Eosinophilic Esophagitis and Celiac Disease
The search strategy yielded 197 references; 153 documents were excluded after examining the title and abstract because they did not fulfil the inclusion criteria. Of the remaining 43 references, four abstracts were excluded either because they were subsequently published as full papers or because they had been presented multiple times at different conferences. For the remaining 39 references considered to be potentially relevant, the full text was retrieved for detailed evaluation. Of these, nine were excluded because they did not include data suitable for calculations. In the end, 30 studies were included in the systematic review (Figure 1).
(Enlarge Image)
Figure 1.
Flow chart for the process of identifying studies that were included in and excluded from the systematic review.
The main characteristics of each study are summarised in Table 1. Of the 30 documents analysed, 10 were full text articles and 20 were abstracts. Overall, data from 7043 patients (3809 children, 2850 adults and 384 not determined) were retrieved, with the size of the various study populations ranging from 1 to 1439 cases.
Fifteen studies reported on the prevalence of EoE among CD patients; most of these were short, retrospective case reports of paediatric populations (Table 1). Overall, six case reports included 13 patients with both EoE and CD. The remaining documents were predominantly short case series, including a total of 78 EoE patients among 4131 patients with CD.
The prevalence of EoE among CD patients ranged from 0% to 10.71% (I = 78.9%) (Figure 2). Differences were observed when paediatric and adult patients were analysed separately (0–0.87% vs. 0.87–10.71) (Figure 3a, b, respectively), although in both cases a wide variability was found (I = 63.2% and 75.2%, respectively).
(Enlarge Image)
Figure 2.
Studies evaluating the prevalence of EoE among patients with CD, which ranged from 0% to 10.71%. A summary estimate of combined prevalence rates was not performed due to clinical and methodological heterogeneity. The I value of 78.9% indicates that intra-study differences (statistical heterogeneity) account for 78.9% of the variability in the overall effect size.
(Enlarge Image)
Figure 3.
Subgroup analysis of studies evaluating the prevalence of EoE among patients with CD; (a) prevalence in paediatric coeliac population; (b) prevalence of EoE in adult CD patients; (c) prevalence in studies published as full articles, and (d) prevalence in studies published as abstracts. I denotes intra-study differences in statistical heterogeneity.
Funnel plot analysis revealed a significant publication bias (P value for the Egger test <0.0001; P value for the Harbord test = 0.0048) (Figure 4a); studies that included small numbers of coeliac patients with an increased prevalence of EoE (thus favouring the existence of an association between both disorders) were predominant.
(Enlarge Image)
Figure 4.
Begg funnel plot of studies evaluating publication bias of articles on the relationship between EoE and CD. (a) Studies assessing the prevalence of EoE among coeliac patients; (b) studies assessing the prevalence of CD among patients with EoE, and (c) studies evaluating the ability of a gluten-free diet to achieve histological remission of EoE in patients who also presented CD. Statistically significant publication bias was demonstrated by the fact that studies including small numbers of coeliac patients with an increased prevalence of EoE predominated.
Ten studies reported on the prevalence of CD among EoE patients (Table 1), including the aforementioned six documents which described 13 patients sharing both disorders. Most of the remaining studies were retrospective case series reports and included a total of 1905 EoE patients, of whom 49 also presented CD. The vast majority of patients were children. All of these studies were judged as having low methodological quality.
The combined prevalence of CD among EoE patients varied from 0.16% to 57.7% (I = 89%) (Figure 5). However, one high-quality, prospective, randomly selected, population-based study carried out on 1000 adult patients identified 11 coeliac subjects, none of whom presented with EoE.
(Enlarge Image)
Figure 5.
Studies evaluating the prevalence of EoE among patients with CD, which ranged from 0.16% to 57.7%. A summary estimate of combined prevalence was not calculated because of clinical and methodological heterogeneity among selected studies. The I value of 89% indicates that intra-study differences (statistical heterogeneity) account for 89% of the variability in the overall effect size.
Once again, funnel plot analysis identified a clear publication bias in favour of studies showing a positive association between EoE and CD (Figure 4b), a finding that was statistically confirmed (P value for the Egger test = 0.0101; P value for the Harbord test = 0.021).
The combined efficacy documented in the 15 studies evaluating the efficacy of a GFD in reversing EoE among coeliac patients (11 of which were carried out on paediatric populations and four on adult patients) was 32.1% (95% CI, 14.9–52.2%). Combined results from 61 patients (53 children and eight adults) were highly heterogeneous (I = 64.7%), regardless of the age of the population being assessed (68.3% and 60% for children and adults, respectively) (Table 2 and Figure 6).
(Enlarge Image)
Figure 6.
Overall combined effects of a gluten-free diet (GFD) for inducing remission of oesophageal eosinophilic infiltration in coeliac patients who also suffered from eosinophilic oesophagitis. Percentage of histological improvement after following a GFD was extracted from each article/abstract and 95% confidence intervals were calculated using the exact binomial method. A random-model effect was used to calculate the overall effect size. The I value of 64.7% indicates that intra-study differences (heterogeneity) account for 64.7% of the variability in the overall effect size.
A significant publication bias in favour of reporting a positive effect of a GFD in achieving EoE remission was documented with every statistical test performed (P value for the Begg–Mazumdar test = 0.0009; P value for the Egger test = 0.0143; P value for the Harbord test = 0.0147).
Two studies have determined the presence of HLA-DQ2 and DQ8 conferring risk for CD in patients with EoE: One Australian case series of 10 correlatively diagnosed EoE patients reported that eight of the 10 expressed the HLA-DQ2 haplotype (with a frequency affecting approximately 45% of the local population), with another patient presenting DQ8. This increased frequency of the HLA-DQ2 and DQ8 alleles predisposing for CD was not documented in a multicentre, observational study carried out in Spain on 78 adult EoE patients as compared with the allelic frequency among healthy individuals. Hence, a common genetic basis for EoE and CD cannot be established.
The clinical and methodological variability in the studies examined prevented us from calculating summary estimates; however, a tentative analysis of subgroups categorised according to quality and type of document was carried out (Table 3). In the hypothetical case that the retrieved studies had shown enough consistency to have undergone meta-analysis, subgroup analyses would have indicated a stronger association between EoE and CD in low-quality studies compared with those considered of mild/high quality. Regarding the type of publication, a higher association between CD and EoE would have been observed in research published as an abstract compared to that published as a full paper (Figure 3c, d, respectively). Finally, the coexistence of EoE and CD would have been more common in children than in adult patients.
Results
The search strategy yielded 197 references; 153 documents were excluded after examining the title and abstract because they did not fulfil the inclusion criteria. Of the remaining 43 references, four abstracts were excluded either because they were subsequently published as full papers or because they had been presented multiple times at different conferences. For the remaining 39 references considered to be potentially relevant, the full text was retrieved for detailed evaluation. Of these, nine were excluded because they did not include data suitable for calculations. In the end, 30 studies were included in the systematic review (Figure 1).
(Enlarge Image)
Figure 1.
Flow chart for the process of identifying studies that were included in and excluded from the systematic review.
The main characteristics of each study are summarised in Table 1. Of the 30 documents analysed, 10 were full text articles and 20 were abstracts. Overall, data from 7043 patients (3809 children, 2850 adults and 384 not determined) were retrieved, with the size of the various study populations ranging from 1 to 1439 cases.
Prevalence of EoE Among CD Patients
Fifteen studies reported on the prevalence of EoE among CD patients; most of these were short, retrospective case reports of paediatric populations (Table 1). Overall, six case reports included 13 patients with both EoE and CD. The remaining documents were predominantly short case series, including a total of 78 EoE patients among 4131 patients with CD.
The prevalence of EoE among CD patients ranged from 0% to 10.71% (I = 78.9%) (Figure 2). Differences were observed when paediatric and adult patients were analysed separately (0–0.87% vs. 0.87–10.71) (Figure 3a, b, respectively), although in both cases a wide variability was found (I = 63.2% and 75.2%, respectively).
(Enlarge Image)
Figure 2.
Studies evaluating the prevalence of EoE among patients with CD, which ranged from 0% to 10.71%. A summary estimate of combined prevalence rates was not performed due to clinical and methodological heterogeneity. The I value of 78.9% indicates that intra-study differences (statistical heterogeneity) account for 78.9% of the variability in the overall effect size.
(Enlarge Image)
Figure 3.
Subgroup analysis of studies evaluating the prevalence of EoE among patients with CD; (a) prevalence in paediatric coeliac population; (b) prevalence of EoE in adult CD patients; (c) prevalence in studies published as full articles, and (d) prevalence in studies published as abstracts. I denotes intra-study differences in statistical heterogeneity.
Funnel plot analysis revealed a significant publication bias (P value for the Egger test <0.0001; P value for the Harbord test = 0.0048) (Figure 4a); studies that included small numbers of coeliac patients with an increased prevalence of EoE (thus favouring the existence of an association between both disorders) were predominant.
(Enlarge Image)
Figure 4.
Begg funnel plot of studies evaluating publication bias of articles on the relationship between EoE and CD. (a) Studies assessing the prevalence of EoE among coeliac patients; (b) studies assessing the prevalence of CD among patients with EoE, and (c) studies evaluating the ability of a gluten-free diet to achieve histological remission of EoE in patients who also presented CD. Statistically significant publication bias was demonstrated by the fact that studies including small numbers of coeliac patients with an increased prevalence of EoE predominated.
Prevalence of CD Among Patients With EoE
Ten studies reported on the prevalence of CD among EoE patients (Table 1), including the aforementioned six documents which described 13 patients sharing both disorders. Most of the remaining studies were retrospective case series reports and included a total of 1905 EoE patients, of whom 49 also presented CD. The vast majority of patients were children. All of these studies were judged as having low methodological quality.
The combined prevalence of CD among EoE patients varied from 0.16% to 57.7% (I = 89%) (Figure 5). However, one high-quality, prospective, randomly selected, population-based study carried out on 1000 adult patients identified 11 coeliac subjects, none of whom presented with EoE.
(Enlarge Image)
Figure 5.
Studies evaluating the prevalence of EoE among patients with CD, which ranged from 0.16% to 57.7%. A summary estimate of combined prevalence was not calculated because of clinical and methodological heterogeneity among selected studies. The I value of 89% indicates that intra-study differences (statistical heterogeneity) account for 89% of the variability in the overall effect size.
Once again, funnel plot analysis identified a clear publication bias in favour of studies showing a positive association between EoE and CD (Figure 4b), a finding that was statistically confirmed (P value for the Egger test = 0.0101; P value for the Harbord test = 0.021).
Efficacy of a Gluten-free Diet on the EoE Remission
The combined efficacy documented in the 15 studies evaluating the efficacy of a GFD in reversing EoE among coeliac patients (11 of which were carried out on paediatric populations and four on adult patients) was 32.1% (95% CI, 14.9–52.2%). Combined results from 61 patients (53 children and eight adults) were highly heterogeneous (I = 64.7%), regardless of the age of the population being assessed (68.3% and 60% for children and adults, respectively) (Table 2 and Figure 6).
(Enlarge Image)
Figure 6.
Overall combined effects of a gluten-free diet (GFD) for inducing remission of oesophageal eosinophilic infiltration in coeliac patients who also suffered from eosinophilic oesophagitis. Percentage of histological improvement after following a GFD was extracted from each article/abstract and 95% confidence intervals were calculated using the exact binomial method. A random-model effect was used to calculate the overall effect size. The I value of 64.7% indicates that intra-study differences (heterogeneity) account for 64.7% of the variability in the overall effect size.
A significant publication bias in favour of reporting a positive effect of a GFD in achieving EoE remission was documented with every statistical test performed (P value for the Begg–Mazumdar test = 0.0009; P value for the Egger test = 0.0143; P value for the Harbord test = 0.0147).
Genetic Associations Between EoE and CD
Two studies have determined the presence of HLA-DQ2 and DQ8 conferring risk for CD in patients with EoE: One Australian case series of 10 correlatively diagnosed EoE patients reported that eight of the 10 expressed the HLA-DQ2 haplotype (with a frequency affecting approximately 45% of the local population), with another patient presenting DQ8. This increased frequency of the HLA-DQ2 and DQ8 alleles predisposing for CD was not documented in a multicentre, observational study carried out in Spain on 78 adult EoE patients as compared with the allelic frequency among healthy individuals. Hence, a common genetic basis for EoE and CD cannot be established.
Subgroup Analysis
The clinical and methodological variability in the studies examined prevented us from calculating summary estimates; however, a tentative analysis of subgroups categorised according to quality and type of document was carried out (Table 3). In the hypothetical case that the retrieved studies had shown enough consistency to have undergone meta-analysis, subgroup analyses would have indicated a stronger association between EoE and CD in low-quality studies compared with those considered of mild/high quality. Regarding the type of publication, a higher association between CD and EoE would have been observed in research published as an abstract compared to that published as a full paper (Figure 3c, d, respectively). Finally, the coexistence of EoE and CD would have been more common in children than in adult patients.
SHARE
