As our understanding and recognition of eosinophilic esophagitis (EE) advances, several significant unmet needs emerge (1). One of these unmet, important, and relatively urgent needs is the ability to assess disease activity using nonendoscopic, that is, noninvasive, tools. Currently, patients with EE sometimes undergo repeated endoscopies during the course of their disease (2). Esophagogastroduodenoscopy (EGD) with biopsies is needed to establish diagnosis, assess response to therapy, document relapse, and potentially monitor disease quiescence. Repeated procedures exert significant physical, psychological, and financial tolls on the patient, the family, the health care provider, and the overall health care system (3). From a clinical standpoint, noninvasive markers may help providers determine whether nonspecific symptoms truly reflect poorly controlled EE versus an unrelated non-EE illness. From a scientific standpoint, the development of noninvasive markers would help improve understanding of the pathogenesis of EE.
A multitude of cytokines and chemokines serve as inflammatory mediators, and several have been examined as potential markers of EE disease activity on endoscopically obtained esophageal mucosal biopsies (4–6); however, published data on noninvasively obtained biomarkers is sparse. EE is thought to be a TH2-mediated disease, with roles for interleukin (IL)-5, IL-13, eotaxin, and eosinophils (1). IL-5 overexpression in mouse models induces significant esophageal eosinophilic inflammation, and IL-5-deficient mice have total absence of allergen-induced esophageal eosinophilic infiltration (7). IL-5 is upregulated in esophageal mucosal biopsies obtained from children and adults with EE (4,6). Eosinophils contain cytoplasmic granules that release cationic toxic proteins upon eosinophil stimulation (8,9). One of these proteins, the eosinophil-derived neurotoxin (EDN) (10), a basic ribonuclease eosinophil granular protein, which is cytotoxic, is released from eosinophils upon activation and degranulation (11). EDN is expressed in liver, spleen, neutrophils, and monocytes/macrophages (12). In addition to mediating tissue injury, EDN may promote antigen uptake and processing by recruiting dendritic cells to inflamed tissue, thereby helping sustain the inflammatory process (12). Elevated levels of eosinophil granule proteins have been reported in the bronchoalveolar fluid, serum, and urine of asthmatics, and in whole-gut lavage fluid and stool of patients with inflammatory bowel disease (7,11,13,14).
In view of the unmet need for a noninvasive biomarker for EE, we conducted this study with the aim of identifying a marker able to assess disease status in lieu of repeated endoscopies. We examined 3 noninvasive biomarkers, serum IL-5, serum EDN, and stool EDN, and correlated the biomarker results with disease phenotype and activity (symptoms and histology) in a longitudinal study conducted in children with EE.
SUBJECTS AND METHODS
This study was conducted as a secondary analysis of data from an open-label, prospective, longitudinal study examining the response of EE to either of 2 medications (fluticasone or prednisone) (15). Using data from that treatment study, we analyzed the ability of noninvasive biomarkers to diagnose or assess EE disease status. Patients were enrolled following endoscopic evaluation confirming the diagnosis of EE (defined as ≥15 eosinophils [eos]/hpf in esophageal mucosal biopsies without gastric or duodenal mucosal inflammation, and lack of response to acid suppression and/or a negative 24-hour pH monitoring study). Following consent, subjects were randomized to receive study medication (either swallowed fluticasone or prednisone) for 12 weeks (4-week induction and 8-week taper), followed by a 12-week follow-up period off medication. Serum and stool samples were collected at diagnostic EGD or at the time of enrollment (ie, week 0), at week 4 (with endoscopy), and at each visit thereafter (weeks 12, 18, and 24). Endoscopy after week 4 was performed only if clinically indicated. These samples were used to quantify EDN and IL-5 levels in subjects with EE. Symptoms (0 = none, 1 = present) were recorded at each visit. Esophageal histology (0 = normal, 1 = mild, 2 = moderate, 3 = severe) was recorded at each endoscopy and was based on eosinophil count and basal cell hyperplasia recorded as percentage of total esophageal epithelial thickness (15).
This study was approved by the institutional review board of Indiana University/Purdue University at Indianapolis (IUPUI) and conducted by the Division of Pediatric Gastroenterology, Hepatology, and Nutrition at Riley Hospital for Children in Indianapolis, Indiana. Eligible patients were ages 1 to 18 years and diagnosed as having EE. Exclusion criteria included any coexisting esophageal condition, Helicobacter pylori infection, inflammatory bowel disease, or inability to tolerate corticosteroids. Twenty normal patients were also enrolled for the purpose of 1-time measurements of the 3 biomarkers. These normal controls were children who underwent EGD for a variety of nonspecific reasons including recurrent abdominal pain and were clinically not believed to have an allergic intestinal disorder. All of them had histologically normal EGDs and biopsies of the esophagus, stomach, and duodenum. Serum and stool specimens were collected from each of the controls.
Serum and stool samples were collected from the subjects throughout the study, specifically at times with the highest likelihood of active disease (initial EGD or week 0), highest likelihood of response to drug (week 4), and highest probable chance of relapse (during the weaning period, or weeks 12, 18, and 24). Serum and stool samples were analyzed at the Allergic Diseases Research Laboratory at the University of Utah in Salt Lake City (G.J.G.). EDN was measured using a double antibody radioimmunoassay described in detail previously (16). IL-5 was measured using an enzyme-linked immunosorbent assay from BD Biosciences Pharmingen (San Diego, CA) per the manufacturer's instructions.
Subjects without a baseline measurement of a particular biomarker were excluded from all of the analyses of that biomarker. A few of the measured results for the biomarkers were several orders of magnitude higher than any other measurement in this study (and orders of magnitude higher than have been reported in the literature, even for patients with asthma or Graves disease) (17,18). In our primary analysis, we considered these high values to be possible artifacts rather than true reflections of biologic levels and excluded them. The outlier thresholds were determined by visual inspection of histogram data. Subjects with any outlier value in any week were excluded from analyses of that biomarker. Because it is also possible that the outlier values are true reflections of biologic levels, we did a secondary analysis in which we included all of the outlier values. Mean values of the biomarkers at different time points were compared pairwise using nonparametric (Wilcoxon) methods. Biomarker levels across the various categorical values of histology were compared at week 4 using Kruskal-Wallis tests.
We examined the relation among serum IL-5, serum EDN, and stool EDN levels, and sex, presence/absence of symptoms, corticosteroid therapy, histological changes, serum immunoglobulin E (IgE) levels, and eosinophil counts, and described serum EDN levels among those with and without atopy and food allergy.
The study cohort included 61 subjects with EE (46 boys [75%] and 15 girls [25%], mean age 7.5 ± 4.4 years) assigned to treatment (N = 31 with fluticasone and N = 30 with prednisone). There were 20 normal control subjects (10 boys [50%] and 10 girls [50%], mean age 6.7 ± 4.1 years).
Given that the normal ranges of serum IL-5, serum EDN, and stool EDN are not well established, we plotted the range of results for the study cohort and then visually defined the thresholds for outlier values: serum IL-5 > 300 pg/mL, serum EDN > 200 ng/mL, and stool EDN > 800 ng/mL. (For serum IL-5, there were 3 values we considered to be outliers: 336, 482, and 485 pg/mL—the highest value kept in the analysis was 153 pg/mL. For serum EDN, there were 5 values we considered to be outliers: 207, 376, 584, 985, and 1037 ng/mL—the highest value kept in the analysis was 132 ng/mL. For stool EDN, there were 14 values we considered to be outliers, ranging from 855 to 6287 ng/mL—the highest value kept in the analysis was 671 ng/mL.) Based on these procedures, among those with EE, 2, 3, and 7 subjects were excluded from the serum IL-5, serum EDN, and stool EDN analyses, respectively. In addition, 2 of the control subjects were excluded because of outlier values from stool EDN analyses. All of the outlier values were kept in the secondary analysis.
Longitudinal measurements of the 3 biomarkers across 24 weeks are shown in Table 1 and Figure 1. Table 1 also describes the results of the 1-time measurements made from the 20 control subjects. The subjects’ biomarker levels in each time period were compared with the preceding time period and the 1-time measurements made in control subjects.
The primary statistical analysis focused on a comparison of levels at baseline compared with week 4 (after 4 weeks of corticosteroid therapy at full dose). For IL-5 and stool EDN, almost all pair-wise comparisons of the cohort's mean level at each time point compared with the subsequent time point were not statistically significant (Table 1). In contrast, significant changes in serum EDN (significant decrease from baseline to week 4, and then rebound from week 4 to week 12) occurred. Serum EDN levels were stable after week 12. Serum IL-5 and stool EDN levels in subjects with EE were not statistically different from those of the control subjects when each time point for the cases was compared with the controls’ 1-time measurement.
Subjects with EE had serum EDN levels significantly higher than controls at baseline and at weeks 12, 18, and 24; the only time when this difference was not statistically significant was at week 4 (immediately after full therapy) when subjects with EE had mean serum EDN 6.4 ng/mL (±11.4) versus 2.7 ng/mL (±1.7) among controls.
The results for subgroups (boys, girls, fluticasone therapy, prednisone therapy) were generally parallel to the overall results: no pattern of statistically significant changes over time for IL-5 and stool EDN, but a statistically significant decline in serum EDN from baseline to week 4 in all subgroups, and some evidence for rebound in serum EDN values from week 4 to week 12 (Table 2). There were no statistically significant differences in temporal serum EDN or IL-5 patterns between the fluticasone and prednisone therapy subgroups. Temporal stool EDN patterns did differ between these groups, with stool EDN levels initially decreasing (baseline to week 4) in the prednisone group (Table 2).
In the secondary analysis, in which all of the values including outliers were included (data not shown), the presence of the outlier values increased mean values across the board for subjects with EE—and in fact by more than 2-fold for serum EDN in weeks 12, 18, and 24, and for stool EDN at baseline and at weeks 18 and 24—and more than doubled the mean stool EDN among the control subjects. The inclusion of the outliers, however, did not change the patterns of statistical significance for serum EDN, and did not alter the unremarkably flat pattern for serum IL-5. For stool EDN, the results with the outliers included showed a significant decrease in mean levels from baseline to week 4, and a significant increase from week 12 to week 18; the clinical utility of these results is not clear, because there was no significant difference in stool EDN between subjects with EE at baseline (305 ng/mL) and control subjects (285 ng/mL).
Symptoms and Histology
All of the subjects with EE in the analyses were symptomatic at baseline; most were asymptomatic from week 4 on. Statistical analyses (Wilcoxon) of the 9 subjects with symptoms after week 4 (comparing their most recent serum EDN value when symptoms were absent vs their serum EDN value when symptoms reappeared) were not significant. The same was true of analyses of IL-5 and stool EDN.
Similar results were noted in regard to histological changes and esophageal eosinophil density comparing weeks 0 and 4. The distribution of histology scores for the 55 subjects with EE in the serum EDN analyses is shown in Table 3. Analyses of bivariate correlation of biomarker level (each of the 3 biomarkers—serum IL-5, serum EDN, and stool EDN—analyzed separately) at week 0 against number of eosinophils per high-power field on esophageal biopsies (middle and distal esophagus, each analyzed separately) showed no correlation between eosinophils per high-power field and either stool EDN or serum IL-5; when eosinophil numbers on biopsy were compared against serum EDN, the Spearman ρ was 0.26 with P values (for middle and distal esophagus) between 0.05 and 0.10. When eosinophil numbers on biopsy at week 4 were compared against serum EDN at week 4, there was no significant correlation (and most of the subjects at week 4 had 0 eosinophils per high-power field). Among those with histological results at week 4, we also compared mean biomarker levels across the 4 histology categories (Table 3). Kruskal-Wallis tests were not significant for any of the biomarkers.
Correlation Between EDN and Other Parameters
At baseline, there was a significant correlation between higher serum EDN and higher serum IgE (r = 0.48, P < 0.01), whereas there was a significant correlation between lower stool EDN and higher serum IgE (r = 0.16, P = 0.05). At baseline, there was a significant correlation between higher serum EDN and higher eosinophils on complete blood count (r = 0.45, P < 0.01), whereas there was no significant correlation between stool EDN and blood eosinophils. Slightly more than half of the subjects had atopy (primarily asthma and/or eczema); slightly less than half had food allergies. The biomarker levels were similar among subjects with and without atopy, and with and without food allergy; patterns between week 0 and week 4 among these various groups also were similar. As an example, Table 4 shows serum EDN levels by the presence or absence of atopy and of food allergy.
Prediction of Relapse
Nonparametric statistical analyses of the 9 subjects with symptoms after week 4 (comparing their most recent serum EDN value when symptoms were absent vs their serum EDN value when symptoms reappeared) were not significant; the same was true of analyses of IL-5 and stool EDN. Analyses of these 9 subjects versus all of the other subjects with EE showed no statistically significant difference in the change in serum EDN from baseline to week 4, or from week 4 to week 12. The same was true of analyses of IL-5 and stool EDN. Four of 9 subjects with symptomatic relapse were in the fluticasone group, and 5 of 9 were in the prednisone group (P = NS). These results could have been affected by the small number of relapsed patients.
Although we did not make a formal Bonferroni (or other) correction for the multiple comparisons in this study, we nevertheless interpret our results cautiously in light of the multiple comparisons. Hence, although there were a few comparisons that were statistically significant for IL-5 or stool EDN, it was only in the analyses of serum EDN that a pattern of statistical significance emerged, involving both the intertemporal comparisons and the comparisons between subjects with EE versus the control subjects.
Management of EE can be costly, especially because of the need for repeated EGDs. This longitudinal study of 3 noninvasive biomarkers collected from children during a prospective, randomized trial explored the possibility that EE may be monitored without as many repeated EGDs (15). IL-5, a significant modulator of eosinophil function, and EDN, a marker of eosinophil degranulation, were measured.
The results of the biomarker analysis suggest a possible future role for the measurement of serum EDN in establishing the diagnosis of EE, assessing response to therapy, and/or monitoring for relapse or quiescence. Although this small study does not definitively establish a clear-cut clinical role for serum EDN, it does suggest that additional, larger studies may help delineate such a role. We found that serum EDN levels were significantly higher in subjects with EE than in normal subjects. The serum EDN levels of subjects with EE decreased significantly 4 weeks into the trial (after full therapy with prednisone or fluticasone), to a level close to that of normal subjects, and then rebounded significantly between week 4 and week 12. From week 12 to week 24, the serum EDN levels remained significantly high in subjects with EE. These trends appeared to be consistent in girls and boys, and in those who received fluticasone or prednisone, although we temper our interpretation of the subgroups with the knowledge that the numbers of subjects in these subgroups were small. Serum EDN levels, however, did not appear to vary with esophageal histology (at week 4) or with the presence/absence of atopy or food allergy (at baseline or week 4). There was no clear relation between serum EDN levels and EE symptoms. Although serum EDN levels and EE symptoms followed a similar trend from baseline to week 4 (all of the subjects with EE were symptomatic at baseline and had high serum EDN levels; no subjects were symptomatic at week 4 and mean serum EDN levels were significantly lower than at baseline), serum EDN levels were not associated with the subsequent recurrence of symptoms; however, only 9 patients in the study had symptomatic relapse. Additional, larger studies of the relation between symptoms and serum EDN may be necessary to further explore this relation.
Serum IL-5 and stool EDN did not appear to have any relation to the course of EE as was also found in the cross-sectional study by Konikoff et al (7). The results are intriguing because both IL-5 and eotaxin-3 are strong eosinophil chemoattractants (19). Additionally, and more important, IL-5 is upregulated in esophageal biopsies from children and adults with EE and controls (4,6). The ongoing clinical trials of anti-IL-5 agents in children with EE will further dissect this hypothesis. It would be logical that fecal EDN would be elevated in EE because it is a luminal eosinophil product. Eosinophils are a normal component of various parts of the gastrointestinal tract, and in the cecum, the normal density can be as high as 50 eos/hpf (20). We hypothesize that the normal eosinophil load of the GI tract mucosa is more than enough to obscure any alternations in fecal EDN, which may be encountered with changes in esophageal eosinophil load.
It is important to note that changes in EDN may reflect the disease activity not only of EE but also of other diseases involving infiltration of eosinophils. Many of the subjects in this study had atopy or food allergies. The patterns observed from week 0 to week 4 were not appreciably different between those with and without these other conditions. Still, we recommend that future studies of larger cohorts further explore the potential role of biomarkers among patients with EE with and without other illnesses in which EDN may play a role. These data reflect one of the first analyses of these biomarkers in a longitudinal study of EE. The study has limitations. For all 3 of the biomarkers studied there were some far outliers. In the primary analysis, we excluded those few subjects based on visual inspection of the results, but the true “normal range” for these biomarkers is not well defined. Subjects with an outlier value for 1 biomarker sometimes had no outlier values for the other biomarkers. It is possible that the few high outlier values of the biomarkers resulted from various aspects of specimen collection, storage, or assay; all of the samples were handled in similar fashion. It is also possible that the high outlier values are true subject levels. The patterns over time for serum EDN (which shows potential value as a biomarker) and for serum IL-5 (which does not) were essentially the same with or without inclusion of the outliers. Stool EDN had a pattern more suggestive of potential value as a biomarker when the outliers were added in, but this conclusion is not clear-cut. The mean stool EDN level was similar among control subjects and subjects with EE at baseline. Despite the limitations, the results for serum EDN seem to clearly reflect differences between subjects with EE and controls, and among subjects with EE over time, in relation to when therapy was given.
Our results contribute to the search for a noninvasive biomarker, which could serve to help monitor disease status in patients with EE. Clearly more needs to be done. To what should putative biomarkers be compared? Is histological resolution of the inflammation the correct criterion standard? There is a growing body of literature that suggests that a subset of patients with severe esophageal eosinophilia respond histologically to acid-suppressive therapy (21,22). Our results carry additional significance considering the studies, suggesting that anti–IL-5 agents may be useful in treating EE and other eosinophilic disorders (20,23,24). Although in this study there was no association between blood eosinophil count and stool EDN, there was a correlation between serum EDN and elevated serum IgE and between serum EDN and blood eosinophil count.
In summary, there is an unmet need in defining noninvasive biomarkers in patients with EE. In this longitudinal study, among the first to examine noninvasive biomarkers in children with EE, we noted some encouraging trends specifically for serum EDN levels in evaluating EE disease activity. Further exploring the role of biomarkers in EE will help us better understand this enigmatic disease, enhance development of targeted therapies, improve the ability to differentiate clinical symptoms as being caused by EE or an unrelated illness, and potentially allow less invasive care, while not totally eliminating the need for some endoscopies and biopsies.
1. Furuta GT, Liacouras CA, Collins MH, et al. Eosinophilic esophagitis in children and adults: a systematic review and consensus recommendations for diagnosis and treatment. Gastroenterology 2007; 133:1342–1363.
2. Spergel JM, Brown-Whitehorn TF, Beausoleil JL, et al. 14 years of eosinophilic esophagitis: clinical features and prognosis. J Pediatr Gastroenterol Nutr 2009; 48:30–36.
3. Gupta SK. Noninvasive markers of eosinophilic esophagitis. Gastrointest Endosc Clin N Am 2008; 18:157–167.
4. Straumann A, Kristl J, Conus S. Cytokine expression in healthy and inflamed mucosa: probing the role of eosinophils in the digestive tract. Inflamm Bowel Dis 2005; 11:720–726.
5. Blanchard C, Wang N, Stringer KF, et al. Eotaxin-3 and a uniquely conserved gene-expression profile in eosinophilic esophagitis. J Clin Invest 2006; 116:536–547.
6. Gupta SK, Fitzgerald JF, Kondratyuk T, et al. Cytokine expression in normal and inflamed esophageal mucosa: a study into the pathogenesis of allergic eosinophilic esophagitis (AEE). J Pediatr Gastroenterol Nutr 2006; 42:22–26.
7. Konikoff MR, Blanchard C, Kirby C, et al. Potential of serum eosinophils, eosinophil-derived neurotoxin, and eotaxin-3 as biomarkers of eosinophilic esophagitis. Clin Gastroenterol Hepatol 2006; 4:1328–1336.
8. Hogan SP, Foster PS, Rothenberg ME. Experimental analysis of eosinophil-associated gastrointestinal diseases. Curr Opin Allergy Clin Immunol 2002; 2:239–248.
9. Rothenberg ME, Hogan SP. The eosinophil. Annu Rev Immunol 2006; 24:147–174.
10. Slifman N, Loegering DA, McKean DJ, et al. Ribonuclease activity associated with human eosinophil-derived neurotoxin and eosinophil cationic protein. J Immunol 1986; 137:2913–2917.
11. Saitoh O, Kojima K, Sugi K, et al. Fecal eosinophil granule-derived proteins reflect disease activity in inflammatory bowel disease. Am J Gastroenterol 1999; 94:3513–3520.
12. Yang D, Chen Q, Rosenberg H, et al. Human ribonuclease a superfamily members, eosinophil-derived neurotoxin and pancreatic ribonuclease, induce dendritic cell maturation and activation. J Immunol 2004; 173:6134–6142.
13. Levy AM, Yamazaki K, Van Keulen VP, et al. Increased eosinophil infiltration and degranulation in colonic tissue from patients with collagenous colitis. Am J Gastroenterol 2001; 96:1522–1528.
14. Menies-Gow A, Robinson DS. Eosinophil chemokines and their receptors: an attractive target in asthma? Lancet 2000; 355:1741–1743.
15. Schaefer ET, Fitzgerald JF, Molleston JP, et al. Comparison of oral prednisone and topical fluticasone in the treatment of eosinophilic esophagitis: a randomized trial in children. Clin Gastroenterol Hepatol 2008; 6:165–173.
16. Abu-Ghazaleh RI, Dunnette SL, Loegering DA, et al. Eosinophil granule proteins in peripheral serum granulocytes. J Leukoc Biol 1992; 52:611–618.
17. Kim KW, Lee KE, Kim ES, et al. Serum eosinophil-derived neurotoxin (EDN) in diagnosis and evaluation of severity and bronchial hyperresponsiveness in childhood asthma. Lung 2007; 185:97–103.
18. Hidaka Y, Kimura M, Izumi Y, et al. Increased serum concentration of eosinophil-derived neurotoxin in patients with Graves’ disease. Thyroid 2003; 13:129–132.
19. Rothenberg ME, Klion AD, Roufosse FE, et al. Treatment of patients with the hypereosinophilic syndrome with mepolizumab. N Engl J Med 2008; 358:1215–1228.
20. Lowchik A, Weinberg AG. A quantitative evaluation of mucosal eosinophils in the pediatric gastrointestinal tract. Mod Pathol 1996; 9:110–114.
21. Ngo P, Furuta GT, Antonioli DA, et al. Eosinophils in the esophagus—peptic or allergic eosinophilic esophagitis? Case series of three patients with esophageal eosinophilia. Am J Gastroenterol 2006; 101:1666–1670.
22. Dranove JE, Horn DS, Davis MA, et al. Predictors of response to proton pump inhibitor therapy among children with severe esophageal eosinophilia. J Pediatr 2009; 154:96–100.
23. Stein ML, Collins MH, Villanueva JM, et al. Anti-IL-5 (mepolizumab) therapy for eosinophilic esophagitis. J Allergy Clin Immunol 2006; 118:1312–1319.
24. Straumann A, Conus S, Simon H, et al. Mepolizumab, a humanized monoclonal antibody to IL-5, for severe eosinophilic esophagitis in adults: a randomized, placebo-controlled double-blind trial. Gastroenterology 2007; 132:2586.