One Food versus Six Food Elimination Diet Therapy for Treatment of Eosinophilic Esophagitis: A Multicenter Randomized Clinical Trial

Summary

Background:

Empiric elimination diets are effective for achieving histologic remission in eosinophilic esophagitis (EoE), but randomized trials comparing diet therapies are lacking. We aimed to compare a six food elimination diet (6FED) with a one food elimination diet (1FED) for treatment of adults with EoE.

Methods:

We conducted a randomized, open label trial across 10 sites in the Consortium of Eosinophilic Gastrointestinal Disease Researchers in the USA. Symptomatic adults aged 18—60 years with active EoE were centrally randomized 1:1 to 1FED (animal milk) or 6FED (animal milk, wheat, egg, soy, nuts, seafood) for 6 weeks. Randomization was stratified by age, enrolling site, and gender. The primary endpoint was the proportion achieving histologic remission (peak esophageal count < 15 eosinophils per high power field). Key secondary endpoints included change from baseline in peak eosinophil count, EoE Histology Scoring System (EoEHSS), Endoscopic Reference Score (EREFS), EoE Symptom Activity Index (EEsAI), and quality of life (EoE-QoL-A and PROMIS) scores. Non-responders to 1FED and 6FED could proceed to 6FED and swallowed fluticasone propionate 880 mcg twice daily (unrestricted diet) respectively for 6 weeks. Efficacy and safety analyses were performed on the intention-to-treat (ITT) population. This trial is registered at ClinicalTrials.gov, NCT02778867 and is completed.

Findings:

Between May 23, 2016 and March 6, 2019, 129 patients were enrolled and randomized to 1FED (n=67) and 6FED (n=62). After 6 weeks, histologic remission was 23 (34%) of 67 in 1FED vs 25 (40%) of 62 in 6FED (difference 6% [95% CI −11 to 23]; $\text{p}=\mathrm{0·58}$). Peak eosinophil counts declined in both groups (difference in geometric mean ratios 0.72 [0.43 to 1.20]; p = 0.21). Mean changes in EoEHSS ($-\mathrm{0·15}$ vs $-\mathrm{0·23}$; difference $-\mathrm{0·08}$ [$-\mathrm{0·21}$ to 0·05]; $\text{p}=\mathrm{0·23}$), EREFS ($-\mathrm{0·6}$ vs $-\mathrm{1·0}$; difference $-\mathrm{0·4}$ [$-\mathrm{01·1}$ to 0·03]; $\text{p}=\mathrm{0·28}$), EEsAI ($-\mathrm{3·0}$ vs $-\mathrm{8·2}$; difference $-\mathrm{5·2}$ [$-\mathrm{11·2}$ to 0·8]; $\text{p}=\mathrm{0·09}$) were not significantly different between 1FED vs 6FED. Changes in quality of life scores were small and comparable between groups. No adverse event was observed in more than 5% of patients following either diet. For 1FED and 6FED non-responders proceeding to 6FED or fluticasone, 9 (43%) of 21 and 9 (82%) of 11, respectively, had histologic response.

Interpretation:

1FED and 6FED achieved similar remission rates and comparable improvements in histologic and endoscopic features in adults with EoE. The 6FED and topical steroids were effective therapies for almost half of 1FED failures and most 6FED failures, respectively. Our findings indicate that eliminating animal milk alone is an acceptable initial dietary therapy for EoE.

Funding:

National Institutes of Health

Introduction

Eosinophilic esophagitis (EoE) is a chronic disease marked by dense esophageal mucosal eosinophilia and inflammation leading to structural changes in the esophagus including stricture formation.¹ The clinical presentation varies with age, initially reflecting an inflammatory phenotype manifesting with symptoms such as abdominal pain, nausea and vomiting in children, and progressing to a fibrostenotic form signaled mostly by dysphagia in adolescence and adulthood.² The etiology is complex with contributions from a genetic predisposition,^3,4 an altered protease dependent epithelial barrier associated with loss of anti-protease,⁵ a likely altered microbiome⁶ and key environmental exposures.⁷ On a cellular basis, EoE is associated with a Th-2 mediated immune pathway and the contribution of key cytokines such as IL-4, −5 and −13, and cells in addition to eosinophils including T and B lymphocytes, epithelial and mast cells.⁸

The mainstay of medical therapy for EoE includes swallowed topical corticosteroids and proton pump inhibitors.⁹ Biologic therapies have also been shown to be effective in clinical trials.¹⁰ These therapies overall have favorable efficacy but are limited by concern for short and long term side effects. An effective therapy with minimal side effects would be ideal. Diet therapy has the potential to fulfill this goal. A large body of data support food antigen exclusion in EoE. For example, the pathogenesis of EoE closely fits the model of an allergic response to the exposure of food allergens,⁸ and an elemental diet, which is devoid of all food antigens, leads to a >90% response in adult and pediatric EoE patients.¹¹ Similarly, elimination of empiric food allergens such as milk, wheat, egg, soy, nuts and seafood, the six most common triggers of EoE, has been shown to induce histologic remission in about 70% of patients in observational and interventional studies.¹¹ As a result, a cornerstone of therapy for EoE has been empiric elimination diet based on the most common food triggers.

One of the controversies in diet-based therapy for EoE is whether to initially exclude a long list of possible food triggers or start with a smaller number of the most common food antigens that initiate esophageal eosinophilia.¹² Advantages to the former are the greater chance that histologic improvement will be achieved but at the cost of a markedly restricted diet when eliminating six or more commonly consumed foods. Avoiding fewer foods may facilitate better compliance but presumably lessens the chance for response. Although single and multicenter studies have been performed evaluating more and less restrictive diet exclusion therapy in EoE, there are no randomized trials in adults comparing diet strategies.^12–15 While one study used a modelling approach to address this question, patient data comparing diet therapies are lacking.¹⁶ It is also unclear if topical steroid therapy is a successful option for patients who fail to respond to food elimination. In addition, it has not yet been determined if response to diet therapy in adults is associated with molecular improvement in the esophagus such as those measured with the EoE Diagnostic Panel (EDP).³ The value of other biomarkers such as serum food specific immunoglobulins in predicting response to diet therapy also remains uncertain.¹⁷ Therefore, the aims of this study were to compare the efficacy of a one food elimination diet (1FED; animal milk) and six food elimination diet (6FED; animal milk, egg, wheat, soy, fish/shellfish, peanut/tree nuts) on clinical and histologic endpoints of EoE and determine the association of biomarkers (milk-specific immunoglobulins, T-cells and esophageal transcripts) to response to diet. We also aimed to assess if steroid therapy is effective in diet non-responders. The primary hypothesis was that 6FED would be superior to 1FED in achieving histologic remission.

Methods

Study design and participants

This was a multicenter, randomized, non-blinded, clinical trial conducted at 10 sites (appendix p 6) in the Consortium of Eosinophilic Gastrointestinal Disease Researchers (CEGIR) in the USA. This study protocol (https://clinicaltrials.gov/ProvidedDocs/67/NCT02778867/Prot_002.pdf) was performed in accordance with the principles of the Declaration of Helsinki. It was approved by Cincinnati Children’s Hospital Medical Center central Institutional Review Board (IRB) and IRBs at participating sites. All participants provided written informed consent.

Patients aged 18–60 years with a confirmed diagnosis of EoE as per consensus guidelines¹⁸ in place at the time of study initiation, which included non-response to a proton pump inhibitor (PPI) trial, were eligible for enrollment. Patients were included if they had histologically active disease, defined as ≥15 eosinophils per high-power field (eos/hpf) in at least one segment of the esophagus, and active symptoms of EoE. Key exclusions were treatment with topical swallowed steroids within 2 months of enrollment; pathological eosinophilia in the GI tract other than the esophagus; GI malabsorption disorders; and current avoidance of animal milk. Patients were not required to be consuming all other food antigens as some patients may have concomitant IgE mediated food allergy to these foods. Medications (PPI, allergy medications) prescribed prior to study entry were required to be maintained at the same dose. Full eligibility criteria are listed in the appendix (p 2).

Randomization and masking

Patients were centrally randomized 1:1 to 1FED or 6FED using a restricted randomization protocol. The allocation sequence was generated by the Data Management and Coordination Center (DMCC) at the University of South Florida using a SAS random number generator. The DMCC provided clinical trial data/project management support and was not involved in enrolling patients. The treatment allocation sequence was stored at the DMCC and was not available to enrolling investigators. Randomization was stratified by age (≤ 30 years old, > 30 years old), study site, and gender (male, female) with a block size of 4. At randomization, the DMCC generated an electronic notification (email) with the treatment assignment to the study coordinator and principal investigator at the study site. Study site investigators and staff enrolled participants and completed study assessments. As this was an open-label design, site investigators, staff, and participants were aware of treatment assignment after randomization. Research pathologists assessing biopsies were blinded to allocation. Lab investigators were blinded to treatment response until after samples were analyzed. Statisticians were not blinded to treatment assignment when conducting analysis.

Procedures

Patients were screened for eligibility and underwent an esophagogastroduodenoscopy (EGD) during a 12 week screening period. Historical biopsies were allowed if collected within 4 weeks of the screening period. After randomization in Phase 1 of the study, participants followed the 1FED (animal milk elimination) or 6FED (animal milk, egg, wheat, soy, fish/shellfish, peanut/tree nuts elimination) for 6 weeks. The length of the intervention was based on prior diet studies in adults who achieved histologic remission at 6 weeks.^13–15 Patients were counseled to avoid consumption of all foods containing these ingredients. All animal milk was eliminated due to potential cross-reactivity of cow, goat, and sheep milk proteins.¹⁹ Milk was chosen as the single food to avoid based on prior diet studies indicating milk was one of the most common food trigger in adults with EoE^14,15 and the most frequent single causative food antigen in a majority of children.¹¹ While wheat was identified as the most common trigger in one study,¹³ the available evidence at the time of study initiation also supported milk as the most common single causative food in adults.¹⁵ Patients assigned to 6FED eliminated wheat but were not required to follow a gluten-free diet (i.e., the elimination of barley and rye, in addition to wheat, was not required). After 6 weeks, an EGD with biopsy was performed to assess histologic response. Responders (<15 eos/hpf) completed the study. Treatment non-responders could opt to continue into Phase 2 where 1FED non-responders advanced to 6FED and 6FED non-responders received topical swallowed steroids (fluticasone propionate, 880 mcg twice daily).²⁰ Of note, participants returned to an unrestricted diet prior to beginning topical swallowed steroid therapy. After 6 weeks, a repeat endoscopy was performed and outcomes re-assessed. The study schematic is provided in the appendix (p 16).

Histologic, endoscopic, and patient-reported data were collected at baseline and end of Phase 1 and Phase 2. Biopsies were evaluated in a blinded fashion by central pathologists for histologic count (eos/hpf) and features using the EoE Histology Scoring System (EoEHSS).²¹ Endoscopic evaluations were completed using the EoE Endoscopic Reference Score (EREFS).²² Patient reported outcomes were collected using the EoE Activity Index (EESAI),²³ EoE quality of life (EoE-QOL-A),²⁴ and PROMIS Global Health Short Form questionnaires. Details of these assessments are in the appendix (p 3). At baseline and end of Phase 1, whole blood and serum were collected for assessment of T-cells and milk specific immunoglobulins, respectively and biopsies were collected for transcriptome analysis using the EoE Diagnostic Panel (EDP)³ (appendix p 4). Participants underwent skin prick testing for food allergens and those with negative prick test to cow’s milk underwent milk patch testing at baseline (appendix p 4). Adverse events were assessed approximately weekly.

Standardized instructions (appendix p 24 – 43) for following the elimination diets were created by Registered Dietitians with expertise in food elimination therapies and were used across all sites. Patients were instructed on reading food labels to identify allergens, dining out while on the diet, and avoiding cross contact with allergens. Substitution guides for eliminated allergens were also provided. A dietary questionnaire to help detect whether food antigens were successfully eliminated, 3-day food diaries completed during week 5 of the intervention, and dietitian input were used to determine compliance with the elimination diets. Participants who excluded all foods containing the allergens (milk or milk, egg, wheat, soy, fish, and nuts) and those who had only rare (less than once a week) exposure to excluded allergens were considered compliant with the assigned diet. The rare exposures could include one or more allergens. Standardized instructions for administering topical swallowed steroids (fluticasone propionate) during Phase 2 of the study can be found in the appendix (p 3).

Outcomes

The primary endpoint was the proportion of participants achieving histologic remission (< 15 eos/hpf)²⁵ post therapy at 6 weeks (Phase 1). Key secondary endpoints in Phase 1 included the proportion achieving complete remission (≤1 peak eos/hpf) and partial remission (≤10 and ≤6 peak eos/hpf) and change from baseline in peak eosinophil count, EoEHSS, EREFS, EESAI, EoE-QOL-A, and PROMIS Global Health scores. Other secondary endpoints included EDP total score; the association of baseline characteristics, skin prick and patch test results, and EDP with response to diets; correlation of patch test results with T-cells; and the association of T-cells and serum levels of milk specific immunoglobulins with response to 1FED. The association of T-cells with dietary response in all participants and the association of avoiding just wheat vs all gluten (wheat, barley, and rye) with response to 6FED were added as post-hoc endpoints. The Phase 2 secondary endpoint included the rate of histological remission (< 15 peak eos/hpf) of 6FED non-responders receiving fluticasone propionate and 1FED non-responders following 6FED. Post-hoc endpoints in Phase 2 included additional rates of remission (≤10, ≤6, and ≤1 peak eos/hpf) and change in peak eosinophil counts, EoEHSS, and EREFS. All histology outcomes, including peak eosinophil count for the primary outcome, were assessed by the consortium’s Central Review Pathology Committee.

Statistical Analysis

The sample size was based on the primary endpoint – the proportion of patients in histologic remission at the end of Phase 1. It was estimated that the remission rates for the 1FED and 6FED groups would be 45% and 70%, respectively.^11,13–15 A sample size of 60 patients per group (120 total patients) provides at least 80% power to detect a difference in rates between groups of 25% at the two-sided $\text{α}=\mathrm{0·05}$ using PASS¹² (NCSS, Kaysville, Utah). To ensure sufficient power even after dropout, the targeted sample size was increased to 136 total participants which would allow 15% drop out. Near the end of the study, the dropout rate was evaluated and found to be much lower than 15%, so recruitment was stopped at 129 participants (1FED, n=67 and 6FED, n=62) with 5 participants not completing the study (1FED, n=65 and 6FED, n=59).

The analysis population for the primary and key secondary endpoints in Phase 1 was the intention-to-treat population which included all patients that completed the enrollment visit and were randomized into the study. Enrolled patients that withdrew from the study were imputed as treatment failures. Data missing at end of treatment for continuous outcomes were imputed using the last measure carried forward. A pre-specified complete cases analysis for primary and key secondary endpoints was conducted as sensitivity analysis to compare to the intention-to-treat population. Complete cases included patients that completed Phase 1 with observed data at both time points (baseline and week 6) for the outcome of interest. The analysis population for secondary and post-hoc endpoints in Phase 2 was the intention-to-treat (ITT) population which included all patients that opted into Phase 2. Patients that withdrew from Phase 2 were imputed as treatment failures and data missing at end of treatment for continuous outcomes were imputed using the last measure carried forward. The association of baseline characteristics and skin prick and patch test results with response to diets was also conducted in the ITT population. Molecular biomarker analysis (EDP, T-cells, and serum immunoglobulins) was completed on the subset of patients with intact samples. Safety analysis was conducted on data collected from the ITT populations (all patients who were randomized into Phase 1 and all patients opting into Phase 2).

Baseline characteristics were summarized for the two treatment groups using mean (standard deviation) or median (IQR) for continuous variables and frequency (percentages) for categorical variables. Comparisons between groups were made using the two-sample t-test (data normally distributed) for continuous variables and the Fisher’s exact test for categorical variables. Comparisons within groups in Phase 2 were assessed using paired t-test. Distributions were assessed using SAS univariate procedure evaluating Kolmogorov-Smirnov, Cramer von Mises, and Anderson-Darling. Log transformation was used in the analysis for data following log normal distribution. To assess associations of allergy testing, EDP score, T-cell analysis, and avoiding just wheat vs all gluten (wheat, barley, and rye)with response to treatment, comparisons between histologic responders and non-responders were made using two-sample t-test for continuous variables and Fisher’s exact test for categorical variables. Logistic regression was used to test the association of age, gender, race, baseline esophageal peak count, treatment, and atopy with week 6 remission. Statistical significance was indicated at the nominal two sided $\text{α}=\mathrm{0·05}$. For continuous variables, 95% confidence intervals for means and mean differences were calculated using t-distribution; for proportions and proportion differences, calculations were based on Wald confidence intervals. Statistical analyses were performed with SAS software, version 9.4 (SAS Institute, Inc.; Cary, NC). A data and safety monitoring board, independent of trial investigators, reviewed all safety data and the conduct of the trial annually. This study was registered with clinicaltrials.gov, NCT02778867.

Role of the funding source

The funder had a role in study design via a National Institutes of Health appointed medical monitor and research nurse. The funder did not have a role in data collection, data analysis, data interpretation, or writing of the report.

Results

Patients were enrolled between May 23, 2016 and March 6, 2019. Of the 143 patients screened for the trial, 129 completed enrollment and were randomized. Of these, 67 patients were allocated to 1FED and 62 patients to 6FED. Recruitment was stopped at 129 patients (with 124 completing the study) as the estimated sample size to meet the target power was achieved. Allocation, discontinuation, and analyses are summarized in figure 1. The mean age was 37·0 years (SD 10·3). Gender was distributed approximately evenly, while 124 (96%) patients were white. Patient demographics and baseline characteristics were comparable between the two treatment arms (table 1) except for the median peak esophageal eosinophil count which was higher in 1FED compared to 6FED.

Figure 1.

Trial profile

Table 1:

Baseline characteristics (intention-to-treat population)

	1FED (n = 67)	6FED (n = 62)
Age, years, mean (SD)	36·4 (10·2)	37·8 (10·4)
≤30 years	18 (27%)	15 (24%)
Gender
Male	37 (55%)	33 (53%)
Female	30 (45%)	29 (47%)
White	63 (95%)	61 (98%)
EoE history
Years since EoE diagnosis, median (IQR)	0·3 (0·1–2·1)	0·2 (0·1–1·8)
Prior dilation	29 (43%)	26 (42%)
Concomitant PPI treatment	41 (62%)	43 (69%)
Symptoms
Food impaction	32 (48%)	30 (48%)
Dysphagia	57 (85%)	55 (89%)
Chest pain	31 (46%)	17 (27%)
Heartburn	40 (60%)	34 (55%)
Abdominal Pain	21 (31%)	14 (23%)
Symptom and quality of life scores, mean (SD)
EEsAI	29·3 (19·3)	30·1 (18·2)
EoE QoL-A	68·9 (15·5)	64·2 (17·0)
Atopic conditions
Asthma	11 (16%)	17 (27%)
Food allergies	16 (24%)	8 (13%)
Eczema	2 (3%)	3 (5%)
Allergic rhinitis/sinusitis	22 (33%)	28 (45%)
Urticaria	0 (0%)	1 (2%)
Oral allergy syndrome	2 (3%)	0 (0%)
Endoscopic findings present
Edema	43 (74%)	39(80%)
Rings	41 (71%)	42 (86%)
Exudates	33 (57%)	33 (67%)
Furrows	46 (79%)	40 (82%)
Stricture	20 (34%)	15 (31%)
Stricture Diameter, mm	15·0 (2·4)	14·7 (3·0)
EREFS*, mean (SD)
Total score	3·7 (1·8)	4·2 (1·8)
Inflammatory score	2·4 ( 1·2)	2·5 (1·2)
Fibrostenotic score	1·3 (1·0)	1·7 (1·0)
Histologic features
Peak eosinophil count, eos/hpf, median (IQR)	58 (25–79)	38 (25–61)
EoEHSS**, mean (SD)
EoEHSS total score	0·83 (0·26)	0·81 (0·25)
EoEHSS grade	0·43 (0·16)	0·42 (0·14)
EoEHSS stage	0·39 (0·13)	0·38 (0·13)
Inflammatory score	0·62 (0·34)	0·59 (0·34)
Architectural score	1·09 (0·26)	1·10 (0·23)

^*Maximum score between distal and proximal esophagus with 1FED (n=58), 6FED (n=49).

^**Maximum score among distal, mid, and proximal esophagus. Mean (SD), median (IQR) and frequency (proportion) are presented according to distribution of the data.

The proportion of patients who achieved the primary endpoint of histologic remission (< 15 eos/hpf) after 6 weeks of diet therapy was 23 (34%) of 67 in the 1FED group vs 25 (40%) of 62 in the 6FED group, a difference of 6% [95% CI –11 to 23] that was not statistically significant ($\text{p}=\mathrm{0·58}$). In contrast, the proportion achieving complete remission (≤ 1 eos/hpf) was 12 (19%) in 6FED compared to 4 (6%) in 1FED (difference 13% [2 to 25]; $\text{p}=\mathrm{0·031}$) (table 2).

Table 2:

Proportion of patients in histologic remission (intention-to-treat population)

	1FED (n = 67)		6FED (n = 62)		Difference**% (95% CI)	p-value
	n	% (95% CI)	n	% (95% CI)
Histologic Response
<15 eos/hpf*	23	34 (23 to 46)	25	40 (28 to 53)	6 (−11 to 23)	0·58
≤ 10 eos/hpf	20	30 (19 to 41)	23	37 (25 to 49)	7 (−9 to 24)	0·46
≤ 6 eos/hpf	12	18 (9 to 27)	20	32 (21 to 44)	14 (−0 to 29)	0·07
≤ 1 eos/hpf	4	6 (0 to 12)	12	19 (10 to 29)	13 (2 to 25)	0·031

Data are compared between groups using Fisher’s exact test.

^*Primary endpoint

^**6FED vs 1FED

Patients following both 1FED and 6FED had improvement compared to baseline for maximum esophageal eosinophil count and the magnitude of the reductions were similar in 1FED and 6FED (table 3; appendix p 17).

Table 3:

Histologic, endoscopic, symptom and quality of life response to 1FED and 6FED (intention-to-treat population)

	1FED (n = 67)			6FED (n = 62)			Difference (95% CI)	P-Value†
	Baseline	Week 6	Change	Baseline	Week 6	Change
Max peak count	50·3 (42·2 to 60·0)	20·8 (15·0 to 28·9)	0·41 (0·29 to 0·57)	38·4 (32·8 to 44·9)	10·9 (7·3 to 16·5)	0·29 (0·20 to 0·43)	0·72 (0·43 to 1·20)	0·21
EoEHSS	0·83 (0·77 to 0·90)	0·68 (0·60 to 0·76)	$-\mathrm{0·15}$ ($-\mathrm{0·25}$ to $-\mathrm{0·06}$)	0·81 (0·74 to 0·88)	0·58 (0·50 to 0·65)	$-\mathrm{0·23}$ ($-\mathrm{0·32}$ to $-\mathrm{0·14}$)	$-\mathrm{0·08}$ ($-\mathrm{0·21}$ to 0·05)	0·23
EoEHSS grade	0·43 (0·39 to 0·47)	0·34 (0·30 to 0·38)	$-\mathrm{0·09}$ ($-\mathrm{0·14}$ to $-\mathrm{0·04}$)	0·42 (0·39 to 0·46)	0·30 (0·26 to 0·33)	$-\mathrm{0·13}$ ($-\mathrm{0·17}$ to $-\mathrm{0·08}$)	$-\mathrm{0·04}$ ($-\mathrm{0·11}$ to 0·03)	0·26
EoEHSS stage	0·39 (0·36 to 0·42)	0·33 (0·29 to 0·37)	$-\mathrm{0·06}$ ($-\mathrm{0·11}$ to $-\mathrm{0·01}$)	0·39 (0·35 to 0·42)	0·28 (0·24 to 0·32)	$-\mathrm{0·11}$ ($-\mathrm{0·15}$ to $-\mathrm{0·06}$)	$-\mathrm{0·04}$ ($-\mathrm{0·11}$ to 0·02)	0·21
EREFS total	3·7 (3·3 to 4·2)	3·0 (2·5 to 3·4)	$-\mathrm{0·6}$ ($-\mathrm{1·0}$ to $-\mathrm{0·2}$)	4·2 (3·7 to 4·7)	2·8 (2·3 to 3·3)	$-\mathrm{1·0}$ ($-\mathrm{1·5}$ to $-\mathrm{0·4}$)	$-\mathrm{0·4}$ ($-\mathrm{1·1}$ to 0·3)	0·28
EEsAI	29·3 (24·5 to 34·2)	26·1 (21·3 to 30·9)	$-\mathrm{3·0}$ ($-\mathrm{7·2}$ to 1·2)	30·1 (25·4 to 34·7)	21·7 (17·5 to 25·9)	$-\mathrm{8·2}$ ($-\mathrm{12·6}$ to $-\mathrm{3·8}$)	$-\mathrm{5·2}$ ($-\mathrm{11·2}$ to 0·8)	0·09
EoE-QoL-A	68·9 (65·0 to 72·7)	67·1 (62·7 to 71·5)	$-\mathrm{0·9}$ ($-\mathrm{3·5}$ to 1·6)	64·2 (59·9 to 68·6)	63·9 (59·7 to 68·1)	$-\mathrm{0·3}$ ($-\mathrm{3·3}$ to 2·7)	0·6 ($-\mathrm{3·3}$ to 4·5)	0·76
PROMIS physical	49·4 (48·0 to 50·9)	50·8 (49·3 to 52·2)	1·3 (0·4 to 2·2)	50·6 (48·8 to 52·4)	52·2 (50·7 to 53·8)	1·6 (0·5 to 2·8)	0·4 ($-\mathrm{1·0}$ to 1·7)	0·61
PROMIS mental	50·0 (48·4 to 51·5)	51·6 (50·0 to 53·2)	1·5 (0·4 to 2·7)	51·7 (49·8 to 53·5)	52·5 (50·6 to 54·4)	1·1 ($-\mathrm{0·3}$ to 2·5)	$-\mathrm{0·4}$ ($-\mathrm{2·2}$ to 1·3)	0·62

EoEHSS and EREFS score are the maximum score among evaluated esophageal sections (distal, mid, and proximal). Max peak count is the highest eosinophil count among distal, mid, and proximal esophagus and is shown as geometric mean (95% CI) with change as the geometric mean ratio. Other variables are shown as mean (95% CI).

^†p-value from two-sample t-test

Baseline mean EoEHSS total scores were 0·83 [95% CI 0·77 to 0·90] and 0·81 [0·74 to 0·88] in the 1FED and 6FED groups, respectively. Scores were improved from baseline in both groups on average decreasing $-\mathrm{0·15}$ [95% CI $-\mathrm{0·25}$ to $-\mathrm{0·06}$] in 1FED compared to $-\mathrm{0·23}$ [$-\mathrm{0·32}$ to $-\mathrm{0·14}$] in 6FED (difference 0·08 [$-\mathrm{0·21}$ to 0·05], $\text{p}=\mathrm{0·23}$) (table 3). Improvement in EoEHSS inflammatory feature sub-scores in 1FED closely matched 6FED ($-\mathrm{0·20}$ [$-\mathrm{0·31}$ to $-\mathrm{0·08}$] vs $-\mathrm{0·24}$ [$-\mathrm{0·35}$ to $-\mathrm{0·13}$]) with only a small difference ($-\mathrm{0·05}$ [$-\mathrm{0·20}$ to 0·11]; $\text{p}=\mathrm{0·56}$) between scores. Mean reductions in EoEHSS histological architectural feature sub-scores in 6FED were more than twice those in 1FED ($-\mathrm{0·23}$ [$-\mathrm{0·32}$ to $-\mathrm{0·13}$] vs $-\mathrm{0·10}$ [$-\mathrm{0·21}$ to 0·0]) but this difference ($-\mathrm{0·13}$ [$-\mathrm{0·27}$ to 0·01]; p=0.08) did not reach statistical significance (appendix p 18).

Total maximum EREFS scores decreased from baseline in both groups improving by $-\mathrm{0·6}$ [95% CI $-\mathrm{1·0}$ to $-\mathrm{0·2}$] in 1FED compared to $-\mathrm{1·0}$ [$-\mathrm{1·5}$ to $-\mathrm{0·4}$] in 6FED (difference $-\mathrm{0·4}$ [$-\mathrm{1·1}$ to 0·3]; $\text{p}=\mathrm{0·28}$) (table 3). Mean reductions in EREFS fibrostenotic feature sub-scores were $-\mathrm{0·09}$ [$-\mathrm{0·28}$ to 0·11] in 1FED and $-\mathrm{0·35}$ [$-\mathrm{0·55}$ to $-\mathrm{0·14}$] in 6FED (difference $-\mathrm{0·26}$ [$-\mathrm{0·54}$ to 0·02]; $\text{p}=\mathrm{0·07}$). Reductions in EREFS inflammatory abnormalities were comparable between 1FED and 6FED ($-\mathrm{0·52}$ [$-\mathrm{0·86}$ to $-\mathrm{0·17}$] vs $-\mathrm{0·63}$ [$-\mathrm{1·09}$ to $-\mathrm{0·17}$]) with a difference of $-\mathrm{0·12}$ [$-\mathrm{0·67}$ to 0·44], $\text{p}=\mathrm{0·68}$ (appendix p 18). The appendix (p 7) summarizes changes in individual endoscopic feature scores.

Mean total EEsAI score changed by $-\mathrm{3·0}$ points [95% CI $-\mathrm{7·2}$ to 1·2] in 1FED compared to $-\mathrm{8·2}$ points [12·6 to $-\mathrm{3·8}$] in 6FED (difference $-\mathrm{5·2}$ [$-\mathrm{11·2}$ to 0·8]; $\text{p}=\mathrm{0·09}$) (table 3). Sub-scores measuring pain associated with swallowing decreased by a greater extent in 6FED compared to 1FED ($-\mathrm{2·4}$ [$-\mathrm{3·8}$ to $-\mathrm{1·0}$] vs 0·0 [$-\mathrm{1·6}$ to 1·6]; $\text{p}=\mathrm{0·025}$). Changes for other sub-scores however were not statistically different between groups (appendix p 8); however, patients following 6FED showed a trend for improvement in frequency of trouble swallowing compared to 1FED (difference $-\mathrm{3·4}$ [$-\mathrm{6·9}$ to 0·1]; $\text{p}=\mathrm{0·06}$).

Post-treatment mean EoE-QoL-A total scores were changed by $-\mathrm{0·9}$ points [95% CI $-\mathrm{3·5}$ to 1·6] in 1FED and $-\mathrm{0·3}$ points [$-\mathrm{3·3}$ to 2·7] in 6FED (difference 0·6 [$-\mathrm{3·3}$ to 4·5], $\text{p}=\mathrm{0·76}$). The total scores did not meaningfully change within either arm compared to baseline (table 3). Eating/diet impact sub-scores worsened in both groups but the small changes were not different between groups ($\text{p}=\mathrm{0·58}$) (appendix p 9). There were small improvements in mean PROMIS Global Health short form mental T-scores (1FED, 1·5 [95% CI 0·4 to 2·7]; 6FED, 1·1 [$-\mathrm{0·3}$ to 2·5]) and physical T-scores (1FED, 1·3 [0·4 to 2·2]; 6FED, 1·6 [0·5 to 2·8]) after treatment, but differences between 1FED vs 6FED were small (table 3).

Histologic responders (< 15 eos/hpf) to diet showed normalization of gene expression and significantly higher mean EDP score compared to non-responders after diet therapy (figure 2A). EDP scores did not vary by diet (figure 2B) and post-treatment transcriptomes were qualitatively similar in 1FED and 6FED (figure 2C). When comparing responders and non-responders to either diet, there was a marked bidirectional change in the expression of 75 genes in diet responders compared to a change in 4 genes in diet non-responders (appendix p 19). An evaluation of the value of baseline EDP results in predicting response to diet was not undertaken due to the small number of intact baseline biopsies.

Figure 2.

Analysis of gene expression in esophageal biopsies pre- and post-diet therapy (A) EDP score in histologic responders and non-responders. (B) EDP score in histologic responders and non-responders by diet therapy. Scores above 333 (dotted line) represent normalization of the transcriptome. (C) Comparison of esophageal transcriptomes by response to 1FED and 6FED. Data are presented as means with 95% CI. P-values are calculated with the two sample t-test. EDP, eosinophilic esophagitis diagnostic panel. Baseline, n=21; 1FED Non-responder, n=10; 6FED Non-responder, n=10; 6FED Non-responder, n=10; 1FED Responder, n=10; 6FED Responder, n=10.

Among patients with evaluable diet records completing the study, 63 (98%) of 64 and 57 (97%) of 59 adhered to 1FED and 6FED, respectively. Whether one patient adhered to 1FED was indeterminable.

A complete cases analysis (patients completing the study without missing data) was conducted as sensitivity analysis for Phase 1. Among these patients, 23 (35%) of 65 in the 1FED group compared to 25 (42%) of 59 in the 6FED group were in histologic remission (difference 7% [95% CI, −10 to 24], $\text{p}=\mathrm{0·46}$). Sensitivity analysis for the key secondary outcomes also showed no important differences compared to the intention-to-treat population (appendix pp 10–11).

Among 1FED and 6FED histologic non-responders, 21 patients opted to follow 6FED and 11 patients opted to receive swallowed fluticasone propionate respectively in Phase 2. One patient following 6FED and 2 patients on swallowed fluticasone dropped out before completing Phase 2 (figure 1). After 6 weeks, 9 (43%) of 21 1FED non-responders on 6FED and 9 (82%) of 11 6FED non-responders treated with swallowed fluticasone were in histologic remission (< 15 eos/hpf). A total of 6 (29%) on 6FED and 6 (55%) on fluticasone achieved complete remission (≤ 1 eos/hpf). All levels of remission are shown in the appendix (p 12). Post-hoc analysis showed mean EoEHSS and EREFS score significantly decreased from Week 6 in patients on 6FED and those receiving swallowed fluticasone therapy (table 4).

Table 4:

Histologic and endoscopic response in Phase 2 (intention-to-treat population)

	6 Food Elimination Diet (n=21)
	Week 6	EOT	Change	P-value*
Max peak count	47·7 (34·5 to 66·1)	10·7 (4·6 to 25·2)	0·23 (0·12 to 0·43)	0·0001
EoEHSS	0·81 (0·69 to 0·94)	0·60 (0·43 to 0·77)	$-\mathrm{0·22}$ ($-\mathrm{0·35}$ to $-\mathrm{0·08}$)	0·0028
EREFS total	3·0 (2·1 to 3·8)	2·2 (1·5 to 2·8)	$-\mathrm{0·8}$ ($-\mathrm{1·5}$ to $-\mathrm{0·1}$)	0·035
	Topical Swallowed Steroids (n=11)
	Week 6	EOT	Change	P-value*
Max peak count	40·4 (28·0 to 58·3)	3·0 (1·0 to 9·6)	0·08 (0·03 to 0·20)	0·0002
EoEHSS	0·82 (0·70 to 0·94)	0·40 (0·18 to 0·61)	$-\mathrm{0·42}$ ($-\mathrm{0·61}$ to $-\mathrm{0·23}$)	0·0007
EREFS total	2·8 (1·7 to 3·9)	1·7 (0·7 to 2·8)	$-\mathrm{1·1}$ ($-\mathrm{2·2}$ to $-\mathrm{0·03}$)	0·045

Max peak count is shown as geometric mean (95% CI) with change as the geometric mean ratio. Other variables are shown as mean (95% CI).

^*p-value from paired t test; EOT, end of treatment (patients followed therapy for 6 weeks).

Gender, age, race, treatment, and baseline peak eosinophil count were not associated with remission (data not shown). When 1FED and 6FED were combined, no significant atopy effect was observed. When stratified by treatment, 16 (62%) of 26 non-atopic patients compared to 9 (25%) of 36 atopic patients were in remission after 6FED ($\text{p}=\mathrm{0·008}$). This association was not seen in patients following 1FED ($\text{p}=\mathrm{0·44}$).

For baseline allergy skin prick testing (SPT) for eliminated foods, neither positive nor negative prick tests to milk were associated with histologic response in the 1FED group. Among patients in 6FED, 7 (70%) of 10 patients with positive SPT to egg were in remission compared to 6 (27%) of 22 with negative SPT to egg ($\text{p}=\mathrm{0·049}$). SPT results are shown in the appendix (p 13). Of the 67 patients who underwent milk patch testing, only one patient had a borderline positive result thus no additional analyses was performed on this outcome.

At baseline, serum levels of cow’s milk specific IgG4 proteins Bos d 4 and Bos d 5 were significantly higher in 1FED histological responders compared to non-responders (p=0.03, p=0.04 respectively; appendix p 20). After 1FED, IgG4 levels to Bos d 4, Bos d 5, and Bos d 8 decreased from baseline in histologic responders and non-responders (appendix p 20). Reductions were 15 – 20% greater in responders than in non-responders but the differences were not statistically significant (appendix p 21). Baseline IgE levels to cow’s milk proteins were similarly low in responders and non-responders (appendix p 20) and levels were not changed from baseline after 1FED (appendix p 21). Ratios of IgG4 to IgE for milk components were high – ranging from 9,000 to 37,000:1 – and were not different between 1FED responders and non-responders at baseline or after diet therapy (appendix p 14).

Ex vivo expression of CD4⁺CD154⁺ T cells and Th-2 cytokine expression (IL-4, IL-5, and IL-13) in CD4⁺CD154⁺ cells were similar in histologic responders and non-responders to 1FED or 6FED prior to treatment in our post-hoc analysis (appendix p 22). After 6 weeks of diet therapy, IL-4 expression in CD4⁺CD154⁺ cells was lower in responders compared to non-responders to diet therapy ($\text{p}=\mathrm{0·041}$). In a subset of patients (n=12) with both pre- and post-measures, expression of CD4⁺CD154⁺ T cells and Th-2 cytokine expression in CD4⁺CD154⁺ cells were improved from baseline only in responders (appendix p 23). Given the absence of positive patch test results, patch testing was not correlated with circulating T cells. Due to small sample size, an assessment of T cells stratified by diet was not undertaken.

As post-hoc analysis, the effect of eliminating only wheat compared to gluten (i.e., wheat, barley, and rye) on histologic remission was assessed. Among patients following 6FED, none reported consuming rye. A total of 5 (29%) of 17 patients who reported consuming barley and 20 (48%) of 42 who avoided barley were in remission ($\text{p}=\mathrm{0·25}$). The primary source of barley intake was beer.

No adverse events were reported at a frequency greater than 5% in patients in either diet group in Phase 1 (appendix p 15). However, mean weight loss in 6FED was significantly higher than in the 1FED group ($-\mathrm{2·2}$ ± 2·5 kg vs. $-\mathrm{1·1}$ ± 2·6 kg; $\text{p}=\mathrm{0·027}$). During Phase 2, one participant who had normal serum cortisol levels (14 mcg/dL) before treatment experienced a low serum cortisol level (<1 mcg/dL) after 6 weeks of swallowed fluticasone, but did not have clinical symptoms of adrenal insufficiency. Repeat levels 10 days after stopping fluticasone were normal (16 mcg/dL). A second patient treated with fluticasone experienced a serious adverse event (suicidal ideation) which was assessed as unrelated to therapy.

Discussion

In this study, we failed to demonstrate that the standard six food elimination diet was superior to the dietary elimination of animal milk alone in the treatment of adults with EoE. Both diets were similarly effective across multiple metrics. Notably, our findings were derived from the first multicenter randomized trial of dietary elimination therapy in adults with EoE and similar results were found in both intention-to-treat and complete cases analyses in a series of primary and secondary outcomes which consistently demonstrated efficacy of 1FED. Among followers of both diets, we found comparable improvements in endoscopic appearance (EREFS), histologic features (EoEHSS), and peak eosinophil counts. Response to both diets reversed the molecular signature associated with active EoE. Symptoms improved from baseline in 6FED but were not statistically different than 1FED and did not lead to improvements in quality of life. Encouragingly, nearly half of patients who failed to respond to 1FED attained histologic remission after therapy with the more restrictive 6FED. Further, more than 80% of the non-responders to 6FED achieved remission with swallowed fluticasone therapy. Taken together, this study shows that 1FED is a reasonable first-line diet therapy option in adults with EoE and that there are effective therapies for those who fail limited and restrictive diet therapy.

Diet exclusion therapy has been a cornerstone of treatment for EoE since attenuation of esophageal eosinophilia was demonstrated with an elemental diet.²⁶ Following the discovery that food antigens were associated with esophageal injury, researchers directed attention to identifying common food triggers and evaluated the effectiveness of excluding six common triggers (milk, egg, wheat, soy, fish, and nuts) on esophageal eosinophilia in children and adults.¹¹ Due to the high response rate (~70%) found in these studies, the 6FED became a frequently employed dietary approach to manage EoE. In more recent years, less restrictive diets that remove one to four of the most common food antigens were evaluated in non-randomized studies and shown to improve esophageal eosinophilia in a meaningful number of adults and children.^12,15,27,28 Our study extends two decades of research by comparing a minimally restrictive elimination diet (milk) to the highly restrictive 6FED in a multicenter randomized trial of adults with EoE using newly developed validated instruments to comprehensively determine the relative effectiveness of these diets, the mechanisms involved, and the value of skin testing and circulating biomarkers in predicting outcome.

In this largest randomized trial of dietary therapy in EoE to date, our findings were unexpected. While studies of elimination diets for children with EoE suggest that 30 – 60% may achieve remission with 1FED,^{11,12,27, 29} a recent study in European (Spanish, Italian) adults with EoE found that less than 20% have milk as the only EoE trigger.¹² Differences in immune response by geographic region may explain why 1FED remission rates are different.³⁰ Non-randomized prospective studies in adults with EoE suggest ~70% may achieve remission with 6FED,^12–14 a rate higher than found in the present study. Two of the previous studies^12,14 also eliminated legumes and gluten which could account for the higher response rate. In the present study, the remission rate was lower in patients who consumed gluten (29%) than in those who did not (48%), although the difference was not significant. Adherence to 6FED was high; however, rare exposure to antigens (known or unknown) may have contributed to the lower response rate. In one study, patients with known antigen exposure underwent a food wash-out period prior to the post intervention endoscopy which resulted in higher response rates than in our study.¹³ The only prospective U.S. study evaluating 6FED in adults was conducted at a single center.¹³ Because the present study included 10 sites across the U.S., it likely included a broader (perhaps more representative) range of the adult EoE population. Further, recruits who suspected egg, wheat, soy, fish or nuts were potential EoE triggers were randomly assigned to 1FED or 6FED in the present study which reduced selection bias that may have contributed to the higher response rates in the non-randomized studies. Finally, while the length of intervention in this study matched other 6FED studies,^13,14 a recent study found 6 weeks of diet therapy may not be sufficient to obtain histologic improvement for a subset of EoE patients.³¹ Histologic non-responders with significant improvement in symptoms, endoscopic appearance, and eosinophil counts after 6 weeks of therapy achieved histologic remission when diet therapy was extended in this study.³¹ Improvement in these three features occurred in approximately 10% of our 6FED non-responders (n=4) suggesting a longer intervention may have only marginally improved our overall remission rate under this hypothesis. The present study results should temper the previous response rates quoted to patients for 6FED from non-controlled studies.

Using validated tools (EREFS and EoEHSS) developed to assess endoscopic and histologic abnormalities in EoE, we found both diets improved abnormal endoscopic features and the extent and severity of histologic findings. Other studies of 1FED and 6FED have relied on subjective endoscopic review¹³ or evaluation of the presence of endoscopic abnormalities.^14,27 Additionally, this is the first diet elimination study to use EoEHSS in adults. The granular review of endoscopic and histologic features in the present study revealed mostly comparable improvement, particularly in the inflammatory features, in 1FED and 6FED. However, improvement in scores for histologic and endoscopic features associated with remodeling were larger in 6FED than 1FED although the differences were not significant. It is likely that the elimination of multiple foods in 6FED increased the chance that at least one trigger was eliminated which may have yielded some benefit even in the absence of remission.

The 6FED served as an effective therapy for almost half of 1FED failures in the presented study. Additionally, topical swallowed steroids were effective for the majority of 6FED failures. Molina-Infante et al. first found step-up diet therapies were effective in achieving remission.¹² The study presented herein builds on this finding to show that for patients who do not initially respond to a milk elimination diet, escalating to 6FED not only reduces esophageal eosinophil counts but also significantly improves histologic and endoscopic abnormalities. Given the presumably long-term sustainability of less restrictive diets and the need for fewer endoscopies compared to more restrictive diets,¹² starting with fewer foods and escalating as necessary may be a worthwhile strategy for many patients. Such a strategy may also reduce cases of sub-optimal nutritional intake observed with more restrictive diets, although even patients following a milk only elimination diet should be monitored for optimal intake of calcium and vitamin D. There are few data examining a change from dietary therapy to topical steroids, and it is reassuring to have data to support this strategy in adults although the sample size was small to be firm about this finding.

We evaluated clinical and laboratory characteristics of patients for their association with response to therapy. Non-atopic patients were more likely to achieve remission on 6FED in this study. This contradicts previous food elimination studies in adults^12–14 where atopic background did not predict response to diet and further replication is needed. Elevated serum IgG4 levels to major cow’s milk proteins were associated with histologic response to 1FED. IgG4 to $\text{α}$-lactalbumin [Bos d 4] and $\text{β}$-lactoglobulin [Bos d 5] were higher in 1FED responders than non-responders suggesting milk components could potentially serve as a predictive marker of milk reactivity in EoE. In contrast, IgE antibodies to milk proteins were low at baseline and unchanged after 1FED. These findings are consistent with prior data supporting a strong role for IgG4, but not IgE, in EoE.¹⁷ Studies have found, for example, increased levels of tissue and serum IgG4 to milk in patients with EoE. ^17,32 In contrast, failure of patients to respond to omalizumab, an anti-IgE agent,³² and lack of predictive value of skin prick testing to identify EoE antigens ³³ indicate EoE is largely independent of IgE. Indeed, in our study, skin prick and patch tests were also not useful in predicting response to diet. While our study supports not assessing skin tests in EoE patients, immunological or allergy testing should still be considered for the management of the other atopic phenomena that are sometimes associated such as asthma, rhinitis or dermatitis, where treatment of these conditions can be aided by such testing. Finally, the small dataset in the present study suggest low or markedly decreasing levels of activated T-cells and Th-2 cytokine expression in activated T-cells should be further investigated as a potential blood biomarker of response to treatment in EoE.

This is the first study in adults to evaluate the transcriptome response to diet. In responders to diet therapy, there was a pronounced reversal (normalization) in the expression of 75 genes associated with EoE. These findings demonstrate that diet therapy has a profound effect on reversing the pathobiology of EoE.

Our study had limitations. First, excluding patients who respond to PPIs may have excluded a subset of EoE. However, recent studies suggest that PPI responsive patients respond to diet therapy^34,35 similar to patients in this study. We also cannot completely exclude a synergistic effect of continued PPIs with diet in affecting the outcomes. Second, as in any clinical study, whether patients completely adhered to dietary exclusions is not certain despite our efforts to detect non-compliance. Additionally, we cannot be certain that rare antigen exposure (in our definition of compliance) did not influence efficacy results. Third, symptom response may have been biased in this unblinded study. Fourth, in the event the dropouts were related to poor treatment outcome, results might be biased. Fifth, while there was a lack of statistical difference between 1FED vs 6FED in nearly all secondary endpoints, the study was not powered for secondary endpoints so we cannot be certain that differences do not exist, especially since 6FED trended to be improved (appendix p 11). Finally, median peak eosinophil count at baseline was higher in 1FED compared to 6FED. If a factor, this likely should have made the response to 1FED less robust therefore further supporting the efficacy of this diet when compared to 6FED. The major strength of this study was the multicenter randomized design with adults recruited from 10 sites throughout the U.S., and the rigorous methodology, including regulatory oversight by the NIH, used for the study. Consequently, selection bias that marred other studies may have been minimized and the presented findings may be more generalizable to the adult EoE population. The study also included EoE disease-specific validated instruments for measuring symptom, histologic, and endoscopic response. Further, diet education and skin prick/patch testing were standardized across all sites.

Despite the data reported herein, important gaps in knowledge related to diet therapy in EoE remain. The optimal duration of diet therapy to maximize the chance of achieving remission, for example, remains uncertain. Additionally, despite decades of diet therapy research, there are few data describing nutritional and psychological effects of elimination diets, especially with long-term use. Finally, although it is recognized that food triggers may vary by geographical location, the extent of these differences is uncertain, and this current study’s findings may be limited to the USA. Future diet studies – whether interventional or observational – should aim to answer these questions and extend to broader geographical areas.

In summary, the findings reported herein show that 1FED and 6FED are similarly effective at achieving remission and improving multiple metrics of response in this first comparison randomized trial of dietary elimination in adults with EoE. The data also show that 6FED serves as an effective therapy for almost half of 1FED failures, and steroids are effective for most 6FED failures. The study not only demonstrates the benefit of diet therapy as assessed by histologic and endoscopic metrics, but also demonstrates that diet therapy reverses the underlying disease pathogenesis as demonstrated by its ability to correct the EoE associated molecular transcriptome. The findings also direct attention away from the value of skin prick and patch testing and call attention to the association of food specific IgG4 with remission but not IgE. Thus, our study indicates that elimination of food and beverage containing animal milk may be an acceptable initial dietary treatment choice, especially if a step-up dietary strategy is being employed, and further informs clinicians about effective therapies for diet treatment failure.