点击显示 收起
1
2 Supported by grants from the Swedish Council for Forestry and Agricultural Research, the Swedish Foundation for Health Care Sciences and Allergy Research, the Swedish Foundation for Research on Asthma and Allergy, the Swedish Medical Research Council (05708), and the Västerbotten County Council. 3 Address reprint requests to A Ivarsson, Department of Public Health and Clinical Medicine, Epidemiology, Umeå University, SE-901 85 Umeå, Sweden. E-mail anneli.ivarsson{at}epiph.umu.se.
ABSTRACT
Background: Celiac disease, or permanent gluten-sensitive enteropathy, is an immunologic disease strictly dependent on exposure to wheat gluten or related proteins in rye and barley.
Objective: The aim of this study was to explore whether breast-feeding and the mode of introducing dietary gluten influence the risk of celiac disease in childhood.
Design: A population-based incident case-referent study of Swedish children, 627 cases with celiac disease and 1254 referents, was conducted; 78% of the matched sets were included in the final analyses. A questionnaire was used to assess patterns of food introduction to infants. Models were built, based on current epidemiologic and immunologic knowledge of celiac disease, to study the potential influence of dietary patterns on disease risk and were evaluated by conditional logistic regression in multivariate analyses.
Results: The risk of celiac disease was reduced in children aged <2 y if they were still being breast-fed when dietary gluten was introduced [adjusted odds ratio (OR): 0.59; 95% CI: 0.42, 0.83]. This effect was even more pronounced in infants who continued to be breast-fed after dietary gluten was introduced (OR: 0.36; 95% CI: 0.26, 0.51). The risk was greater when gluten was introduced in the diet in large amounts (OR: 1.5; 95% CI: 1.1, 2.1) than when introduced in small or medium amounts. In older children, these risk factors were of no or only minor importance.
Conclusions: The gradual introduction of gluten-containing foods into the diet of infants while they are still being breast-fed reduces the risk of celiac disease in early childhood and probably also during the subsequent childhood period.
Key Words: Breast-feeding • case-referent study • celiac disease • childhood • diet • epidemiology • gluten • Sweden • prevention
INTRODUCTION
Celiac disease is recognized as a common but largely undiagnosed health problem in many European countries (1–4) and the United States (5,6). This situation is challenging because untreated celiac disease is associated with high morbidity (7).
The etiology of celiac disease is not fully understood. Genetic susceptibility is a prerequisite, although the genes involved have not yet been identified (8). The condition is considered to be an immunologic disease, possibly of autoimmune type with tissue protein-glutamine -glutamyltransferase (transglutaminase) suggested as a major autoantigen (9). The result of the disease process on the small-intestinal mucosa (ie, villous atrophy), however, is strictly dependent on the individual's exposure to wheat gluten or related proteins in rye and barley, and efforts have been made to identify the precise structure of crucial epitopes in these cereals (10,11).
In general, environmental exposures—including the dietary patterns of infants—might influence the immunologic process, resulting in tolerance or intolerance to a food constituent (12). There might be an age interval during which humans have a decreased ability to develop oral tolerance to a newly introduced antigen. The amount of antigen given might also influence the process. The type of food used could be important if the process of food preparation alters antigenicity (13). Furthermore, it is likely that the response of the immune system to an antigen may be modified by other exposures, eg, breast-feeding, because of its immune-modulating effect (14,15).
The causal role of these different aspects of infant dietary patterns in the etiology of celiac disease has not been shown consistently (16–18). Sweden recently experienced an epidemic of symptomatic celiac disease in children aged <2 y, and we showed that both the abrupt increase and decline in the incidence of the disease coincided with changes in the dietary patterns of infants (19). During the peak years of the epidemic, we performed a population-based incident case-referent study to explore the risk factors for celiac disease. We analyzed whether the risk of developing celiac disease was affected by 1) the age at which gluten was introduced in the diet, 2) the amount of gluten introduced, 3) the type of gluten-containing foods introduced, and 4) breast-feeding status (ie, whether the infant was still breast-feeding when dietary gluten was introduced).
SUBJECTS AND METHODS
Subjects
A prospective register of all incident cases of celiac disease in children aged <15 y was established in 1991. The surveillance covered 40% of the Swedish population in this age group, ie, 623439 children (19). From 1 November 1992 to 30 April 1995, a total of 714 children with suspected celiac disease were reported to the register, of whom 665 (93%) had a characteristic small-intestinal mucosal abnormality on histologic examination of a biopsy specimen taken while they were consuming a gluten-containing diet, followed by clinical remission during a gluten-free diet (20). After further diagnostic measures (described below), celiac disease was confirmed in 627 children (94%). In the diagnostic evaluation, all information available through 31 December 1998 was considered. The age at diagnosis was defined as the age when the initial small-intestinal biopsy was obtained.
All of the initially invited children (n = 665) fulfilled the established diagnostic criteria for celiac disease (20); however, the diagnosis was ascertained further (21) by obtaining a biopsy specimen from 524 of the children after they had consumed a gluten-free diet; 509 of the children showed an improvement in the small-intestinal mucosa. Subsequently, 391 children were challenged with gluten, and a third biopsy specimen was taken; 368 of the children showed a deterioration of the mucosa. Thus, a total of 38 children were excluded from the study on the basis of these additional diagnostic measures because a longer follow-up would have been necessary to confirm a diagnosis of celiac disease. Accordingly, had all 665 children initially invited to participate had a second and third biopsy specimen taken, the diagnosis may have, theoretically, been questioned in another 3% (n = 19) of the children.
Two referents matched with each celiac disease case for date of birth, sex, and area of residence (n = 1254) were selected through the national population register. A questionnaire (see below) was answered by 601 cases (96%) and 1124 referents (90%). Complete information on key variables concerning infant feeding was available for 524 cases (84%) and 919 referents (73%); 491 cases and 781 referents were also part of matched sets, corresponding to 78% of the eligible case-referent triplets (Table 1).
View this table:
TABLE 1 . Overview of participation for the celiac disease cases and referents in the 2 age groups
The Swedish Data Inspection Board and the Research Ethics Committees of all Swedish Medical Faculties approved the study, and informed consent was obtained from the families.
Questionnaire
A questionnaire about the general health and diet of the participating children was mailed to the families; our special interest in celiac disease was not indicated. The questionnaire asked whether the child had ever been breast-fed, and if so, the age (in mo) at the last breast-feeding. The food-frequency component of the questionnaire, which contained semiquantitative information on portion sizes (3 levels) was used on the basis of experience from an earlier study (22). The pattern of introduction for each food item was assessed on the basis of the age (in mo) of the infant at the time when the first portion was given, on the size of the first portion, and on the average portion size and frequency 2 wk later. Dietary intake at 7 mo of age was assessed on the basis of the frequency of consumption and portion sizes of each food item.
Definitions
Breast-feeding duration was defined as the period of time when infants were exclusively or partially breast-fed. The age at which dietary gluten was first introduced was defined as the first month postpartum, during which time flour from wheat, rye, or barley was given to the infant. Breast-feeding status during introduction of gluten-containing flour was constructed from these variables and categorized into 3 groups: breast-feeding that discontinued during the month preceding introduction of gluten into the diet, breast-feeding that continued until the month when dietary gluten was introduced or even into the following month, and breast-feeding that continued even longer than that. Furthermore, the age at introduction of gluten-containing flour was regrouped into what was the nationally recommended age at introduction (ie, age 5–6 mo) and separate categories for earlier and later introduction. The types of foods used during introduction of flour into the diet were categorized into 2 groups: 1) solid foods, including bread, biscuits, porridge, and pasta; and 2) gluten-containing follow-up formula, used exclusively or in combination with solid foods. In Sweden, gluten-containing follow-up formula is widely used but only after the age of 6 mo. The amount of gluten-containing flour in home-prepared foods was calculated on the basis of standard Swedish recipes, and the amount in industrially produced foods was obtained from the manufacturers. The amount of gluten-containing flour consumed during introduction was assessed on the basis of the food item contributing the largest amount of flour during the first 2 wk of consumption, whereas the amount consumed at 7 mo of age was based on all food items consumed at that age. The amount of flour consumed in each feeding situation was initially divided into thirds on the basis of the distribution of flour intake by the referents. In the final analyses, the 2 lower thirds (ie small and medium amounts) were analyzed together because their risk estimates were comparable, whereas the upper third (ie, large amounts) was analyzed separately. This resulted in a cutoff of 7.0 g flour for the first portion of gluten-containing food introduced and in a cutoff of 16 g flour/d for the consumption of gluten-containing food consumed 2 wk after the first portion. At 7 mo of age, the analysis resulted in a cutoff of 58 g flour/d for the total amount of gluten-containing food consumed, in a cutoff of 39 g flour/d for solid foods, and in a cutoff of 17 g flour/d for follow-up formula.
Statistical analyses
Matched sets of cases and 1–2 referents were included in the final analyses if there was complete information on breast-feeding duration, age at introduction of flour into the diet, and the average daily amount of flour consumed 2 wk after the first portion was introduced.
The children were analyzed in 2 separate groups on the basis of their age at diagnosis of celiac disease (0–1.9 or 2.0–14.9 y), mainly because the epidemic of celiac disease only affected the younger age group. The basic characteristics and dietary patterns of these children as infants were described. Associations between covariates were evaluated, and stratified analyses were performed to identify possible effect-modification or confounding. Models were developed for the tentative influence of dietary patterns on the risk of celiac disease and were evaluated with the use of conditional logistic regression in bivariate and multivariate analyses. The population attributable fraction (AFp) was estimated as follows:
RESULTS
Characteristics of the children
Celiac disease was diagnosed before 2 y of age in most of the cases (80%), and most of the cases were girls. The median delay between the onset of the first symptoms compatible with celiac disease and diagnosis was 3.0 mo in the group aged 0–1.9 y; however, a reliable estimate could not be made in the group aged 2.0–14.9 y (Table 2).
View this table:
TABLE 2 . Characteristics of celiac disease cases and referents
Validity of feeding information
The procedure by which cases were matched with referents on the basis of date of birth, sex, and area of residence had the potential of introducing selection bias; however, the dietary exposures did not differ significantly by sex or area of residence (data not shown). Compared with referents in the older age group, the younger age group had a longer breast-feeding duration and were introduced to dietary gluten at a later time (Table 3). This finding agrees with national trends in infant-feeding practices in Sweden (19). The analyses were stratified by age partly for this reason.
View this table:
TABLE 3 . Dietary patterns during infancy for the celiac disease cases and referents
In the group aged 0–1.9 y, 96% of the cases and 90% of the referents answered the questionnaire (Table 1). Information on breast-feeding duration was reported for 99% of both groups. Information on the age at which dietary gluten was first introduced was available for 94% of the cases and for 90% of the referents. Information about the average daily amount of gluten-containing flour given 2 wk after the first portion was introduced was available for 91% of the cases and for 85% of the referents. The median breast-feeding duration was not significantly different between children with complete data and those with missing data for the variables described above.
Infant feeding and risk of celiac disease
Dietary patterns in infancy differed significantly between cases diagnosed before 2 y of age and their referents (Table 3). Breast-feeding duration was significantly longer in the referents than in the cases, and more referents than cases were being breast-fed when dietary gluten was introduced and thereafter. Most children were introduced to flour at 5–6 mo of age, more so in the cases than in the referents. Cases received larger initial amounts of flour, assessed as the averaged daily amount consumed 2 wk after the first portion. Furthermore, more cases than referents had been given their first portion of flour as follow-up formula than as solid foods. At 7 mo of age, cases consumed larger amounts of flour than did referents, with follow-up formula being the major source.
Associations between the dietary exposures of the referents are shown in Table 4. Continued breast-feeding at the time gluten-containing flour was introduced was associated with the provision of small to medium amounts of gluten, whereas if breast-feeding had already been discontinued the amounts given at introduction ranged from small to large amounts. A small to medium amount of flour was usually given as solid foods, and a large amount was usually given as follow-up formula. A similar pattern of associations was found for the cases (data not shown). Stratified analyses showed no effect-modification (data not shown) but verified the above-described confounding. This was addressed by multivariate analyses.
View this table:
TABLE 4 . Associations between the dietary exposures for the referents to the celiac disease cases aged <2 y
The first multivariate model (model 1) was restricted to exposures characterizing introduction of gluten-containing flour into the diet (ie, the antigen) with respect to the age of the infant, the average daily amount consumed 2 wk after the first portion was introduced, and the type of food given (Table 5). The introduction of large amounts of flour was significantly associated with an increased risk of celiac disease. There was a significant bivariate association with higher risk of celiac disease if dietary gluten was introduced between 5 and 6 mo of age, but in this adjusted model this association was no longer significant. The type of gluten-containing food given was not an independent risk factor.
View this table:
TABLE 5 . Dietary patterns during infancy and the risk of celiac disease before 2 y of age1
In the next model (model 2), breast-feeding status at the time flour was introduced in the diet was added because of its immune-modulating potential (Table 5). Continued breast-feeding at the time gluten-containing flour was introduced had a protective effect against celiac disease. The risk pattern was similar to that with model 1, and the risk estimates changed only slightly.
The final model (model 3) included all variables of potential importance (Table 5). A protective effect was found if the infant was still being breast-fed when gluten-containing foods were introduced (OR: 0.59; 95% CI: 0.42, 0.83), and this effect was even more pronounced if breast-feeding continued beyond the time of gluten introduction (OR: 0.36; 95% CI: 0.26, 0.51). Furthermore, there was a greater risk of celiac disease associated with a large average daily consumption of flour 2 wk after the first portion was introduced (OR: 1.5; 95% CI: 1.1, 2.1) than with consumption of a small to medium amount. The risk estimates were largely unchanged after adjustment for the amount of flour consumed at 7 mo of age (data not shown).
Model 4 was developed by reanalyzing the data from model 3 with a multiple imputation procedure to account for missing values (Table 5). The risk estimates were only slightly different from those with model 3, and the conclusions did not change.
In the older age group, no significant differences between the cases and referents were found with respect to the discussed dietary variables. Furthermore, the final model, as described above for the younger children (model 3), was not significant (model likelihood ratio statistics on 5 df = 4.4, P = 0.5).
Public health effect
AFp was estimated as 31% for the independent effect of discontinued breast-feeding at the time flour was first introduced into the diet compared with continued breast-feeding at the time flour was introduced. The AFp for introduction of flour in large compared with small to medium amounts was estimated at 16%. The combined effect resulted in an AFp of 45%, suggesting that almost one-half of the cases would have been spared had the infant-feeding practices been as favorable as possible according to the results obtained in the present study (Figure 1).
FIGURE 1. . The number of celiac disease cases aged <2 y with respect to breast-feeding (BF) status (continued or discontinued) at the time gluten-containing flour was introduced into their diets, the amount of flour given (small to medium or large amounts), and the number of cases attributable to the independent or combined effect of these exposures. Risk estimates were based on conditional logistic regression with 392 matched sets of cases and referents and were adjusted for the age of the infant at the time flour was introduced in the diet. The ORs (and 95% CIs) were as follows: A, 1.0; B, 2.0 (1.4, 3.0); C, 2.8 (1.9, 4.0); and D, 3.3 (2.3, 4.8).
DISCUSSION
A major finding of this study was the lower risk of celiac disease in infants who were still being breast-fed than in infants who had discontinued breast-feeding at the time when gluten-containing foods were introduced into the diet. The risk was even lower in those infants who continued breast-feeding also beyond the time at which gluten was introduced. Larger amounts of gluten at the time of gluten introduction increased the risk, but for any given amount the type of food given was not important. The effect of age at the time of gluten introduction on the risk of celiac disease was not conclusive.
This is the first study of the association between celiac disease and environmental risk factors in which an attempt was made to develop an analytic model for causality based on immunologic and epidemiologic knowledge. Multivariate analyses were used to adjust risk estimates for confounding and to suggest causal relations. A multiple imputation approach was used to assess the possible effect of missing data on the risk estimates, and the conclusion was that missing data had no significant effect. Because the study was population based and had a high participation rate, we are confident that the results represent the country of Sweden at large. Although the use of gluten-containing follow-up formula is typical in Sweden, the findings should be relevant to any country because the type of food used could be excluded as an independent risk factor.
Moreover, the accuracy of the diagnosis of celiac disease was ascertained by well-established criteria (20,21); however, a small number of cases still may have been misclassified, resulting in a marginal underestimation of risks. Referents were chosen from the same population that generated the cases and were matched to the cases for date of birth and sex to decrease confounding and for area of residence to make another substudy feasible. Because dietary exposures did not differ significantly on the basis of sex or area of residence, no selection bias was introduced. To reduce differential misclassification of exposure, the study protocol was introduced to the participants without explicitly mentioning the focus on celiac disease. Only incident cases were included to reduce the recall period; however, inevitably, this was longer for older children. Moreover, the relatively few participants in the older age group reduced the statistical power.
It was suggested as early as the 1950s that breast-fed infants have a later onset of celiac disease (27), and this view was later shared by others (28,29). Furthermore, it was shown in case-referent studies based on prevalent cases that children with celiac disease had been breast-fed for a significantly shorter duration than had referents (30–32). However, the question was raised of whether the association of celiac disease with breast-feeding is direct and causal or indirect through postponed introduction of infant formula (30) or a reduced amount of dietary gluten consumed (32).
In the present study, we showed that breast-feeding had an independent protective effect against celiac disease if the infants were breast-fed at the time when gluten-containing foods were introduced. This effect was even more pronounced in infants who continued to be breast-fed also beyond the time of gluten introduction. Confounding by other known dietary factors was unlikely because the risk estimates were influenced only slightly by adjustments for covariates, eg, the amount of gluten consumed. Furthermore, note that our study was performed in a population in which most infants were breast-fed for 6 mo. Thus, most of the infants were introduced to cow milk while they were still being breast-fed, and consequently through sources other than infant formula. Furthermore, the cessation of breast-feeding did not usually coincide with the introduction of infant formula. It is also biologically likely that the presence of breast milk at the time gluten is introduced increases the chance of developing oral tolerance for the antigens of importance (14,15).
A larger consumption of wheat gluten was reported in healthy infants in Sweden and Italy than in infants in Finland, Denmark, and Estonia, and the former countries reported a higher occurrence of celiac disease than did the latter country (17,33,34). In addition, a dose-related response on the small-intestinal mucosa was noted when individuals were given increasing amounts of dietary gluten (35–37). Our study showed that introduction of gluten-containing foods in large amounts rather than in small to medium amounts, was an independent risk factor. The cases consumed larger amounts of flour than did the referents also at 7 mo of age.
In most countries, the recommended age for introducing gluten-containing foods has gradually increased over time to delay the manifestations of celiac disease and possibly also to reduce the risk of disease (38). Previous case-referent studies reported that the age of the infant at the time dietary gluten is introduced has no bearing on the risk of celiac disease (30–32). In our study the bivariate association between age at introduction of dietary gluten and the risk of celiac disease was not significant in the adjusted models. Thus, it is still unclear whether the ability to develop oral tolerance to gluten is decreased if the antigen is introduced into the diet during a certain period in infancy.
During the Swedish epidemic of celiac disease in children aged <2 y, the incidence increased 3-fold to levels that were higher than previously reported from any country, followed by an abrupt decline to the preceding level (19). Using an ecologic study approach, we found that the increase in incidence was preceded by an increase in the amount of gluten consumed, by a postponement of dietary gluten introduction, and by no change in breast-feeding habits. The decrease in incidence was preceded by several years of an increasing average breast-feeding duration and coincided with a decrease in the amount of gluten consumed. Thus, both the increase and decrease in incidence were most likely accompanied by a change in the proportion of infants introduced to gluten in small amounts while the infants were still being breast-fed. In other words, and on the basis of the findings in the present study, Swedish infant-feeding practices have shifted over time from a favorable to an unfavorable pattern and back to a favorable pattern. On the basis of the AFp, about one-half of the cases during the epidemic might have been spared celiac disease if the favorable dietary pattern had been followed.
It has been suggested that environmental exposures, including the dietary patterns of infants, merely affect the clinical expression of celiac disease and not the process resulting in the small-intestinal celiac lesion (18). However, there is increasing evidence to suggest that the etiology of celiac disease is multifactorial, both with regard to genetic (8) and environmental (39–41) factors. Furthermore, the immune-modulating potential of breast milk is supported by studies of the composition of breast milk (14,15) and by the increasing number of epidemiologic studies suggesting an inverse association between breast-feeding and the risk of autoimmune diseases, such as type 1 diabetes and inflammatory bowel disease (15,42), and of other diseases, such as leukemia (43). Thus, it seems likely that the dietary patterns of infants, in addition to the mere presence of gluten in the diet, interact with an individual's genetics, resulting in an immunologic process that may or may not result in the small-intestinal celiac lesion.
An important question is whether favorable infant dietary patterns postpone the onset of celiac disease until later in life, or in fact reduce the overall lifetime risk of the disease. In the present study, the dietary patterns of the infants had an effect on the risk of celiac disease in the children aged <2 y but had no or only a minor effect on risk in the older children. Thus, introduction of small to medium amounts of gluten-containing foods in the diet of infants while they are still being breast-fed clearly reduces the risk of celiac disease in early childhood and probably also reduces the risk during the remaining childhood period. It is tempting to speculate that this dietary pattern also reduces the lifetime risk of celiac disease; however, further studies are needed to confirm this notion.
ACKNOWLEDGMENTS
We thank Susanne Walther (Department of Public Health and Clinical Medicine, Epidemiology, Umeå University) for her daily work with the study; Roger Stenling (Department of Medical Biosciences, Pathology, Umeå University) for diagnostic support; Lars Stenhammar (Norrköping), Henry Ascher (Göteborg), Bertil Cavell (Lund), Lars Danielsson (Stockholm), Anders Dannaeus (Uppsala), Tor Lindberg (Umeå), and Bo Lindquist (Örebro) of the Swedish Pediatric Association's Working Group on Celiac Disease for invaluable support during the planning and performance of the study; the collaborators at participating pediatric departments in Växjö, Skellefteå, Huddinge, Uddevalla, Linköping, Jönköping, and Västerås for reporting incident cases; and Semper Foods AB and Svenska Nestlé AB for kindly providing information on the composition of industrially produced foods.
REFERENCES