Prospective Endoscopic Activity Assessment for Eosinophilic Gastritis in a Multisite Cohort.

This study utilized data that was prospectively collected as part of the Consortium for Eosinophilic Gastrointestinal Disease Researchers (CEGIR), a national collaborative network of 16 academic centers caring for adults and children with eosinophilic gastrointestinal disorders supported by a U54 grant ( AI117804 ) as part of the Rare Disease Clinical Research Network (RDCRN), an initiative of the Office of Rare Diseases Research (ORDR), and funded through collaboration among NCATS, NIAID and NIDDK. CEGIR is also supported by the Division of Intramural Research (NIAID) and patient advocacy groups including the American Partnership for Eosinophilic Disorders (APFED), Campaign Urging Research in Eosinophilic Disease (CURED) and the Eosinophilic Family Coalition (EFC). 9 , 10 The CEGIR clinical study, Outcomes Measures in Eosinophilic Gastrointestinal disorders Across the ages (OMEGA), is a longitudinal cohort study aimed at understanding the natural history of EoE, EG, and EC during routine clinical care. 11 Patient reported demographics, clinical, endoscopic and histologic data were prospectively collected starting in 2015. 9 , 10 , 12 Clinical features of subjects were recorded during standard of care evaluation with intake and follow up forms. This study was approved by the institutional review boards of the participating institutions via a central institutional review board at Cincinnati Children’s Hospital Medical Center. Patient consent was obtained by the participating medical center in accordance with the approved study protocol. All subject data collected for this study were stored at the Data Management and Coordinating Center (DMCC) at the University of South Florida in Tampa, FL. Atopy was defined based on self-report of allergic rhinitis, dermatitis, asthma and/or food allergy. For the OMEGA study, EG was defined by the presence of upper gastrointestinal symptoms combined with the histologic finding of ≥30 eos/hpf in 5 high power fields in any part of the gastric mucosa with exclusion of secondary causes of gastric eosinophilia. 7 , 13 During the course of standard of care endoscopic examinations, endoscopic features in patients with EG were prospectively recorded in real-time using a classification and grading system specifically developed for EG. The system was developed through collaborative input from both pediatric and adult gastroenterologists with expertise in EGID as part of an annual CEGIR meeting held in 2015. A comprehensive list of endoscopic features was assembled, and severity grading was proposed through an iterative process achieved by correspondence with a working group to refine the instrument. The final instrument termed the Eosinophilic Gastritis endoscopic Reference System (EG-REFS), modeled on the EoE endoscopic scoring metric EREFS, 14 included features of erosion/ulceration, granularity, raised lesions, erythema, friability, fold thickness and pyloric stenosis ( Table 1 , Figure 1 ). EG-REFS scores were separately assessed in the gastric fundus, body and antrum. A composite EG-REFS score was calculated as the sum of the EG-REFS scores for each feature from the 3 locations. Physician overall global assessment of endoscopic severity was scored on a 5 point Likert scale (0, normal to 5, most severe). Biopsies were obtained using standard of care protocols at each institution. Use of a systematic location and number of biopsies was not required for the diagnosis of EG. Whole slide images of gastric biopsies (400X magnification) obtained within ±30 days of PRO completion were reviewed by pathologists comprising the CEGIR Pathology Core (MHC, KEC, GYY). Pathologists were blinded to treatment status and therapy at the time the biopsies were procured. Peak and mean eosinophil densities were based on review of 5 high power fields that were selected with the greatest inflammation. Symptoms were prospectively assessed using the Severity Of Dyspepsia Assessment (SODA) instrument. Symptom data were included only for the subset of patients with completed questionnaires. For the patients with EG-REFS, only SODA data completed within 30 days of the endoscopic exam were included. Symptoms of abdominal pain, nausea, heartburn and dyspepsia were reported. Demographic and clinical characteristics were summarized using frequency and percent for categorical variables and median (interquartile range) for the continuous variables. For analyses that investigate the association between histology and endoscopy data, visits were identified where histology and endoscopy data were both available. If there was more than one visit meeting this criterion, the visit with the most severe inflammatory features based on histologic evaluation was selected for the purpose of evaluating EG-REFS scores. For analyses focused on endoscopy data only, the endoscopy visit date with the highest total score in the fundus, body and antrum combined was selected if there was more than one visit for the participant. Spearman nonparametric correlations (Spearman r) were used to assess the relationship between the following pairs of measurements: (1) eosinophil density and EG-REFS scores based on region of stomach; (2) physician global assessment of endoscopic severity and EG-REFS scores based on region of the stomach; (3) physician global assessment of endoscopic severity and individual features of EG-REFS based on regions of the stomach; (4) duration of disease at time of endoscopy and EG-REFS scores. The Wilcoxon Rank Sum test was used to compare EG-REFS scores between patients with isolated EG versus those with EG combined with esophageal and/or colonic involvement. The Wilcoxon Signed Rank test for paired data with the Hochberg-Benjamini multiple testing adjustment was used to test for differences in scores among the three regions of the stomach. P- values<0.05 were considered statistically significant. Statistical analyses were performed using SAS software (version 9.4; SAS, Cary, NC).

There were ninety-eight patients with EG enrolled in the CEGIR OMEGA study at the time of this study ( Table 2 ). These patients were derived from 9 CEGIR adult and pediatric sites. Among the overall group of 98 patients, 58 patients (59%) had gastric involvement in addition to involvement of other regions of the gastrointestinal tract (esophagus and/or colon). Fifty one patients (53%) had concurrent EG and esophageal involvement (>15 eosinophils/hpf). The median age of 17 years reflected a predominantly pediatric subgroup, with 59% under the age of 18 years at the time of endoscopy. Similar to the demographic characteristics found in the general EoE population, the EG patients were predominantly Caucasian (85%). Approximately half were male (53%). Atopy was present in 55%. No patient had evidence of concomitant Helicobacter pylori on immunohistochemical staining of gastric biopsies. The majority of patients were on active medical or diet therapy at the time of evaluation. Available symptom data using the SODA instrument demonstrated that 63% had abdominal pain while approximately half the patients had nausea, heartburn or bloating. Sixty-five patients underwent real-time, prospective evaluation of endoscopic features using EG-REFS ( Tables 2 and 3 ). Demographic and disease characteristics were similar between the overall EG cohort of 98 patients and the 65 patients who had EG-REFS. Notably, 82% were on active therapy for EG at the time of the endoscopic assessment. Endoscopic abnormalities (EG-REFS score > 0) were identified in 53 of the 65 patients (82%). The most common abnormalities included erythema (72%), raised lesions (49%), erosions (46%) and granularity (35%) ( Figure 2 ). Thickened gastric folds and pyloric stenosis were the least prevalent features, identified in less than 17% and 2%, respectively. EG-REFS scores spanned from 0 to 24 out of a maximal score of 46 ( Figure 2 ). The median composite EG-REFS score for the cohort was 4 (IQR 1,7). The severity of the composite EG-REFS scores strongly correlated with the endoscopic physician global assessment ( Figure 3 , Spearman r = 0.84, P< 0.0001). No significant differences were found in the EG-REFS scores when comparing endoscopic activity in patients with isolated EG (5.0; IQR 1.5, 6.5) compared to those with a combination of EG and esophageal or colonic eosinophilic involvement (3.0; IQR 1.0, 7.5) ( P= 0.65). Pairwise comparisons demonstrated significant differences in EG-REFS scores between the antrum, body and fundus, with the greatest severity in the antrum ( P< 0.001 for all pairwise comparisons) ( Figure 4 , Table 3 ). Erosions or ulcerations were identified in the antrum in 42% but only 16% in the body and 3% in the fundus. Raised lesions or nodules were present in 42% in the antrum, 28% in the body and 6% in the fundus. Similar gradients of endoscopic activity from the antrum to body to fundus were noted for granularity, erythema and friability. Comparing the individual features of EG-REFS with physician global assessment of endoscopic severity demonstrated the strongest correlations for raised lesions (nodularity), erosion/ulceration, erythema, friability and granularity, particularly in gastric antrum ( Table 4 ). The weakest correlations were noted for thickened folds and pyloric stenosis. Baseline demographic and clinical characteristics were compared with endoscopic severity. Age at baseline showed a weak correlation with the composite EG-REFS scores (r=0.26, P= 0.04). Gender, atopy, and presence of concomitant eosinophilic GI disease outside of the stomach (i.e. esophageal or colonic involvement) did not influence the endoscopic activity. Duration of disease defined by date since histologic diagnosis of EG to date of endoscopy for CEGIR was moderately associated with the composite EG-REFS score ( Figure 5 , Spearman r=0.48, P< 0.001). Physician global assessment of endoscopic severity also showed significant but lower correlation with duration of disease (r=0.33; P= 0.02). Histologic data were available for 57 of the 65 patients with EGREFS data. Eight patients had missing histologic data. Forty six percent (26/57) of patients had active pathology at the time of endoscopy, defined as ≥30 eos/hpf in 5 hpf. Endoscopic abnormalities were demonstrated in 92% of patients with active pathology, and in 61% of patients with inactive pathology ( P= 0.0126). Composite EG-REFS scores were significantly higher in patients with active (median 5.0 (IQR 3.0, 7.0)) compared to those with inactive pathology (median 2.0 (IQR 0.0, 3.0)); P= 0.0002. Furthermore, peak and mean eosinophil densities demonstrated moderate correlations with EG-REFS activity in the fundus, body, antrum and overall locations ( Table 5 ). Global endoscopy scores also showed a significant association with active pathology with median scores in patients with active pathology of 4.5 (IQR 2.0, 7.0) vs inactive pathology of 1.0 (IQR 0.0, 4.0); P= 0.0013.

The EG-REFS classification and grading system prospectively assessed the presence and severity of endoscopically identified gastric abnormalities in a cohort of patients with EG. The most common features identified included erythema, raised lesions, erosions and granularity that were notably more pronounced in the gastric antrum. In the subset of patients with active histopathology, 92% of patients had one or more endoscopic abnormalities identified by EG-REFS. This prevalence is notably higher than previously reported in several retrospective pediatric and adult series, where only approximately 50% of patients had abnormalities. 3 , 4 , 7 . We postulate that the increased detection is related to the prospective data acquisition, experience of the CEGIR investigators, and systematic inclusion of multiple features. It should be noted that one Japanese study that carefully reassessed endoscopic images of the stomach reported higher prevalence of endoscopic abnormalities compared to studies relying upon endoscopic reports. 6 Of note, this study also identified mucosal cracks (fissures), rings and white exudate similar to features described in EoE that were not included in EG-REFS. 6 The heterogeneity in the prevalence of endoscopic findings may be related to varied demographics, diagnostic criteria, and use of active medical/diet therapies at the time of the endoscopies across studies. The EG-REFS demonstrated a strong correlation with physician global assessment of endoscopic activity, supporting the face validity of the system to capture the relevant aspects of disease activity determined by the endoscopist. The correlation of EG-REFS with overall severity was primarily due to features of erosion/ulceration, granularity, raised lesions/nodules, erythema, and friability. Abnormalities were significantly more common in the gastric antrum relative to the body or fundus, which correlated with histologic activity. The reason for antral predominance of disease activity is unclear but could be related to distinct mucosal cell types, concordant duodenal-gastric reflux, or distinct functional differences in the different regions of the stomach. While endoscopic severity did not appear to vary with age, gender or atopic status, severity was notably greater in patients with a longer duration of disease. Development of endoscopic assessment tools will allow not only standardization of nomenclature but also may provide a disease metric to complement symptom and histologic features in the assessment of therapeutic response in clinical practice and in medical/dietary trials. Currently, the clinical management of EGID focuses on symptom and histopathologic outcomes. The multitude of symptom manifestations of EG that include abdominal pain, nausea, vomiting, anorexia and early satiety create challenges in assessment. Due to adaptive behaviors of food avoidance and variations in sensory perception, symptoms may not accurately reflect disease activity. Moreover, EG may present with iron deficiency anemia, occult bleeding, nutrient malabsorption, or protein-losing enteropathy in the absence of overt symptoms. 15 For these reasons, symptom improvement may be an inadequate indicator of response to therapeutic intervention. Histologic assessments are limited as a sole therapeutic marker of disease control especially since EG inflammation can be focal with variable severity, and biopsies sample only a small fraction of the overall gastric mucosa. While data to support this notion are not available for EG, in EoE as well as in inflammatory bowel disease, correlation between validated patient-reported symptom assessment instruments and histologic activity has been only modest. 11 , 16 , 17 Such dissociations point to the importance of objective measures of disease activity using metrics such as endoscopic activity. In EoE, endoscopic features are more strongly correlated with symptoms and the EoE genetic transcriptome than with eosinophil density. 10 , 18 , 19 . A case series correlated gene expression patterns using microarray analyses and endoscopic features amongst 8 children with EG. 20 Substantial overlap in gene expression profiles was found when comparing patients with endoscopic findings of nodules compared to those with ulcerations. In a previously published CEGIR study of EG, we evaluated the association between mucosal eosinophilia and endoscopic activity using EG-REFS and several genes including CCL26 (eotaxin-3), CLC (Charcot Leyden crystal, an eosinophil specific marker), IL13RA2 (IL-13 receptor alpha 2), IL5 and SST (somatostatin). 21 CCL26 showed the strongest correlation with any endoscopic features, most notably nodularity and granularity, followed by IL33 , which inversely correlated most notably with granularity and friability and bleeding. 21 Interestingly, clustering of EG gene expression profiles separated endoscopic features into two general groups. The first group, which was associated with friability/bleeding and erythema, correlated with downregulation of molecular signatures ( ATP4A, IL33, and SLC26A7 ), while the second group was associated with nodularity and granularity correlated with upregulation of type 2 immunity and eosinophil-associated pathways ( CCL26, IL13RA2, and IL5 ). Further studies are needed to better define the clinical and pathophysiologic implications of these findings. The current study has several important limitations. Acknowledging that EG is a rare disease, the sample size of the current study was small, limiting the power to detect potentially meaningful associations. The EG-REFS system was based on expert opinion and literature review and is not a validated instrument but represents a first step towards this end. Endoscopic assessments were not obtained in a blinded manner. The sensitivity and specificity of EG-REFS in the diagnosis of EG could not be established due to the lack of validation as well as heterogeneity of the patient population and endoscopist experience. Indeed awareness of an existing diagnosis of EG may bias reporting of endoscopic features. As most patients were on active medical or diet therapy at the time of the index endoscopy, the burden of endoscopic abnormalities was likely underestimated as reflected by the low composite EG-REFS scores. Active therapy may also have affected observations regarding regional differences and temporal progression of EG. Since most patients were children, further studies are needed to confirm the generalizability of the results to an adult population. In addition, our study only evaluated EG patients with mucosal involvement. Muscular and serosal variants of EG without mucosal pathology were not included. Finally, diagnostic criteria for the histopathology and definition of histologic activity used in the study were based on limited data and may change with further consensus and research. Strengths of the study include the prospective, multicenter collection of baseline demographic, clinical, and endoscopic data that enhanced the accuracy and completeness of data capture for correlation with endoscopic features. Use of real-time endoscopic evaluation is likely a more robust means of assessment of disease activity compared with retrospective review of endoscopy reports or digital still images. Prior studies have described erythema, nodularity and ulceration as features of EG but none has systematically and prospectively assessed these features. 3 , 4 , 7 Additional features not previously well characterized in the literature were included based on the experience of a working group of pediatric and adult gastroenterologists with expertise in EGID. These included granularity, nodules, thickened folds, friability and pyloric stenosis. Finally, the current study is the largest prospective study to date of patients with EG. Identification of patients was greatly facilitated by the multi-center and collaborative CEGIR program, as well as the standardized data collection and management across the consortium. 9 Future directions should focus on refinement and validation of the EG-REFS. Studies that correlate specific endoscopic features with detailed histopathology beyond eosinophil density would be of interest. Based on the results from this study, specific features such as thickened folds, that show limited correlation with the overall severity assessment may be unnecessary. Similarly, the severity scales assigned were based on expert opinion and not clinical outcomes. The low prevalence of specific grades of certain features, such as extensive gastric ulceration, could provide rationale for the simplification of the scales. Studies defining the inter and intra-observer reliability of the EG-REFS should also be conducted. Furthermore, application of the EG-REFS to clinical trials will assess the responsiveness of the instrument to therapy and allow for additional refinement. In summary, we have described the presence and frequency of endoscopic findings in patients with EG prospectively collected through a multicenter study. Furthermore, we have used this data to assess an endoscopic scoring system for the characterization of endoscopically identified, gastric features in a relatively large cohort of children and adults with EG. Prospective application of this endoscopic outcome tool revealed that the most common endoscopic abnormalities in EG include erythema, raised lesions, erosions and granularity, and that the antrum has the most visible endoscopic changes. The developed tool was evaluated across pediatric and adult institutional sites, with content validity supported by correlation with global physician assessment of endoscopic findings. Validated outcome measures are increasingly relevant given the recent rise in scientific interest and investigations in EGIDs. Further refinement and validation of the EG-REFS should improve its potential utility in clinical studies and therapeutics in EGID.

Eosinophilic gastritis (EG) is a Th2-associated inflammatory disease of the stomach characterized by eosinophil-predominant, gastric mucosal inflammation and gastrointestinal symptoms. Reported symptoms of EG include abdominal pain, nausea, vomiting, early satiety, diarrhea and weight loss. EG may occur concomitantly with features consistent with EoE, eosinophilic enteritis (EEn) or eosinophilic colitis (EC), and falls under the broader category of eosinophilic gastrointestinal disease (EGID). 1 While less common than eosinophilic esophagitis (EoE), an accurate prevalence of EG is difficult to estimate due to the absence of accepted diagnostic criteria. Based on ICD-9 coding in a health insurance database, Jensen et.al. estimated the prevalence of EG as 6.3/100,000 and eosinophilic gastroenteritis as 8.4/100,000 in the United States. 2 Endoscopic features are used as clinically relevant outcomes for the assessment of disease activity in chronic, inflammatory, gastrointestinal diseases that include EoE, reflux esophagitis, peptic ulcer disease, and inflammatory bowel disease. For these diseases, validated instruments that quantify mucosal injury are used to characterize disease severity and assess therapeutic efficacy in clinical trials. Endoscopic features of EG reported in retrospective studies have included erythema (24–72%), erosions/ulceration (28–39%), and nodularity (0–28%). 1 , 3 – 8 A normal appearance has been observed in up to 62% of cases. 8 To date, no endoscopic assessment tool exists for EG. The aim of this study was to prospectively evaluate endoscopic features in a relatively large series of children and adults with EG for the purpose of standardizing nomenclature, improving understanding of the disease manifestations and developing the foundation for an endoscopic outcome instrument for clinical studies. A secondary aim was to assess the correlation between the severity of endoscopic features and gastric eosinophilia.