Prevalence and Predictors of Compromised Bone Mineral Density in Pediatric Eosinophilic Esophagitis

Abstract

Objectives:

The aim of the study was to determine whether patients with eosinophilic esophagitis (EoE) have lower bone mineral density (BMD) than expected and if bone deficits are more pronounced in subgroups of patients according to comorbidities (atopic disease and joint hypermobility) or treatments (dietary restriction, medication exposure).

Study Design:

Retrospective chart review was performed to obtain clinical data, including length of diagnosis, comorbidities, and methods of treatment for patients with EoE ages 3 to 21 years who had a lumbar spine dual-energy x-ray absorptiometry scan performed between 2014 and 2017. BMD was standardized by calculation of age, sex, and race-specific z scores.

Results:

A total of 269 patients met study criteria. The mean BMD z score (−0.55, 95% confidence interval: −0.68, −0.42) was lower than expected (P<0.0001), and the prevalence of low BMD z score (≤−2.0) was higher than expected (8.5%, 95% confidence interval: 5.2%–11.9%, P < 0.0001). In multivariable regression models, BMD z scores were −0.27 lower among those following an elimination diet and −0.65 lower among those with any lifetime use of a proton pump inhibitor (93% of the sample). There was no association with swallowed steroid use.

Conclusions:

In our sample, pediatric patients with EoE had a slightly lower BMD z score compared to peers, and the prevalence of low BMD was higher than expected. Taken cautiously given the limitations of our sample, risk factors for bone deficits included any lifetime use of proton pump inhibitor and a restrictive elimination diet, but not swallowed steroid use. Larger prospective studies are needed to better characterize risk factors for low BMD to help inform screening, selection of therapies, and provide appropriate anticipatory guidance for patients with EoE.

Eosinophilic esophagitis (EoE) is a chronic disease of the esophagus caused by an abnormal immunologic response to food antigens (1,2). EoE is diagnosed in patients with characteristic clinical history and histology including esophageal eosinophilia, defined by ≥15 intraepithelial eosinophils per high power field (eos/HPF). The most commonly prescribed therapies include proton pump inhibitors (PPIs), swallowed steroids, and food elimination diets (1-3). Steroid therapy is typically limited to swallowed asthma preparations (budesonide or fluticasone) that have extensive first pass metabolism and thus function topically with a lower systemic effect than systemic steroids such as prednisone. Many patients with EoE suffer from other atopic diseases, including asthma and eczema, that may lead to additional steroid exposure via inhaled, dermatologic, and/or systemic routes.

Glucocorticoid-induced bone disease is a well-known side effect of systemic steroids (4). The long-term effects of swallowed topical steroids and the combined effects of exposure to steroids via multiple routes are unknown. In addition to steroid exposure, patients with EoE have additional risk factors for low bone mineral density (BMD) including dietary restrictions, proton pump inhibitor (PPI) use (5,6), and frequent coexistence of connective tissue disease (7,8). The prevalence and predictors of low BMD in pediatric patients with EoE is unknown, limiting the ability to organize screening efforts and identify means to optimize BMD.

We aimed to determine whether pediatric patients with EoE have lower BMD than their peers and determine whether deficits are more pronounced in subgroups of patients according to degree of dietary restriction, steroid and PPI exposure, atopic history, and history of connective tissue disease.

METHODS

Study Design

This was a retrospective medical record review of patients with EoE cared for in the Cincinnati Children’s Hospital Medical Center (CCHMC) gastroenterology (GI) department. The CCHMC Institutional Review Board approved the study. Inclusion criteria were age between 3 and 21 years, diagnosis of EoE and having had a lumbar spine dual-energy x ray absorptiometry (DXA) scan performed between January 1, 2014 and December 31, 2017. Diagnosis of EoE was determined by chart review of clinical notes, problem lists, and biopsy results. Exclusion criteria included other causes of esophageal eosinophilia, diagnoses known to cause decreased BMD including inflammatory bowel disease, celiac disease, liver and/or small bowel transplant, and recent total parental nutrition use. The patient’s first DXA scan within the study period was used in our analysis. Our typical practice is to obtain a DXA scan of patients ≥3 years of age at time of initial EoE evaluation at the Cincinnati Center for Eosinophilic Disorders (CCED). The CCED serves local patients and as a national and international referral center.

To investigate whether patients who had a lumbar spine DXA scan were representative of all patients with EoE at our institution, we exported demographic and general medical information from electronic medical records (EMR) on patients diagnosed with EoE seen at a CCHMC GI appointment within the study period. We compared patients with and without a DXA scan using identical sources of EMR data. Recognizing the limitations of data extraction from EMR fields, we then performed a detailed, manual chart review of patients with a DXA scan to more thoroughly capture clinical and demographic data.

Bone Mineral Density

BMD (g/cm²) of the lumbar spine was measured by a Hologic Horizon (Hologic Inc, Marlborough, MA) DXA scanner. To account for expected differences in BMD with age, sex, and race, we calculated BMD z scores with use of reference data from healthy populations. The Bone Mineral Density in Childhood Study reference data were used to calculate age-, sex- and race-specific BMD z scores for children ages 5 to 20 years (9), and the Hologic reference data for children ages 3 to <5 years (10).

Demographics

Sex, ethnicity (Hispanic/non-Hispanic), race, and date of birth were abstracted for each patient with a DXA scan. The patient’s age at the time of the DXA scan was calculated. Weight and height measurements within 2 months of the DXA scan were used to calculate body mass index (BMI, kg/m²). Z scores for weight-for-age, height-for-age (HAZ), and BMI were calculated using the Centers for Disease Control and Prevention growth reference.

Eosinophilic Esophagitis Disease Burden

EoE disease activity was defined histologically by esophageal eosinophilia within 6 months before the DXA scan. Inactive disease was defined as <15 eos/HPF and active disease as ≥15 eos/HPF, in agreement with the current consensus guidelines (11). If the date of diagnosis was unavailable, disease duration was calculated from age at diagnosis.

Comorbid Disease History

History of comorbid atopic disease, connective tissue disease, and eosinophilic gastrointestinal disease beyond the esophagus was established based on the problem list and clinical notes. IgE-mediated food allergy was defined using the problem list, clinical notes, and/or medication prescription for emergency injectable epinephrine. Connective tissue disease was limited to joint hypermobility or Ehlers-Danlos syndrome.

Medication Exposures

We collected information on type, dose, and length of therapy with swallowed steroids. If start and end dates were unavailable, we estimated length of therapy by using the midpoint of a range estimated by chart review. We categorized use of PPI and steroids (systemic, oral, nasal, topical, and inhaled) as lifetime (ie, ever), within the year preceding the DXA scan, or at the time of the DXA scan. We used these broad ranges as information on dosage and length of these therapies was not consistently recorded in the medical record.

Diet Exposures

Elimination diets included a 6-food elimination diet (6FED-milk, soy, wheat, egg, peanuts/treenuts, fish/shellfish), or 5FED (6FED except fish/shellfish). Dairy elimination was defined as eliminating dairy separate from a 6- or 5FED. Patients were considered to be on an elemental diet if their main nutritional source was elemental formula only (complete) or with ongoing food trials to add additional foods (partial). Serum 25-OH vitamin D level was abstracted if the value was obtained within 1 year before the DXA scan.

Statistical Analysis

Data were analyzed using SAS, version 9.4 (SAS Institute, Cary, NC). Continuous data were summarized as median with interquartile range (25th and 75th percentiles) or mean ± standard deviation. Categorical data were summarized as frequency counts and percentages. Chi-square and Kruskal-Wallis tests were used for comparisons of patients with EoE with and without DXA scans. One sample t tests or binomial proportion tests were used to examine whether the mean BMD z score differed from the expected value of 0.0 and if the prevalence of low BMD (z score ≤−2) differed from the expected value of 2.5%, respectively (12). We used multivariable generalized linear regression models to identify factors associated with BMD z scores (log transformation was applied to cumulative swallowed corticosteroid use duration due to skewing). Variables were tested for association with spine BMD z score as a continuous variable, as well as the dichotomous variable, low BMD z score. Both models controlled for BMI z score and HAZ. Variables from these models that had a P ≤ 0.20 were considered in multivariable models. Stepwise selection was used to create 2 final models, 1 containing variables significantly associated with the continuous variable BMD z score and 1 containing variables significantly associated with the dichotomous variable, low BMD z score. A P<0.05 was considered significant.

RESULTS

Study Participants

During the study period, 365 unique patients had a DXA scan. Ninety-six patients were excluded: 50 who lacked confirmed EoE diagnosis, 3 did not meet age criteria, and 43 had an exclusionary diagnosis. Data from the remaining 269 individuals were used in analyses.

We compared patients with EoE that had a DXA scan with those that did not have a DXA scan (n=1945) using data extracted from the EMR. Patients with and without a DXA scan were similar in terms of sex, race, ethnicity, HAZ, and atopic history (Table 1). Patients with a DXA scan had a younger age at their first visit within the study interval, lower weight and BMI for age z scores, higher prevalence of comorbidities, and increased frequency of swallowed and inhaled steroids compared to those without a DXA scan (Table 1). A surprisingly large percentage of patients in both groups had a history of systemic steroid use, most commonly due to single doses of dexamethasone given prophylactically with anesthesia during procedures rather than longer courses to control disease.

TABLE 1.

Characteristics of eosinophilic esophagitis patients that had a dual-energy x-ray absorptiometry scan and eosinophilic esophagitis patients without a dual-energy x-ray absorptiometry scan

	EoE patients with DXA N=269*	EoE patients without DXA N=1945*	P
Age at first visit, y	9.5 (6.3–12.7)	11.8 (7.8–15.4)	<0.0001
Sex (male)	204 (76%)	1391 (72%)	0.14
Race			0.44
White/Caucasian	237/259 (92%)	1728/1931 (89%)
Black/African American	10/259 (4%)	112/1931 (6%)
Other	12/259 (5%)	91/1931 (5%)
Ethnicity (Hispanic)	12/267 (4%)	35/1936 (2%)	0.004
BMI z score at visit	0.0 (−0.7, 0.6)	0.2 (−0.7, 1.1) n=1379	0.01
HAZ at first visit	−0.2 (−0.8, 0.5)	−0.1 (−0.9, 0.7) n=1382	0.61
WAZ at first visit	−0.2 (−0.8, 0.5)	0.1 (−0.8, 1.0) n=1380	0.003
Comorbid disease history
Asthma	157 (58%)	912 (47%)	0.0004
Eczema	89 (33%)	459 (24%)	0.001
Allergic rhinitis	128 (48%)	712 (37%)	0.0005
IgE-mediated food allergy	167 (62%)	983 (51%)	0.0004
Hypermobility/EDS/connective tissue disease	54 (20%)	243 (12%)	0.001
Mitochondrial disorder	11 (4%)	53 (3%)	0.21
Medication use
Proton pump inhibitor, ever	246 (91%)	1814 (93%)	0.27
Swallowed topical steroid, ever	187 (70%)	1085 (56%)	<0.0001
Oral systemic corticosteroid, ever	260 (97%)	1804 (93%)	0.02
Inhaled steroid, ever	228 (85%)	1355 (70%)	<0.0001
Nasal steroid, ever	110 (41%)	713 (37%)	0.18
Topical steroid, ever	88 (33%)	612 (31%)	0.68

Data extracted from EMR system records. Data presented as median (25th, 75th percentile) or n (%).

BMI = body mass index; EDS = Ehlers-Danlos syndrome; HAZ = height-for-age z score; WAZ = weight-for-age z score.

^*Sample size unless otherwise specified.

Patient characteristics, EoE disease burden, medication and dietary exposures, and vitamin D levels based on detailed chart review for the patients with a DXA scan are shown in Table 2. Data in Table 2 differ from Table 1 as they reflect the comprehensive review of the medical record of patients having DXA scans.

TABLE 2.

Eosinophilic esophagitis patient characteristics, dietary, and medication exposures and their association with bone mineral density z score in regression models

Characteristic	N	Median (IQR) or frequency (%)	P value for predicting BMD z score	P value for predicting low BMD z score (<−2)
Age at EoE diagnosis, y	231	5.9 (2.4–9.5)	0.15*	0.26
Age at DXA, y	269	10.1 (6.5, 13.2)	0.03*	0.08*
Sex	269		0.04*	0.08*
Male		204 (76%)
Female		65 (24%)
Race	259		0.73	0.86
White/Caucasian		237 (92%)
Other		22 (8%)
Ethnicity	267		0.68	0.27
Hispanic		12 (4%)
Non-Hispanic		255 (96%)
BMI z score	269	0.1 (−0.7, 0.7)	<0.0001*	0.002*
Height-for-age z score	269	−0.2 (−0.8, 0.5)	<0.0001*	<0.0001*
EoE disease history
Disease activity:	258		0.80	0.10*
Inactive (<15 Eos/HPF)		120 (47%)
Active (≥15 Eos/HPF)		138 (53%)
EoE duration before DXA, y	257	3.0 (1.3–5.7)	0.20*	0.33
Disease burden	246
Disease activity (active)			0.63	0.26
EoE duration before DXA, y			0.30	0.43
Comorbid disease history
Asthma	269	136 (51%)	0.85	0.79
Eczema	269	150 (56%)	0.82	0.76
Allergic rhinitis	269	201 (75%)	0.20*	0.04*
IgE-mediated food allergy	269	134 (50%)	0.0495*	0.38
Hypermobility	269	44 (16%)	0.05*	0.51
EGID	269	26 (10%)	0.81	0.28
Mitochondrial disorder	269	7 (3%)	0.07*	0.07*
Medication exposures
Proton pump inhibitor
Ever (any lifetime use)	269	251 (93%)	0.02*	0.58
Within 1 year before DXA	259	198 (76%)	0.87	0.14*
At time of DXA	269	173 (64%)	0.52	0.17*
Swallowed topical corticosteroid
Ever	269	202 (75%)	0.88	0.74
Within 1 year before DXA	264	160 (61%)	0.78	0.40
At time of DXA	269	125 (46%)	0.26	0.33
Cumulative duration, mo	186	13.5 (5.0–31.4)	0.42	0.98
Oral systemic corticosteroid
Ever	269	66 (25%)	0.50	0.88
Within 1 year before DXA	269	32 (12%)	0.69	0.69
Inhaled corticosteroid use
Ever	269	77 (29%)	0.99	0.97
Within 1 year before DXA	269	60 (22%)	0.78	0.99
Nasal/topical corticosteroid
Ever	269	89 (33%)	0.11*	0.13*
Within 1 year before DXA	269	57 (21%)	0.47	0.69
Combined inhaled + swallowed steroid use	269
Inhaled ever			0.99	0.99
Swallowed steroids ever			0.88	0.74
Dietary exposures
Elimination diet (5 or 6 food):
Ever	269	142 (53%)	0.03*	0.19*
Within 1 year before DXA	263	93 (35%)	0.80	0.25
At time of DXA	269	78 (29%)	0.61	0.27
Excluded dairy separate from elimination diet
Ever	269	128 (48%)	0.19*	0.74
Within 1 year before DXA	267	96 (36%)	0.46	0.79
At time of DXA	269	76 (28%)	0.37	0.65
Elemental diet use:
Ever	269	103 (38%)	0.22	0.06*
Within 1 year before DXA	261	82 (31%)	0.27	0.07*
At time of DXA	269	76 (28%)	0.34	0.06*
Serum 25-(OH) vitamin D (ng/dL)	245	33 (26–39)	0.32	0.60

Data collected from manual chart review separate from exported EMR data used in Table 1.

BMD = bone mineral density; BMI = body mass index; DXA = dual-energy x-ray absorptiometry; EGID = eosinophilic gastrointestinal disease; EoE = eosinophilic esophagitis; HPF = high powered field; IQR = interquartile range, 25th, 75th percentiles.

^*Variables considered in final multivariable regression model.

Italics were used to make P values <0.05 stand out.

Mean Bone Mineral Density z Score and Prevalence of Low Bone Mineral Density z Score

The mean lumbar spine BMD z score of −0.55 (95% confidence interval [CI]: −0.68, −0.42) was lower than expected and significantly differed from 0 (P<0.0001). The prevalence of a clinically low BMD z score (≤−2.0) was 8.6% (23/269; 95% CI: 5.2–11.9) and was statistically higher than the predicted value of 2.5% based on the normal distribution (P<0.0001). Examining the prevalence of low BMD is important as these children are at greatest risk of fracture.

Predicting Bone Mineral Density z Score

Disease and medication exposures associated with BMD z score with a P value of ≤0.2 after adjusting for BMI and HAZ were included in multivariable models (Table 2). Of note, each variable was negatively associated with BMD z score. The final model predicting BMD z score included age at EoE diagnosis, IgE-mediated food allergy, PPI use (ever), and elimination diet (ever) with PPI use having the largest regression coefficient (Table 3). Using this model, we can predict the BMD z score of patients with a combination of risk factors. For example, a patient with a BMI z score of 0.1, HAZ score of −0.2, diagnosed with EoE at age 6 years, without IgE-mediated food allergy, and with any lifetime use of PPI and elimination diet has a predicted BMD z score of: 0.7299 + (0.1 × 0.23) + (−0.2 × 0.27) + (6 × −0.04) + (0 × −0.28) + (1 × −0.65)+(1 × −0.27) = −0.46.

TABLE 3.

Final multivariable model predicting spine bone mineral density z score in pediatric patients with eosinophilic esophagitis

	Spine BMD z score N=231 *
	Regression coefficient ± Standard Error	P
Intercept	0.7299±
BMI z score	0.23±0.05	<0.0001
Height-for-age z score	0.27±0.06	<0.0001
Age at EoE diagnosis, y	−0.04±0.02	0.007
IgE-mediated food allergy	−0.28±0.13	0.04
PPI use, ever (any lifetime use)	−0.65±0.28	0.02
Elimination diet (5FED or 6FED), ever	−0.27±0.14	0.045

BMD = bone mineral density; BMI = body mass index; EoE = eosinophilic esophagitis; 6FED = 6-food elimination diet; PPI = proton pump inhibitor.

^*The sample size was <269 owing to missing data in some predictors.

To help determine whether the lower than expected lumbar spine BMD z score in patients with EoE with a DXA was potentially due to their lower BMI and height z scores compared to those patients with EoE without a DXA, we estimated the lumbar spine BMD z score assuming the median BMI z score and HAZ values from Table 1 for patients who did not have a DXA scan and coefficients from a regression model (see above). The estimated BMI z-score and HAZ adjusted mean BMD z score for those without a DXA was −0.48 versus −0.55 (the mean BMD z score of our EoE sample with a DXA) accounting for the differing BMI z-score and HAZ.

Predicting Low Bone Mineral Density

Although our sample size of patients with low BMD (n=23) was small, greater BMI and HAZ-scores were both associated with reduced risk of low BMD as expected (Table 4). Male sex was associated with an increased risk (odds ratio 4.84, 95% CI [1.12–20.9], P=0.03) of low BMD whereas active EoE was associated with a reduced risk of low BMD (odds ratio 0.34, 95% CI [0.12–0.95], P=0.04).

TABLE 4.

Final multivariable logistic regression model predicting low spine bone mineral density z score (<−2) in pediatric patients with eosinophilic esophagitis

	Low spine BMD z score (<−2) N=258*
	Odds ratio (95% CI)	P
BMI z score	0.43 (0.28–0.68)	0.0002
Height-for-age z score	0.33 (0.20–0.57)	<0.0001
Sex (male vs female)	4.84 (1.12–20.91)	0.03
Disease activity (active vs inactive)	0.34 (0.12–0.95)	0.04

BMD = bone mineral density; BMI = body mass index.

^*The sample size was <269 owing to missing data in some predictors.

DISCUSSION

In this retrospective study we found that BMD of pediatric patients with EoE was lower than expected for children of similar age and sex, and the prevalence of low BMD z score (≤−2.0) was significantly higher than expected. The lower than expected BMD z score for patients with EoE with DXA scans was only modestly accounted for (−0.55 vs −0.48) by the lower BMI and height z scores of the DXA patients compared to those that did not have a DXA scan.

The deficit in BMD was greatest in those with any lifetime use of PPI and a history of an elimination diet. The effect of PPIs on bone health remain controversial (13,14), and our findings regarding PPI use should be considered cautiously. Only 7% (18 patients) of our study sample did not have a history of PPI use; thus, it is difficult to make strong conclusions about PPI use predicting BMD z score in the greater EoE population. We speculate that finding an association with any lifetime PPI use, but not PPI use in the prior year, was because any lifetime use reflected a longer duration of exposure. Additional studies in this patient population are needed to verify or refute our findings as PPIs are commonly used in patients with EoE.

A history of a restrictive elimination diet also was associated with a small decrease in BMD z score. Similar to the results presented for PPI use, only historical elimination diet use was associated with BMD, whereas elimination diet at the time of or within the year before the DXA scan were not associated with BMD, likely because historical use captures the complete duration of dietary elimination more accurately. Although our standard of care includes regular dietician visits for patients on restrictive diets, patients on elimination diets still had slightly lower BMD z scores. Many institutions have less access to dieticians, which may accentuate this risk. These data stress the importance of dietary counseling in pediatric EoE and consideration of minimally restrictive diets.

An additional concern is the effect of swallowed topical steroids and the total cumulative steroid burden on bone health. We did not find an association between historical use or cumulative duration of use of swallowed topical steroids and BMD. Given the high percentage of pediatric EoE patients who also have asthma and require combined inhaled and swallowed steroids, it is reassuring that the combination of the 2 preparations did not have a detectable association with BMD z scores. Our results are complementary to a recent study showing a negligible effect on linear growth from swallowed topical steroid use in pediatric patients with EoE (15). Interestingly, we also did not find an association with systemic steroid use, which is known to negatively affect bone health (4,16). This lack of effect was likely due to the short (5 days or less) systemic steroid courses the patients in this cohort were prescribed, rather than prolonged use as is commonly seen in patients with chronic inflammatory diseases.

An older age at diagnosis was inversely associated with BMD z scores, but the regression coefficient (−0.04 per year, Table 3) was not clinically significant. In contrast, patients with active EoE at time of DXA had a decreased risk of low BMD (BMD ≤−2.0) but given the small number of patients with low BMD it is difficult to draw firm conclusions. Prospective longitudinal studies are needed to further assess these risks. It was difficult to assess disease burden in this retrospective chart review, as esophageal eosinophilia does not correlate well with symptom burden in this disease (17). CCHMC is an international tertiary referral center for eosinophilic disease that often sees children with a severe, refractory disease phenotype. The decreased BMD in our cohort may in part reflect this more severe population. Therefore, the lack of association between the degree of esophageal inflammation and BMD may be clinically less meaningful as our entire cohort is of higher acuity.

Unexpectedly, a history of IgE-mediated food allergy was a risk factor for lower BMD z score. This may be due in part an increased likelihood of restricted diets in patients with IgE-mediated allergies, but we did not have sufficient information to examine this.

LIMITATIONS

Although there was no written protocol at the time of the study, our practice was to obtain a baseline lumbar spine DXA scan of patients ≥3 years old at time of initial evaluation at the CCED. We acknowledge that DXA scans could have been obtained on the more severely affected, and therefore, we investigated the potential for selection bias. Even though patients with EoE who had a DXA scan had a higher proportion of swallowed topical and inhaled steroid use as well as a higher proportion of atopic comorbidities compared to patients with EoE that did not have a DXA scan, none of these characteristics had an appreciable impact on bone density in our analysis. Regardless, the regression model used to assess risk factors with bone deficits is only valid for this specific sample that had DXA scans and may not be representative of the general pediatric EoE population.

Another limitation is potentially incomplete information in the medical record and resultant misclassification of exposures, which is inherent in a retrospective chart review. Importantly, we were unable to capture duration and type of PPI use. In addition, we were unable to capture cumulative swallowed steroid dose, and it remains unclear if patients taking higher doses of a swallowed steroid are at higher risk for lower BMD. Our findings are thus a starting point and should be explored in prospective studies.

We identified a higher prevalence of low BMD in pediatric EoE than expected (8.5% vs 2.5%). Due to the small number of individuals classified as having a low BMD, we, however, had limited power to identify predictors of low BMD in this population. Data regarding long bone fractures could not be captured accurately, and thus we are unable to comment on this clinical consequence of the lower BMD in our population. Larger prospective investigations are needed to better identify pediatric patients with EoE at greatest risk of bone deficits.

CONCLUSION

In our sample, pediatric patients with EoE had a slightly lower BMD z score compared to their peers, and the prevalence of low BMD in these patients is higher than expected. There are no clear guidelines to identify pediatric patients with EoE at risk for low BMD and its consequences. To help fill this gap, we examined clinical risk factors related to BMD in a large retrospective pediatric EoE cohort. The biggest risk factors included any lifetime use of PPI, history of IgE-mediated food allergy and restrictive elimination diets. The effects of PPI use on BMD should be taken cautiously given limitations in our sample. Importantly, swallowed steroid use was not associated with lower BMD. Our study supports the need for prospective research to identify treatments associated with clinically significant reductions in BMD, to develop screening tools to identify pediatric patients with EoE at greatest risk of bone density deficits and to design strategies to mitigate these risks.

What Is Known

• Many patients with eosinophilic esophagitis are atopic and are at risk for a large combined glucocorticoid burden, including the use of inhaled, nasal, topical, and systemic steroids.

• Glucocorticoid-induced bone disease is a known side effect of systemic steroids, but the long-term effects of swallowed topical steroids, combined steroid burden, and elimination diets are unknown.

What Is New

• Pediatric eosinophilic esophagitis patients have a slightly lower lumbar spine bone mineral density z score compared to children of similar age, sex, and race.

• Prevalence of low bone mineral density is slightly higher in the pediatric eosinophilic esophagitis population than the healthy population.

• Risk factors associated with lower bone mineral density z score, which should be taken with caution given limitations in our sample, include any history of proton pump inhibitor use, elimination diet, and IgE-mediated food allergy, and older age at eosinophilic esophagitis diagnosis, but not swallowed topical steroid use.