 000 persons-year; 95% CI 3-11.7; I 2 = 99.8%), but 5.6; 95% CI, 2.6-9.8; I 2 = 96.6% when only the most recent studies were considered, thus supporting stable incidence rates for EoE over time (Figure 4B). Subgroup analysis showed no significant differences in incidence rates according to risk of bias or origin of data (Table 2).

Incidence of EoE in Adults

Five studies reported the incidence of EoE in adults (Table S1)—one from the US[29] and four from Europe[33–35,7] with collective data for the period 1989 to 2017. Overall, the summary estimates for the whole period were 7.7/100 000 person-year (95% CI, 1.8-17.8), with no significant differences between US and European figures (Figure 4C). According to the most recent studies using current diagnostic criteria for EoE,[41,7] the summary incidence for the period 2017 and beyond was 8.5 new cases per 100 000 persons-year (95% CI 7.5-9.6).[41,7] Subgroup analyses according to risk of bias and source of data (administrative or insurance database vs hospital-based cases series) showed no significant differences in incidence rates (Table 2).

Prevalence Rate Ratio for Genderss

The prevalence rate ratio by gender was reported in nine studies.[20,23–25,27–29,41,7] Although a significant heterogeneity in the results was observed, the pooled prevalence of EoE among male patients was 72.1 (95% CI, 41.3-111.5; I 2 = 99.9%) patients per 100 000 inhabitants, while in females it was 29.4 (14.8-48.8; I 2 = 99.9%). Males were thus at greater risk for presenting EoE compared to females, with an OR of 2.22 (95% CI: 2-2.46; Figure S1).

Supplementary Figure 1.

Summary estimates for population-based prevalence of EoE in male and female patients, expresses as Odds ratio (OR). Summary estimates are expressed as the number of EoE patients/100,000 inhabitants. An I2 value (statistical heterogeneity) of 94.1% indicates a high variability.

Publication Bias Assessment

Funnel plot analyses of studies assessing the prevalence of EoE revealed no significant publication bias (Begg-Mazumda's rank test P = 0.676; Harbord test P = 0.980). Likewise, studies reporting on the incidence of EoE exhibited no significant publication bias (Begg-Mazumda's rank test P = 0.549; Harbord test P = 0.091). Funnel plots assessing size effect against study precision are shown in Figure S2.

Supplementary Figure 2.

Begg funnel plot of population-based studies assessing the prevalence and incidence of eosinophilic esophagitis. The solid line in the centre is the natural logarithm of pooled prevalence (A) or incidence (B) rates remission rates, and the two oblique lines are pseudo 95% confidence limits.