Confocal Laser Endomicroscopy With Duodenal Food Challenge Does Not Predict Triggering Food Items in Eosinophilic Esophagitis: Results From a Pilot Trial

Abstract

Background/Aims

Targeted elimination of triggering food items is an attractive treatment option in eosinophilic esophagitis (EoE). However, no reliable biomarkers exist to identify culprit foods. In a previous uncontrolled study, 6/9 responded to duodenal food challenge (DFC), visualized by confocal laser endomicroscopy (CLE). This study aims to assess efficacy of a targeted elimination diet based on identification of nutrients that trigger acute mucosal alterations in the duodenum compared to a sham diet.

Methods

Randomized double-blind cross-over study in EoE. After baseline endoscopy, patients with peak eosinophil count of > 15 eosinophils/high power field underwent CLE. Following baseline assessment, a solution of foods (grains/milk/egg/soy/fish/nuts) was sprayed on the duodenal mucosa, with repeat imaging thereafter. Mucosa was judged as altered in case of fluorescein leakage and particle shedding. Based on the results, patients underwent a blinded randomized 6-week cross-over elimination diet with follow-up endoscopy. Exploratory measurements were done for small intestinal permeability, eosinophil- and mast cell activation.

Results

Seven patients completed the study. At CLE, 1 patient displayed baseline alterations, 5 reacted to 1 to 2 nutrients in 2 sessions (milk = 4, grains = 2, nuts = 2, fish = 1). There was no association between CLE results, and presence of food-specific IgE (P > 0.999). Both dietary arms resulted in a similar histological response (P > 0.999), with histological remission in only 1 patient (excluding gluten-containing grains). Permeability, eosinophil or mast cell activation were not different between arms.

Conclusions

In this pilot blinded cross-over trial, a targeted elimination diet based on the results of CLE with DFC did not generate a superior response compared to a sham diet.

INTRODUCTION

Since its first description in the early nineties, the number of patients diagnosed with eosinophilic esophagitis (EoE) has seen an enormous increase.¹ This has been paralleled by an impressive growth in knowledge regarding clinical presentation, underlying pathophysiology, and management strategies.^1-4 Nevertheless, an adequate diagnostic tool or biomarker to identify the causing allergens—typically food antigens—remains to be discovered. Up to now, the gold standard for dietary management of EoE—and the only way to eliminate the cause of the inflammation—remains an exclusion diet, empirically excluding the 6-, 4-, 2-, or even the single most common food antigen(s) from the diet,^5-7 followed by a reassessment with esophagogastroduodenoscopy (EGD) and esophageal biopsies. These diets can be effective in 34% to 70% percent of patients but are cumbersome and require repeat endoscopies with biopsies after each elimination and reintroduction step.⁸ Less invasive methods such as cystosponge or string test are being developed, however not yet available for routine clinical practice.¹

The initial hypothesis that immunoglobulin E (IgE) could play a key role in pathogenesis, was invalidated as treatment with an anti-IgE antibody resulted in a remission rate of only 33% in a non-controlled setting^9,10 and no differences were found in a randomized controlled trial.¹¹ Also targeted elimination diets based on IgE sensitization to food allergens only led to improved symptoms and histology in a minority of adult patients.^9,12 The same study that tested an anti-IgE antibody suggested a role for IgG4.11 However in a recent study, the fall in median peak esophageal eosinophil count (PEC) on a targeted exclusion diet based on food-specific IgG4 was not statistically significant, and the methodology needs further evaluation.¹³

Because of the limitations of current test methods and the intensity of an empiric food elimination diet, dietary treatments are not frequently used in adult patients with EoE. Therefore, developing reliable diagnostic tests to identify causative foods remains an unmet need in EoE.¹

In patients with active EoE, the esophageal mucosal integrity is impaired and intercellular spaces are dilated.^14-16 In other atopic diseases, barrier dysfunction not only affects the target organ, but also the small intestine, the most immunogenic organ in the gut. It has been postulated that duodenal permeability might also be increased in EoE patients.^14-17 An increase in small intestinal permeability was shown in a study measuring the lactulose to mannitol ratio in urine,¹⁴ however not confirmed in another.¹⁵ It remains however plausible that increased permeability and sensitization for antigens (nutrients, microbiota etc) in the duodenum is present, but that we lack the right tools to measure it.¹⁷

Confocal laser endomicroscopy (CLE) is a technique that allows for real-time microscopic visualization of the mucosa during upper endoscopy.¹⁸ In patients with irritable bowel syndrome (IBS), assessment of the duodenal mucosa using CLE has shown acute food-induced mucosal alterations within minutes after intraluminal food administration.¹⁹ Acute mucosal alterations were defined as fluorescein leakage and the appearance of particles in the lumen. 18 This definition has shown substantial intra- and inter-observer agreement rates in a blinded assessment of experienced investigators.²⁰

This technique, including duodenal food challenge (DFC), has been used in a proof-of-concept study in 9 patients with histologically proven EoE. One patient showed baseline mucosal alterations and of the remaining 8, 61% showed alterations upon DFC. These were, in order of frequency: wheat, milk, soy, and egg. The patients were treated with established medical treatment options for EoE (including proton pump inhibitor [PPI], swallowed topical corticosteroids [STC] or endoscopic dilation) and an additional open-label food exclusion of the triggering nutrient. All patients reported a reduction in symptoms, however the study did not report endoscopic/histological responses.²¹ Assessment of the epithelial barrier using CLE in the esophagus has been undertaken,²² but consists of a tight barrier and is not subject to acute alterations after food application as seen in the duodenum.

The aim of our study is first to assess nutrient-triggered acute mucosal alterations in the duodenum of EoE patients, using CLE. Secondly, in patients in whom these alterations were observed, the aim is to assess the efficacy of a targeted elimination diet based on the identification of nutrients triggering acute mucosal reactions in the duodenum in a blinded, randomized controlled trial. In addition, we aim to explore permeability and markers of inflammation in the duodenum.

MATERIALS AND METHODS

Trial Design and Study Population

We performed a double-blind randomized controlled trial in patients diagnosed with EoE based on symptoms of esophageal dysfunction, endoscopic features and histological confirmation,² with non-response or intolerance to PPIs, and motivated to start an elimination diet. An EGD with duodenal and esophageal biopsies was performed at baseline, along with blood samples, urine samples, and baseline questionnaires. In patients with active EoE based on symptoms and histology, this was followed by a second EGD with probe-based CLE at least 2 weeks later. Three days prior to the procedure, patients followed the six-food elimination diet (SFED), to avoid interactions during the measurement. If acute mucosal alterations were visualized during the CLE procedure (see criteria below), patients were randomized to receive a blinded crossover diet for 6 weeks; excluding either the nutrient(s) triggering alterations during CLE, or a sham diet excluding (a) nutrient(s) that did not cause acute alterations. In patients where no alterations were visualized, or in case of no adequate negative comparator for the sham diet, a control diet was started using consecutively gluten-containing grains and milk in random order. At the end of each diet period, baseline measurements were repeated (Fig. 1, study flowchart). Primary outcome was the difference in histological response rate, defined as a PEC of < 15 eosinophils per high power field (PEC < 15/HPF) in patients undergoing duodenal CLE-targeted elimination diet compared to the sham diet. Secondary outcomes included difference in deep histological remission (PEC < 6/HPF); in PEC before and after the targeted elimination diet and symptomatic and endoscopic changes in patients undergoing duodenal CLE-targeted elimination diet compared to the sham diet. Measures of permeability and duodenal inflammation were assessed as exploratory outcomes. The study was conducted according to the Declaration of Helsinki and Good Clinical Practice regulations after approval by an Ethics Committee (S66223, NCT NCT05695456). Written informed consent was obtained from all study participants before inclusion in the study.

Figure 1.

Study flowchart. *V3 with second confocal laser endomicroscopy (CLE) was performed in case of early alterations seen at first CLE (at baseline, or after administration of 1 of the first 3 foods). Blood samples and urine samples were taken at baseline, and after each elimination diet. Esophagogastroduodenoscopy (EGD) was repeated after each elimination diet. Questionnaires were filled out at baseline, and at regular timepoints during the diet periods. A phone-call visit was scheduled half-way each diet period.

Study Procedures

Probe-based confocal laser endomicroscopy

The procedure was performed according to guidelines¹⁸: During EGD with propofol sedation, a CLE probe was passed through the working channel of the endoscope and gently placed against the mucosa in the second part of the duodenum (D2). After administration of IV fluorescein (2.5 mL, 10% solution) as a contrast agent, baseline assessment of the mucosa was performed for 2 minutes (Fig. 2). Subsequently, the 6 foods of the SFED (10 mL solutions of milk/egg/soy/lyophilized fish and shellfish/nuts/a mixture of grains containing wheat, barley, and rye) were applied sequentially in a semi-random order. The endoscopist (J.T.) was blinded for the applied nutrient. The order was semi-randomized starting with either soy or egg, followed by either grains or milk, and ending with fish and nuts. This order was chosen to preferentially test the nutrients that are known to be the more common culprits in EoE. After application of the nutrient, evaluation of the mucosa was repeated for 4 minutes, and the mucosa was judged as altered in case of presence of both intraluminal fluorescein and luminal particles at least 2 minutes after food application (Fig. 2).^18,20 In case alterations were observed, the exam was terminated as changes can no longer reliably be assessed after acute alterations. When this occurred after the application of one of the first 3 foods, the procedure was repeated at least 2 weeks later to allow for restoration of the mucosal wall, and the remaining foods were tested. Biopsies from D2 were collected at the end of the examination.

Figure 2.

Confocal laser endomicroscopy (CLE) assessment before (A, B) and after (C, D) nutrient exposure in the duodenum. (A) Baseline CLE image with presence of epithelial cells in the villi made visible through intravenously administered fluorescein. At baseline the lumen is dark. (B) Some fluorescein in the lumen (circle) but no shedding of particles. (C) Positive CLE image with extravasation of fluorescein into the lumen and particles (arrow) previously termed “cell-shedding.” (D) Positive CLE image: contrast inversion with the lumen appearing white, and the epithelium dark.

Biological Material

Blood samples

Baseline blood samples were used to assess full blood counts, high-sensitivity C-reactive protein (hsCRP), serum tryptase, total IgE, and food-specific IgE (FS-IgE) to each of the tested nutrients (measured using ImmunoCAP on Phadia 1000; ThermoFisher Scientific, Waltham, MA, USA: egg white, cow’s milk, soy, a mixture of grains, a mixture of nuts, and a mixture of fish and shellfish). Repeat serum tryptase level was measured at least 30 minutes after a reaction occurred during CLE. HsCRP was repeated after each exclusion diet (real/sham).

Urinary lactulose/mannitol ratio

Participants ingested a fixed dose of 5g lactulose and 1g mannitol dissolved in 250 mL of water, and urine was collected for 2 hours after ingestion in a dry container containing 250 mg neomycin in powder form. The containers were brought to the laboratory and after filtration through a 0.45 μm filter, 1 mL aliquots were frozen until analysis. Urinary lactulose and mannitol concentrations were quantified with gas chromatography-mass spectrometry (GC-MS).²³ Limits of quantification were 15 mg/L and 5 mg/L for mannitol and lactulose, respectively. The lactulose-mannitol ratio (LMR) was calculated as the fractional lactulose concentration divided by the fractional mannitol concentration.

Ussing chamber experiments

The protocol used for Ussing chamber experiments was reported previously.²⁴ In summary, biopsies were mounted in modified 3-mL Ussing chambers (Mussler Scientific Instruments) within 30 minutes of collection in open-circuit conditions. The surface of exposed biopsies was 0.017 cm². For each chamber, transepithelial electrical resistance (TEER) was calculated from voltage deflections induced by bipolar constant-current pulses of 16 μA every 60 seconds with duration of 200 msec and was recorded every 30 minutes over 2 hours. The average TEER of timepoints 60-90-120 minutes was calculated and presented as Ω x cm². Paracellular flux was measured using fluorescein (molecular mass of 400 Da, 20 mg/200 μL; Sigma-Aldrich, St Louis, Missouri, USA), which was added to the mucosal compartment of the 3 Ussing chambers containing esophageal biopsies. A fluorescent labeled dextran (FD4 molecular mass of 4 kDa; Sigma-Aldrich) was used for the 3 duodenal biopsies. Serosal samples were collected every 30 minutes for 2 hours and the passage of fluorescein or FD4 was measured using a fluorescence reader (FLUOstar Omega; BMG Labtech, Ortenberg, Germany). The average was calculated for each chamber at 90 and 120 minutes presented as pmol/cm². Paracellular flux could only be measured at baseline, and after each diet period, as the IV administration of fluorescein during CLE procedure interfered with the measurement.

Histology

A total of 6 esophageal biopsies were taken at 3 levels (proximal, mid- and distal esophagus) at baseline, and after each exclusion diet of 6 weeks. All biopsies were assessed for PEC by one and the same expert pathologist (G.D.H.). The pathologist was blinded to the timepoint of the study (ie, baseline or during dietary intervention) and to the nature of the diet (ie, targeted diet or sham diet).

Duodenal eosinophil and mast cell counting of baseline biopsies was performed blinded after H&E and c-kit staining, respectively. Stained duodenal biopsy sections were scanned with a ZEISS Axioscan 7 slide scanner in a random and blinded fashion and visualized using QuPath software (version 0.5.1, UK).²⁵ Duodenal eosinophils (bilobar nucleus and eosinophilic granules on H&E staining) and mast cells (visible nucleus and cytoplasmic c-kit staining) were counted according to the Leuven Intestinal Counting Protocol (LICP).²⁶ The threshold was set at 400 eosinophils/mm² and 700 mast cells/mm² for abnormal duodenal infiltration.²⁶

Biopsy supernatants

Two duodenal biopsies were obtained using standard size biopsy forceps and placed in culture medium (Roswell Park Memorial Institute containing glutamine, supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin 10,000 U/mL, (all from Gibco, Life Technologies). After 1 hour, biopsies were transferred to a 48-well plate containing the same culture medium and incubated for 24 hours at 37°C. Supernatants were obtained after 24 hours, snap frozen and kept at –80°C until analysis.

Beta-2 tryptase and eosinophil derived neurotoxin (EDN) in biopsy supernatants were measured using commercially available ELISA kit (TPSB2 ELISA, ThermoFisher Scientific, Life Technologies; EDN ELISA, Diagnostics Development, Uppsala, Sweden, respectively). Resulting beta-tryptase and EDN content was corrected for biopsy weight and reported as μg/mg tissue.

Mucosal protein extracts

Eosinophilic cationic peptide (ECP), alpha tryptase and IgE were measured on mucosal protein extracts with Immunocaps, using Phadia 250 Instrument (ThermoFisher Scientific) and corrected for biopsy protein content measured using BC Assay (Interchim, Montluçon, France). Results for IgE are reported in kU/g, results for ECP and alpha tryptase are reported as ng/mg protein.²⁷

Comparison with healthy volunteers

We used available data from a cohort of healthy volunteers to compare baseline measurements (TEER, mucosal IgE, ECP, and alpha-tryptase). The cohort of healthy volunteers was recruited in our center in the context of a similar study protocol (NCT05097872), with main exclusion criteria being acute or chronic gastrointestinal symptoms, any gastrointestinal comorbidity such as coeliac disease, EoE or inflammatory bowel disorder, or history of gastro-intestinal surgery. All healthy volunteers were negative for serum IgE specific for any of the administered foods.

Questionnaires

At baseline, and at regular timepoints during the diet periods, symptoms were assessed using dysphagia symptom questionnaire (DSQ), Eosinophilic Esophagitis Activity Index (EEsAI), and eosinophilic esophagitis quality of life questionnaire for adults (EoEQoL-A) (Fig. 1). The DSQ is a short, validated, 4-question assessment of dysphagia symptoms over the last 24 hours, which was used as a daily diary for 2 weeks at baseline and during the entire diet periods.²⁸ The EEsAI is a symptom-based severity index designed and validated for EoE; and evaluates symptoms related to food intake over a recall period of 7 days.²⁹ The EoE-QoL-A is a reliable disease-specific health-related quality of life measure for adult EoE patients.³⁰ The latter 2 questionnaires were filled out at baseline, and every 2 weeks during the diet. All questionnaires were endorsed by the international multidisciplinary consensus as part of the core outcome set for therapeutic studies in Eosinophilic Esophagitis (Coreos).²⁸ The Monash University Comprehensive Nutrition Assessment Questionnaire was filled in at baseline, a short food frequency questionnaire was filled in by participants weekly during both diet periods to assess compliance with the exclusion diet. FIRE questionnaire (food induced immediate response), was filled in at baseline and at the end of each diet period.

Statistical Analysis

This is a pilot study using a new technique (CLE) in EoE. Since no studies were published on the use of CLE in EoE patients, the sample size calculation was based on data available from dietary studies not using CLE. Assuming that the CLE directed diet will be effective in 75% of patients (similar to the 6FED), and the sham diet in 10% of patients, using a 2-tailed McNemar test with an α of 0.05, 1-β of 0.8, we calculated a sample size of 15. Adjusting for one drop-out, we aimed to include 16 CLE positive patients. A futility analysis was performed after 50% of planned participants were included (ie, 8 patients) due to the invasive nature of the study. As we did not reach our anticipated sample size, the power of the study did not reach 0.8.

Descriptive statistics were used for baseline characteristics. Data are represented as median (interquartile range). A McNemar test was used to assess the primary outcome. Non-parametric tests (Wilcoxon signed rank test, Kruskal Wallis rank sum test) were used for secondary and exploratory outcomes, as well as Spearman/Pearson correlations where applicable. A logistic regression model was used to test for carry-over effect. P-value of < 0.05 was considered statistically significant.

RESULTS

Baseline Characteristics and Measurements

We included 8 patients, of which 1 was later excluded due to inactive EoE at baseline (PEC < 15/HPF). Median age at inclusion was 41.5 (28.0-47.0) years, 5 (62.5%) participants were male, with a median body mass index of 27.9 (26.3-29.8) kg/m².

Of the 7 patients who had active EoE at baseline (PEC > 15/HPF), median PEC was 80/HPF (50-95), median endoscopic reference score (EREFS) score was 3 (3-4), with moderate symptoms and significant impact on quality of life (Table 1).

Table 1.

Demographics and Baseline Measurements

Variable	value
Demographics
N	7
Age (yr)	41.5 (28-47)
Sex (%)	62.5% male
BMI (kg/m2)	27.9 (26.3-29.8)
PPI use	7 (100%)
STC use	5 (71%)
History of atopy	5 (71%)
Antihistamine use	2 (29%)
Baseline EoE severity and impact
PEC (max eosinophils/HPF)	80 (50-95)
EREFS	3 (3-4)
EEsAI	42 (33-59.5)
EoE-Qol-A (/96)	52 (37.5-74)
DSQ (/84)	12 (0-25.5)

EoE, eosinophilic esophagitis; BMI, body mass index; PPI, proton pump inhibitor; STC, Swallowed topical corticosteroids; PEC, peak eosinophil count; HPF, high-power field; EREFS, Endoscopic reference score; EEsAI, Eosinophilic Esophagitis Activity Index; EoE-Qol-A, eosinophilic esophagitis quality of life score for adults; DSQ, dysphagia symptom questionnaire.

All patients had previously taken PPI as first line treatment, which was ineffective in six, and was not tolerated by the seventh patient (skin rash). Five patients were previously treated with STC, with no response in 1 patient, only short-term and suboptimal use of a flixotide inhaler in another patient, and response in the 3 other patients. These latter patients wanted to stop medication and switch to diet-based therapy, for which STC was interrupted at least 2 months before inclusion.

A personal history of atopy was present in 5 of 7 patients; most common were hay fever and allergic rhinitis. One patient was treated with anti-TNF therapy (certolizumab subcutaneously) for severe psoriasis. Another patient had severe atopic dermatitis, food allergy to soy, and oral allergy syndrome for stone fruits, kiwi and banana. None of the patients however, had a history of anaphylaxis. Two patients took antihistamines at baseline, which was continued on a stable dose throughout the study.

Four out of seven patients experienced FIRE (food-induced immediate response of the esophagus) symptoms at baseline, with an acute reaction to wheat mentioned by all 4 of them (specifically beer, bread, and pasta), raw fruits or vegetables mentioned by 3 (raw fruit and vegetables, fresh fruit juice, banana), and dairy products by 1 (cheese, yoghurt).

Laboratory measurements

Total serum IgE was elevated in all but 1 patient, with a median of 346 (148-505) kU/L. FS-IgE were also elevated in these 6 patients, with FS-IgE for milk elevated in 5 patients, for nuts in 4, for grains, soy, egg and fish in 3 patients (Supplementary Table 1).

Baseline white blood cell counts were normal in all patients, with a slightly increased peripheral eosinophil count in 2 patients (0.4 × 109/L and 1.0 × 109/L, respectively). Baseline serum tryptase was low in all patients with a median of 6.8 (4.6-7.2) μg/L.

Confocal Laser Endomicroscopy

At least 2 weeks after baseline EGD, 7 patients underwent a second EGD with CLE. During the first CLE examination, 1 patient displayed baseline mucosal alterations, all other patients reacted to one of the foods applied in the duodenum. In 3 patients, a second CLE examination was performed with mucosal alterations after food application in all 3. We were able to test at least 3 nutrients in all patients (3 in 1 patient, 4 in 3 patients, and 5 nutrients in 3 patients). As a result, 5 patients reacted to 1 nutrient, and 2 patients reacted to 2 nutrients in 2 different sessions (milk = 4, grains = 2, nuts = 2, fish = 1) (Table 2).

Table 2.

Overview of Alterations as Assessed With Confocal Laser Endomicroscopy

Patient	Baseline	Grains	Milk	Soy	Eggs	Fish	Nuts	N CLE	N applied nutrients
1	neg	neg	neg	neg	neg	pos		1	5
2	neg	neg	pos	neg	neg		pos	2	2-3
4	neg	neg	pos	neg	neg			1	4
5	neg	pos	neg	neg	neg			1	4
6	neg	neg	pos	neg	neg			1	4
7	neg	neg	pos	neg	neg		pos	2	2-2
8	pos	pos		neg	neg			2	0-3
pos/tested	1/10	2/7	4/6	0/7	0/7	1/1	2/2

N CLE, number of confocal laser endomicroscopy (CLE) examinations performed per subject; N applied nutrients, number of applied nutrients per examination; neg, negative; pos, positive; pos/tested, total amount of alterations seen divided by the times each nutrient was applied in all examinations.

In subjects where 2 CLE procedures were performed, the numbers for each procedure are separated by -.

There was no association between reaction on CLE, and the presence of serum specific IgE for the tested nutrients (P > 0.999). Moreover, we saw no association between reaction on CLE, and baseline intake of the tested nutrients (P = 0.230) (Supplementary Table 2). Also, there was no significant correlation between baseline intake and FS-IgE, except for fish (ρ = 0.87; 95% CI, 0.33-0.98; P = 0.010), although FS-IgE for fish were only elevated in 2 patients.

Response to Targeted Exclusion Diet

Five patients started with a blinded diet for the trigger nutrients (real diet), 2 patients started with a sham diet (Supplementary Table 3).

The primary endpoint of PEC < 15/HPF, was only reached by 1 patient after the real diet (elimination of gluten-containing grains), the response rate was not different between diet arms (P > 0.999). One other patient had a peak eosinophil count of 15/HPF after the real diet (grains), and of 18/HPF after the sham diet (soy). A third patient had symptomatic improvement and a reduction in PEC from 100/HPF at baseline to 25/HPF after the real diet (milk and nuts), and a resurgence after the sham diet (grains and fish) to 230/HPF. None of the other patients showed histological response to either of the diets. Adherence to the diet was excellent except for 2 patients, who had consumed lactose-free milk where they had to exclude all dairy products containing animal-protein (Supplementary Table 2). A logistic regression model was built to test for carry-over effect, however this could not be formally tested because no responses were observed in the second diet phase.

Regarding the secondary endpoints, PEC did not change (baseline: 80.0 (50.0-95.0), real diet: 35.0 (21.5-88.0), sham diet: 115.0 (26.5-176.0), P = 0.450, Fig. 3A). Median EREFS did not change after either of the diets (baseline: 3.0 (3.0-4.0), real diet: 3.0 (2.5-4.0), sham diet: 3.0 (3.0-4.5), Fig. 3B). DSQ over a period of 2 weeks (maximal total score of 84) was compared at baseline versus the last 2 weeks of each diet period. There was no significant change in DSQ score from baseline after any of the diet periods (P = 0.180). Median EEsAI was 58.0 (21.0-68.5) after the real diet, and 42.0 (13.5-68.5) after the sham diet. Both were not different from baseline (P = 0.400, Fig. 3C). Also median EoE-Qol-A was not different from baseline (real: 50.0 (41.0-82.5), sham: 69.0 (47.0-85.5), P = 0.740, Fig. 3D).

Figure 3.

Evolution of peak eosinophil count (PEC), endoscopic reference score (EREFS), Eosinophilic Esophagitis Activity Index (EEsAI), and eosinophilic esophagitis quality of life score for adults (EoE-Qol-A) from baseline and after cross-over diet periods. (A) PEC at baseline, after the real diet and sham diet. Median PEC remained above the threshold for histological remission of 15/HPF, 1 patient was in histological remission after the real diet. There was no significant change from baseline after the diet periods. (B) EREFS at baseline, after the real diet and sham diet. There was no significant change from baseline after the diet periods. (C) EEsAI score at baseline, after the real diet and sham diet. There was no significant change from baseline after the diet periods. (D) EoE-Qol-A score at baseline, after the real diet and sham diet. There was no significant change from baseline after the diet periods.

Exploratory Measurements of Intestinal Permeability and Inflammation

Exploratory results can be found in Table 3.

Table 3.

Exploratory Measurements

Measurement	Eosinophilic Esophagitis	Healthy volunteers		P-value
Total IgE in duodenal biopsy (kU/g)	5.99 (3.18-9.11)	1.31 (1.10-2.75)		0.050
ECP in duodenal biopsy (ng/mg)	41.6 (29.6-52.7)	37.5 (22.7-51.3)		0.610
Alpha tryptase in duodenal biopsy (ng/mg)	492 (249-662)	177 (133-322)		0.050
TEER (Ω × cm2)	23.6 (22.2-24.1)	23.3 (22.6-27.5)		0.650
Lactulose mannitol ratio	1.16 (1.07-1.27)	0.96 (0.63-1.29)		0.480

	Baseline	CLE		P-value
Duodenum
Eosinophil count (/mm2)	254 (207-350)
Mast cell count (/mm2)	627 (569-713)
TEER (Ω × cm2)	23.6 (22.2-24.1)	23.7 (22.5-25.3)		0.840
Beta-tryptase from biopsy supernatant (μg/mg)	0.64 (0.44-0.81)	0.67 (0.27-1.25)		0.640
EDN from biopsy supernatant (μg/mg)	5.15 (3.33-5.98)	5.62 (4.77-6.64)		> 0.999
Serum tryptase (μg/L)	6.8 (4.6-7.2)	4.7 (3.35-5.3)		0.040

	Baseline	Real diet	Sham diet	P-value
Esophagus
TEER (Ω × cm2)	62.8 (39.4-102)	55.3 (36.0-89.8)	44.3 (25.5-72.5)	0.460
Duodenum
TEER (Ω × cm2)	23.6 (22.2-24.1)	26.2 (24.5-34.1)	25 (22.7-26.4)	0.290
FD4 flux (pmol/cm2)	1788 (1286-2356)	1884 (1866-2440)	1884 (1798-2233)	0.870
Total IgE in duodenal biopsy (kU/g)	5.99 (3.18-9.11)	4.30 (2.30-7.87)	5.99 (4.43-15.2)	0.600
ECP in duodenal biopsy (ng/mg)	41.6 (29.6-52.7)	31 (21.6-46.4)	59.1 (27.0-95.1)	0.440
Alpha tryptase in duodenal biopsy (ng/mg)	492 (249-662)	263 (182-544)	504 (212-676)	0.810
Lactulose mannitol ratio	1.16 (1.07-1.27)	0.913 (0.88-1.46)	0.946 (0.73-1.10)	0.340
High-sensitivity CRP	1.02 (0.49-1.92)	0.98 (0.83-1.08)	0.42 (0.28-0.75)	0.420

IgE, immunoglobulin E; ECP, eosinophil cationic peptide; TEER, transepithelial electrical resistance; CLE, confocal laser endomicroscopy; EDN, eosinophil derived neurotoxin; FD4, fluorescent labeled dextran.

Healthy volunteers

Data from 8 healthy volunteers were used (6/8 female), with a median age of 22.0 years (20.8-24.0), and median body mass index of 23.4 (21.8-25.3).

Baseline

As a marker for duodenal permeability, TEER was compared between included EoE patients with data from healthy volunteers who underwent the same protocol, and was not significantly different (P = 0.650). Paracellular flux of FD4 was not performed in our cohort of healthy volunteers, however when comparing with a previous cohort, results were similar.³¹ As a marker of small intestinal permeability, baseline LMR was assessed, which was again not different from healthy volunteers (P = 0.480).

At baseline, duodenal mast cells and eosinophils were counted using the LICP.²⁶ Well-oriented slides were available for all but 1 patient. Median duodenal eosinophil count was 254/mm² (207-350), which falls within the normal range as previously defined (< 400 eosinophils/mm²).²⁶ Median mast cell count at baseline was 627/mm² (569-713), with the highest value being 826/mm², which is on the high end as compared with data from healthy volunteers (normal < 700 mast cells/mm²).²⁶

When comparing local total IgE levels in duodenal biopsies with healthy volunteers, we found a trend towards higher total IgE levels in active EoE (P = 0.050). Also local alpha-tryptase levels in duodenal biopsies tended to be higher in EoE patients than controls (P = 0.050). We found no difference in presence of ECP (P = 0.610).

Before vs after confocal laser endomicroscopy

After an acute mucosal reaction during CLE, TEER was not different from baseline (P = 0.840, Fig. 4A). We saw no difference in secretion of EDN (P = 0.640) or beta-tryptase (P > 0.999) in duodenal biopsy supernatant after CLE, as compared with baseline. Also systemically measured tryptase in serum, taken 30-60 minutes after acute alterations were observed during CLE, was not higher than baseline, and was even lower than baseline (P = 0.040, Fig. 4B).

Figure 4.

Transepithelial electrical resistance (TEER) and serum tryptase at baseline versus after confocal laser endomicroscopy (CLE). (A) Comparison of TEER (Ω × cm²) in ussing chambers at baseline, versus after CLE. There was no significant difference between conditions. (B) Comparison of serum tryptase levels (μg/L) at baseline versus after CLE. There was no increase in serum tryptase after CLE, instead there was a significant decrease in serum tryptase levels taken 30-60 minutes after mucosal alterations were seen during CLE examination (P = 0.040).

Comparison with exclusion diets

LMR was not significantly different between conditions (P = 0.340). Duodenal TEER (P = 0.290), as well as paracellular flux of FD4 (P = 0.870) did not significantly change from baseline after any of the diets. Esophageal TEER was low at baseline, with a median of 62.8 Ω × cm² (30.3-115.0), and did not improve after any of the diet periods (P = 0.460). Trans- and paracellular flux of fluorescein in esophageal biopsies could not reliably be assessed because of the severely impaired barrier function in most patients.

HsCRP remained stable during follow-up (P = 0.420). There was no change in any of the markers of inflammation tested: alpha-tryptase (P = 0.810), ECP (P = 0.440), and total IgE (P = 0.600) in mucosal protein extracts; EDN (P = 0.800) and beta-tryptase (P = 0.680) in duodenal biopsy supernatants.

DISCUSSION

In this pilot randomized controlled trial, a targeted exclusion diet was evaluated, using the results of acute mucosal alterations to food exposure in the duodenum, as assessed by confocal laser endomicroscopy. This technique allows for live, in vivo assessment of the intestinal mucosa during endoscopy on a microscopic level, and has shown acute mucosal alterations to food exposure in the duodenum in IBS patients.^19,32

In our study, we found that all 7 participants reacted to at least 1 food applied to the duodenal mucosa, and when the examination was repeated at least 2 weeks later, 2 patients reacted to a second food item (1 patient showed baseline alterations in 1 of the 2 exams). In the landmark paper in IBS patients, an acute reaction occurred in 70% of patients after duodenal food challenge,¹⁹ and a similar observation was reported for 6 out of 8 in a pilot study in patients with EoE (with a ninth patient showing baseline alterations).²¹ It is unclear why the response rate to CLE was as high as 100% in our study. Similar to previous studies, our patients followed a hypo-allergenic diet 3 days prior to the CLE examination. It is unlikely that this was too short, as we noted first unaltered mucosa during baseline assessment in all but 1 patient, and an acute change after application of a certain nutrient. In addition, most of our patients were exposed to several nutrients triggering no CLE reaction, before alterations were observed. Other than in previous studies, most of our patients had positive serum FS-IgE to at least 1 of the tested nutrients, however, we did not find an association between the presence of serum IgE antibodies, and a reaction to the same nutrient, applied in the duodenum. The nutrients tested in this study were slightly different from those in previous studies, where wheat, milk, soy, yeast, and egg white were tested. However, we chose to test the 6 foods of the SFED, including nuts (positive in 2 patients) and fish and shellfish (positive in 1 patient). Yeast was not tested in this study. In addition, we used probe-based CLE, whereas most previous studies used endoscopy-based CLE (which is no longer available), making direct comparisons difficult. For probe-based CLE, a recent blinded assessment of CLE images by experienced investigators from 3 European expert centers showed substantial intra- and interobserver agreement rate.²⁰ Lastly, we cannot exclude other mechanisms for the alterations we see, such as duration of the exam or physiological changes after application of nutrients to the intestinal mucosa, and therefore results of the application in healthy volunteers should be awaited.

When putting the patients on a blinded cross-over exclusion diet, excluding either the trigger nutrient(s), or a sham diet for 6 weeks, only 1 patient achieved histological remission, as defined by PEC < 15/HPF. There was no difference in response between the real diet (1 patient, excluding gluten-containing grains) and the sham diet (none) (primary endpoint). Two other patients had a partial response, one to the real diet (milk & nuts), and the other to both the real diet (grains) and the sham diet (soy). When looking at the evolution of PEC over time, we saw no difference between the real and the sham diet. Also, endoscopic findings (EREFS) and symptom scores (DSQ, EEsAI, and EoE-QoL-A) did not change over the course of the study. In the previous study mentioned above, where 6 out of 8 EoE patients reacted to DFC as assessed by CLE, the effect of a targeted exclusion diet was not assessed in a blinded, controlled manner as was the case in this study. The patients followed an open-label exclusion diet, alongside established treatment options for EoE. Furthermore, there is no mentioning of a histological response after the exclusion diet.²¹ This is thus the first study evaluating the results of DFC during CLE in a blinded, controlled targeted elimination diet, which was not more effective than a sham diet, excluding nutrients that did not trigger acute alterations. When comparing with empirical elimination diets, a one-food elimination diet was shown to be effective in 34% of adults,⁵ and a 2-food elimination diet (excluding wheat and milk) in 43% of adults,³³ and even more in a systematic review and meta-analysis also including children.⁸ The 1 participant that achieved histological remission in this study, excluded gluten-containing grains, one of the most common triggering food antigens in EoE.⁸ In summary, from the 7 patients included in this study, only 1 responded to a targeted diet based on CLE results, and there was no difference in primary and secondary outcomes between real and sham diet. To be able to generalize these results, they should be confirmed in a larger sample.

We explored several markers of permeability and of inflammation in the duodenum. As previously documented, we saw a highly increased permeability ex-vivo of the esophageal mucosal wall in patients with active EoE.^15,34 As we had only one patient in histological remission after treatment, we could not compare with EoE in remission.

Katzka et al¹⁴ showed an increase in LMR in urine, a measurement correlating to small bowel permeability, of active EoE patients as opposed to the ones in remission. However, Warners et al¹⁵ found no difference in LMR. Furthermore, the latter also used duodenal impedance measurement in vivo, and TEER measurement and transepithelial flux in vitro in Ussing chambers, and found no difference in duodenal integrity between EoE patients and controls. Also in our study, we could not confirm an increased permeability in the small intestine of EoE patients with active inflammation.³¹ In addition, we found normal eosinophil counts and ECP values, as compared with healthy volunteers. As a conclusion, in this small cohort of patients with active EoE at baseline, we did not find signs of small intestinal increased permeability or eosinophil activation, however the sample size is too small to exclude involvement of the small intestine in EoE.

We did find a trend towards an increased level of local IgE in mucosal protein extracts of duodenal biopsies, as compared with healthy volunteers. Even though only exploratory, higher levels of local total IgE have previously been demonstrated in patients with confirmed food allergy²⁷; also in our cohort, the majority had positive FS-IgE in serum for the 6 tested nutrients, and possibly others which were not tested. There was a trend towards higher alpha-tryptase concentrations in mucosal biopsies, and also duodenal mast cell counts were on the high end, as compared with a previous cohort of healthy volunteers using the same counting protocol. 26 This suggests an increased duodenal mast cell infiltration and/or activation in the duodenum of EoE patients, similarly to patients with atopic dermatitis,³⁵ but should be confirmed. With only seven patients concluding the study, it is underpowered to draw firm conclusions. In addition, in this cross-over study, more patients started with the real diet than the sham diet, inducing a possible order effect. However, we did not see any difference between both diet arms, and we relied on the objective measurement of PEC, therefore it is unlikely that a more balanced distribution would have changed our results. A formal test for carry-over effect could not be used because of no responses in the second diet phase. Three of our patients were in histological remission on STC before the start of the study, and discontinued medication at least 2 months prior to baseline measurements. Even though all patients had a PEC > 15, these patients might not have been fully relapsed at the start of the study, inducing a possible bias in further evaluations. Only 1 of these patients responded to 1 of the diets, suggesting that this did not affect our results. Lastly, most of the ex vivo measurements are exploratory and should be interpreted with caution.

In conclusion, in this blinded cross-over pilot trial, the results of a duodenal food challenge during CLE were not associated with response to a targeted exclusion diet. Targeted elimination diets using CLE are unlikely to provide a benefit over established empirical exclusion diets in EoE. In addition, in this study, no small intestinal increased permeability was found.

SUPPLEMENTARY MATERIALS

Note: To access the supplementary tables mentioned in this article, visit the online version of Journal of Neurogastroenterology and Motility at http://www.jnmjournal.org/, and at http://doi.org/10.5056/jnm25132.