Benralizumab for eosinophilic gastritis: A phase 2, randomized, double-blind, placebo-controlled trial

SUMMARY

Background:

The role of eosinophils in disease pathogenesis and the effect of eosinophil depletion on patient outcomes in eosinophilic gastrointestinal diseases remain unclear. Benralizumab, an eosinophil-depleting monoclonal antibody targeting the interleukin-5 receptor alpha, may eliminate gastric tissue eosinophils and improve outcomes in patients with eosinophilic gastritis. We aimed to assess the efficacy and safety of benralizumab in patients with eosinophilic gastritis.

Methods:

We conducted a single-site randomized, double-blind, placebo-controlled phase 2 clinical trial in the USA. Individuals aged 12 to 60 years with symptomatic, histologically-active eosinophilic gastritis (EoG) and blood eosinophilia > 500 cells/μL were randomized (1:1) to benralizumab 30 mg or placebo, stratified by use of systemic and swallowed glucocorticoids for gastric disease. Treatments were administered subcutaneously every 4 weeks (3 total injections). The primary endpoint was the percentage of patients who achieved histological remission defined as peak gastric count <30 eosinophils per high-power field (eos/hpf) at week 12. Key secondary endpoints were change from baseline to week 12 in peak gastric eosinophil count, blood eosinophil count, EoG Histology Score including total, inflammatory, and structural feature scores, EoG Endoscopy Score, patient-reported outcome symptom measures (Severity of Dyspepsia Assessment [SODA] and Patient Reported Outcome Measurement Information System [PROMIS] short form), and gene expression patterns (EoG Diagnostic Panel [EGDP]). After the double-blind period, patients were eligible for continuation in two open label extensions for a total of 88 weeks. Efficacy analysis was done in the intention-to-treat (ITT) population and safety analysis in all patients receiving at least one dose of drug. This trial is registered on ClinicalTrials.gov, NCT03473977, and is completed.

Findings:

Between April 23, 2018 and January 13, 2020, 34 patients were enrolled, and 26 were subsequently randomly assigned to benralizumab (n=13) or placebo (n=13), and included in the ITT population (mean age 19·5 [SD 7·3] years, 73% male, 27% female). At week 12, 10 (77%) of 13 patients who received benralizumab and 1 (8%) of 13 receiving placebo achieved remission (difference 69% [95% CI 32 to 85]; p=0·0010). Changes from baseline to week 12 were greater in benralizumab vs placebo for gastric counts (−137 vs −38; difference −99 eos/hpf [−169 to−28]; p=0·0080), EoG histology total score (−0·31 vs −0·02; difference −0·29 [−0·46 to - 0·12]; p=0·0016), histology inflammatory score (−0·46 vs −0·04; difference −0·41 [−0·63 to −0·20]; p=0·0006) and blood eosinophil levels (−1060 vs −160; difference −910 cells/µL [−1570 to −350]; p=0·0044). By contrast, changes in EoG histology structural score (−0·07 vs 0·03; difference - 0·10 [−0·26 to 0·06]; p=0·23), endoscopy score (−1·0 vs −0·5; difference −0·5 [−2·4 to 1·4]; p=0·62), SODA pain intensity score (−3·8 vs −2·1; difference −1·8 [−9·3 to 5·7]; p=0·63), SODA non-pain symptom score (−2·5 vs −0·9; difference −1·6 [−4·3 to 1·0]; p=0·22), SODA satisfaction score (1·8 vs 2·1; difference −0·2 [−3·6 to 3·1]; p=0·88), PROMIS pain intensity (0·1 vs −0·8; difference 0·9 [−1·6 to 3·4]; p=0·45) and PROMIS T-scores were not significantly different between groups. Genes comprising the EGDP were not normalized by benralizumab. Headache was the most common adverse event, reported in 6 (46%) receiving benralizumab and 2 (15%) on placebo during the double-blind. Three serious adverse events (dizziness, aspiration, and rhabdomyolysis) unrelated to study treatment occurred during the open label extensions.

Interpretation:

Benralizumab treatment induced remission, as defined by absence of tissue eosinophilia, in most patients with eosinophilic gastritis, but the persistence of histologic, endoscopic, and gene expression features of the disease suggests a co-existing, eosinophil-independent pathogenic mechanism.

Funding:

AstraZeneca and Division of Intramural Research (National Institute of Allergy and Infectious Diseases, US National Institutes of Health).

Introduction

Eosinophilic gastrointestinal diseases (e.g., eosinophilic esophagitis and eosinophilic gastritis), are chronic inflammatory disorders characterized by eosinophilic infiltration in segments of the gastrointestinal tract.^1–3 Emerging evidence suggests these are food antigen–driven diseases involving type 2 immunity, mediated by memory T helper 2-cells, which drive eosinophilia via interleukin-5 (IL-5) and IL-13.^4,5 Despite the dominant presence of eosinophils in these diseases, it remains unclear whether eosinophils are pathogenic cellular mediators, bystander cells, or regulatory cells.

Previous studies of eosinophilic gastrointestinal diseases have primarily focused on eosinophilic esophagitis, which is the most prevalent of these diseases.⁶ Eosinophilic gastritis is less understood in part due to its lower prevalence which is estimated to occur in 6.3 cases per 100,000, based on a US population-based study.⁷ Patients with eosinophilic gastritis have a high-symptom burden exacerbated by delayed diagnosis,⁸ treatments with significant drawbacks (e.g., risks of acute/chronic complications with glucocorticoids),^9,10 and limited efficacy and/or tolerance of elimination diets.^11,12

Benralizumab is a humanized, afucosylated monoclonal antibody against the alpha chain of the IL-5 receptor (IL-5Rα), which is expressed by mature eosinophils, eosinophil progenitor cells, and basophils.^13,14 Benralizumab-induced engagement of natural killer cells results in enhanced antibody-dependent cell-mediated cytotoxicity, yielding near complete depletion of eosinophils.^15,16 Benralizumab is approved for treatment of severe eosinophilic asthma for ages 12 years and older and has shown promise in other eosinophil-associated diseases.^17–21 In this study, we aimed to assess the efficacy and safety of benralizumab in patients with eosinophilic gastritis and evaluate the effects of eosinophil depletion on features of the disease.

METHODS

Study design and participants

We conducted a single-site, investigator-initiated, phase 2, randomized, double-blind, placebo-controlled trial to assess the efficacy of benralizumab compared with placebo in adolescents and adults with eosinophilic gastritis. The study was completed at Cincinnati Children’s Hospital Medical Center (Cincinnati OH, USA), performed in accordance with the principles of the Declaration of Helsinki, and approved by the hospital institutional review board. Patients were recruited from clinical practice within Cincinnati Children’s Hospital and via referrals from other clinical practices throughout the USA. All participants or legally authorized representative for patients under age 18 provided written informed consent.

Patients eligible for enrolment were aged 12 – 60 years with symptomatic eosinophilic gastritis and were histologically active at time of screening, defined as a peak gastric eosinophil count ≥30 per high power field (hpf) in ≥5 hpfs (a diagnostic cut-off proposed by disease experts at study initiation^22,23 and utilized in recent trials^24–26). Both adolescents and adults were included in the study as published data at the time of study initiation suggested disease presentation and unmet treatment needs were common to both age groups.^27–29 Clinical symptoms were required to affect daily activities ≥2 times in 3 of the 4 weeks prior to enrollment. Patients were further required to have a history of blood eosinophilia, defined as an absolute eosinophil count >500 cells/µL, at least once in the 6 months prior to enrollment based on prior findings indicating benralizumab is beneficial in patients with pre-existing blood eosinophilia.¹⁷ Pre-existing treatment for eosinophilic gastritis or gastroenteritis (duodenal involvement) was permitted at a stable dose. Key exclusion criteria included historical diagnosis of other gastrointestinal disorders such as Crohn’s disease, inflammatory bowel disease, or Celiac disease; and use of any investigational biologic drug within 6 months prior to screening. Full eligibility criteria are available in the appendix (p 2 – 3).

Randomization and masking

Patients were randomized 1:1 to treatment with benralizumab 30 mg or placebo via block randomization (block size of four). Randomization was stratified by use of glucocorticoids (systemic or swallowed) for gastric disease. The allocation sequence was generated and held by the investigational drug services at Cincinnati Children’s Hospital who were not involved in enrollment of patients. The sequence was generated using the random number generator in randomization.com and stored in a locked location in investigational drug services which was not accessible to study staff. Treatment assignment was completed by the unblinded pharmacist by selecting the next available slot in the subject’s stratum. The investigators, study staff and study participants were blinded to treatment assignment throughout the trial. To maintain masking, the placebo syringes and cartons were matched in appearance to benralizumab. Study site personnel were blinded to blood eosinophil results and biopsy (histology) results until all patients completed the double-blind period. Laboratory investigators were masked to response to treatment until after all samples were analyzed. The study pathologist was masked to treatment assignment for histology analyses (primary and key secondary outcomes). Statisticians were not blinded to assignment when analyzing the data.

Procedures

Patients were screened during an up to 8-week screening period. Eligible patients were randomized to receive benralizumab 30 mg or placebo every 4 weeks for a total of three doses during a 12-week double-blind treatment period (appendix p 6). After the double-blind period, patients were eligible for continuation in a 24-week, open-label extension period (OLEI), during which all patients received a total of 4 doses of benralizumab. Patients who completed OLEI were eligible to continue benralizumab treatment every 4 or 8 weeks for 52 weeks (OLEII). Dosing intervals in OLEII were determined based on patient symptoms and coronavirus disease 2019 (COVID-19) restrictions. No other dose changes were permitted. Benralizumab or placebo was administered as subcutaneous injections in the upper arms, thighs, or abdomen and rotated at each visit. Individual patient dosing schedules throughout the study are shown in the appendix (p 13). Patients could be discontinued from the study at their request or investigator discretion; if lost to follow-up; due to severe complications from their eosinophilic gastrointestinal disorder or a serious adverse event related to therapy; or due to initiation of rescue medications for their eosinophilic gastrointestinal disorder.

Esophagogastroduodenoscopy with biopsy was conducted during screening (baseline) and following each treatment period (end of double-blind (week 12), OLEI (week 36), and OLEII (week 88)). Biopsies were collected from the gastric antrum and body, distal esophagus, and duodenal bulb and segment D3. Historical biopsies for assessment of eligibility were allowed if collected within the screening period. A single blinded pathologist with expertise in eosinophilic gastrointestinal disorders evaluated the gastric antrum and body biopsies for peak eosinophil count and histologic features using the Eosinophilic Gastritis Histology Score.³⁰ Additional gastric antrum biopsies collected at these time points were subjected to transcriptome analysis with the Eosinophilic Gastritis Diagnostic Panel [EGDP].³⁰ At the endoscopies conducted during screening and at the end of the double-blind, OLE1, and OLEII periods, research endoscopists assessed gastric endoscopic features using the Eosinophilic Gastritis Endoscopy Score.³¹ Blood was collected at screening, randomization (baseline), every 4 to 8 weeks, and at the end of each treatment period (week 12, week 36, and week 88). Blood was used to measure levels of eosinophils at these time points, for biomarker analysis (chemokines/cytokines) during the double-blind period, and for assessment of anti-drug antibodies during the double-blind and open label periods as described in appendix p 4. As no validated questionnaires for eosinophilic gastritis were available at the time of the study, patients self-administered the Severity of Dyspepsia Assessment (SODA)³² at screening, randomization (baseline) and every 4 weeks for the evaluation of gastric symptoms during the double-blind and OLEI periods and every 24 weeks during OLEII. Patient Reported Outcome Measurement Information System (PROMIS) short-form questionnaires were also completed by the patients at these time points to evaluate overall well-being. Esophageal and duodenal histologic, endoscopic, and symptom assessments (Pediatric Eosinophilic Esophagitis (EoE) Symptom Score (PEESS v2.0), PedsQL^™ EoE Module, EoE Activity Index (EEsAI), and EoE Quality of Life-A (EoE-QoL-A)) were completed during screening, randomization (baseline) and following each treatment period for the evaluation of patients with concomitant eosinophilic esophagitis and/or duodenitis. Sex, as part of a demographic assessment, was self-reported as male, female, or undifferentiated at baseline. As the elimination of foods could affect outcome, patients’ diets (foods actively eliminated) were documented at baseline and reassessed for changes at the end of treatment periods and periodically throughout the study. Details for all of the assessments are in the appendix p 3 – 5. Adverse events were collected from the time of consent until the end of patient study participation or until 30 days after patient withdrawal from the study. Adverse events were assessed every 4 weeks at study visits by the site investigator and/or site staff and were reported using the National Cancer Institute’s Common Terminology Criteria for Adverse Events (version 4.0). Safety assessments (vitals, weight, laboratory findings, and physical exams) were conducted at baseline and regular intervals (generally every 4 – 8 weeks) during the double-blind and open label periods.

Outcomes

The primary endpoint was induction of disease remission at week 12, defined as the percentage of patients who achieved histological remission (i.e., peak gastric eosinophil count <30 eos/hpf). Key secondary efficacy endpoints included change from baseline to week 12 in peak gastric eosinophil count, blood eosinophil count, EoG histology score (including total, inflammatory, and structural feature scores), EoG endoscopy score, and patient-reported outcome measures (SODA, PROMIS). Other secondary endpoints included changes in gastric gene expression patterns as assessed by the EGDP. The EGDP was utilized in lieu of whole genome RNA sequencing due to its specificity to genes shown to be reflective of EoG activity and disease pathogenesis. Secondary endpoints for patients with concomitant eosinophilic esophagitis (≥15 eos/hpf in esophageal biopsy) and duodenitis (≥52 eos/hpf in duodenal biopsy) included change from baseline to week 12 in esophageal and duodenal histological scores and peak eosinophil counts; esophageal endoscopy score; and symptom and quality of life scores for eosinophilic esophagitis (PEESS v2.0, PedsQL^™ EoE, EEsAI, and EoE-QoL-A). Due to the small number of patients with comorbid eosinophilic esophagitis and an even smaller subset of patients completing these pediatric and adult questionnaires, this data was not analyzed as no inference could be made. Post-hoc endpoints in the double-blind period included changes in blood biomarkers. Post-hoc endpoints in the open label periods included induction of remission after each period and change from baseline to week 36 and week 88 in histologic, endoscopic, clinical, and molecular endpoints as described for the double-blind period. Safety outcome measures included rates of adverse events and changes in vital signs, physical examination, and laboratory findings.

Statistical Analysis

A power calculation showed a sample size of 24 patients (12 per treatment arm) was needed to achieve at least 80% power to detect a difference in histological remission rates between treatment arms, assuming a response rate of 10% in the placebo group and 70% in the benralizumab group¹⁶ at week 12 with a two-sided Fisher’s exact test at alpha level of 0.05. Sample size calculations were done using nQuery (Graphpad software, Boston, MA, USA). We planned to enroll 26 patients to account for potential attrition.

Efficacy outcomes were assessed in the intention-to-treat population which included all patients who underwent randomization regardless of major protocol deviations. For the two major protocol deviations in which treatment was extended (figure 1), the last efficacy measures taken prior to extension were imputed for end of treatment (week 12) values. Missing data for secondary efficacy outcomes were imputed using the last measure carried forward. A pre-planned per-protocol analysis of efficacy for primary and key secondary endpoints was also performed which included all randomized patients without major protocol deviations. For the subgroup of patients with available biopsies, transcriptome analyses were performed. A historical control group recruited from the study site³⁰ was used for comparison. For the open-label extension, analyses for post-hoc endpoints were performed on complete cases (patients with baseline and open-label data for the outcome of interest). No imputation for missing data was performed for the open-label extension periods. Safety and tolerability were evaluated in all patients who received ≥1 dose of study drug (safety population).

Figure 1.

Trial profile

The primary outcome was evaluated using Fisher’s exact test. Secondary outcome changes from baseline to week 12 were compared between groups using Wilcoxon rank-sum test for blood eosinophil counts and PROMIS scores and two-sample t-tests for other continuous variables (normally distributed data). Data distributions were evaluated using the Kolmogorov-Smirnov test and Q-Q plot with the univariate procedure in SAS (version 9.4). 95% confidence intervals for means and proportions were determined using the Student’s t distribution and confidence limits were calculated by the Wilson score method, respectively. For equal variance, p values and 95% CI using the pooled method were reported; otherwise, Satterthwaite method was used. Hodges–Lehmann method was used to estimate median differences with 95% confidence intervals. Logistic, linear, and quantile regression analyses were used to examine the interaction and confounding between age and baseline endoscopy score and treatment for the primary (pre-specified sensitivity analysis) and key secondary (post-hoc sensitivity analyses) endpoints (with the exception of PROMIS scores). The open-label extension efficacy endpoint changes from baseline were analyzed using paired t-tests or Wilcoxon signed-rank tests. A p-value of <0·05 indicated statistical significance. All statistical analyses were performed with SAS software, version 9.4 (SAS Institute, Inc.; Cary, NC). A data and safety monitoring board reviewed the safety data once yearly and was responsible for prompt review of serious adverse events for recommendations for continuation of the study. This trial was registered with ClinicalTrials.gov, NCT03473977.

Role of the funding source

The funders of the study were not involved in the design or conduct of the study, data collection, data analysis, data interpretation, writing of the report, or the decision to submit the manuscript for publication.

RESULTS

Between April 23, 2018 and January 13, 2020, a total 34 patients were screened, of whom 26 were randomized to treatment with benralizumab (n=13) or placebo (n=13) and included in the intention-to-treat and safety populations (figure 1). All patients completed the double-blind period and continued into OLEI, 21 completed OLEI, and 13 completed OLEII. During the double-blind period, one patient in each group had an 8-week extension (2 additional injections) due to endoscopy delays related to COVID-19 restrictions, and one patient in the placebo group received benralizumab rather than placebo at week 4; these patients were excluded from the per-protocol analyses.

Patient demographics and baseline characteristics were similar between groups, except for patient age and eosinophilic gastritis endoscopy score, which were both lower in the benralizumab group versus placebo (table 1). Of the 26 patients randomized, 19 (73%) were male and 25 (96%) were white. The mean age was 19·5 years (SD 7·3). Overall, 22 (85%) were atopic with concomitant active eosinophilic esophagitis, eosinophilic duodenitis, or both present in 10 (38%), 4 (15%), and 1 (5%) patient, respectively. The mean peak gastric eosinophil count was 165 (SD 70) eos/hpf, and median blood eosinophil count was 1350 (IQR 710 – 2630) cells/µL. Throughout the double-blind treatment period, all patients reported adherence to their baseline medications and diet.

Table 1.

Baseline characteristics (intention-to-treat and safety population)

	Benralizumab	Placebo
Characteristic	(n=13)	(n=13)
Age, years	14·9 (2·0)	24·1 (7·9)
Sex
Male	9 (69%)	10 (77%)
Female	4 (31%)	3 (23%)
Race
White	12 (92%)	13 (100%)
Black	1 (8%)	0
Medical History
Atopic†	10 (77%)	12 (92%)
Food allergy	6 (46%)	8 (62%)
>5 years since eosinophilic gastritis diagnosis	4 (31%)	8 (62%)
Body mass index	21.2 (5.2)	23.5 (5.8)
Active disease
Eosinophilic gastritis only	7 (54%)	4 (31%)
Eosinophilic gastritis and esophagitis	5 (38%)	5 (38%)
Eosinophilic gastritis and duodenitis	1 (8%)	3 (23%)
Eosinophilic gastritis, esophagitis, and duodenitis	0	1 (8%)
Symptoms
Abdominal pain	10 (77%)	13 (100%)
Nausea	11 (85%)	11 (85%)
Vomiting	9 (69%)	10 (77%)
Blood eosinophil count, cells/µL	1100 (830–2610)	1640 (710–2800)
Histologic features
Gastric peak eosinophil count, eos/hpf*	160 (72)	171 (71)
Eosinophilic gastritis histology score
Total score	0·53 (0·16)	0·51 (0·14)
Inflammatory score	0·53 (0·19)	0·51 (0·18)
Structural score	0·53 (0·19)	0·50 (0·14)
Eosinophilic gastritis endoscopy total score¶	4·2 (1·5)	6·2 (2·4)
Severity of Dyspepsia Assessment
Pain intensity score	25·2 (7·5)	25·3 (5·3)
Non-pain symptom score	15·8 (4·1)	14·9 (2·5)
Satisfaction score	10·0 (4·2)	8·0 (4·8)
PROMIS short form
Pain intensity rating	4·3 (2·4)	4·2 (1·9)
Biomarkers
Interleukin-5, pg/mL	3.8 (1.1–10.9)	6.9 (2.5–15.4)
Eotaxin, pg/mL	470 (299)	692 (285)

Data are n (%), mean (SD), or median (IQR).

^†History of asthma, eczema, or allergic rhinitis/seasonal allergies.

^*Highest gastric eosinophil count among 5 high-power fields evaluated.

^¶Maximum score among fundus, body, and antrum with data available in benralizumab (n=11)

In the intention-to-treat population, 10 (77%) of 13 patients randomized to benralizumab versus 1 (8%) of 13 patients receiving placebo experienced disease remission (peak gastric count <30 eos/hpf) at week 12 representing a marked difference of 69% (95% CI 32 to 85; p=0·0010; table 2). Notably, remission rate was not markedly different between those who used steroids (5 out of 14) compared to those who did not (6 out of 12; p=0.69) and therefore we did not adjust for stratification variables. Excluding those with major protocol deviations, 10 (83%) of 12 patients in the benralizumab group achieved remission compared with 0 (0%) of 11 in the placebo group (difference 83% [45 to 95]; p=0·0001; appendix p 14). Gastric and blood eosinophils were eliminated at week 12 in most patients who received benralizumab (figure 2A, 2B). The mean change from baseline in peak gastric eosinophil count in patients receiving benralizumab was - 137 eos/hpf compared to −38 eos/hpf in the placebo group (mean change difference, −99 eos/hpf [−169 to −28]; p=0·0080). Median reductions in blood eosinophil count were also considerably greater in the benralizumab group versus placebo (median change difference, - 910 cells/µL [−1570 to −−350]; p=0·0044; table 2).

Table 2.

Primary and secondary endpoints following 12 weeks of treatment (intention-to-treat population)

	Benralizumab (n=13)	Placebo (n=13)	Difference vs Placebo	P value
Primary endpoint
Remission†
Peak gastric eosinophil count < 30 eos/hpf	10 (77%)	1 (8%)	69%	0·0010
	(50 to 92)	(1 to 33)	(32 to 85)
Key secondary endpoints	Change from baseline
Gastric eosinophil counts	−137 (−186 to −88)	−38 (−94 to 18)	−99 (−169 to −28)	0·0080
Blood eosinophil counts	−1060 (−1740, −830)*	−160 (−710, 120)*	−910 (−1570 to −350)**	0·0044
Eosinophilic gastritis histology
scores
Total score	−0·31 (−0·42 to −0·20)	−0·02 (−0·16 to 0·12)	−0·29 (−0·46 to −0·12)	0·0016
Inflammatory score	−0·46 (−0·60 to −0·31)	−0·04 (−0·22 to 0·13)	−0·41 (−0·63 to −0·20)	0·0006
Structural score	−0·07 (−0·19 to 0·05)	0·03 (−0·09 to 0·15)	−0·10 (−0·26 to 0·06)	0·23
Eosinophilic gastritis endoscopy score¶	−1·0 (−2·3 to 3)	−0·5 (−2·0 to 1·0)	−0·5 (−2·4 to 1·4)	0·62
Severity of Dyspepsia
Assessment
Pain intensity symptom score	−3·8 (−9·5 to 1·8)	−2·1 (−7·7 to 3·5)	−1·8 (−9·3 to 5·7)	0·63
Non-pain symptom score	−2·5 (−4·7 to −0·4)	−0·9 (−2·6 to 0·8)	−1·6 (−4·3 to 1·0)	0·22
Satisfaction score	1·8 (−0·8 to 4·5)	2·1 (−0·2 to 4·3)	−0·2 (−3·6 to 3·1)	0·88
PROMIS short form
Pain intensity rating	0·1 (−1·7 to 1·8)	−0·8 (−2·8 to 1·1)	0·9 (−1·6 to 3·4)	0·45
Other endpoints
Interleukin-5, pg/mL	27·5 (2·2, 133·2)	0·8 (−0·0, 7·6)	20·7 (0·2 to 126·5)	0.031
Eotaxin, pg/mL	662 (331 to 993)	142 (−107 to 390)	520 (128 to 913)	0.012

Data are n (%) (95% CI) or mean (95% CI) unless otherwise stated.

^†Remission is defined as <30 eos/hpf for the 5 high-power fields with the highest eosinophil counts.

^*Data are median (IQR)

^**Hodges Lehmann Estimate (95% CI)

^¶Data were available for 11 and 13 patients in the benralizumab and placebo groups, respectively.

Figure 2. Effects on tissue, blood, and transcriptome after 12 weeks of treatment (intention-to-treat population).

(A) Changes in gastric eosinophil counts. (B) Changes in blood eosinophil counts. (C) Expression of genes associated with eosinophilic gastritis. Control samples and active eosinophilic gastritis are reference data from Shoda et al.³⁰ (D) Changes in plasma concentrations of interleukin-5 (IL-5) (E) Changes in plasma concentrations of eotaxin.

An eosinophilic gastritis-specific histology score³⁰ was used to quantify changes in eosinophil-associated inflammatory parameters and structural features. Indicators of inflammation (e.g., lamina propria eosinophil sheets, periglandular circumferential collars, eosinophils in surface epithelium, eosinophil glandulitis, and eosinophils in muscularis mucosa) were improved from baseline in the benralizumab group in all but 2 patients. Structural abnormalities (reactive epithelial changes) improved in 6 of 13 in the benralizumab group but worsened or were unchanged in 7 of 13 patients, and lamina propria smooth muscle hyperplasia persisted in all patients on benralizumab (appendix p 7; appendix p 8 for representative images of biopsies). These findings are reflected by greater improvements from baseline to week 12 in EoG histology total and inflammatory scores in the benralizumab group compared with placebo (between-group mean change differences: −0·29 [95% CI −0·46 to −0·12]; p=0·0016 and −0·41 [95% CI −0·63 to −0·20]; p=0·0006, respectively), whereas there was a small difference in changes in structural scores between treatment groups (mean change difference −0·10 [−0·26 to 0·06]; p=0·23; table 2).

Changes from baseline to week 12 in EoG endoscopy score were minimally different between groups (mean change difference −0·5 [95% CI, −2·4 to 1·4]; p=0·62; table 2). Additionally, only small differences were observed in mean changes from baseline in SODA pain intensity, non-pain symptoms, and satisfaction scores between benralizumab and placebo (between-group mean change differences: −1·8 [−9·3 to 5·7], p=0·63; −1·6 [−4·3 to 1·0], p=0·22; −0·2 [−3·6 to 3·1], p=0·88; respectively, table 2). Consistent with these findings, changes in PROMIS T-scores for health domains (appendix p 15) and pain intensity rating (table 2) were similar between benralizumab and placebo groups.

The effect of eosinophil depletion on cardinal features of eosinophilic gastritis pathogenesis were evaluated using transcriptomics. EGDP results, comprising 48 genes dysregulated in eosinophilic gastritis,³⁰ suggest the overall abnormal expression pattern did not improve with benralizumab (figure 2C). Notably, this gene signature normalizes in patients with disease remission induced by diet and/or steroids.³⁰ The only genes that improved with benralizumab were eosinophil-specific markers CLC and CCR3 (figure 2C). Comparisons to a historical control group revealed an expression signature consistent with persistent pathology (figure 2C, left lane). Post-hoc evaluation of blood chemokines/cytokines showed changes from baseline in plasma levels of IL-5 and eotaxin were significantly greater in the benralizumab group compared with placebo (median change difference 20·7 pg/mL [95% CI 0·2 to 126·5]; p=0·031 and mean change difference 520 pg/mL [128 to 913]; p=0·012 respectively) (figure 2D, 2E). Changes in plasma levels of other chemokines/cytokines relevant to eosinophilic gastritis (e.g., IL-4, IL-13, and IL-33) were similar between treatment groups (data not shown). Anti-drug antibodies developed in two patients in the benralizumab group during the double-blind; however, in spite of this, benralizumab gastric and blood eosinophils were eliminated (appendix p 16). Baseline and week 12 values for key secondary and biomarker outcome measures are shown in the appendix (p 17).

Analyses of the per-protocol population yielded similar results to the intention-to-treat analyses for the key secondary endpoints (appendix p 14). Because age and baseline endoscopy score were statistically different between treatment groups, we examined the interaction and potential confounding between age/endoscopy score and treatment for major endpoints. Regression analyses in the per-protocol and intention-to-treat populations supported the significant impact of treatment on remission, peak gastric eosinophil count, blood eosinophil count, and total and inflammatory histology scores independent of age and endoscopy score (appendix p 18).

In patients with comorbid eosinophilic esophagitis receiving benralizumab, eosinophils were depleted from esophageal tissue and histology inflammatory sub-scores were reduced, although improvements were not statistically different compared to placebo (mean change difference −29 eos/hpf [95% CI −95 to 38]; p=0·36 and −0·37 [−1·18 to 0·44]; p=0·34 respectively) likely due to the small number of cases. Histology structural sub-scores increased (worsened) in both groups. In the one patient with comorbid eosinophilic duodenitis, eosinophils were depleted from duodenal tissue. A summary of changes in esophageal and duodenal histology and endoscopy scores is shown in appendix p 19. Patient reported outcomes for symptoms of eosinophilic esophagitis were not analyzed due to small sample size when stratified by adult and pediatric questionnaires and symptoms were also not analyzed for the one patient with comorbid eosinophilic duodenitis.

As post-hoc analysis, rates of remission and histologic, endoscopic, molecular, and clinical improvements were evaluated after completion of OLEI and OLEII. At week 36 (end of OLEI), 8 (80%) of 10 patients who received benralizumab from the double-blind period through OLEI (benra/benra) achieved disease remission as defined by gastric eosinophil levels. Among patients who received placebo in the double-blind period followed by benralizumab in OLEI (placebo/benra), 7 (64%) of 11 achieved histologic remission. Levels of anti-drug antibodies were elevated (positive) in one non-responder in the benra/benra group and one non-responder in the placebo/benra group (appendix p 16). At the end of OLEII, 7 (100%) of 7 and 4 (67%) of 6 patients in the benra/benra and placebo/benra treatment groups achieved remission, respectively. Changes from baseline in gastric and blood eosinophil counts and histology inflammatory scores were maintained in the benra/benra group and improvements were observed in patients who transitioned from placebo to benralizumab (figure 3A, 3B; appendix p 9 and p 20). By week 88, histology structural score, endoscopy score, and gastric transcripts were largely unchanged from baseline (figure 3C, 3D; appendix p 10), and histologic structural abnormalities persisted in most patients (appendix p 11 – 12). In patients who completed OLEII, gastric mRNA levels of key cytokines/chemokines related to eosinophilic gastritis [CCL26 (eotaxin-3), CCL18, IL13RA2, and IL-5] remained upregulated, although expression of genes associated with fibrosis (COL2A1, BMP3) showed improvement. Levels of CLC and CCR3 remained down-regulated (normalized) in patients receiving benralizumab from week 12 through week 88 (appendix p 10). Notably, mean SODA pain intensity score and non-pain symptom score improved from baseline among the limited number of patients completing the open-label extensions (figure 3E, 3F; appendix p 20).

Figure 3. Histologic, endoscopic, and symptom response over time (all observations).

(A) Peak gastric eosinophil count. (B) Scores for inflammatory features from gastric histology assessments. (C) Maximum (worst) endoscopic feature scores from the gastric body, antrum, or fundus. (D) Scores for structural features from gastric histology assessments. (E) Scores for Severity of Dyspepsia Assessment (SODA) pain intensity (F) Scores for SODA non-pain symptoms. All patients with data at the given visit time points including patients whose treatment was extended are included. Individual patient measures (unbolded) and group means (bold) are shown. No imputations for missing data were performed. BL, baseline; DB, double-blind period; OLE, open label extension.

During the double-blind period, 11 (85%) of 13 patients randomized to benralizumab, and 6 (46%) of 13 patients randomized to placebo experienced ≥1 treatment-emergent adverse events (appendix p 20). The most common adverse events during this period were headache, nausea, vomiting, and dizziness, which were mild to moderate in intensity; no serious adverse events occurred. Rates of treatment-emergent adverse events in OLEI and OLEII were similar to the double-blind period (appendix p 21). One patient discontinued treatment during OLEI due to an adverse event. The patient experienced a moderate allergic reaction after receipt of the first study drug injection for OLEI. Approximately 1 hour prior to treatment, the patient accidentally consumed a known offending food allergen. As study drug administration could not be excluded as a contributing factor, further study treatment was discontinued for this patient. Serious adverse events (dizziness, aspiration, and rhabdomyolysis) were reported by 2 patients in OLEI and OLEII. Two events of rhabdomyolysis and one event of dizziness occurred in the same patient; both events of rhabdomyolysis were preceded by unaccustomed strenuous exercise. A second patient experienced aspiration post endoscopy. The serious adverse events were considered unrelated to study treatment. No patients required dose reductions or were discontinued due to drug-related toxicity during the study and no treatment-related deaths occurred. Finally, no clinically significant changes in vital signs or physical findings were observed throughout the study.

Discussion

A significantly greater proportion of patients with active eosinophilic gastritis treated with benralizumab in this phase 2 clinical trial achieved remission, as measured by gastric eosinophil counts, compared with placebo. Benralizumab depleted eosinophils in blood and tissue and improved histologic measures of inflammation in the gastrointestinal tract. However, structural histologic abnormalities, endoscopic severity, molecular markers of disease, and patient-reported measures of pain and non-pain symptoms, persisted during short-term treatment. Post-hoc analysis of the histologic, endoscopic, and molecular endpoints in the open-label period supported the double-blind results. These findings suggest depletion of blood and gastric eosinophils is not sufficient to normalize histologic and endoscopic abnormalities or improve symptoms in many patients with eosinophilic gastritis.

Our findings suggest that the pathogenesis of eosinophilic gastrointestinal diseases involve mechanisms that are independent of eosinophilia in large part. Given the presumed role of eosinophils in these diseases, several biologic therapies that reduce eosinophils to varying degrees and in conjunction with different arrays of other immune cell types have been explored. Neutralization of IL-5, which is involved in eosinophil production, activation, and survival,³³ with reslizumab and mepolizumab in patients with eosinophilic esophagitis, has not shown favorable clinical effects,^34,35 possibly because reslizumab and mepolizumab reduce but do not deplete eosinophils.^36,37 A phase 2 clinical trial demonstrated that the anti-Siglec-8 antibody, lirentelimab, dramatically reduced gastrointestinal eosinophil counts and improved clinical symptoms in patients with eosinophilic gastritis and eosinophilic duodenitis.²⁴ However, patients in phase 3 clinical trials of lirentelimab with eosinophilic gastritis, duodenitis, and esophagitis achieved improvement in histologic but not symptomatic aspects of the disease, although this is not published.³⁸ The findings reported herein are supported by another recent open-label study of 7 patients with hypereosinophilic syndrome and gastrointestinal involvement which demonstrated persistent disease flares in some patients and histological findings despite eosinophil depletion by benralizumab.³⁹ It is important to note that none of these studies examined the impact of eosinophil reduction on a diverse set of outcome measures as we employed in this study (including broad endoscopic, histologic, and molecular markers). Our findings call attention away from eosinophils as a direct therapeutic target for eosinophilic gastrointestinal diseases, consistent with the recently reported benefit of anti-type 2 cytokine therapy, which more broadly suppress type 2 immunity, for eosinophilic esophagitis.⁴⁰ It is notable that some type 2 cytokines actually increased in the benralizumab treated patients. Furthermore, our results suggest that defining disease remission in eosinophilic gastritis by eosinophil counts should be reconsidered given that core disease features persist despite eosinophil depletion.

This study was limited by the small sample size and involvement of only a single site. While treatment groups differed by age and endoscopy score (due to the small sample size), analysis indicated that the treatment effect was not confounded by these parameters. Furthermore, there is no evidence to suggest that the disease pathogenesis would be different as a function of age between adolescents and young adults (which comprised the majority of our adult enrollees). Consistent with the marginally older age in the placebo group, it is notable that the duration of disease was also longer in this group; if anything, these differences are predicted to favor benralizumab, yet no benefit was observed. The double-blind period of this study was only 12 weeks duration, which may not have been an adequate period of time to uncover the eosinophil dependent effects of the disease. Consequently, a post-hoc analysis of the two open label extensions for up to 88 weeks was completed and reported herein which did not reveal dramatic changes in any additional key secondary endpoints. The modest improvements observed in symptoms in the open-label extension phases were confounded by subjective scoring and patient selection (especially in OLEII where large patient drop out occurred). While the assessments employed in this study were not disease-specific validated instruments as none were existent when this study was undertaken, the symptoms evaluated were consistent with the core symptoms of eosinophilic gastritis.⁴¹ Finally, this study had methodological limitations including the use of last measure carried forward to impute the efficacy outcomes for the two individuals who had received additional injections before the end of the double blind period due to COVID-19. Further, during the open label extension, data may be affected by selection bias. Collectively, despite limitations, these findings provide compelling evidence for a largely eosinophil independent disease mechanism.

In conclusion, this study demonstrated that treatment with benralizumab resulted in remission, as defined by absence of tissue eosinophilia, in most patients with active eosinophilic gastritis. Nonetheless, the persistence of signs, symptoms, and biomarkers of eosinophilic gastritis suggests the co-existence of an eosinophil-independent pathogenic mechanism in many patients. These findings provide compelling evidence for a changed paradigm, shifting attention away from eosinophils as the main contributor and biomarker in eosinophilic gastrointestinal diseases. As such, successful management of eosinophilic gastritis, may require inhibiting pathways that more broadly reduce type 2 inflammation rather than only targeting eosinophils.⁴⁰

Research in context.

Evidence before this study

Eosinophilic gastritis is one of several chronic inflammatory diseases characterized by eosinophilic infiltration of the gastrointestinal tract. These eosinophilic gastrointestinal diseases, which also include eosinophilic esophagitis and enteritis, are primarily driven by immunological hypersensitivity to food antigens. The gastrointestinal eosinophilia is driven by the interplay of innate and adaptive immunity involving type 2 immune responses, especially T helper type 2 cell production of interleukin (IL)-5 and IL-13, and epithelial cell production of eotaxins. Eosinophilic gastritis affects about 6.3 persons per 100,000 in the US population and its symptoms commonly include abdominal pain, nausea and vomiting. Although there are no approved drugs for the treatment of eosinophilic gastritis, therapy often focuses on food elimination diets as well as glucocorticoids, which have significant side effects (e.g., complications with long-term systemic glucocorticoids use), limited efficacy and/or low tolerability and related poor compliance rates. A major advance in treating eosinophilic diseases has been the development of eosinophil-depleting antibody therapeutics which have now been approved for eosinophilic asthma, hypereosinophilic syndrome and other eosinophilic diseases depending upon the specific agent. These drugs provide an opportunity to improve care of patients with eosinophilic gastritis and to assess the role of eosinophils in disease pathogenesis. Early studies with neutralizing antibodies against IL-5 have not shown clinical efficacy in eosinophilic gastrointestinal diseases but the results are hard to interpret as anti-IL-5 only partially reduces gastrointestinal eosinophil levels and only a limited number of outcome measures have been examined. Benralizumab (an FDA approved anti-IL-5 receptor antibody) and lirentelimab (a non-FDA approved anti-Siglec-8 antibody), more extensively deplete eosinophils via antibody dependent cellular cytotoxicity mechanisms, and hence provide even better opportunities to treat eosinophilic gastritis and uncover disease mechanisms. While a phase II reported study with lirentelimab showed promising results, phase III results failed to demonstrate clinical improvement despite improved histology; the released results have been limited in scope and remain unpublished. In a small cohort of benralizumab treated patients with hypereosinophilic syndrome involving gastrointestinal eosinophilia, a heterogeneous responses was seen in an open label period. These mixed and limited results indicate a need for further (randomized) studies to comprehensively evaluate the efficacy of eosinophil-depleting agents (like benralizumab) in improving disease-specific features of eosinophilic gastritis. Such a study may help clarify the role eosinophils have in disease pathogenesis.

We searched PubMed using the terms “eosinophilic gastritis” and “benralizumab” for articles published from database inception to March 1, 2023. No language restriction was applied. Our literature search showed no randomized trials have evaluated benralizumab in eosinophilic gastritis.

Added value of this study

In this phase 2 randomized trial, we compared patients with eosinophilic gastritis receiving benralizumab or placebo using a broad set of endoscopic, histologic, molecular, and symptom outcomes after 12 weeks of therapy. These adolescents and adults where subsequently followed for up to 88 weeks of additional treatment with benralizumab in an open label extension. Benralizumab depleted eosinophils in the blood and gastric tissue and improved histologic measures of inflammation in the gastrointestinal tract but had limited effect on structural histologic abnormalities, endoscopic severity, and molecular markers of disease when compared to placebo. These findings were largely unchanged even after long-term use of benralizumab and only a subset of patients experienced symptomatic improvement in the trial. Thus, key signs, symptoms, and molecular biomarkers of eosinophilic gastritis persisted despite elimination of tissue eosinophils. This was the first randomized trial of benralizumab to evaluate a diverse set of outcome measures in eosinophilic gastritis.

Implications of all the available evidence

This trial and previous non-randomized trials and/or unpublished data show that the depletion of eosinophils in the treatment of eosinophilic gastritis – and potentially other eosinophilic gastrointestinal diseases – has limited efficacy in improving cardinal features of the disease in many patients. These findings suggest that the pathogenesis of eosinophilic gastrointestinal diseases likely involve co-existing, eosinophil-independent pathogenic mechanisms and successful therapies may require inhibiting pathways that more broadly reduce type 2 inflammation rather than only targeting eosinophils. Furthermore, our results suggest that defining disease remission in eosinophilic gastritis by eosinophil counts should be reconsidered given that core disease features persist despite eosinophil depletion.