Early Infant Diet and Islet Autoimmunity in the TEDDY Study

The interplay between genes and environmental factors, such as diet, has been hypothesized to play an important role in triggering type 1 diabetes (T1D) (1,2), the incidence of which continues to increase globally (3).

The ability of breast milk to provide the required nutrients becomes limited among older infants. Therefore, timely introduction of complementary foods is essential for the baby’s well-being and growth (4). To examine certain infant feeding practices, several studies have been carried out in populations with genetically increased T1D risk by focusing on an arbitrary chosen age range in relation to islet autoimmunity (IA) and/or T1D. It has been reported that a shorter duration of breastfeeding (5–7), certain “age windows” (≤3 and ≥7 months vs. 4–6 months) for introducing cereals (8,9), as well as early (<3–4 months) exposure to cow’s milk (10,11), gluten-containing cereals (12,13), fruits and berries (9,14), and potatoes and root vegetables (14,15) may increase the risk of IA and/or T1D. However, a recent study (16) showed that this association between age at introduction of complementary foods and IA may decrease/disappear when the follow-up is extended, including older children. The association between the timing of food introduction and IA remains inconclusive because of inconsistent results (17). Types of early foods linked to IA/T1D seem to vary at least partly by country.

The mechanisms linking early feeding practices and the development of IA/T1D are not very well known, but those suggested to play a crucial role include immature and adverse immunological responses of the gut to complementary food (18,19), mucosal inflammation, and increased gut permeability (20,21).

This study aims to examine breastfeeding duration and the timing of initiating infant formula, regular cow’s milk, and solid food in relation to the risk of IA in The Environmental Determinants of Diabetes in the Young (TEDDY) study. We investigated the overall association between those dietary exposures and the risk of IA. To our knowledge, this is the first attempt to analyze these dietary exposures without a predetermined categorization. Additionally, we examined whether a categorization of breastfeeding duration or age at initiating a food is statistically justified based on the association with the risk of IA in TEDDY.

TEDDY is a prospective observational cohort with the primary aim to identify environmental causes of T1D. The study includes the following six clinical research centers: three in the U.S. (Colorado, Georgia/Florida, and Washington) and three in Europe (Finland, Germany, and Sweden). A total of 424,788 newborns were screened in hospitals affiliated with the study centers between September 2004 and February 2010, identifying 21,589 HLA-eligible infants. The HLA typing has been previously described in detail (22–24). Of the 8,676 enrolled subjects, 8,263 singleton babies were identified carrying one of the eligible HLA types with determined IA status. Of those, we included 7,572 subjects who were followed for at least 9 months to obtain complete information on the duration of breastfeeding and the timing of the introduction of complementary foods. After excluding nine subjects who lacked information on early feeding, a total of 7,563 children were analyzed in this study. Their median (Q₃–Q₁) follow-up time was 92 months (114–54).

Written informed consent was obtained for all children in the study from a parent or primary caretaker, separately, for genetic screening and participation in prospective follow-up. The study was approved by local institutional review or ethics boards and was monitored by an External Evaluation Committee formed by the National Institutes of Health.

Blood samples using serum separation tubes were drawn every 3 months between 3 and 48 months of age and every 6 months thereafter, unless the autoantibodies developed in the child, in which case the child continued to be followed, including blood draws every 3 months. Serum was stored in two 0.5-mL cryovials for autoantibody measurements and were frozen within 2 h from collection. Persistent IA (any IA) was defined as confirmed positive insulin autoantibodies (IAAs), GAD antibody (GADA), specifically to isoform GAD₆₅, or insulinoma antigen-2 autoantibody, which were analyzed by radiobinding assays (25,26) on at least two consecutive study visits. All positive and 5% of negative islet autoantibodies were confirmed in the following central autoantibody laboratories: Barbara Davis Center for Childhood Diabetes, University of Colorado, in the U.S. and University of Bristol in the U.K., which both previously have found high sensitivity and specificity (27) and concordance. Positive results due to maternal IgG transmission when defining the child’s IA status were omitted from the IA-positive group. The date of persistent IA was defined as the draw date of the first of two consecutive samples confirmed positive for a specific autoantibody, with which the child was deemed persistent. In addition to any IA, we separately studied children who had either IAAs alone or GADAs alone as their first appearing autoantibody. The median age (Q₃–Q₁) of children at IA seroconversion was 33 months (62–16) (n = 703). The median (Q₃–Q₁) values were 21 months (40–11) for IAAs (n = 272) and 46 months (77–25) for GADAs (n = 299), respectively.

Demographic characteristics, family history of diabetes, and infant feeding practices were obtained from various questionnaires and have been explained previously (28). Information about infant feeding (breastfeeding and food introductions) was recorded by parents in a diary (“TEDDY Book”) at home and collected every 3 months during the clinic visits or over the phone starting at 3 months of age. This information was recorded in the TEDDY Book until the clinic visit at 24 months except for any breastfeeding, which was followed up to 5 years of age. If the breastfeeding duration was >5 years, the child was not introduced to a complementary food by 24 months of age, or the correct timing of food introduction was not available, the information was regarded as unknown. The duration of breastfeeding that corresponded to the age when breastfeeding ended (exclusive and any) and the ages at which consumption of infant formula and solid foods started were examined in relation to IA. A baby was considered to be exclusively breastfed when his/her diet included only breast milk and possibly small amounts of non-nutritious drinks (e.g., water). Any breastfeeding could also be accompanied by other foods in the diet. The infant formula in this study included the following: commercial infant formulas that contain intact cow’s milk proteins or cow’s milk proteins hydrolyzed to any degree, soy formula, elemental formula, regular cow’s milk and other animals’ milks, and vegetarian “milks.” The solid foods that were studied separately in this study included the following: any type of cereal (wheat, rye, barley, oat, rice, or any other nongluten cereal), gluten-containing cereals (wheat, rye, and barley), rice, potatoes, root vegetables, fruits and berries, meat (beef, pork, game, and/or poultry), eggs, fish, and other seafood (Table 1). In addition to these foods, the “any solid food” included milk products (e.g., cheese, yogurt), sausages, and various vegetables. The age at the introduction of any solid food was defined as the earliest time when any of the aforementioned solid foods were introduced. “Selected foods” included foods found to be associated with IA/T1D in the earlier studies, as follows: cereals (IA), including rice/oat (T1D) and gluten-containing cereals (IA); potatoes (IA); root vegetables (IA); fruits and berries (IA/T1D); and eggs (IA) (8,9,12,14,15).

Cox proportional hazards models were used to study the association between dietary exposures (duration of breastfeeding and age of initiating a food, as defined above) and the risk of IA, after adjusting for country, T1D family history (first-degree relative [FDR]), sex of the child, HLA (DR3/4 vs. other genotypes), and exposure to probiotics at <28 days of age. Time of seroconversion was the age when the first blood sample for persistent IA was drawn. Time for right censoring was the age when the last blood sample in the follow-up was determined to be negative for IA. A proportional cause-specific hazard model for first-appearing IAA or first-appearing GADA was used by treating events other than the one of interest as censored observations. In each risk set, including those who experienced the event of interest and those who were event free by a certain age, the age of initiating consumption of a food was analyzed in those who had initiated consumption of the food at an age younger than that of the risk set.

The functional form of each dietary exposure and IA association was explored by plotting martingale residuals with a loess smoothing parameter of 0.4. Additionally, we applied the change-point method, based on the log-rank statistic, in order to find a cut point for each dietary exposure dichotomization in relation to the risk of IA (29).

Two-sided P values <0.05 were considered to determine a statistical significance. All analyses were performed using the Statistical Analysis System Software (version 9.4; SAS Institute, Cary, NC).

Characteristics and potential confounders associated with risk of any IA are presented in Table 2. After adjusting for those factors, we found that later introduction of gluten-containing cereals was associated with increased risk of any IA (hazard ratio [HR] for 1-month delay 1.05; 95% CI 1.01, 1.10; P = 0.02) and with increased risk of IAA (HR for 1-month delay 1.08; 95% CI 1.00, 1.16; P = 0.04) (Table 3). When examining the durations of exclusive breastfeeding and any breastfeeding, the timing of any infant formula, the timing of single foods other than gluten-containing cereals, or any solid food introduction as a combined variable of exposures of solid foods, we could not detect any association between them and the risk of outcomes (Table 3). There were 959 children (12.7%) who moved straight from breast milk to solid food (e.g., milk-based thin porridges) and therefore had no values for age of introduction of infant formula.

The martingale residual analysis showed changes in the association between age of introduction of gluten-containing cereals and the risk of any IA. There was an increasing trend of risk between 0 and 4 months, a plateau from 4 to 9 months, and increasing risk again at introductions from 9 months on (Supplementary Fig. 1). The application of the change-point method revealed a significant dichotomization in the duration of any breastfeeding (at 7 months of age with any IA and at 6 months of age with GADA), age of introducing cow’s milk (at 5 months of age with any IA), cereals (at 4 months of age with any IA), rice (at 7 months of age with any IA and at 6 months of age with GADA), fruits and berries (at 4 months of age with any IA), potato (at 4 months of age with any IA), meat (at 8 months of age with any IA), egg (at 9 months of age with any IA), and fish and seafood (at 9 months of age with GADA).

We applied the data-driven categorizations of dietary exposures in evaluating the risk of IA. When compared with the introduction at 4–9 months of age (Supplementary Fig. 1), introduction of gluten-containing cereals before 4 months of age showed decreased risk of any IA (HR 0.68; 95% CI 0.47, 0.99) but increased risk of any IA (HR 1.57; 95% CI 1.07, 2.31) if introduced after 9 months of age. The HRs remained similar in the introduction of gluten-containing cereals before 4 months of age when adjusted for country of residence, HLA, FDR with T1D, sex of the child, and early exposure to probiotics (HR 0.67; 95% CI 0.54, 0.98; P = 0.04). When the dichotomizations were applied, the risk difference between the two timing categories of duration of breastfeeding and food introductions was not very noticeable. However, the introduction of egg at or before 9 months of age showed consistently lower risk of any IA compared with introduction after 9 months of age both in the unadjusted analysis (HR 0.86; 95% CI 0.74, 0.99) and in the adjusted analysis (HR 0.84; 95% CI 0.72, 0.99) (Table 4).

Data from the multinational prospective TEDDY Study suggests that later introduction of gluten-containing cereals is associated with increased risk of any IA and IAA. The residual plot suggested a plateau in risk at introduction between 4 and 9 months of age, and the results from categorized analysis supported that, as well as the overall finding that later introduction of gluten-containing cereals is associated with increased risk of IA.

The major strength of the study was its consistently collected data using the same protocol and questionnaires across four TEDDY countries. Including larger geographical areas in the study made it possible to consider the importance of varying feeding habits. Additionally, both continuous and statistically derived categorized exposures were used in the investigation of associations. As a limitation, we did not record the amounts of the introduced food or count the initial frequency of feeding the new food. Thus, the cumulative exposure of a new food or foods was not possible to study.

When studying introductions of solid foods, we found that later introduction of gluten-containing cereals was associated with increased risk of IA. Later introduction of foods into a diet overall may be associated with larger initial amounts of food given to the children. Larger amounts can be challenging to their immature immune system and can therefore hamper the development of tolerance to foreign antigens. Very few studies have investigated the amounts of food at early age and a risk of disease. Aronsson et al. (30) suggested that a larger amount of gluten consumed during the first 2 years of life was associated with increased risk of celiac disease. However, we did not study the amounts of foods given in this study. A positive association between the early introduction of gluten-containing foods (<3 months vs. later introduction with exclusive breastfeeding until 3 months of age) (12), as well as early (≤3 months vs. 4–6 months) and late (≥7months vs. 4–6 months) introductions of any cereals (8), and the risk of IA has been reported among children with increased risk of T1D. Our results did not support the finding related to early gluten introduction and the risk of IA by Ziegler et al. (12).

The preferred first solid foods varies among the countries, for example, cereals in the U.S. and fruits, potato, and root vegetables in Finland (28). It appears that the first solid food a child consumed was most often associated with IA risk (8,14,15). This could be interpreted in a way that the type of complementary food first introduced may be of less importance to the disease risk than the timing of introduction of any first solid food. However, we did not observe an association between any first solid food consumed, as defined in Table 1, and IA in the TEDDY cohort.

The finding related to early introduction (≤9 months vs. at a later age) of egg and decreased risk of IA contradicted the finding by Virtanen et al. (15), who suggested that the early introduction of egg (<8 months vs. later) was linked to increased risk of IA during the first 3 years of life. In a recent study in the same population (16), this association was no longer found after children older than 3 years of age were included in the analysis. However, the association in the current study was quite weak, and no association was observed when the egg exposure was investigated as a continuous variable.

Our findings related to the timing of gluten-containing cereals and egg are not consistent with the findings from earlier studies, and the reasons for that can be speculated. Previous studies have been carried out in populations within small geographical areas. The type and timing of first complementary foods as well as the length of the follow-up have varied between the study populations. Moreover, timing has been studied using arbitrary categorization of dietary exposures. There have been differences in the use of dietary supplements (28,31) as well as in types of infant formula (32). Variations in HLA genotype eligibility between the studies may also have contributed to the discrepant findings. It is important to recognize that infant feeding habits change over time. New types of processed foods and dietary supplements are continuously adopted. Use of probiotics during the first year of life has become more common among children in the TEDDY study (31), and they are often given concurrently with new solid foods. Wheat is the main source of gluten in an infant diet (30) and is also an important source of prebiotics (9). Early exposure of both probiotics and prebiotics in gluten-containing cereals like wheat may provide a favorable base for beneficial gut microbiota. However, the role of dietary gluten and wheat in the etiology of T1D remains controversial in animal models (33–35). It has been suggested that wheat may result in mimicry-induced autoimmune disorders given that its peptide sequence is similar to that of human tissue, such as human islet cell tissue (36).

This was the first international study where duration of breastfeeding and timing of the introduction of new foods and their relationship with T1D-related autoantibodies were studied. Overall, we could not confirm the previously published findings between early infant feeding and the risk of IA. Nevertheless, the timing of gluten-containing cereals and the appearance of islet autoantibodies should be studied further. New dietary recommendations for early infant feeding cannot be made based on the current results.

Funding. The TEDDY Study is funded by the National Institute of Diabetes and Digestive and Kidney Diseases grants U01-DK-63829, U01-DK-63861, U01-DK-63821, U01-DK-63865, U01-DK-63863, U01-DK-63836, U01-DK-63790, UC4-DK-63829, UC4-DK-63861, UC4-DK-63821, UC4-DK-63865, UC4-DK-63863, UC4-DK-63836, UC4-DK-95300, UC4-DK-100238, and UC4-DK-106955, and Contract No. HHSN267200700014C; the National Institute of Allergy and Infectious Diseases; the National Institute of Child Health and Human Development; the National Institute of Environmental Health Sciences; JDRF; and the Centers for Disease Control and Prevention. This work was supported in part by the National Institutes of Health/National Center for Advancing Translational Sciences Clinical and Translational Science Awards to the University of Florida (UL1-TR-000064) and the University of Colorado (UL1-TR-001082).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. U.U. contributed to the study design and the acquisition, analysis, and interpretation of data and drafted the article. H.-S.L. performed statistical analysis and contributed to the interpretation of data and the drafting of the manuscript. C.A.A., K.V., J.Y., S.H., and K.S. contributed to the acquisition and interpretation of the data and critically reviewed the manuscript. Å.L., M.R., W.H., J.-X.S., O.S., J.T., A.-G.Z., B.A., J.K., S.M.V., and J.M.N. contributed to the study concept and design and the acquisition and interpretation of data and critically reviewed the manuscript. U.U. and H.-S.L. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.