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


Description of Studies Included in the Review

Twenty-eight potentially eligible studies were initially identified. During a repeat search (July 2012), one additional study was identified. Eventually, 12 studies were included, the characteristics of which are summarised in Table S1.[8,15–25] Eleven included trials were of observational design. Two studies[21,25] used healthy children as controls. In those two studies, CD, based on positive serology but not biopsy-proven CD, was assessed. Three of the studies were cohort studies. Studies by Norris et al.[21] and Ziegler et al.[25] followed children at genetic risk of CD or type 1 diabetes. Welander et al.[24] performed a population-based cohort study. Only one interventional study was identified.[8] This study recruited 34 infants at risk of CD (positive for HLA DQ2 and/or HLA DQ8). The families of four of these infants refused to participate. Therefore, from 6 to 12 months of age, 30 infants were randomly assigned to receive either a gluten-free diet (delayed exposure group, n = 13) or a gluten-containing diet from 6 months of age (early exposure group, n = 17). In all infants, a normal, gluten-containing diet was administered at 12 months. While the study was reported as a randomised, double-blinded trial, it lacked adequate information to assess the overall risk of bias (unclear randomisation, allocation concealment and blinding). The researchers assessed CD development and serological evaluation. In addition, the stool samples of eight randomly selected infants in each group were collected for microbiota and metabolome analyses from day 7 to 24 months of age.

The protocol of one ongoing multicenter, randomised, double-blind, placebo-controlled study exploring the role of early infant feeding on CD development (PREVENTCD) has been published.[4]

In addition, two systematic reviews were identified. Akobeng et al.[26] conducted a systematic review of the literature that explored the effect of BF compared with no BF, the effect of the duration of BF and the effect of BF at the time of the introduction of dietary gluten. The MEDLINE, EMBASE and CINAHL databases were searched (until May 2004) as well as reference lists. No language restrictions were applied. An attempt to identify unpublished data was made. Two reviewers independently assessed the methodological quality of the included trials using the Critical Appraisal Skills Programme tool for case-control studies. Studies were assigned an overall rating of A (low risk of bias), B (moderate risk of bias) or C (high risk of bias). A total of six case-control studies,[15,16,18–20,22] all of which were also identified by us, were included. All included studies were graded B. To assess the effect of BF at the time of gluten introduction, a meta-analysis of all included case-control studies was performed using a fixed effect model. For other outcomes, a meta-analysis was not feasible, so only a systematic review was performed.

The second systematic review by Nash et al.[27] was aimed at determining if exclusive BF reduced the risk of CD. The MEDLINE, EMBASE and CINAHL, databases were searched (presumably in 2003) for cohort studies and case-control studies, if published in English and available in the library of the reviewer. No attempt was made to identify unpublished data. The methodological quality of included trials was described, although not formally assessed. Three case-control trials[16,19,22] were included in the review.

For the current review, 16 publications were excluded (Table S2). Among them, four studies were of retrospective design with no control group, 10 were reviews and one was a letter without a description of the methodology provided. In addition, one trial was not included because it explored the changing practices of early infant feeding in relation to the incidence of CD.

A summary of the results for all clinical questions is presented in Table 1 , Table 2 and Table 3 .

Breastfeeding and Coeliac Disease

Exclusive Breastfeeding vs. Any Breastfeeding One systematic review[27] assessed the possible relationship between exclusive BF and a reduction in the risk of CD. Three case-control studies[16,19,22] were included in the review (n = 2935; 560 cases and 2375 controls). All of the studies were retrospective and open to recall bias. There was no evidence suggesting that exclusive BF compared with formula or mixed feeding either reduces the risk of CD or delays the onset of symptoms.

Ever Breastfed vs. Never Breastfed Two studies reported this exposure.[17,22] In the study by Decker et al.,[17] more children with CD were ever breastfed (OR 1.99, 95% CI 1.12–3.51; P = 0.015).[17] Peters et al.[22] reported a lower risk of CD in ever-breastfed children compared with those who were never breastfed. Moreover, the longer the breastfeeding duration compared with no breastfeeding at all, the lower the risk of CD.

Duration of Breastfeeding and Coeliac Disease Eleven studies[15–25] evaluated the relationship between the duration of BF and CD ( Table 1 ). Data from six of these trials[15,16,18–20,22] were analysed in the systematic review by Akobeng et al.[26] Based on the findings from five of the studies, the reviewers stated that protection against CD with longer duration of BF was reported. However, more current evidence did not show that short-term BF was associated with an increased risk for CD.[17,21,23–25] With few exceptions (e.g. Welander[24]), most of the data were collected retrospectively.

Breastfeeding at the Time of Gluten Introduction and Coeliac Disease

Five studies[15,18,20–22] explored the role of BF at the time when gluten-containing products were introduced into the infants' diet (Table 2). Three of them reported a significantly reduced risk of CD in children who were breastfed when they started receiving gluten. In one small study[15] and in one large prospective trial,[21] no statistically significant difference was observed between the case and control groups. In addition to the prospective design, one additional feature of the study by Norris[21] that differentiates it from earlier studies is that it focused on children at high risk for CD (in a cohort originally designed to study children at high risk of developing type 1 diabetes). All studies but the latter study[21] were considered in the meta-analysis by Akobeng et al.[26] In the latter meta-analysis, the pooled risk for developing CD in children breastfed compared with those who were not breastfed at the time of gluten introduction was reduced by almost 50% (OR 0.48; 95% CI 0.40–0.59).

Timing of Gluten Introduction

Six studies[18,20–22,24,25] explored whether the timing of the first introduction of gluten-containing products may influence the risk of CD (Table 3). Most of these defined the moment of gluten introduction as a time interval (e.g. <3 months or 7–8 months). Therefore, the definitions used by the authors were different, thus, results were difficult to compare.

Out of five trials, two[21,22] reported a significantly increased risk of developing CD (or CD-associated autoantibodies) related to the timing of gluten introduction. The prospective, observational, cohort study by Norris et al.[21] revealed that both early (less than 3 months) and late (more than 7 months of age) introduction of gluten to children at increased risk of CD and type 1 diabetes mellitus was associated with an increased risk of CD. Children exposed to gluten before 3 months of age had a fivefold higher risk of developing CD than those with gluten introduced between 4 and 6 months of age. The risk was slightly higher when gluten was first given at the age of 7 months or later compared with when it was first given at 4–6 months of age (see Table 3 for details). The strength of this study is its prospective design; however, it has several limitations (e.g. the small number of subjects in whom the outcome measure occurred, the use of CD autoantibodies as a surrogate for biopsy-diagnosed CD). Moreover, the amount of gluten given during introduction was not assessed, thus, remains as a potential confounder. According to Peters et al.,[22] no difference was found in the risk of CD depending on the time of gluten introduction for the majority of time intervals they defined. However, when adjusted for age, sex, number of inhabitants of residence and family predisposition to CD, the OR for >4 months vs. ≤4 months was 0.66 (95% CI 0.44–1.00). The remaining studies did not show a relationship between the timing of gluten introduction and the risk of developing CD.

In the only included RCT,[8] the researchers reported data on CD development and serological evaluation in a subgroup of eight infants in each study group. At 24 months, no significant difference was found in CD development, defined by the appearance of CD anti-tissue transglutaminase antibodies, onset of CD-related symptoms and/or evidence of autoimmune enteropathy, in the delayed exposure to gluten group compared with the early exposure to gluten group (0/8 vs. 1/8, respectively, relative risk 0.33, 95% CI 0.02–7.1). Similarly, there was no difference in anti-gliadin antibodies of the class IgG between groups except at 12 mo when the difference between the delayed exposure to gluten group compared with the early exposure to gluten group was of borderline statistical significance (0/12 vs. 8/13, relative risk 0.06 (95% CI 0.00–0.99).

Amount of Gluten at Weaning (and Later) and CD

Only one study[20] analysed the amount of gluten that children received. In children younger than 2 years of age, the risk of developing CD was greater when gluten was introduced into the diet in large amounts than when introduced in small or medium amounts (adjusted OR 1.5, 95% CI 1.1–2.1). In older children, there was no effect.

Administration of Probiotics and/or Prebiotics

No studies that have addressed these issues were identified.