Esophageal Epithelium and Lamina Propria are Unevenly Involved in Eosinophilic Esophagitis

Abstract

Background and Aims:

The nature of the involvement of esophageal tissue in eosinophilic esophagitis (EoE) is unclear. We estimated the intra-biopsy site agreements of the histologic scoring system (EoEHSS) scores for the grade (degree) and stage (extent) of involvement of the esophageal epithelial and lamina propria and examined if the EoE activity status influenced the intra-biopsy site agreement.

Methods:

Demographic, clinical, and EoEHSS scores collected as part of the prospective OMEGA (Outcome Measures for Eosinophilic Gastrointestinal Diseases Across Ages) study were analyzed. Weighted Cohen’s kappa agreement coefficient (k) was used to calculate the pairwise agreements for proximal:distal, proximal:mid, and mid:distal esophageal biopsy sites, separately for grade and stage scores, for each of the eight components of EoEHSS. k of > 0.75 was considered uniform involvement. Inactive EoE was defined as <15 eosinophils per high-powered field.

Results:

EoEHSS scores from 1263 esophageal biopsies were analyzed. The k for the stage of involvement of the dilated intercellular spaces (DIS) across all three sites in inactive EoE was consistently > 0.75 (range: 0.87-0.99). The k for lamina propria fibrosis was > 0.75 across some of the biopsy sites but not across all three. Otherwise, the k for all other features, for both grade and stage, irrespective of the disease activity status, was ≤ 0.75 (range: 0.00-0.74).

Conclusion:

Except for the extent of involvement of DIS in inactive EoE, the remaining epithelial features and lamina propria are unevenly involved across biopsy sites in EoE, irrespective of the disease activity status. This study enhances our understanding of the effects of EoE on esophageal tissue pathology.

Graphical Abstract

Introduction

Eosinophilic esophagitis (EoE) is an increasingly prevalent, allergen-mediated inflammatory disorder affecting children and adults.¹² Histologically, EoE is characterized by an intense intraepithelial eosinophilic inflammation [defined as a peak eosinophil count (PEC) of ≥15 eosinophils per high power field (eos/HPF)] in at least one of multiple biopsies obtained from proximal, middle, or distal esophagus.^3,4 Additionally, histologic abnormalities can also be observed in the esophageal epithelium, and fibrotic transformation can occur in the subepithelial lamina propria (LPF). The Histology Scoring System (EoEHSS) can be used to determine the grade (degree) and stage (extent) of histologic abnormalities in the EoE-relevant epithelial features and lamina propria.⁵ This system has been validated in pediatric and adult EoE patients and has excellent inter- and intra-observer reliability.^6,7 It also performs better than PEC alone in assessing treatment response^5,8

EoE is regarded as a patchy disease because of an uneven eosinophilic infiltration in the esophageal epithelium across the biopsy sites.^9–13 However, it is unknown if the other EoE-relevant histologic features in the esophageal epithelium and lamina propria are evenly or unevenly involved. Defining the nature of the involvement of the esophageal epithelial tissue and the lamina propria in EoE can add to our understanding of its pathobiology, inform biopsy protocols and improve clinical practice.

To address this knowledge gap, we examined the agreement among histologic abnormalities in the EoE-relevant esophageal epithelial features and lamina propria across the esophageal biopsy sites, separately for the grade and stage of involvement, assessed per the EoEHSS. In addition, we also examined the impact of EoE activity status on the intra-site agreements. Based on our current understanding of the patchy distribution of the esophageal intraepithelial eosinophilia and an improvement of PEC in inactive EoE, we hypothesized that histologic abnormalities in the EoE-relevant esophageal epithelial features and lamina propria are unevenly distributed across the biopsy sites and that the EoE activity status influences the distribution of these histologic abnormalities.

Methods

Data Source

This study involved secondary analysis of the demographic, clinical, and EoEHSS data collected prospectively as part of the Outcome Measures for Eosinophilic Gastrointestinal Diseases across Ages (OMEGA) study between 2015-2019. The OMEGA study is an observational study conducted by the Consortium of Eosinophilic Gastrointestinal Disease Researchers (CEGIR). As a national collaborative network of academic centers caring for adults and children with eosinophilic gastrointestinal diseases (EGIDs), CEGIR is focused on understanding the natural history of EoE and other non-EoE eosinophilic gastrointestinal disorders (non-EoE EGIDs).^14–16

Subjects

Subjects provided consent to participate in the OMEGA study and for future use of their samples and data, per central Institutional Review Board (IRB) and local IRB requirements. The present study was approved by the IRBs at Cincinnati Children’s Hospital Medical Center (CCHMC) and Vanderbilt University Medical Center.

In the OMEGA study, patients three years and older met the established EoE criteria of having esophageal symptoms, and a PEC of ≥ 15 eos/HPF in at least one of their esophageal biopsies was included.^3,4 Exclusion criteria included EoE patients with a history of gastrointestinal surgery other than gastric tube placement, planned or recent enrollment in blinded investigational studies, esophageal stricture < 3 mm, and other identifiable potential causes for esophageal eosinophilia. Full details of the study have been presented elsewhere.^17,18

Collecting and processing esophageal biopsies

Esophageal biopsies were acquired from the proximal, mid, and distal esophagus at the discretion of the treating gastroenterologist during a clinically indicated esophagogastroduodenoscopy (EGD). All biopsies were processed at the local participating study sites. The hematoxylin and eosin (H&E) stained slides of these biopsies were then scanned using the 2-dimensional Aperio Digital Pathology Slide Scanner (Leica Biosystems, Illinois) and stored on a secure server at CCHMC. The scanned slides were accessed and assessed per EoEHSS by the CEGIR Central Review Pathology committee comprised of 3 pathologists (M.H.C, N.C.A, and G-Y.Y) with expertise in EoE and eosinophilic gastrointestinal diseases.

The EoEHSS was used to assess eosinophilic inflammation (EI), basal zone hyperplasia (BZH), eosinophilic abscess (EA), eosinophilic surface layering (SL), dilated intercellular spaces (DIS), surface epithelial alteration (SEA), dyskeratotic epithelial cells (DEC), and LPF when adequate lamina propria was available for evaluation (Supplementary Figure 1). Each feature was scored on a 4-point scale (0 to 3) for severity (grade) or extent (stage) of the abnormality, with 0 representing normal features and 3 denoting most severe or extensive pathology.⁵

Statistical analysis

We used descriptive statistics to summarize our cohort, including the counts and percentages for categorical variables and medians and quartiles for continuous variables. The pairwise intra-site agreements [proximal:distal (P:D), proximal:mid (P:M), and mid:distal (M:D)] for each of the eight histologic features were calculated using the Cohen’s weighted kappa coefficient (k).¹⁹ The k was estimated for grade score and stage score separately. Since a k of > 0.75 indicates of an excellent agreement beyond chance, we used it as an a priori threshold to define even (or uniform) involvement. Conversely, a k of ≤ 0.75 was regarded as sub-optimal agreement and suggestive of uneven (or variable) involvement.^20,21 Sub-group analysis was performed to examine the impact of EoE activity status (active vs. inactive EoE, defined per the accepted threshold of 15 eos/HPF) on the intra-site agreements.

As this study aimed to assess the agreement of EoE-relevant histologic abnormalities in the esophageal epithelium and the lamina propria across the biopsy sites, we used non-paired 0 (or normal) scores in our analyses and excluded paired 0 (or normal) scores. For example, if the grade score for BZH for P:D was 0:3 (or normal:abnormal) then it was included in the analysis, and only the instances when the EoEHSS score was 0:0 (or normal:normal) it was excluded. This approach also allowed us to eliminate the possibility of a falsely high k which can occur while assessing perfectly matched pairs [such as 0:0 (or normal:normal)].

Since some subjects had undergone EGD with biopsies on 2 or more occasions, we initially assessed the intra-site agreement for the EoEHSS scores of their first biopsies recorded in the database (initial biopsy analysis). Using EoEHSS scores of their first biopsies allowed us to eliminate any potential impact of intra-site correlation within an individual and mitigate the effects of non-uniform data. Next, to use information from all available biopsies, we explored the esophageal intra-site agreement irrespective of the number of times the biopsies were obtained from an individual (sensitivity analysis).

All analyses were performed in R Statistical Software (version 4.0, R Foundation for Statistical Computing, Vienna, Austria).

Results

Cohort characteristics

In all, 397 EoE patients from 13 academic centers were included in this secondary analysis. The average age at the time of their EGD resulting in the first set of biopsies scored by the CEGIR pathologists was 24 ± 17 years. Participants were predominantly male (67%) and self-identified as White (93%). A higher proportion had active EoE (58%) (Table 1). They had undergone between 1-6 EGDs with biopsies, with 272 having had one, 88 having had two, and 37 having had three or more EGDs with biopsies.

Table 1:

Characteristics of individuals included in the secondary analysis

		Number of subjects (N = 397)
Age (years)#
		24 ± 17
Male*		267 (67)
Race*	White	370 (93)
	Non-white	26 (7)
Disease status*	Active EoE	229 (58)
	Inactive EoE	168 (42)
Participants*
	Cincinnati Children’s Hospital	70 (18)
	Children’s Hospital Colorado	61 (15)
	Children’s Hospital of Philadelphia	2 (1)
	Laurie Children’s Hospital of Chicago	11 (3)
	Northwestern University	34 (9)
	Rady Children’s Hospital	36 (9)
	Riley Children’s Hospital	3 (1)
	Tuft’s Medical Center	34 (6)
	University of California San Diego	8 (2)
	University of Colorado Denver	23 (6)
	University of Illinois at Peoria	11 (3)
	University of North Carolina	68 (17)
	University of Pennsylvania	46 (12)

EoE: Eosinophilic esophagitis; CEGIR: Consortium of Eosinophilic Gastrointestinal Disease Researchers

^#Mean ± Standard deviation,

^*n (%)

Biopsies analyzed, and the distribution of the histologic features

The grade and stage scores of the EoE-relevant histologic abnormalities assessed per the EoEHSS from a total of 1263 biopsies. Of these, 534 were from the proximal, 153 from the mid, and 576 were from the distal esophagus. Over 75% of biopsies were scored as 0 (or normal) for SL, SEA, and DEC, suggesting that these features were relatively infrequently affected compared to EI, BZH, and DIS which were more commonly affected. The LPF could not be assessed for the grade of involvement for 58% of proximal, 87% of mid, and 50% of distal biopsies. Likewise, the stage of involvement of LPF could not be assessed for 58% of proximal, 87% of mid, and 50% of distal esophageal biopsies. The LPF could not be assessed in these biopsies due to inadequate sampling or crush artifacts rendering the specimens uninterpretable (Figure 1).

Figure 1:

Number of biopsies analyzed and the distribution of EoEHSS scores (a) grade, and (b) stage. [PE: Proximal esophagus, ME: Middle esophagus, DE: Distal esophagus, EI: Eosinophilic inflammation, BZH: Basal zone hyperplasia, EA: Eosinophilic abscess, SL: Surface layering, DIS: Dilated intercellular spaces, SEA: Surface epithelial alteration DEC: Dyskeratotic epithelial cells, LPF: Lamina propria fibrosis]

Initial biopsy Analysis

Intra-site agreement

For grade, a total of 365 P:D (with LPF=117), 62 P:M (with LPF=23), and 89 M:D (with LPF=34) pairs, and for stage 363 P:D (with LPF=117), 62 P:M (with LPF=23), and 87 M:D (with LPF=35) pairs were analyzed. The k for grade scores was below our a priori threshold for each of the features except for M:D pair of LPF (k = 0.81). For the stage score, the k exceeded our a priori threshold for DIS (range: 0.77 – 0.79) and LPF (M:D, k = 0.78) and not for any of the other histologic features (Figure 2).

Figure 2:

Pairwise analysis of the EoEHSS scores of the initial biopsies. Weighted kappa coefficient for (a) grade, and (b) stage. [P:D: Proximal:Distal, P:M: Proximal:Middle, M:D: Middle:Distal, EI: Eosinophilic inflammation, BZH: Basal zone hyperplasia, EA: Eosinophilic abscess, SL: Surface layering, DIS: Dilated intercellular spaces, SEA: Surface epithelial alteration, DEC: Dyskeratotic epithelial cells

LPF: Lamina propria fibrosis] [Weighted kappa coefficient (k): optimal agreement (> 0.75 or even), suboptimal agreement (⩽ 0.75 or uneven)]

Impact of EoE activity status

For the grade of involvement, the k for each of the features in active (total number of pairs, P:D= 211, P:M= 40, and M:D = 54) and inactive EoE (total number of pairs, P:D= 154, P:M= 22, and M:D = 35) was below the threshold, except for one pair of LPF (k for M:D is 0.76-0.80) and DIS (k for P:M inactive EoE = 0.93) each. Due to insufficient pairs, some of the k could not be calculated for EA and DEC in inactive EoE. However, for the stage of involvement, in inactive EoE (total number of pairs, P:D= 153, P:M= 22, and M:D = 35), the k was consistently above the threshold only for DIS (range: 0.87 – 0.99). These agreements were irrespective of EoE activity status (Figures 3).

Figure 3:

Pairwise analysis of the EoEHSS scores of the initial biopsies to assess the impact of EoE activity on the intra-site agreement. Weighted kappa coefficient for (a) grade, and (b) stage. [P:D: Proximal:Distal, P:M: Proximal:Middle, M:D: Middle:Distal, EI: Eosinophilic inflammation, BZH: Basal zone hyperplasia, EA: Eosinophilic abscess, SL: Surface layering, DIS: Dilated intercellular spaces, SEA: Surface epithelial alteration, DEC: Dyskeratotic epithelial cells, LPF: Lamina propria fibrosis] [Weighted kappa coefficient (k): optimal agreement (> 0.75 or even), suboptimal agreement (⩽ 0.75 or uneven)]

Sensitivity analysis

For the grade of involvement, 532 proximal-distal (with LPF: 167), 109 proximal-mid (with LPF: 35), and 151 mid-distal (with LPF: 49) pairs were analyzed. For the stage of involvement, 528 proximal-distal (with LPF: 167), 108 proximal-mid (with LPF: 34), and 149 mid-distal pairs (with LPF: 51) were analyzed.

The k for the grade score of each of the epithelial features and LPF was below our set threshold, and for the stage score, only instances of k for DIS (k for P:D = 0.80 and k for M:D = 0.77) were over our threshold, and that for the remaining features were below our threshold. Overall, the agreement was comparable between the grade and stage scores (Supplementary Figure 2).

Impact of disease activity

For the grade scores, the k for DIS (P:M = 0.86) and LPF (M:D = 0.81) in inactive EoE, and the k for the remaining epithelial was below our threshold of 0.75 in both active and inactive EoE. The k for the grade of LPF was over the a priori threshold only for M:D (k = 0.81) in inactive EoE. Due to inadequate pairs, the k could not be assessed for EA and some of the DEC pairs. Regarding the stage scores, the k for the involvement of DIS was consistently above our threshold (range: 0.88 – 0.95) in inactive EoE across all sites. The k for the stage of LPF was over the a priori threshold only for P:M (k = 0.88) and M:D (k = 1.00) in inactive EoE. Otherwise, the k for the stage of involvement of the remaining features was below the threshold for both active and inactive EoE (Supplementary Figure 3).

Discussion

We investigated whether a broad range of esophageal histologic abnormalities on biopsies as assessed per the EoEHSS are evenly or unevenly involved in EoE and explored the impact of EoE activity on the nature of their involvement. Our results suggest that almost all EoE-relevant esophageal epithelial features are unevenly involved for grade (degree) and stage (extent) of involvement, irrespective of disease activity status. The only exception was DIS stage (extent), which was uniformly involved in inactive EoE. Regarding lamina propria, even though the scores for LPF did not consistently meet our a priori cut-off across all three esophageal biopsy sites, there were indications that the grade (degree) and stage (extent) of its involvement, maybe even across one or more esophageal biopsy sites in both active and inactive EoE. Our results provide a deeper understanding of the impact of EoE on the esophageal epithelium and lamina propria and extend our prior knowledge related to variability in intraepithelial eosinophilia. In addition, these observations can inform clinical practice, biopsy protocols, and pathologic assessment. In addition, the findings are consistent with a recent analysis of OMEGA samples focused on transcriptomic profiling, which demonstrated that gene expression pattern is consistent throughout the esophagus.²²

Of the epithelial features, EI, BZH, and DIS were more commonly affected when compared to EA, SL, SEA, and DEC, which were mostly unaffected (or scored 0). Esophageal intraepithelial eosinophilia is an essential histologic feature of EoE. The PEC is traditionally used to diagnose EoE and to classify EoE activity status.²³ BZH is a hyperplastic process related to ongoing inflammation and is associated with persistent symptoms in patients with inactive EoE and has the potential to contribute towards the development of LPF^24,25 In our secondary analysis, we observed a sub-optimal agreement for EI and BZH, for both grade and stage scores, across the biopsy sites. These findings suggest that EI and BZH abnormalities at one biopsy site cannot be a proxy for their involvement at other sites. Furthermore, our observation regarding EI concurs with the previous reports indicating that the eosinophils are unevenly distributed across the esophageal epithelium in EoE. Like EI and BZH, DIS is also a prominent histological feature in EoE²⁶ It is associated with increased esophageal epithelial permeability²⁷ and can be reversed with successful therapy.²⁸ Interestingly, only the stage (extent) of involvement of DIS had excellent intra-site agreements in inactive EoE. Therefore, DIS may serve as a marker of treatment response in EoE, given its uniform distribution after successful treatment. It may also provide additional rationale for measuring mucosal impedance to differentiate treatment responders from non-responders in EoE.^29–31 More than half of the esophageal samples acquired using conventional biopsy forceps can have inadequate lamina propria^32,33, and our data corroborate this. When present, the LPF appeared to be inconsistently involved, in both grade and stage scores across the biopsy sites in active and inactive EoE.

Our study has limitations. Though these data were collected prospectively at multiple clinical sites, we did not have control over the esophageal biopsy collection process. Furthermore, variability in the number of biopsy specimens per location and size of biopsy forceps may have contributed to the variation in the detection of features. Next, we had fewer mid-esophageal samples in the analyses than proximal and distal esophageal samples. While this might represent the prevailing practice, it resulted in a smaller sample size for our pairwise comparisons involving mid-esophageal sites and limited our understanding of the degree and extent of mid-esophageal involvement in EoE. We could not examine the effect of interaction between disease activity status and patient factors such as age, sex, symptoms [such as gastroesophageal reflux (GERD)] and treatment approaches on the nature of involvement of esophageal tissue in EoE. Despite these limitations, this study has several strengths. We analyzed a large multi-institutional dataset with prospectively collected biopsies from children and adults with EoE. The histologic alterations in the esophageal biopsies were assessed using the validated EoEHSS by pathologists with expertise in EoE and non-EoE EGIDs. We used weighted kappa scores, which allowed for the disagreements in EoEHSS scores to be weighted differently and eliminated the possibility of estimating falsely high agreement coefficients. Next, we used a stringent criterion of agreement (> 0.75) to define the uniform involvement of histologic features. Finally, we analyzed grade and stage scores separately, allowing us to gain novel insights into the severity and extent of esophageal tissue pathology in EoE.

In summary, we found that EoE-relevant esophageal epithelial features, except for DIS, are unevenly involved in EoE, irrespective of the disease activity status. In addition, although the LPF appears to be largely unevenly involved across all three biopsy sites, it can be evenly involved across some sites in both active and inactive EoE. These observations reinforces the current recommendations to obtain multiple esophageal biopsies from at least two levels in the esophagus. Future prospective studies are warranted to understand the impact of the interaction between the nature of involvement of EoE-relevant esophageal epithelial and subepithelial tissues and the patient factors, clinical presentation, disease activity status and the clinical course of EoE.

Supplementary Material

1

Supplementary Figure 1: These photos are from images in the CEGIR OMEGA archive. (a) In this area numerous intraepithelial eosinophils are present (arrow), the basal zone is expanded (bar), intercellular spaces are dilated (arrowhead), and the lamina propria, which does not appear crushed, shows mildly thickened fibers (asterisk) and exhibits eosinophils (arrows); (b) In the area at the upper right of this image, eosinophils are aligned at the surface forming surface layering (arrows). A dyskeratotic epithelial cell is seen (shaded arrow) in the epithelium; (c) In another area of the biopsy in Figure b, surface epithelial cells stain more pink than the rest of the epithelium (arrows) forming surface epithelial alteration; (d) An eosinophil abscess is seen at the arrow.

Supplementary Figure 2: Sensitivity analysis of the EoEHSS scores involving all biopsies. Weighted kappa coefficient for (a) grade, and (b) stage. [P:D: Proximal:Distal, P:M: Proximal:Middle, M:D: Middle:Distal, EI: Eosinophilic inflammation, BZH: Basal zone hyperplasia, EA: Eosinophilic abscess, SL: Surface layering, DIS: Dilated intercellular spaces, SEA: Surface epithelial alteration, DEC: Dyskeratotic epithelial cells, LPF: Lamina propria fibrosis] [Weighted kappa coefficient (k): optimal agreement (> 0.75 or even), suboptimal agreement (⩽ 0.75 or uneven)]

Supplementary Figure 3: Sensitivity analysis of the impact of EoE activity on the pairwise agreement of EoEHSS scores of all biopsies. Weighted kappa coefficient for (a) grade, and (b) stage. [P:D: Proximal:Distal, P:M: Proximal:Middle, M:D: Middle:Distal, EI: Eosinophilic inflammation, BZH: Basal zone hyperplasia, EA: Eosinophilic abscess, SL: Surface layering, DIS: Dilated intercellular spaces, SEA: Surface epithelial alteration, DEC: Dyskeratotic epithelial cells, LPF: Lamina propria fibrosis] [Weighted kappa coefficient (k): optimal agreement (> 0.75 or even), suboptimal agreement (⩽ 0.75 or uneven)]

STUDY HIGHLIGHTS.

WHAT IS KNOWN

• Eosinophilic esophagitis (EoE) is characterized by patchy eosinophilic infiltration in the esophagus

• Biopsies from the proximal, mid, and distal esophagus are examined for histologic abnormalities to diagnose and monitor EoE

• The grade (degree) and stage (extent) of histopathologic abnormalities in the esophageal epithelium and lamina propria can be assessed by the Histology Scoring System (EoEHSS).

WHAT IS NEW HERE

• Dilated intercellular spaces are consistently and unevenly involved across the esophageal biopsy sites in inactive EoE.

• Lamina propria fibrosis can be inconsistently affected across biopsy sites.

• The EoEHSS scores reflecting the grade and the stage of involvement of esophageal epithelium and lamina propria are inconsistently and unevenly distributed across the biopsy sites in EoE.

Background:

Eosinophilic esophagitis (EoE) is a patchy disease based on the even distribution of intraepithelial eosinophilia in the esophagus.

We examined the agreement for the involvement of esophageal epithelial features and lamina propria across the esophageal biopsy sites in EoE and investigated if the disease activity status influenced the agreements.

Findings:

The severity (or grade) and the extent (or stage) of involvement of the esophageal epithelial features and lamina propria are unevenly distributed across esophageal biopsy sites in EoE.

Only the extent (or stage) of involvement of dilated intercellular spaces was evenly involved in inactive EoE.

Implications for patient care:

Our findings reinforce the current recommendations to obtain multiple esophageal biopsies from at least two levels in the esophagus.