Effect of Hydrolyzed Infant Formula vs Conventional Formula on Risk of Type 1 Diabetes: The TRIGR Randomized Clinical Trial

Question

Does weaning to an extensively hydrolyzed formula decrease the cumulative incidence of type 1 diabetes in children at risk?

Findings

In this randomized clinical trial that included 2159 children with human leukocyte antigen–conferred susceptibility to type 1 diabetes and at least 1 affected family member, weaning to a hydrolyzed formula compared with a conventional formula did not significantly decrease the cumulative incidence of type 1 diabetes after a median of 11.5 years (8.4% vs 7.6%).

Meaning

Weaning to a hydrolyzed formula did not reduce the risk of type 1 diabetes in children with an increased disease risk.

Importance

Early exposure to complex dietary proteins may increase the risk of type 1 diabetes in children with genetic disease susceptibility. There are no intact proteins in extensively hydrolyzed formulas.

Objective

To test the hypothesis that weaning to an extensively hydrolyzed formula decreases the cumulative incidence of type 1 diabetes in young children.

Design, Setting, and Participants

An international double-blind randomized clinical trial of 2159 infants with human leukocyte antigen–conferred disease susceptibility and a first-degree relative with type 1 diabetes recruited from May 2002 to January 2007 in 78 study centers in 15 countries; 1081 were randomized to be weaned to the extensively hydrolyzed casein formula and 1078 to a conventional formula. The follow-up of the participants ended on February 28, 2017.

Interventions

The participants received either a casein hydrolysate or a conventional adapted cow’s milk formula supplemented with 20% of the casein hydrolysate. The minimum duration of study formula exposure was 60 days by 6 to 8 months of age.

Main Outcomes and Measures

Primary outcome was type 1 diabetes diagnosed according to World Health Organization criteria. Secondary outcomes included age at diabetes diagnosis and safety (adverse events).

Results

Among 2159 newborn infants (1021 female [47.3%]) who were randomized, 1744 (80.8%) completed the trial. The participants were observed for a median of 11.5 years (quartile [Q] 1-Q3, 10.2-12.8). The absolute risk of type 1 diabetes was 8.4% among those randomized to the casein hydrolysate (n = 91) vs 7.6% among those randomized to the conventional formula (n = 82) (difference, 0.8% [95% CI, −1.6% to 3.2%]). The hazard ratio for type 1 diabetes adjusted for human leukocyte antigen risk group, duration of breastfeeding, duration of study formula consumption, sex, and region while treating study center as a random effect was 1.1 (95% CI, 0.8 to 1.5; P = .46). The median age at diagnosis of type 1 diabetes was similar in the 2 groups (6.0 years [Q1-Q3, 3.1-8.9] vs 5.8 years [Q1-Q3, 2.6-9.1]; difference, 0.2 years [95% CI, −0.9 to 1.2]). Upper respiratory infections were the most common adverse event reported (frequency, 0.48 events/year in the hydrolysate group and 0.50 events/year in the control group).

Conclusions and Relevance

Among infants at risk for type 1 diabetes, weaning to a hydrolyzed formula compared with a conventional formula did not reduce the cumulative incidence of type 1 diabetes after median follow-up for 11.5 years. These findings do not support a need to revise the dietary recommendations for infants at risk for type 1 diabetes.

Trial Registration

clinicaltrials.gov Identifier: NCT00179777

Study Design

A randomized, double-blind study was conducted in 78 study centers from 15 countries as previously described. The study protocol is available in Supplement 1. Newborn infants who had a first-degree relative with type 1 diabetes and defined human leukocyte antigen (HLA) genotypes were recruited between May 2002 and January 2007 and followed up until the youngest participant reached 10 years of age in February 2017. Randomization of the infants who met the inclusion criteria took place before birth or immediately after birth (Figure 1). Randomization was stratified by study center, with a block size of 4. Written informed consent was obtained from the family before enrollment. The study was approved by the ethics committees of all participating centers.

Dietary Intervention

Infants were randomly assigned weaning to either the intervention or control formulas, which were produced specifically for this study. Randomization was carried out in each strata within 4 blocks. The intervention formula was an extensively hydrolyzed casein-based formula, while the control formula was composed of 80% intact cow’s milk protein and 20% hydrolyzed milk protein and formulated so that the taste and smell would be indistinguishable from the intervention formula. Study formulas were prepared and coded with the use of 4 colors by Mead Johnson Nutritional and were blinded to all investigators except the data management unit. Newborn infants requiring supplemental feeding before randomization (eg, infants born at night or on weekends) received banked breast milk or Nutramigen, an extensively hydrolyzed casein-based formula.

Breastfeeding was practiced at the discretion of the participating mothers, and maternal diets were unmodified. Breastfeeding was encouraged and exceeded national averages in both groups. The dietary intervention period lasted until the infant was at least 6 months of age and, if by that time the child had not received the study formula for at least 60 days, study formula feeding was continued until 60 days of study formula exposure was reached, but not beyond 8 months of age. Parents were asked not to feed the children any commercial or other baby foods containing bovine protein during the intervention period. Adherence to the protocol was monitored by means of regular family nutrition interviews (at the age of 0.5, 1, 2, 3, 4, 5, 6, 7, and 8 months) and by the analysis of cow’s milk antibodies in serum samples.

HLA Genotyping

Cord blood or a heel stick blood sample collected on filter paper shortly after birth was immediately sent to the Turku (Europe and Australia) or Pittsburgh (North America) laboratories for HLA genotyping. HLA genotyping for the selected DQB1 and DQA1 alleles was performed using sequence-specific oligonucleotide hybridization, with quality control between the 2 laboratories carefully maintained. The following genotypes were regarded as eligible: (1) HLA DQB1*02/DQB1*03:02 (high risk); (2) HLA DQB1*03:02/x (x not DQB1*02, DQB1*03:01, or DQB1*06:02) (moderate risk); (3) HLA DQA1*05-DQB1*02/y (y not DQA1*02:01-DQB1*02, DQB1*03:01, DQB1*06:02, or DQB1*06:03) (mild risk); and (4) HLA DQA1*03-DQB1*02/y (y not DQA1*02:01-DQB1*02, DQB1*03:01, DQB1*06:02, or DQB1*06:03) (rare mild risk).

β-Cell Autoimmunity

ICAs were detected using indirect immunofluorescence. The other 3 autoantibodies were quantified with the use of specific radiobinding assays in the Scientific Laboratory, Children’s Hospital, University of Helsinki, Helsinki, Finland, with cutoff limits for positivity of 2.5 JDF units for ICAs, 2.80 relative units (RU) for IAA, 5.36 RU for GAD autoantibodies, and 0.77 RU for IA-2 autoantibodies. The disease sensitivity and specificity of the ICA assay were 100% and 98%, respectively, in the fourth round of the international workshops on standardization of the ICA assay. According to the Diabetes Autoantibody Standardization Program and the International Autoantibody Standardization Program workshop results in 2002-2016, the disease sensitivities of the IAA, GAD autoantibody, and IA-2 autoantibody radiobinding assays were 42% to 62%, 70% to 92%, and 62% to 80%, respectively. The corresponding disease specificities were 93% to 99%, 90% to 98%, and 93% to 100%, respectively.

Outcomes

The primary end point was the diagnosis of diabetes according to World Health Organization criteria. According to those criteria, the diagnosis is based on (1) symptoms + a single random plasma glucose level of 200 mg/dL or greater (to convert to mmol/L, multiply by 0.0555) or (2) if no symptoms, the diagnosis requires a raised random plasma glucose reading of 200 mg/dL or greater on 2 occasions, a raised fasting plasma glucose reading of 126 mg/dL or greater, or a diabetic oral glucose tolerance test (OGTT, fasting venous plasma glucose ≥126 mg/dL and/or a 2-hour venous plasma glucose ≥200 mg/dL) on 2 occasions. OGTTs were performed by protocol on all study participants who had not been previously diagnosed at 6 and 10 years of age and at study end. Additional OGTTs were performed as clinically indicated. All diagnosed cases were centrally reviewed.

Adverse Events

Undesirable experiences occurring to a child during the trial, whether or not considered related to the investigational product, were reported as adverse events. Serious adverse events were reviewed centrally by the safety monitoring group for this study and were reported annually in tabular form to an external data safety and monitoring board, which reviewed each serious adverse event individually.

Statistical Analyses

The cumulative incidence of diabetes onset from the time of randomization within each group was estimated using a modified Kaplan-Meier diabetes-free survival function. The difference between groups in the cumulative incidence functions, and the associated hazard functions, was tested using the Mantel–log rank test on discrete time to type 1 diabetes (6-month intervals). The relative risk of diabetes onset between groups was estimated from the discrete Cox proportional hazard model. The proportionality assumption of the Cox proportional hazard model was tested. First, the Schoenfeld residuals were examined to determine whether there was an association with time. Second, the interaction of parameters of interest and time were included in the models and tested for significance. For treatment and the variables used in the adjusted models, the null hypothesis of proportionality failed to be rejected. The analyses were adjusted for HLA risk, duration of breastfeeding, duration of study formula consumption, sex, and region, while treating study center as a random effect. The critical value for the test statistic (P = .047) and confidence intervals in this primary analysis were adjusted for multiple looks, which took place during the trial and were based on the Lan and DeMets spending function. When comparing data between the 2 study groups, the t test was applied for normally distributed variables and the nonparametric Mann-Whitney U test for skewed variables.

The effects of weaning to the casein hydrolysate vs conventional formula were tested using the intention-to-treat principle including all HLA-eligible participants who were randomized to a treatment group. Tests of significance reported herein were 2-tailed. Statistical analyses were performed using SAS version 9.4 (SAS Institute). No imputation for missing values was performed; rather, observations with relevant missing values were excluded from respective analyses. The analysis of diabetes risk was also performed according to treatment received (per-protocol analysis). Participants were included in this analysis if they had exposure to the study formula for 60 days or longer and were not exposed to nonallowed foods. This study was designed such that given a confidence level of 95%, an estimated cumulative incidence of diabetes of 7.6% by the age of 10 years in the control group and an expected dropout rate of 20% by 10 years and a frequency of 10% of exclusive breastfeeding (up to age of 6 months), the study would have 80% power to detect a 40% change in the end point. As a post hoc analysis, the hazard ratio of the treatment groups was also calculated after adjusting for the age at which multiple autoantibodies appeared as an exploratory analysis.

Table 1. Demographic Characteristics, Dietary Exposure, and Autoantibody Status of the Trial Participants.

Study Intervention

Eighty percent of infants in the casein hydrolysate group and 80.9% in the control group were exposed to the study formula during the intervention period. The mean (SD) ages of the infants at the time of study formula introduction were 2.0 (2.3) months in the hydrolysate group and 1.8 (2.2) months in the control group (difference, 0.2 months [95% CI, 0-0.42]). The mean (SD) duration of study-formula feeding was 10.2 (9.3) weeks in the casein hydrolysate group and 11.7 (9.7) weeks in the control group (difference, 1.5 weeks [95% CI, 0.7-2.3]; P < .001). As previously reported, the analysis of cow’s milk antibodies confirmed that the families adhered well to the dietary intervention, resulting in conspicuous differences in the antibody levels between the treatment groups.

Progression to Diabetes

The median age at initial seroconversion was 1.6 years (Q1-Q3, 1.0-3.0 years) in the casein hydrolysate group among those who progressed to clinical diabetes, whereas it was 1.5 years (Q1-Q3, 1.0-3.0 years; P = .38) among the progressors in the control group. The mean duration from seroconversion to clinical diabetes was 4.1 years (median, 3.5 years [Q1-Q3, 1.4-6.6]) in the casein hydrolysate group and 3.9 years (median, 3.1 years [Q1-Q3, 1.1-6.2]) in the control group (difference, 0.2 years [95% CI, −0.8 to 1.1]; P = .76). The number of participants who were positive for each specific autoantibody during the preclinical period is shown in Table 1. Five children (5.5%) in the casein hydrolysate group and 6 (7.3%) in the control group had no detectable autoantibodies before the diagnosis of diabetes (P = .62). At diagnosis, the number of autoantibody-negative participants had dropped to 4 (4.4%) and 5 (6.1%), respectively (difference, 1.7% [95% CI, −6.4% to 10.4%]).

Diabetes

During follow-up, diabetes developed in 91 children in the casein hydrolysate group (8.4%) and in 82 in the control group (7.6%) (difference, 0.8% [95% CI, −1.6% to 3.2%]; P = .47; Figure 2). The hazard ratio for type 1 diabetes adjusted for HLA risk group, duration of breastfeeding, duration of study formula consumption, sex, and region, while treating study center as a random effect, was 1.1 (95% CI, 0.8-1.5; P = .46). There was no significant difference in the median age at diagnosis between the 2 groups (6.0 years [Q1-Q3, 3.1-8.9] vs 5.8 years [Q1-Q3, 2.6-9.1]; P = .75; difference, 0.2 years [95% CI, −0.9 to 1.2]). About one-fourth of the cases in each group were diagnosed without clinical symptoms (Table 2). Five children (5.5%) in the casein hydrolysate group and 3 (3.7%) in the control group presented with diabetic ketoacidosis (difference, 1.8% [95% CI, −6.3% to 9.8%]; P = .57). Comparisons between the treatment groups within HLA risk groups, according to the relationship to the affected family member (father, mother, or sibling with diabetes), geographic region associated with the clinical site of enrollment, or sex were not statistically significant (Table 3).

The prespecified per-protocol analysis was defined to include those who were not exposed to any nonallowed foods containing cow’s milk and had exposure to study formula for at least 60 days. The hazard ratio for type 1 diabetes in this subpopulation (n = 1177), adjusted for HLA risk group, duration of breastfeeding, duration of study formula consumption, sex, and region, while treating study center as a random effect, was 1.1 (95% CI, 0.7-1.7; P = .63).

As noted previously, the 2 treatment groups did not differ according to the characteristics of participants who developed diabetes (Table 2) or when analyzed within prespecified subgroups (Table 3). As a post-hoc analysis, the hazard ratio of the treatment groups was estimated after adjusting for the age at which multiple autoantibodies appeared (median, 3.2 years [Q1-Q3, 1.6-6.3] in the casein hydrolysate group and 3.0 years [Q1-Q3, 1.5-6.1] in the control group; P = .42) with little effect on the overall results (hazard ratio, 0.95 [95% CI, 0.70-1.28]; P = .95).

Adverse Events

The frequency of any infection was 0.90 events/year in the hydrolysate group and 0.93 events/year in the control group. The corresponding frequencies of upper respiratory infections were 0.48 and 0.50, respectively. The rate of other adverse events was of the same magnitude in the 2 groups (eTable in Supplement 2). Similar linear growth and weight gain were observed in both groups.

Limitations

The results of this study are not directly generalizable to the background population because participants were selected based on a positive family history for type 1 diabetes and an HLA genotype conferring risk for type 1 diabetes. In addition, the outcome is not necessarily applicable to children with other HLA genotypes.

• 1.Knip M. Can we predict type 1 diabetes in the general population? Diabetes Care. 2002;25(3):623-625. [DOI] [PubMed] [Google Scholar]
• 2.Knip M, Siljander H, Ilonen J, Simell O, Veijola R. Role of humoral beta-cell autoimmunity in type 1 diabetes. Pediatr Diabetes. 2016;17(suppl 22):17-24. [DOI] [PubMed] [Google Scholar]
• 3.Ziegler A-G, Rewers M, Simell O, et al. Seroconversion to multiple islet autoantibodies and risk of progression to diabetes in children. JAMA. 2013;309(23):2473-2479. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 4.Libman IM, LaPorte RE. Changing trends in epidemiology of type 1 diabetes mellitus throughout the world: how far have we come and where do we go from here. Pediatr Diabetes. 2005;6(3):119-121. [DOI] [PubMed] [Google Scholar]
• 5.Patterson CC, Dahlquist GG, Gyürüs E, Green A, Soltész G; EURODIAB Study Group . Incidence trends for childhood type 1 diabetes in Europe during 1989-2003 and predicted new cases 2005-20: a multicentre prospective registration study. Lancet. 2009;373(9680):2027-2033. [DOI] [PubMed] [Google Scholar]
• 6.Ziegler A-G, Hummel M, Schenker M, Bonifacio E. Autoantibody appearance and risk for development of childhood diabetes in offspring of parents with type 1 diabetes: the 2-year analysis of the German BABYDIAB Study. Diabetes. 1999;48(3):460-468. [DOI] [PubMed] [Google Scholar]
• 7.Kimpimäki T, Kupila A, Hämäläinen A-M, et al. The first signs of β-cell autoimmunity appear in infancy in genetically susceptible children from the general population: the Finnish Type 1 Diabetes Prediction and Prevention Study. J Clin Endocrinol Metab. 2001;86(10):4782-4788. [DOI] [PubMed] [Google Scholar]
• 8.Ilonen J, Kiviniemi M, Lempainen J, et al. ; Finnish Pediatric Diabetes Register . Genetic susceptibility to type 1 diabetes in childhood: estimation of HLA class II associated disease risk and class II effect in various phases of islet autoimmunity. Pediatr Diabetes. 2016;17(suppl 22):8-16. [DOI] [PubMed] [Google Scholar]
• 9.Kostic AD, Gevers D, Siljander H, et al. ; DIABIMMUNE Study Group . The dynamics of the human infant gut microbiome in development and in progression toward type 1 diabetes. Cell Host Microbe. 2015;17(2):260-273. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 10.Virtanen SM, Räsänen L, Aro A, et al. ; Childhood Diabetes in Finland Study Group . Infant feeding in Finnish children less than 7 yr of age with newly diagnosed IDDM. Diabetes Care. 1991;14(5):415-417. [DOI] [PubMed] [Google Scholar]
• 11.Norris JM, Barriga K, Klingensmith G, et al. Timing of initial cereal exposure in infancy and risk of islet autoimmunity. JAMA. 2003;290(13):1713-1720. [DOI] [PubMed] [Google Scholar]
• 12.Ziegler A-G, Schmid S, Huber D, Hummel M, Bonifacio E. Early infant feeding and risk of developing type 1 diabetes-associated autoantibodies. JAMA. 2003;290(13):1721-1728. [DOI] [PubMed] [Google Scholar]
• 13.Virtanen SM, Kenward MG, Erkkola M, et al. Age at introduction of new foods and advanced beta cell autoimmunity in young children with HLA-conferred susceptibility to type 1 diabetes. Diabetologia. 2006;49(7):1512-1521. [DOI] [PubMed] [Google Scholar]
• 14.Knip M, Virtanen SM, Åkerblom HK. Infant feeding and the risk of type 1 diabetes. Am J Clin Nutr. 2010;91(5)(suppl):1506S-1513S. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 15.Knip M, Åkerblom HK, Becker D, et al. ; TRIGR Study Group . Hydrolyzed infant formula and early β-cell autoimmunity: a randomized clinical trial. JAMA. 2014;311(22):2279-2287. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 16.Åkerblom HK, Krischer J, Virtanen SM, et al. ; TRIGR Study Group . The Trial to Reduce IDDM in the Genetically at Risk (TRIGR) Study: recruitment, intervention and follow-up. Diabetologia. 2011;54(3):627-633. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 17.Sorkio S, Cuthbertson D, Bärlund S, et al. ; TRIGR Study Group . Breastfeeding patterns of mothers with type 1 diabetes: results from an infant feeding trial. Diabetes Metab Res Rev. 2010;26(3):206-211. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 18.Parkkola A, Härkönen T, Ryhänen SJ, Ilonen J, Knip M; Finnish Pediatric Diabetes Register . Extended family history of type 1 diabetes and phenotype and genotype of newly diagnosed children. Diabetes Care. 2013;36(2):348-354. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 19.Alberti KG, Zimmet PZ. Definition, diagnosis and classification of diabetes mellitus and its complications, part 1: diagnosis and classification of diabetes mellitus provisional report of a WHO consultation. Diabet Med. 1998;15(7):539-553. [DOI] [PubMed] [Google Scholar]
• 20.Goggins WB, Finkelstein DM. A proportional hazards model for multivariate interval-censored failure time data. Biometrics. 2000;56(3):940-943. [DOI] [PubMed] [Google Scholar]
• 21.Lan KK, DeMets DL. Changing frequency of interim analysis in sequential monitoring. Biometrics. 1989;45(3):1017-1020. [PubMed] [Google Scholar]
• 22.Knip M, Virtanen SM, Seppä K, et al. ; Finnish TRIGR Study Group . Dietary intervention in infancy and later signs of beta-cell autoimmunity. N Engl J Med. 2010;363(20):1900-1908. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 23.Warram JH, Krolewski AS, Gottlieb MS, Kahn CR. Differences in risk of insulin-dependent diabetes in offspring of diabetic mothers and diabetic fathers. N Engl J Med. 1984;311(3):149-152. [DOI] [PubMed] [Google Scholar]
• 24.Familial risk of type I diabetes in European children: the Eurodiab Ace Study Group and the Eurodiab Ace Substudy 2 Study Group. Diabetologia. 1998;41(10):1151-1156. [DOI] [PubMed] [Google Scholar]
• 25.Lund-Blix NA, Dydensborg Sander S, Størdal K, et al. Infant feeding and risk of type 1 diabetes in two large Scandinavian birth cohorts. Diabetes Care. 2017;40(7):920-927. [DOI] [PMC free article] [PubMed] [Google Scholar]
• 26.Vaarala O, Knip M, Paronen J, et al. Cow’s milk formula feeding induces primary immunization to insulin in infants at genetic risk for type 1 diabetes. Diabetes. 1999;48(7):1389-1394. [DOI] [PubMed] [Google Scholar]
• 27.Vaarala O, Ilonen J, Ruohtula T, et al. Removal of bovine insulin from cow’s milk formula and early initiation of beta-cell autoimmunity in the FINDIA pilot study. Arch Pediatr Adolesc Med. 2012;166(7):608-614. [DOI] [PubMed] [Google Scholar]

Supplementary Materials