Systematic Review: Early Infant Feeding and the Prevention of Coeliac Disease

H. Szajewska; A. Chmielewska; M. Pieścik-Lech; A. Ivarsson; S. Kolacek; S. Koletzko; M. L. Mearin; R. Shamir; R. Auricchio; R. Troncone


Aliment Pharmacol Ther. 2012;36(7):607-618. 

In This Article

Materials and Methods

The systematic review of the literature was initially performed in May 2011 and was updated in January 2012, and again in July 2012. The electronic searches were based on the content of the Cochrane Central Register of Controlled Trials (CENTRAL, the Cochrane Library), MEDLINE and EMBASE. Several searches were performed separately for all categories of clinical questions listed above. In addition, the reference lists from identified studies and key review articles were searched. Researchers working in the field, primarily partners in PREVENTCD, were contacted for any unpublished data. Certain publication types (i.e. letters to the editor, abstracts, proceedings from scientific meetings) were excluded, unless a full set of data was available from the authors. No language restriction was imposed. The search was carried out independently by two reviewers (AC, MP). The most recent update was carried out by one reviewer (HS). The following search terms were used separately for each clinical question:

  1. celiac or coeliac or CD or sprue or gluten enteropathy.

  2. breast-feeding or breastfeeding or breast feeding or breastfed.

  3. child or childhood or children or child* or infant* or toddler or early.

  4. gluten and (timing or time) and introduction.

  5. amount or quantity.

  6. probiotic* or prebiotic*.

Types of Studies

Studies of any design, preferentially randomised controlled trials (RCTs), investigating the potential association between early feeding practices and risk of CD were eligible for inclusion. In addition, previously published systematic reviews/meta-analyses were considered for inclusion.

Types of Participants

Participants involved in the prospective studies had to be infants at population risk or increased risk of developing CD (defined by HLA status, first-degree relative with CD or type 1 diabetes mellitus). For retrospective studies, participants had to be children or adults with CD diagnosed by small bowel biopsy or presenting with positive serology indicative of CD.

Types of Intervention (Interventional Studies)

Interventions used had to be a gluten-containing product meeting any definition (e.g. cereals, flour or any other foods containing gluten, preparations manufactured for research purposes). In addition, studies that assessed the effect of probiotics and/or prebiotics compared with control (placebo or no treatment) were considered.

Outcome Measures

The primary outcome measure was CD or the development of CD-associated autoantibodies (i.e. anti-TG2 or EMA). The studies should have assessed the risk of CD in people who were:

  1. Ever breastfed compared with those never breastfed.

  2. Exclusively breastfed compared with those receiving any human milk.

  3. Breastfed for different periods of time (short compared with long breastfeeding according to the definition given by the authors).

  4. Breastfed at the time of gluten introduction compared with those who were not.

  5. Given gluten for the first time at different ages (early compared with late introduction according to the definition given by the authors).

  6. Given gluten in different amounts (by any quantity units or thresholds used by the authors).

Data Collection and Analysis

An initial screening of the title, abstract and keywords of every record identified was performed. The next step was retrieval of the full text of potentially relevant trials. Two reviewers (AC, MP) independently assessed the eligibility of each potentially relevant trial with the use of inclusion criteria. If they had different opinions, these were resolved by discussion with the third reviewer (HS).

Assessment of Methodology of Included Studies

The reviewers independently, but without being blinded to the authors or journal, assessed the risk of bias in the studies that met the inclusion criteria. For interventional studies, The Cochrane Collaboration's tool for assessing risk of bias was used, which includes the following criteria: adequacy of sequence generation, allocation concealment and blinding of participants, personnel and outcome assessors; incomplete outcome data are addressed, free of selective outcome reporting and free of other sources of bias.[12] There is no tool for assessing the quality of nonrandomised trials that would be widely recognised as most effective.[13] Again, we chose to use The Cochrane Handbook for Systematic Reviews of Interventions.[14]

Statistical Methods

Although a meta-analysis of the available data was initially planned, after data collection it turned out not to be feasible for any of the outcomes. The reason was the different definitions of outcomes of interest used by different authors. For example, breastfeeding was defined as the duration of breastfeeding by some authors or time intervals when it was ceased by others. In addition, the timing of gluten introduction was reported as either a point in time or a certain time interval. Whenever possible, we report the binary measure for individual studies as the odds ratio (OR) between the experimental and control groups with 95% CI or as the hazard ratio (HR), as presented by the authors of individual trials. Continuous outcomes are presented as the mean with standard deviation (s.d.) or the median with ranges, again as reported by the authors. For outcomes of interest that have previously been reviewed systematically, we have summarised the findings from those reviews.