What effect does breastfeeding have on coeliac disease? A systematic review update
- 1Department of Biosciences and Nutrition, Karolinska Institutet/Stockholm University, Stockholm, Sweden
- 2Division of Nursing, Department of Neurobiology, Care Science and Society, Karolinska Institutet, Stockholm, Sweden
- 3Department of Geriatric Medicine, Danderyd Hospital, Stockholm, Sweden
- 4Division of Obstetrics and Gynaecology, Department of Clinical Sciences, Danderyd Hospital, Karolinska Institutet, Stockholm, Sweden
- Correspondence to: Dr Ingela E Wiklund
Department of Clinical Sciences, Danderyd Hospital, Karolinska Institutet, 182 88 Stockholm, Sweden;
Objective To update the evidence published in a previous systematic review and meta-analysis that compared the effect of breastfeeding on risk of coeliac disease (CD).
Material and methods A systematic review of observational studies published between 1966 and May 2004 on the subject was conducted in 2005. This update is a systematic review of observational studies published between June 2004 and April 2011. Pubmed, EMBASE and Cinahl were searched for published studies that examined the association between breastfeeding and CD.
Results After duplicates were removed 90 citations were screened. Four observational studies were included in the review. Two of three studies which had examined the duration of breastfeeding and CD reported significant associations between longer duration of breastfeeding and later onset of CD (OR ranged from 0.18 to 0.665). Breastfeeding during the introduction of gluten to the infant was reported to have a protective effect in two studies.
Conclusions Our findings support previous published findings that breastfeeding seems to offer a protection against the development of CD in predisposed infants. Breastfeeding at time of gluten introduction is the most significant variable in reducing the risk. Timing of gluten introduction may also be a factor in the development of CD.
Coeliac disease (CD), also known as gluten sensitive enteropathy, is a chronic autoimmune disease that affects the small intestine in genetically predisposed individuals. CD is induced by dietary intake of wheat gluten and related prolamines, for example, in rye and barley. Most of the individuals with CD carry a human leucocyte antigen (HLA)-DQ2 or DQ8 haplotype gen but so does about 30% of the general population.1 The prevalence of CD is difficult to determine when the disease can be asymptomatic for a long time and the diagnosed prevalence varies from 3 to 14 / 1000 children which means that the environmental factors in the aetiology play a crucial role.1 The disease is multifactorial and CD seems to act as a combination of adaptive and innate immune response to gluten. Studies show that one of the most critical environmental factors is early infant feeding practices with focus on breastfeeding.2 ,3
Breastfeeding is a subject for constant debate and strong opinions, not whether or not it has positive effects for the infant, but more as a question of how good it is and for what. As early as in the 1950s, breastfeeding was suggested to have a protective effect and that breastfed infants had a later onset of CD. For the last decades epidemiological studies have shown breastfeeding to be a crucial element in allergy development.4 The mechanisms underlying the aetiology are not fully understood and therefore nor the mechanisms behind the protective effect. Continuing breastfeeding at time of gluten introduction has been suggested to limit the amount of gluten the child receives and thereby decreasing the chance of developing the symptoms of CD. Another possible explanation is that the milk in itself protects against gastrointestinal infections.2 ,4 Such infections may lead to increased permeability which allows gluten to pass on into the lamina propia triggering the process of CD in susceptible individuals.4 Whatever the mechanism might be, breastfeeding is the most important environmental factor in CD development for an infant. That makes this area of research important since there are indications that breastfeeding rates are declining. For example, in the Stockholm area in Sweden the number of children exclusively breastfeeding have declined by 10.5% in the last 15 years, if this has an effect on prevalence of CD is yet to be seen.5
The WHO declared in a systematic review from 2002 that one of their goals is that 80% of all the children in the world should be exclusively breastfed for 6 months.6 Sweden and many other European countries adapted this recommendation, which means that other foods, such as wheat, should be introduced after 6 months. Lately, it has been questioned whether this is the best way of introducing possibly allergenic foods. Models in animals suggest that oral tolerance is an allergen-driven process and that there is a ‘critical early window’ when exposure to proteins is essential to the development of this process.7 The timing for this window for oral allergen tolerance is not fully established in humans but current studies point towards it being between 4 and 6 months.7 ,8 This makes the timing of gluten introduction and whether it is introduced under the protection of breastfeeding or not a significant question.
In 2005, Akobeng et al2 conducted a systematic review and a meta-analysis with the aim to explore the potential association between breastfeeding and reduced risk of CD. The following outcomes were examined: (1) the effect of breastfeeding compared with no breastfeeding; (2) the effect of duration of breastfeeding and (3) the effect of breastfeeding at time of introduction of dietary gluten. Akobeng et al2 systematically searched various databases for studies examining these effects published from 1966 to May 2004. The result of the study suggested that breastfeeding may offer protection against the development of CD. A meta-analysis of four of the six studies indicated that children being breastfed at the time of gluten introduction had a 52% reduction in risk of developing CD compared with their peers who were not breastfed at the time of gluten introduction. This study aims to update the evidence published in the systematic review with articles published from June 2004 to April 2011 in order to evaluate possible developments in this research field. Following research questions guided the study: (1) the effect of breastfeeding compared with no breastfeeding; (2) the effect of duration of breastfeeding and (3) the effect of breastfeeding at the time of introduction of dietary gluten.
Material and methods
In this review we have used similar review methodology (types of studies, search strategy and assessments) and examined the identical outcomes as in the previous published systematic review.2
Types of studies
Included studies were observational studies and one case–control study with the same inclusion criteria as in the systematic review: (1) compared risk of CD in people who were breastfed with risk in those who were not breastfed, or compared risk of CD according to duration of breastfeeding; (2) had used histological criteria for diagnosing CD; (3) had controlled for potential confounders by matching in the study design or used risk adjustment in the analysis and (4) had provided sufficient data to allow the reconstruction of 2 × 2 tables to determine relative risks or OR with 95% CI. Only articles written in English were included.
Pubmed, EMBASE and Cinahl were systematically searched for articles and studies published between June 2004 and April 2011 with the search strategy presented in table 1. The search was conducted on 25 April 2011.The reference lists of the relevant citations found in this search was examined for further potentially appropriate articles.
Assessment of study eligibility
The first author assessed each article for eligibility using the inclusion criteria above. Unclear citations were discussed with the other authors and agreement reached by consensus.
Assessment of methodological quality
The authors’ independently rated methodological quality of selected studies using the Critical Appraisal Skills Programme (Oxford, UK) tool for observational studies.9 Each study was grade as A (low risk of bias), B (moderate risk of bias) or C (high risk of bias) according to published criteria.10 Disagreements were resolved by consensus.
Information and data on relevant features and results was extracted from each of the included studies. When data was missing or unclear in the published paper we emailed the authors for more information. None of the authors responded on our requests.
In the included studies many different assessments for breastfeeding and statistical tests were used, and the presentation of data prevented us from performing a meta-analysis. Therefore, the findings from the articles are presented in narrative form.
Description of included studies
The initial search resulted in 164 identified records; after duplicates were removed 90 records remained. After reading the abstracts we excluded articles not focusing on the association of breastfeeding and the development of CD. Articles not written in English were also excluded together with articles in the form of comment and reviews. Of these 90 records only 10 were identified as potentially relevant studies (figure 1). Six were excluded for following reasons. Four were review articles and did not report original data.4 ,8 ,11 ,12 One did not confirm CD in a correct way13 and one was an animal study.14 One citation was an abstract in a conference report,15 although the results from this abstract will be included in the update. Thus, four studies were identified that met the inclusion criteria.15–18 Three studies were observational studies16–18 and one a case–control study.15 They were originated in different countries, one in Serbia,16 two in the USA17 ,18 and one in Spain.15 They ranged from 8916 to 173715 participants of varying age. Description of included studies is presented in table 2. No cohort study was found on the subject. All participants had been diagnosed to have CD based on small intestinal biopsy. All studies used questionnaires or interviewing techniques to obtain infant feeding history from parents.
Methodological quality of included studies
The methodological quality of each study was evaluated and a summary of the methodological quality and the grades are shown in table 2.
Association between breastfeeding and coeliac disease
Ever breastfed versus never breastfed
No study was conducted with children who had never been breastfed therefore no comparisons of the groups based on this variable could be conducted.
Duration of breastfeeding
Radlovic et al16 did a retrospective observational study of 89 infants with CD using their medical records to determine the duration of breastfeeding and timing of gluten introduction at the University Children's Hospital in Belgrade. The infants were divided into one group that had been exclusively breastfed at the time of gluten introduction (n=33) and a group that had not been breastfed at the time of gluten introduction (n=56). The infants were also divided into groups based on the timing of gluten introduction: prior to the fourth month (n=22), between the fourth and sixth month (n=36) and after the sixth month (n=4). The researchers conducted a conditional logistic regression to estimate the impact of early feeding practices, that is, duration of breastfeeding, timing of gluten introduction and breastfeeding at the time of gluten introduction on the risk of disease onset in the first year of life. The results revealed that longer duration of breastfeeding significantly reduced the risk to develop CD in the first year of life (OR 0.665; 95% CI 0.481 to 0.891) and the duration of breastfeeding was the only significant variable in delaying the age at the CD diagnosis for this group of infants (B=0.49; SE=0.159; p=0.007).
D'Amico et al17 conducted a retrospective study with 141 children 20 years of age or less with CD and compared infants who where exclusively breastfed (n=40) for at least 6 months with all other infants where the vast majority was both breastfed and bottle fed (n=101). The researchers showed that children who were exclusively breastfed in the first 6 months developed symptoms significantly later (on average 15 months later; p<0.05) and were diagnosed with CD significantly later (at a mean age of 4 years and 10 month versus 2 years and 11 months; p<0.05) than the other children. Exclusively breastfeeding significantly delayed the onset of the symptoms leading to CD (OR 0.18, 95% CI 0.05 to 0.64). The study also showed that the exclusively breastfed children had lower rates of severe CD symptoms (p<0.05).
Norris et al18 conducted a 10-year (1994–2004) prospective observational study in Denver, USA where 1560 children at increased risk for CD participated and 51 developed CD. The 51 children diagnosed with CD were breastfed on average 8.3 (SD 8.8) months compared with the other children who were breastfed 6.7 (SD 6.8) months. No significant difference was found between the groups.
Roman15 conducted a nationwide prospective case–control study in Spain including 993 children with CD and 744 paired controls from June 2006 to May 2007. However, Roman did not report any results on this outcome.
Breastfeeding at the time of gluten introduction
Radlovic et al16 found that the 33 infants who had been breastfed at the time for the gluten introduction were significantly older when diagnosed (mean=16.21±3.31 months) compared with the 56 infants who had not been breastfed (mean=13.04±5.01 months; p=0.029).
D'Amico et al17 did not report results on this outcome.
Norris et al18 did not find a significant difference between the children who were breastfed when first exposed to wheat, barley or rye with those who were not breastfed. Of the 51 CD children, 49% were breastfed at the time of gluten introduction versus 44% among the 1509 children without CD. The researchers showed however that children exposed to gluten in the first 3 months of life had a fivefold increased risk compared with children who were given gluten between the fourth to sixth month. Results also showed a marginally increased risk when gluten was not introduced until the seventh month or later compared with those exposed at 4–6 months.
Roman15 reported that if gluten was introduced while the child was still breastfeeding the risk of developing CD was reduced by 58–62%. These results were presented in a conference in June 2010 but the study has not been published. This was the only case–control study in the update.
To summarise our findings, two of the three studies which had examined the duration of breastfeeding and CD reported significant associations between longer duration of breastfeeding and later onset of CD. Furthermore, three of the included four studies reported that breastfeeding during gluten introduction significantly delayed the onset of CD. Norris et al18 report that timing of introduction of gluten into the infant diet is significantly associated with the appearance of CD.
The aim of this study was to update the previous published evidence in a systematic review and meta-analysis examining the association between breastfeeding and reduced risk of CD.2 The review suggested that breastfeeding may offer protection against the development of CD and that children being breastfed at time gluten introduction had a 52% reduction in risk of developing CD compared with their peers who were not breastfed at the time of gluten introduction.2 In this updated article the results point in the same direction although the evidence is weaker. In the original review2 all included studies were case–control studies, and in this update only one case–control study was available. The remaining studies were observational studies reporting data in various ways preventing us conducting a meta-analysis. During the last 7 years there has been few studies conducted in this research field which prevent a reliable update of the meta-analysis. Two review articles (although not systematic reviews) have been recently published in support of our findings; Silano et al8 confirms a negative correlation between the duration of breastfeeding and development of CD and Selimoğlu et al4 states that the most important preventive strategies include the encouragement of breastfeeding. Our findings in combination with the previous findings from Akobeng et al's2 systematic review and the two recent published review articles make it's safe to say that breastfeeding have an important role in delaying and/or preventing the development of CD.
The findings on timing of gluten introduction and the circumstances at this point is also crucial. It is not clear from these results (or previous ones) if breastfeeding provides a permanent effect against CD or if it delays the symptoms. One study found no correlation neither between the duration of breastfeeding nor if gluten was introduced under the protection of breastfeeding.18 This study had relatively few cases (children with CD; n=51) and this may explain why no correlation was identified. The cases in this study were recruited from a large birth cohort in Denver, Colorado, USA. Cord blood samples were collected and screened and infants with a specific HLA genotype was invited to participate. This means that these children and their parents were selected because of their high risk of developing autoimmune diseases such as CD. No dietary advice was given to these families but it is impossible to rule out that this did not affect the choice of infant diet. It was a prospective observational study and recall bias is therefore a minimal risk.
Breastfeeding, infant feeding, weaning and everything related to it are subjects where everyone has an opinion and where medical facts are combined with traditions. There are national differences as well as cultural and personal beliefs between how parents choose to feed their children. In the previous systematic review the study objects originate from three different countries; Italy, Germany and Sweden.2 In this update the included studies are conducted in three other countries; Serbia, the USA and Spain. Together these six countries represent widely shifting prevalence of breastfeeding with Italy and Serbia with the lowest percentage were only 18% and 23% of the infants are breastfeeding at 4 month and with Sweden in the top where practically all children are breastfed during the first week and 60% still at 4 months of age. USA and Spain are somewhere in-between with rates at 33% and 40% at 4 months.19 ,20 Despite these differences nearly all studies have the same results; when it comes to CD, breastfeeding seems to have a protective effect against the development. The previously published evidence are confirmed and reinforced by the new findings in this update and one could assume that future studies will continue to confirm this finding.
Sweden is a country with one of the highest rates of breastfeeding in the world; however, the rates are dropping. In a report from 2010 conducted in the Stockholm area results showed that breastfeeding rates are declining at both 4 months as well as the share of infants being breastfed at age 1 week.5 The reason for this decline is probably multifactorial. But what is interesting is what effect the decreasing breastfeeding rate will have on CD prevalence. In a study with the aim to use a statistical method in order to calculate a risk factor over a time period the association between breastfeeding and CD among the 596 122 babies born in England and Wales was examined. It was found that CD could be prevented over 7–9 years if ‘no breastfeeding’ as a risk factor was eliminated. The number of cases that could be prevented was 2655 (95% CI 1937 to 3343).21 If we interpret the results of this review and previous studies the prevalence of the disease may increase with higher cost for the individual family and the healthcare system. In Sweden the National Food Administration (NFA) is responsible for diet recommendations for the whole population, including infants.22 Since WHO in 2002 declared its policy to recommend 6 months exclusive breastfeeding NFA has made the same recommendations. As aforementioned, this has been questioned in relation to food allergy and in May 2011 NFA announced in a press release that the infant feeding recommendations are under consideration and will be updated during the year. Norris et al18 showed that timing of gluten introduction was associated with the appearance of CD with an increased risk if introduced in the first 3 months or in the seventh or later. European Food Safety Association and European Society for Paediatric Gastroenterology, Hepatology and Nutrition recommends introduction of wheat not earlier than 4 months and no later than 6 months to reduce the risk of CD.23 However, NFA and its European counterparts stress the importance of gluten being introduced while the infant is still breastfeeding.
We chose to expand the inclusion criteria with the knowledge that the studies would not be totally equivalent to the ones in the first systematic review and meta-analysis. Both the outcome and the exposure were defined differently in the separate studies and the results could not be pooled together. A golden standard is missing for how to define and report exclusive and partial breastfeeding just like there is a clear definition of the CD diagnosis. In different countries and cultures infant feeding is thought of in different ways. This could be depending on everything from duration of maternity leave to traditions passed on from generation to generation regarding opinions about optimal nutrition.
The results of this review are subject to limitations. Akobeng et al conducted their search in May 2005 and we performed our search 6 years later but only four new studies were found. The two of the four studies were assessed to be of moderate risk of bias (grade B) and one study was assessed to be of high risk of bias. Two-included studies used retrospective design which is subject to recall bias. Using interviews and questionnaires, as done in most of these studies, misclassification of infant feeding is likely to occur, both of duration of breastfeeding and age of introduction of gluten. Such misclassification severely affects the results. Retrospective observational studies as a method is always in risk of bias with the consequence that other risk factors could be less countered for. None of the included studies accounted for socioeconomic status to be a confounding factor although this is a crucial factor for diet choice. Furthermore, the published studies provided only data for narrative presentation and we could not conduct a meta-analysis. Thus, the findings from this updated review must be interpreted with caution.
Conclusions and future research
Breastfeeding seems to offer a protection against the development of CD in predisposed infants. Breastfeeding at time of gluten introduction and the total duration of breastfeeding appears to be the two most significant variables in reducing the risk. Timing of gluten introduction may also be a factor in the development of CD. Whether possibly allergenic food should be introduced during a critical early window or not is a subject for future research.