Systematic Review With Meta-analysis

Endoscopic Dilation Is Highly Effective and Safe in Children and Adults With Eosinophilic Oesophagitis

F. J. Moawad; J. Molina-Infante; A. J. Lucendo; S. E. Cantrell; L. Tmanova; K. M. Douglas


Aliment Pharmacol Ther. 2017;46(2):96-105. 

In This Article


We used PRISMA methodology for conducting this systematic review and meta-analysis.[30] This meta-analysis was registered in PROSPERO, the International Prospective Register of Systematic Reviews (CRD42016037658). There was no funding received for this review.

Selection Criteria

To be included in this review, studies needed to involve adult or paediatric EoE patients undergoing oesophageal dilation. Major complications needed to be explicitly reported, as did clinical effectiveness. Major complications were defined as perforation, haemorrhage, hospitalisation, or death. Randomised controlled trials, prospective cohorts, retrospective observational studies, case series, and case reports were eligible for inclusion. Systematic reviews, guidelines, review articles, letters to the editors, editorials, and articles not relevant to EoE were excluded. Studies providing duplicate information, to include subsets of data already published, were excluded.

Search Strategy

A search of the literature was conducted by two of the authors (SC and LT), both professional librarians, in December 2016 without any restriction to language in the following databases: MEDLINE (via OvidSP, 1946 to present), EMBASE (via Elsevier, 1947 to present), and the Cochrane Library (via OvidSP), including Cochrane Register of Controlled Trials, through February 2016; the Cochrane Database of Systematic Reviews, 2005 to March 30, 2016; Database of Abstracts of Reviews of Effects, through 1st Quarter 2016; Health Technology Assessment, through 1st Quarter 2016; Cochrane Methodology Register, through 3rd Quarter 2016; and NHS Economic Evaluation Database, through 1st Quarter 2016. Literature search strategies in these databases used a combination of subject headings and index terms, as well as, key words relating to eosinophilic oesophagitis, endoscopic, dilation, Maloney, Savary, Through-the-scope, diameter, safety, perforation, haemorrhage, hospitalisation, chest pain, treatment outcomes, adverse events, complications and efficacy (full search strategy in Appendix S1). The main search strategy was conducted in MEDLINE. Parallel search strategies were conducted in EMBASE and the Cochrane Library. A third author (AL) searched Scopus, Web of Science, and abstracts from the most relevant gastroenterology and endoscopy conferences to include Digestive Diseases Week, American College of Gastroenterology, and United European Gastroenterology Week.

All references were screened for eligibility both independently and in duplicate. Two review teams, which included two authors on each team (JM/AL and FM/KD), each screened half of the references by review of titles and abstracts. If either reviewer on a team felt that a title or abstract suggested study eligibility, the full text of the reference was retrieved. The authors resolved any discrepancies by discussion.

Data Extraction

We developed a data extraction form for this study and each author agreed upon the variables a priori. All data were extracted independently and in duplicate to minimise error or bias. The variables included first author of the study, design of the study, year and country of publication, demographics (age, race, sex), clinical presentation (dysphagia, food impaction, heartburn, chest pain), allergic history (allergic rhinitis, food allergies, asthma, eczema), endoscopic features (rings, furrows, plaques, strictures), type of stricture (diffusely narrow, dominant, and location), treatment (PPI, steroids, diet), type of dilator used (Maloney, Savary-wire guided, through-the-scope balloon, Celestin, EndoFLIP (Endoluminal Functional Lumen Imaging Probe, Crospon, Inc, Carlsbad, CA, USA), or dilation with the endoscope, number of dilations performed per patient, clinical response and duration of response, patient follow-up, complications (perforation, haemorrhage, hospitalisation, death) whenever they were available. We also collected data on chest pain and mucosal laceration.

Quality Assessment

Studies were ranked according to three metrics of quality: study design, completion of follow-up, and duration of follow-up. We assigned a score of high for randomised controlled trials, moderate for cohort studies, and low for case reports and case series. For completion of follow-up, we assigned a score of high quality if more than 80% of patients had follow-up, moderate quality if between 50% and 80% had follow-up, and low if less than 50% had follow-up or if follow-up was not reported. For duration of follow-up, high quality studies reported follow-up greater than 6 months, studies that reported follow-up between 1 and 6 months were moderate in quality, and low quality was assigned if follow-up was less than 1 month. We considered a study to be of high quality overall if it scored highly on all three metrics and low quality if it scored low on any metric (Tables S1 and S2).


The primary outcomes of this meta-analysis were to assess the safety of dilation in EoE patients by calculating the rate of complications associated with dilation and to assess clinical improvement following dilation. Our secondary outcome was to explore whether differences existed in complication rate by type of dilator [Maloney, Savary, Through-the-scope (TTS) Balloon].

Statistical Analysis

Percentages of patients experiencing an outcome of interest were summarised with the aid of a random effects model for proportions (STATA command metaprop one).[31] The STATA command metaprop one was used to model outcome data. This routine provides procedures for pooling proportions in a meta-analysis and displays the results in a forest plot. The pooled estimate is obtained as a weighted average by fitting the logistic-normal random-effects model without covariates but random intercepts after Freeman-Tukey Double Arcsine Transformation to stabilise the variances.[32] The confidence intervals were based on the exact binomial (Clopper-Pearson) procedure.[33] For sparse event data, breakdown of the modelling procedure is known to occur, in which case calculation of the pooled estimate and confidence interval using the exact binomial method was performed. Statistical significance of heterogeneity was tested by means of the Chi-squared statistic for the likelihood ratio test. Heterogeneity was also quantified using the I-squared measure assigning categories of low, moderate, high or very high for values of 1%-25%, 26%-50%, 51%-75% and 76%-100% respectively.[34] Because methods for assessing publication bias in meta-analytic studies of proportion data are not well-established, we do not present such an analysis.[35,36]A sensitivity analysis was performed with regard to quality by excluding all case series and case reports.

All analyses were carried out using STATA (version-13.1; Statcorp, College Station, TX, USA). There was no funding received for this meta-analysis and all authors approved the final version of the manuscript.