NUTRITIONAL PRIMARY PREVENTION OF TYPE 1 DIABETES IN CHILDREN (TRIGR)

An intriguing question is the mechanism by which CMP may cause the lesion of the pancreatic ß-cells. Several mechanisms have been proposed, e.g. molecular mimicry between a fraction of bovine serum albumin (BSA) and a ß-cell antigen, a possible deleterious effect of other whey protein components of CM, such as ß-lactoglobulin, and of ß-casein, bovine insulin, etc (reviewed in ref. 7). A recent ecological study indicated that the diabetogenicity of CM might bedue to the A1 variant of bovine ß-casein (8). Regardless of the mechanism, we consider the indirect evidence from animal experiments and observations in man, particularly from the experience from our second pilot study in Finland, described below under b.a.c., and the risk linked with early supplementation of CMP reviewed briefly above, to be sufficient to justify a dietary intervention trial. The formula to be used, a casein hydrolysate, does not contain whey proteins, such as BSA or ß-lactoglobulin, and the small peptides of the casein component (<1500 D) are characterized by low immunogenicity. This formula has been shown to be non-diabetogenic in the NOD mouse (9). In the second pilot of the present TRIGR-study in Finland the use of the test formula resulted in lower prevalence of diabetes-related autoantibodies in children at increased genetic risk of type 1 diabetes (10).

We certainly agree with the notion that prospective studies, considering both the timing and duration of exposure to CMPs are worthwhile, preferably also in different populations, as the genetic-environmental interactions may vary from one population to another (12). But even prospective studies do not seem to give uniform answers: the question of a possible effect of BF vs. early supplementation with CMP containing formula on the development of autoimmune markers of type 1 diabetes has been addressed by several research groups. The DAISY group in Denver, the German BABY-DIAB Study Group and the Australian Baby DIAB Study Group have reported the absence of an association between the duration of BF or introduction of CM, and the development of islet autoimmunity in children with first degree relatives (FDR) with type 1 diabetes (13-15). However, in the Finnish Diabetes Prediction and Prevention (DIPP) Study the opposite has been observed: the series consists of children with an increased genetic risk, identified from the general population. In a nested case-control study the first 65 children who seroconverted to ICA positivity (cases) before the age of 4 years and 390 ICA negative control subjects (six controls/case), matched for date of birth, sex and HLA-DQB1-genotype, were compared. The risk of IA-2A positivity or positivity for all four antibodies (ICA, IAA, GADA and IA-2A) was lower in children exclusively breast-fed for more than 3 months compared to those who were breast-fed for 3 months or less. Also infants in whom supplementary milk feeding had started after the age of 3 months were at lower risk of IA-2A positivity or positivity for all four autoantibodies (16). The associations remained significant after adjustment for maternal age, duration of maternal education and 12 month relative weight and height of the child. These observations suggest that short BF and early introduction of CM based infant formula predisposes young children at increased genetic risk for diabetes to progressive signs of ß-cell autoimmunity.

There are many indications by now that the pathogenetic process leading to manifest type 1 diabetes may start in some cases very early in life, even in utero (17,18). In addition, during childhood the incidence of the disease has over the last decade shifted towards younger age in several countries (19). Therefore, studies on environmental factors in the first years of life, among them the role of CMPs are crucial. CMPs are unique in one respect: in industrialized countries they are the first foreign proteins entering the infant gut, since most infant formulas are CM based. In a critical review of the hypothesis, published in August 1999, it was stated that the real debate is about mucosal immune function as to the possible role of CM in the pathogenesis of type 1 diabetes (11), which is in line with a proposal from our group some years ago in relation to the role of CMPs in the pathogenesis of type 1 diabetes (20,21).

In the second pilot study 207 children with an increased genetic risk (first degree relative(s) with type 1 diabetes, HLA DQB1*0302 and/or DQB1*02 positive, absence of DQB1*0602, *0603 and *0301) have been followed for the observation period of 2 years. The distribution of the HLA DQB1*02/0302, *0302/x and *02/y genotypes was rather similar in the two study groups. The infants received after exclusive BF in a double blind, randomized design either a casein hydrolysate formula or a conventional CM-based formula until the age of 6-8 months. Antibodies to GAD65, IA-2 and insulin (IAA) were assayed by radiobinding assays, and islet cell antibodies ( ICA ) by immunofluorescence. By 24 months of age, 5/98 (5.1 %) of the subjects in the casein hydrolysate group had developed at least one autoantibody, whereas 14/109 (12.8 %) in the control group had done so (P=0.09, Fisher's exact test). IAA were the first to appear in the latter group. Theeffect of the intervention on the risk of seroconversion of any of the four autoantibodies was as follows: casein hydrolysate vs. CM-based formula, risk ratio (RR) 0.40 (95% CI 0.13-1.03), p=0.059. As there were some differences between the two groups in the duration of BF and formula feeding, the analysis was done by adjusting for the duration of study formula feeding, the RR being 0.34 (0.11- 0.89), p=0.027. These findings are supportive of the hypothesis and justify moving to a full study described in this proposal.

These preliminary findings suggest that elimination of dietary CMPs and replacement with hydrolyzed casein formula over the first 6-8 months may modulate the appearance of diabetes associated autoantibodies in children at increased risk for type 1 diabetes over the first 2 years of life.They also demonstrate for the first time in man that it may be possible to modify spontaneous ß -cell autoimmunity by dietary intervention.

It is our firm opinion that the CMP and type 1 diabetes issue will not be solved without a properly planned, sufficiently powerful, prospective, double-blinded randomised intervention trial. Association studies have an inherent weakness in identifying a ubiquitous, noxious agent, such as CMP. For instance, the cause of coeliac disease would likely not have been discovered were it not for the unavailability of wheat in the Dutch World War II diet (22).The prevailing controversies make an intervention study even more important than earlier.

b.b. Significance

The consequences and costs of type 1 diabetes presenting in childhood are immense, not only economically and for the society, but also with regard to human life. Microvascular complications develop in a considerable proportion of the patients in the course of time affecting their quality of life. Those individuals manifesting type 1 diabetes in childhood and adolescence may have a severalfold increased risk of macrovascular complications in adult life. It is therefore easy to realize benefits which would result if we could prevent type 1 diabetes even in a part of the cases. The economic burden of type 1 diabetes is admittedly difficult to assess, the costs being composed of three elements: direct, indirect and psychological (23). However, it has been assessed that the life-time cost of an individual contracting type 1 diabetes in childhood may reach 1 million USD. Thus it is evident that the high incidence of type 1 diabetes, like in Finland and several other countries participating in this project not only carries a great social but also a heavy economical impact as well. It is of interest to note the recent increasing attention by WHO in early nutrition in the context of the etiopathogenesis of type 1 diabetes (24), and it has been proposed that research should respond with a new focus on early events (25). Our research proposed does just that. A positive outcome of the present proposal would thus have a major input on the economic and social consequences of type 1 diabetes.

The accumulated data suggest the simple and testable hypothesis that temporary but strict avoidance of CMPs in the vulnerable early infancy period could prevent the development of type 1 diabetes in genetically susceptible children. If the hypothesis is proven correct, primary prevention of type 1 diabetes and avoidance of the associated morbidity, mortality and health care costs would for the first time become a realistic goal. An additional benefit is the possibility to gain important insight in to the immunological events in high risk infants during the first years of life, so poorly understood so far, and into the natural course of those infants who have been on the control formula. As about half of the children in our study are control subjects, we can also learn of the natural history of the development of islet cell related antibodies and other measured variables in this part of the study.

We like to emphasize, that this is an innovative trial, being the first ever started primary prevention trial against type 1 diabetes. In contrast to the current secondary prevention strategies which involve extensive genetic and/or autoimmune marker screening, the innocuous present strategy could be directly applied to the general population from where 90 % of the new cases of type 1 diabetes are derived, most of whom have the increased risk MHC genotype, selected in this proposal.

In the following years several investigators from other countries, especially Canada and Sweden, joined the Finnish group in the planning of a pilot study. Professor Hans-Michael Dosch of Toronto contributed with immunological expertise and established the contact with the Mead Johnson Nutritionals company, the provider of the study formulas for the first and second pilot studies. Epidemiological, nutritional and clinical input was given in the planning by the Hamilton,Ont., Canada group (Dr. John Vandermeulen, Associate Professors Hertzel Gerstein and Stephanie Atkinson), and also by Professors Gisela Dahlquist of Umeå and Johnny Ludvigsson of Linköping, Sweden. The Finnish group consisted of experts in pediatrics (Professors Hans K.Åkerblom and Mikael Knip, 15 local investigators), including e.g. infant nutrition and pediatric gastroenterology (Dr. Erkki Savilahti ), neonatology (Drs. Anna-Liisa Järvenpää and Martin Renlund ), and in obstetrics ( Professor Kari Teramo ), nutrition science (Dr. Suvi Virtanen ), genetics (Dr. Jorma Ilonen), immunology (Professor Mikael Knip, Dr. Outi Vaarala ), epidemiology (Dr. Antti Reunanen ), and virology (Dr. Heikki Hyöty).

In addition to providing practical information regarding study implementation, valuable data on immunological responses to dietary CM antigens in early infancy and the development of islet-cell specific autoantibodies were obtained. We observed that oral exposure to foreign proteins, i.e. CMPs in infancy resulted in both cellular and humoral immune responses. No responses to ß-lactoglobulin or BSA proteins were seen in infants given a hydrolysed formula. In both groups the immune responses to CMPs declined during a 3-year follow-up suggesting the development of oral tolerance (27, unpublished observations). One infant became positive for IAA before the age of 6 months, with increasing levels later, seroconverted to positivity for ICA and GAD65A between 6 and 9 months and presented with clinical type 1 diabetes at the age of 14 months. He carried the HLA DQB1*0302/x genotype which predisposes to type 1 diabetes and had been given the casein hydrolysate formula as supplementary milk. The appearance of IAA before the age of 6 months and of ICA and GAD65A before 9 months shows that signs of autoimmune â-cell damage may emerge at a very young age. He was the only subject in the total series with signs of a neonatal enterovirus infection, and suffered another enterovirus infection at the end of the first year (28,29). Evidently the dietary manipulation did not prevent this child from contracting type 1 diabetes at an early age, the etiological factor very likely being the neonatal and subsequent enterovirus infection.

These observations lead us to devise a larger, second, multicenter pilot study, aimed besides at testing further the practical implementation of the project and the feasibility of genetic screening, particularly at studying the development of immunological variables during the first 2 years of life.

c.a.b. Second pilot study

The second pilot study was carried out in 15 centres in Finland, the recruitment being initiated in April 1995 and finished in November 1997, the last child reaching thus the age of 2 years in November 1999. The experience and results will be presented briefly below. Smaller series were started later on with EC-support in one center in Sweden, two in Estonia and one in Hungary, and in addition genetic studies and network building in Sardinia, Italy. The results on the type 1 diabetes associated autoantibodies from the other countries are not yet available, as the follow-up is still continuing.

Newborns who had FDRs with type 1 diabetes (i.e. a mother, father or sibling), who met the inclusion and who did not meet the excluson criteria were recruited. The inclusion and exclusion criteria were the same as used in the study proper, presented under d.a., p. 14. To facilitate recruitment and to minimize any possibility of inadvertent exposure to CMP, every attempt was made to identify eligible families before the child was born. Written consent was obtained at this time; the child participated after birth if he/she met the inclusion but did not meet the exclusion criteria. Families not identified until just prior to the onset of maternal labor were approached at that time. Affected mothers were identified during pregnancy via hospitals monitoring preg­nant women with type 1 diabetes. Fathers with type 1 diabetes were identified by a) available history or data already in the medical record of pregnant women, and b) interviewing women at prenatal maternity clinic visits.Pregnant mothers already having one or more children affec­ted by diabetes were identified through various diabetes clinics. Children with type 1 diabetes are usually followed by pediatricians or pediatric endocrinolo­gists, who will be aware of a pregnant mother. These mothers were identified through these physicians.

During the recruitment period of 2 years and 8 months, 521 mothers agreed to join the study. Forty-five newborn infants were excluded, and 476 were subjected to genetic screening and they got a study code at birth. After genetic screening, 234 continued in the intervention study as they had the HLA genotypes conferring increased genetic risk. The drop-out rate over the 2-year follow-up period was 21%, the majority of them occurring before the age of 3 months, i.e. before the first follow-up blood sampling. Therefore we have autoantibody data on 207 children, to be reported below. The reason for dropping out was in about half of the cases some family problem, and the next most common was related to the study formula ( refusal by parents to give or by the child to take it ).

Our co-ordinating staff, the co-ordinating nurse and co-ordinating nutrition advisor have put in a lot of effort to improve the recruitment rate at various hospitals in Finland. It has turned out, that in addition to the initial information to the various parties (nurses, physicians, parents etc.) also repeated newsletters with information and reports on the ongoing study and also site visits are necessary to maintain the recruitment rate. After the 2-year follow-up we have sent out to the participating mothers, a questionnaire asking for their opinions of participating in the study. Only about 54 % have so far responded, and the results are presently being analysed.

The subjects were randomized immediately after birth to define the formula allocation. The randomization code was kept at the data base at the Hospital for Children and Adolescents, University of Helsinki, Helsinki. The recruited subjects were randomi­zed to receive either the test formula ( NutramigenTM, Mead Johnson Nutritionals, USA, not containing intact CMP), or a CMP containing control formula which had an addition of Nutramigen to eliminate taste and smell difference between the two study formulas, whene­ver breast milk was not available. Any subject requiring supple­mental feeding prior to randomi­za­tion (e.g. subjects born at night or on weekends) was given banked breast milk or Nutra­migenTM.

All recruited mothers were encouraged to breast feed; newborn infants were, however, randomi­zed as soon as possible after birth so that any elective formula supplementation or weaning by the mother were done with the appropriate formula. The duration of the intervention was until at least 6 months of age. If the mother chose to exclusive­ly breastfeed up to the age of 6 months (official recommenda­tion of both the Finnish and Swedish Pediatric Societies) she was advised, thereafter, to give the study formula for at least 2 months, i.e. until the age of 8 months. Similarly, if exclusive BF lasted for 5 months, the infant received the study formula until the age of 7 months. Infant feeding practices were altered as little as possible by the trial. In particular, BF practice(s) was entirely at the discretion of participating mothers. However, all relevant data were recorded. Compliance with the denial of CM was monitored by assessing formula usage, regular interviews with participating families, and by appropriate serology.

The overall experience by our staff and participating centers has been positive in relation to the parents' and infants' participation in our study. The various study formulas have been accepted well in general except for a few instances. We have stressed the importance of giving enough time to the mothers at the follow-up visits to discuss the participa­tion, and particularly to relieve possible anxiety triggered by the knowledge of an increased genetic risk in the study subject in families with at least one affected number.

c.a.b.a. Genetic screening

Cord blood was collected at birth in an EDTA-tube, and the sample was forwarded to the genotyping centre in Turku, Finland (Dr. Jorma Ilonen) and screening was performed for the presence of HLA DQB1*0302, DQB1*02, DQB1*0602-03, and DQB1*0301. Results were sent to the data centre in Helsinki, Finland within 5-7 days.Only subjects who were HLA DQB1*0302 and/or DQB1*02-positive and HLA DQB1*0602, *0603 and *0301-negative were invited to continue in the study; all other subjects (approxi­mately 50% of randomized subjects) were withdrawn from the study at that time; their parents were told that genetic screening suggests that their child is at low risk for type 1 diabetes. The distribution of HLA -genotypes in the two groups was as follows: DQB1* 02/*0302 22%, *0302/x 39%, and *02/y in 39% in the CM-based formula group, and 21%, 41%, and 38% in the casein hydrolysate group, respectively.

The distribution ofDQB1 genotypesin Finnish familial cases is shown in Table 1. This suggests that the sensitivity of the used selection criteria is 84.6%,when 50.7% of the studied newborns can be excluded from the study. In terms of odds ratios the subjects with selected risk genotypes had an OR value of 5.7 compared to 0.2 of nonselected. When the general familial risk of 3% by the age of 10 years is used as the basis of the calculation, the predictive positive value shows that 5.4 % of selected children could be expected to develop diabetes compared to only 0.8 % within the nonselected group.

Table 1. HLA-DQB1 genotypes in Finnish newborn children recruited to the TRIGR study compared to familial type 1 diabetes cases.

*Data from reference 30

To ensure the usefulness of genetic screening also in the other participating populations type 1 diabetic children andcontrol subjects from Estonia and Sardinia were analysed with the same screening method (Table 2). Information on the genotypes in Hungary was obtained from the Medical University of Pecs (Dr Robert Hermann).

The sensitivity was highest in Sardinia but the number of DQB1*02 positive subjects in the control population is very high there making the screening system relatively unspecific. The peculiar characteristic of the Sardinian population was also very clear when the effect of adding a further step to the screening procedure was tested in samples from different populations. This was made by typing series of samples with DQB1*02/x (x#*0301,*0302,*0602,*0603) genotype for the presence of DQA1*05 (DR3 haplotype) or DQB1*0201 (DR7 haplotype) with an additional PCR and hybridisation reaction. Also in the control population the frequency of DR3 haplotype was very high in Sardinia and and DR7 haplotype was very rare making the value of DQA1 typing useless (Table 2). The additional value obtained was also limited in the Finnish population, because the prevalence ofthe DR7 haplotype was rather low and its effect neutral. Instead, in the Estonian and Hungarian populations this testing step seems to be very useful because of the higher RR value associated with the DR3 haplotype and especially also because the DR7 haplotype in these populations seems to be protective.

Table 2. The risk carried by the by HLA-DQB1*02/x genotype when divided according to the presence of DQA1*05 and/or DQA1*0201 alleles in the haplotype

*Hermann et al. Unpublished data

The additional value of DQA typing in selected cases was also demonstrated in Russian and Latvian populations (31).

Based on these studies we have decided that the study proper will be conducted using a two phase screening where HLA-DQB1 genotyping is the first phase. Those with DQB1*02/*0302 and DQB1*0302/x (x¹*02,*0301 or *0602) genotype are directly selectedfor the study and those with DQB1*02/y (y¹*0301,*0302, 0602 or 603) genotype are analysed for the presence of DQBA1* alleles DQA1*05 and DQA1*0201. Those who are positive for DQA1*05 but not for DQA1*0201 are also selected to continue in the study group.

During the second pilot study we established logistics for sample distribution and swift information transfer between participating hospitals and the laboratory. The DNA hybridisation based method does not have any strict requirements for howthe samples should be sent but the main consideration is the need for rapid genotyping. The use of courier delivery of samples and e-mail or fax in transferring the genotyping result was observed to ensure smooth data processing also in Estonian, Hungarian and Swedish centers participating in the project.

Islet-cell related autoimmune markers were analysed in the Research Laboratory, Department of Pediatrics, University of Oulu under the supervision of Professor Mikael Knip. The methodology for the assays of islet cell antibodies (ICA), insulin antibodies (IAA), antibodies to the 65kD isoform of glutamic acid decarboxylase (GADA) and antibodies to the protein tyrosine phosphatase related IA-2 molecule (IA-2A) was the same as described in the present application under d. RESEARCH DESIGN AND METHODS, d.i.b. Methodology for Humoral Markers of Beta-Cell Autoimmunity.

 The results from the 207 children in Finland are summarized in Table 3.

Table 3. Emergence of ICA, IAA, IA-2A and GADA by the age of 2 years (n=207)

A decrease in the frequency of autoantibodies in the casein hydrolysate vs. control formula group occurredfor all variables except for GADA, the relative decreases being: ICA 63%, IAA 51%, IA-2A 64%, at least one autoantibody 60% and at least two autoantibodies 44%.

Theeffect of the intervention on the risk of seroconversion of any of the four autoantibodies was as follows: casein hydrolysate vs. CM-based formula, risk ratio (RR) 0.40 (95% CI 0.13-1.03), p=0.059. As there were some differences between the two groups in the duration of BF and formula feeding, the analysis was done by adjusting for the duration of study formula feeding, the RR being 0.34 (0.11- 0.89), p=0.027.
(Here a summary of the CM antibody findings from Dr.Erkki Savilahti, when the results are available in March 2000).

c.a.b.d. Child health

The children and their mothers visited the outpatient clinics when the child was 3, 6, 9, 12, 18 and 24 months old, being seen by the study nurse or nutrition advisor, and at certain intervals by the pediatrician. A physical examination was performed, and a venous blood specimen was obtained after using a local analgesic ointment for venipuncture. The children's health was in general good. The height and weight development in the two groups was similar.

By now, after analysing the type 1 diabetes associated autoantibody data, we can rule out option 3, i.e. we have not done harm to the investigated children. We feel strongly that we have the justification and obligation to continue into the study proper phase with the purpose to get a final answer to the study question mentioned under a. SPECIFIC AIMS , p.1.

c.b. Other studies related to the CM and type 1 diabetes hypothesis, and to the prediction of type 1 diabetes

The undersigned P.I. and his associates have over the last 15 years conducted several studies with relevance to the CM and type 1 diabetes hypothesis. Among them several came from the nationwide "Childhood Diabetes in Finland" (DiMe) Study, aimed at elucidating the genetic, immunological and environmental determinants of type 1 diabetes in children in Finland, the undersigned P.I. being the Project Director and one of the two P.I.'s of the DiMe-project. The population-based DiMe-study comprised 750 consecutively diagnosed 0-14-year-old children with newly diagnosed type 1 diabetes, 819 3-19-year-old non-diabetic siblings, and control subjects (34).

c.b.a.Infant feeding studies

In the DiMe-study Virtanen et al. showed an inverse association between the duration of BF and risk of type 1 diabetes in 0-6-year-old (35) and 7-14-year-old (36) subjects. We were the first to show that after adjustment for duration of BF, an increased risk of type 1 diabetes in conjuction with early introduction of dairy products was still present (5).

As others have proposed that an increased weight gain in infancy may act as a risk factor for type 1 diabetes in children, Hyppönen et al. (37) analysed weight measurements from the 1st year of life of 435 full-term subjects with type 1 diabetes and 386 control subjects. The results indicated that an early exposure to CM formula feeding and rapid growth in infancy are independent risk factors of childhood type 1 diabetes.

c.b.b. Studies on the role of CMPs in the pathogenesis of type 1 diabetes

Savilahti et al. of our group in Helsinki were the first to report in man on elevated levels of CM and b-lactoglobulin antibodies at the diagnosis of type 1 diabetes (38). This finding was later confirmed in the DiMe-study, comprising 706 children with newly diagnosed type 1 diabetes, 456 non-diabetic siblings, and 105 unrelated age-matched controls below 7 years of age (39). Multivariate analyses including infant feeding and milk consumption variables, current age and ICA status indicated that young age at the introduction of dairy products and high CM consumption during childhood inrease the levels of CM antibodies and that high IgA CM antibodies are independently associated with an increased risk of type 1 diabetes (40). Features of our data suggest that antibodies to CM and b-lactoglobulin are not derived from the autoimmune process probably responsible for destroying ß-cells and causing type 1 diabetes. Instead, we inferred that a high titre of CM antibodies is an epiphenomenon of an insult and immunization caused by CMPs, which in some genetically susceptible individuals may lead to type 1 diabetes after a long but variable time interval (39).

We have also tested the cellular immunity to several CMPs (b-lactoglobulin, BSA, a-casein, ovalbumin) in patients with newly diagnosed type 1 diabetes, and only T-cell reactivity to b-lactoglobulin was enhanced in affected patients when compared to the non-diabetic children (20).

We have studied the role of the gut immune system in type 1 diabetes by analysing the expression of a gut-specific adhesion molecule (a4b7-integrin) on the islet cell antigen (GAD)-reactive lymphocytes from patients with type 1 diabetes (21).We found a marked decrease in proliferation of PBMC to GAD in four of seven patients with type 1 diabetes, whereas the proliferation response to a parenteral antigen, tetanus toxoid increased, when a4b7-integrin-positive cells were depleted. These findings demonstrate that islet cell-reactive lymphocytes express a gut-specific homing receptor,which emphasizes the role of gut immunity in type 1 diabetes. This regulatory defect in the gut lymphocyte population may be responsible not only for b-cell reactivity but also for enhanced reactivity against oral antigens, e.g. CMPs (21). We also observed that orally fed CMPs induce an elevation in soluble intercellular adhesion molecule 1 (ICAM-1) in infants, possibly reflecting the generation of an immune response against these proteins, because ICAM-1 has an important role in lymphocyte activation (41).These observations, and the findings on humoral and cellular immunity to dietary CMPs can be interpreted as indicative of dysregulated oral tolerance. This regulatory defect of the gut immune system may actually play a fundamental role in the pathogenetic process leading to type 1 diabetes, as proposed earlier by Vaarala et al. in our group (20,21).

Vaarala et al. have recently generated evidence for a new possibility to explain the epidemiological link between the risk of type 1 diabetes and early exposure to CM containing formulas, i.e. by immunization to bovine insulin (42,43). It is known that CM products contain bovine insulin. Initially Vaarala et al. observed that CM-based formula could induce insulin-binding antibodies in children (42). Thereupon, in a larger series it was found that the amount of IgG-antibodies binding to bovine insulin was higher at 3 months of age in infants who were exposed to CM-based formula thanin infants who were exclusively breast-fed at that age ( p< 0.0001) (43). In that series there were 11 subjects who developed signs of b-cell autoimmunity before the age of 2 years. Those children had increased levels ofbovine insulin antibodies over time in contrast to the other children whose antibody levels to bovine insulin decreased after the age of 12 months. These observations imply that the immune response initially induced by bovine insulin may later be diverted into autoaggressive immunity against the b-cells in a few unfortunate individuals, i.e. those who subsequently experience b-cell damage and progress to clinical type 1 diabetes.

c.b.c. Studies on serological risk markers and prediction of type 1 diabetes

The results of the present project are depending not only on the amount of manifest type 1 diabetes cases in the intervention and control groups, but also on the appearance of "surrogate" markers of the disease, i.e. type 1 diabetes associated autoantibodies. Therefore it is relevant to describe below briefly studies of our group on these markers, important in the prediction of the disease.

Knip et al. described the natural history of preclinical type 1 diabetes in a group of 57 siblings positive for ICA and/or IAA when first screened within 6 months after the diagnosis of the proband (44). Those who progressed to clinical diabetes were characterized by young age, strong and increasing signs of islet-cell specific autoimmunity, reduced insulin secretory capacity and emerging glucose intolerance. Kulmala et al. evaluated the predictive characteristics of ICA, IAA, GADA and antibodies to IA-2A in an unselected population of 755 siblings of children with type 1 diabetes, 32 of whom progressed to clinical disease within 7.7 years of the initial sample taken (45). The risk for type 1 diabetes in siblings with four, three, two, one, or no antibodies was 40, 70, 25, 2 and 0,8 %, respectively. The results indicated that the combined screening for IA-2A and GADA may replace the ICA assay, giving comparable sensitivity, specificity, and positive predictive value. In another study on the same series, Mrena et al. evaluated the efficacy of grading the siblings into different risk categories based on the number of positive antibodies alone, or in combination with the first-phase insulin response FPIR (46). The results indicated that it is possible to grade siblings of children with newly diagnosed type 1 diabetes into categories with significant differences in the subsequent risk of clinical disease and time to diagnosis.

Kulmala et al. studied the relationships between genetic markers and disease-associated autoantibodies in an unselected population of 701 siblings in the DiMe-study over a period of 9 years (47). The combination of the genetic and immunological markers increased the positive predictivevalues of all autoantibodies substantially. Because the combination also resulted in reduced sensitivity, autoantibodies alone were recommended as the first-line screening in siblings. Finally, Kimpimäki et al. evaluated the emergence of diabetes-associated autoantibodies in 180 initially unaffected siblings, younger than 6 years at the initial sampling. Fifteen siblings (18%) manifested type 1 diabetes before the age of 10 years. All progressors had increased HLA-defined genetic susceptibility . Thirteen of the 17 siblings testing positive fortwo or more autoantibodies before the age of 6 years became diabetic before the age of 10 years (48). These findings have been utilized in the power calculations of the study proper phase in the present application.

c.c. Rewiew articles on environmental factors in type 1 diabetes and prevention trials

The undersigned P.I. has with his associates in recent years reviewed the putative role of various environmental factors, and particularly the role of CMPs in the etiology of type 1 diabetes (49,7,50-52), the interaction of genetic and environmental factors in the pathogenesis of the disease (11), participated in the discussion on the role of CMPs in the pathogenesis of type 1 diabetes (53,54), and reviewed critically the type 1 diabetes prevention trials in progress (55). Some of these articles are enclosed as Appendix:

d.RESEARCH DESIGN AND METHODS

d.a.Sample Specification

Newborn infants who have FDRs with type 1 diabetes (i.e. a mother, father or sibling), and who meet the inclusion but not the exclusion criteria will be recruited:

d.b.Sample Size

The estimation of the sample size presented in Table 4 is based onexperiences from family studies analyzing the occurrence of autoantibodies in siblings of children with type 1 diabetes or in offspring of affected parents and progression to clinical disease in such young first-degree relatives. The data on the frequency of multiple (> 2) autoantibodies by the age of 6 years and the cumulative incidence of type 1 diabetes by the age of 10 years are based on 82 young siblings from the DiMe Study carrying increased genetic risk for type 1 diabetes as defined according to the criteria in the second TRIGR pilot trial (48).

Nineteen out of 325 offspring (5.9%) with increased genetic risk according to the criteria used in the second TRIGR pilot trial tested positive for at least one autoantibody by the age of 2.4 years in the German BABYDIAB study (56). Seventeen of these 19 subjects (89%) developed multiple autoantibodies during prospective observation.Assuming a constant increase in autoantibody frequency from age 2 to age 6 years, the expected cumulative incidence of at least one antibody would be 14.8%and that of at least two antibodies 13.0% by the age of 6, which is well within the 95% CI of the observed frequency of 20.7% of multiple autoantibodies in the young DiMe siblings.

The cumulative incidence of at least one autoantibody by the age of 4 years was observed to be 10.6% in siblings and offspring of affected subjects in the DAISY study from Denver (57).The frequency was even as high as 37.9% in those carrying the DR3/4, DQB1*0302 combination.Published data are not available allowing an estimation of the frequency of at least one or at least two autoantibodies in subjects with increased genetic risk as defined in the second TRIGR pilot trial.Nevertheless a constant increase in the frequency of at least one autoantibody from age 4 to age 6 years, would result in an autoantibody prevalence of 15.9% by the age of 6 years in siblings and offspring irrespective of HLA-defined genetic risk.

The estimation that 2730 infants should be randomized for the trial (Table 4) is based on the following assumptions:

This figure represents a conservative estimate, since it is based on the lower 95% CI (7.6%) of the observed cumulative incidence of type 1 diabetes by the age of 10 years (15.4%) in young siblings with moderate (DQB1*0302/x) and low increased genetic risk (DQB1*02/y).The background for this conservative estimate is the observation from the second pilot trial that the prevalence of the high risk genotype (DQB1*02/0302) is about two times higher (absolute frequency 20%) among siblings of affected children than among offspring of affected parents (absolute risk 10%).If also the young siblings carrying the high risk genotype in the DiMe Study should be included in the group at risk, the lower 95% CI would be 10.6%, which should result in a recommendation of 1920 infants to be randomized for the trial.

The aim is to randomize 2800 infants with increased genetic risk as defined in section d.e. Genetic Screening.To achieve that number 6220 infants should be screened assuming a frequency of45% of the genotypes conferring increased risk.The observed prevalence of risk genotypes was 46% among the 474 infants screened for the second pilot trial.An enrollment scheme for the trial proper is presented in Appendix IV . This flow sheet indicates that the trial requires initially access to 8000 pregnancies.This will result in 6800 families giving their consent to participation in the trial before the birth of the child provided that the consent rate is85%.With an exclusion rate of 8.5% after birth based on the exclusion criteria approximately6220 infants will be available for genetic screening.The observed exclusion rate in the second pilot was 8.4% .

Table 4. Estimation of the sample size for the trial

d.c.Recruitment

Recruitment is carried out during a 2-year period in six major centers in USA, in 11 centers in Canada, in 10 European countries, in one center in the Dominican Republic, two centers in South Africa, and three centers in the Sydney region, N.S.W., Australia. A list of participating centers is presented in Appendix V. It is estimated that about one half of the subjects will be recruited in USA and Canada, and the other half in Europe and other participating countries. As to stopping rule practice, the recruitment will be continued past the 2 years period, if we have not obtained 2,800 high risk newborn infants to the series by that time. To facilitate recruitment and to minimize any possibility of unintentional exposure to CMP, every attempt will be made to identify eligible families before the child is born. Written consent will be obtained at this time; the child will participate after birth ifhe/she meets the inclusion but not the exclusion criteria. Families not identified until just prior to the onset of maternal labour will be approached at that time.

Mothers with type 1 diabetes will be identified during pregnancy via hospitals monitoring preg­nant type 1 diabetic women. Fathers with type 1 diabetes will be identified by (i) available history or data already in the medical record of pregnant women, and (ii) interviewing women at prenatal maternity clinic visits.Pregnant women already having one or more children affec­ted with diabetes will be identified through various diabetes clinics. Children with type 1 diabetes are usually followed by paediatricians or paediatric endocrinolo­gists, who will be aware of a pregnant mother. The identification of such mothers will be arranged through these physicians. Consent will be obtained during pregnancy.

Experience gained in the second pilot study of this project will be utilised to optimise the efficacy of the recruitment in the proposed study proper (use of diabetes societies, information in mass media etc.).

d.d.Subject allocation

Immediately after the birth of child, the research assistant or investigator will contact the national randomi­sation center to determine the formula allocation. Randomi­sa­ti­on will be done centrally (website). Large enough blocks of randomised code numbers will be provided to each national center. After the birth the newborns who meet the inclusion and do not meet exclusion criteria are randomi­sed as follows: randomisati­on in each strata will be within four blocks. Subjects will receive either the test formula (Nutramigen™, Mead Johnson Nutritionals, not containing antigenic CMP), or a CMP containing control formula which has an addition of Nutramigen, whene­ver breast milk is not available. Any subject requiring supple­mental feeding prior to randomi­sa­tion (e.g. infants born at night or on weekends) will be given banked breast milk or Nutra­migenTM. The randomisation code will be kept at the central data base and it will be opened when the last recruited child has reached the age of 6 years.

Newborn infants will have a heel prick performed as soon as possible after recruitment to the study; some centres may alternatively collect cord blood at birth. Blood will be collected on filter paper; the sample forwarded to the continental genotyping centre and screening performed for the presence of the genotypes listed below under d.e. indicating an increased genetic risk.Results will be sent to the central project office and the hospital in question within 5-7 days.Only subjects with these genotypes will be included inthe nutritional prevention trial; all other subjects (approxi­mately 45% of randomised subjects) will be withdrawn from the study at that time; their parents will be told that genetic screening suggests that their child does not fall into the increased genetic risk group for type 1 diabetes.

d.e. Genetic Screening

The role played by genetics in type 1 diabetes is illustrated by family studies and genetic marker studies. FDRs of patients with type 1 diabetes have a markedly increased risk of type 1 diabetes compared with the general population which, however, is dependent on the risk within the background population (58). For example, siblings of an affected patient may have a 6-8% average rate of type 1 diabetes development by adulthood. This rate may be as high as 12-19% if the sibling and patient have identical HLA haplotypes (59). If they are identical twins,approximately 20-30% of the siblings will develop diabetes (60,61), this rate increasing in the presence of a high risk genotype (62). Children of affected mothers have a 1.5-2.5 % prevalence of type 1 diabetes; the rate in children of affected fathers is 3-6 % (63,64,34,65) .

The most important genetic risk markers for type 1 diabetes are located within class II region of MHC. The protein products encoded by class II genes, which are expressed on antigen presenting cells, are involved in the presentation of processed antigen to the T-cell receptor - a central event in the initiation of an immune response. HLA-DQ heterodimer molecules encoded by HLA-DQA1 and -DQB1 genes are the markers most strongly associated with disease susceptibility (65). DQA1*0301-DQB1*0302 (DR4-DQ8 haplotype) and DQA1*0501-DQB1*0201 (DR3-DQ2 haplotype) are associated with the disease risk, DQA1*0102-DQB1*0602 (DR2-DQ6 haplotype), DQA1*0501-DQB1*0301 (DR5-DQ7 haplotype), DQA1*0301-DQB1*0301 (DR4-DQ7 haplotype) and DQA1*0103-DQB1*0603 (DR13-DQ6 haplotype) with a variable degree of protection. More than 90% of type 1 diabetic patients have at least one of the susceptibility haplotypes, but they are present also about in half of the general population. The highest risk is associated with the genotype containing two different susceptibility genotypes (DR4-DQ8/DR3-DQ2), whereas the risk associated with a double dose of individual haplotypes is not markedly higher than the single dose without protective markers. This synergism between risk haplotypes may be due to the creation of new heterodimer molecules encoded by A1 and B1 genes in trans position. Due to the strong linkage disequilibrium especially between alleles in DQA1 and DQB1 loci, the typing for DQA1 is in most cases not necessary, but is needed to confirm whether DQA1*0501 (DR3 haplotype) or DQA1*0201 (DR7 haplotype) is associated with the DQB1*02 allele (DQB1*0201 and DQB1*0202 associated with DR3 and DR7 haplotypes, respectively, do not differ within the region used to differentiate most DQB1 alleles from each others). In some haplotypes originating in the Mediterranean region or in Africa, DQB1*02 is associated with DQA1*0301 and these together form a diabetes associated heterodimer (66,67).

The HLA-DR4-DQ8/DR3-DQ2 genotype is associated with a risk of about 7% to develop type 1 diabetes before the age of 15 yrs in the Finnish general population, and HLA-DQ genotypes can be arranged into an order according the associated disease risk from high risk to strongly protective genotypes (68). HLA-DQA1*0301-DQB1*0302 is associated with higher relative risk than the DQA1*05-DQB1*02 haplotype, andthe effect of various protective haplotypes is also different when combined with these two alleles. Thus DQB1*0302/DQB1*0603 is still associated with considerable risk but DQA1*05-DQB1*02/DQB1*0603 is not increased among patients with type 1 diabetes (69). It is evident that in addition to HLA-DQ genes also other genes both within and outside HLA are important. HLA-DR4 subtypes among DR4-DQ8 haplotypes are conferring different risk and DRB1*0403 and *0406 alleles seem to be dominantly protective (70). There are several sites around the genome where new risk genes have been preliminary localised but only the effect of the insulin gene region on chromosome 11 has been firmly established (71,72).

d.e.a.Rationale for genetic pre-screening

The risk of type 1 diabetes in newborns with an affected FDR is increased approximately 10-20 fold in comparison with the general population. The recognised association of some HLA-DQ alleles with increased and others with decreased risk, and the ability to easily detect these alleles provides the oppportunity to study FDRs at even higher risk for type 1 diabetes. This can be accomplished by restricting the study to newborn infants positive for DQ genotypes known to be associated with increased risk for type 1 diabetes. This results in an increase in the expected event rate (i.e. the development of type 1 diabetes associated autoantibodies or the disease itself) in control subjects and consequently reduces the sample size and cost of the study and simplifies the logistics of long term follow-up.

The options for genetic screening vary between the extremes of a very high risk population requiring a small sample size (e.g.DQA1*0301-DQB1*0302/DQA1*0501-DQB1*02 heterozygotes), and a lower averagerisk population with larger sample size (e.g. newborn infants with an FDR but with no additional genetic selection applied). Based on anticipated rates of identification of eligible newborn infants and a feasible enrolment period, we propose to base the selection on the presence of the following genotypes:

These genotypes are present in 77.2% ofFinnish children with type 1 diabetes and in 23.1% of background population (69). Theanalysis of 289 diabetic members of multiplex case families confirmed our earlier observation on the higher HLA associated genetic load found in familial cases: 85.8% were positive for the genotypes listed above (30). These risk genotypes were on the other hand present in 46.1% of the 475 newborns with a FDR screenedduring the second pilot study of this project in Finland. If an approximate figure of 10% is used for the risk of a FDR to develop multiple autoantibodies before the age of 6 yrs, a risk of 17.1% can be calculated for those with risk genotypes compared to the risk of 2.8% of those without the risk genotype. This figure is very close to the percentage given in Table 4 for the positivity of at least two autoantibodies by the age of 6 years.

Preliminary data demonstrate that the planned genetic screening strategy would identify approximately 80 % of future cases with childhood type 1 diabetes in Puerto Rico (Santiago, Trucco, Frazer et al, unpublished data).

d.f.Study Formula Composition, Labelling and Distribution

Each study formula will be a nutritionally complete infant formula manufactured by one company (Mead-Johnson Nutritionals). The control formula will contain CMP (including bovine serum albumin, BSA and casein) plus 25 % protein hydrolysate (to blind the taste and smell of the control formula). The test formula is a casein hydro­lysate (NutramigenTM, Mead Johnson Nutritionals), documen­ted to be free of peptides with a molecular weight higher than 1000 D, and being non-diabetogenic in experimen­tal animals. Its non-diabetogenic properties have been confirmed in the BB rat and the NOD mouse models.

The formula will be packed in four different colours - two colours for test formula and two colours for control formula. Only Mead Johnson Nutritionals will know which colours correspond to test and control formulas. Subjects will be allocated to receive one colour formula box, which will be maintained throughout the study. Each centre will maintain a reserve supply of each "colour" formula and will be responsible for ensuring that the home supplies of each participant are maintained. All requests for more formula will be communicated to Mead Johnson Nutritionals through the central study office.The company will provide the coded formulas free of charge, but the shipping costs must be paid by the study grant (Appendix VI).

d.g.Implementation ofIntervention

All recruited mothers will be encouraged to breast feed; the newborn infants will, however, be randomi­sed as soon as possible after birth so that any elective formula supplementation or weaning by the mother will be done with the appropriate formula. The duration of the intervention will be until at least 6 months of age. If the mother chooses to exclusive­ly breastfeed up to the age of 6 months she will be advised, thereafter, to give the study formula for at least 2 months, i.e. until the age of 8 months. Similarly, if exclusive breastfee­ding lasts for 5 months, the infant would receive study formula until the age of 7 months. Infant feeding practices will be altered as little as possible by the trial. In particular, breast feeding practice(s) will be entirely at the discretion of participating mothers. However, all relevant data will be recorded. Compliance with the denial of CMPs will be assessed by the self reports of the families on deviations made to child's diet, diet recall or record, counting of formula usage (amount received substracted by amount returned), and by measuring CMP antibody levels from sera. The diet of the child will be assessed by a telephone interview when the child is 1 month old, and at the 3, 6 and 9 months visits,and data collected on dietary compliance will be continuously checked, cleaned, entered to the computer and analysed.

Dietary advice will be given at the delivery hospital, at the 1 month telephone call, and at the 3 and 6 month visits. The families will receive both written and oral instructions about infant feeding during the intervention period. Parents will be given a pamphlet which describes the sequence and amounts of food recom­men­ded at specific ages, according to local guidelines. It is important to avoid any conta­mination of the dietary intervention with sources of CMP (including milk products, beef, and veal) contained in foods ingested by the infant. Thus, parents will be provided with a list of all solid foods giving choices of brand names which can be given to the infant and which do not contain CMP; they will also be provided with a list of "forbidden" foods which do contain CMP. A food frequency record will be held until the age of 1 year. Post-trial nutrition will follow generally accepted practices. Designated personnel will be available during working hours to deal with urgent nutritional questions by telephone (e.g. feeding intolerance). Psychological expertise is used for the training of study personnel to minimize the risk of anxiety among the mothers.

Dietary advice leaflets, field book, dietary forms and questionnaires will be translated to relevant languages and adapted to national practises before the start of the grant period. The dietary surveillance and advice centre (Dr. Suvi Virtanen, University of Tampere, Finland)will take responsibility for the education and supervision concerning the assessment and maintenance of the dietary compliance and the dietary advice given by the centres involved. A continuous quality control system will be implemented in each center for the data collection and content, especially to ensure that the delivery of dietary advice and the measurement of dietary compliance function properly.

d.h.Study Assessments

d.h.a. Baseline Assessments

At the time of randomisation, medical and perinatal history of the infant and mother (inclu­ding birth weight and gestational age) and the results of the newborn physical examination will be recorded on the case report forms.

d.h.b. Assessments during the Follow‑up Period and after Intervention

At the end of the intervention period (6-8 months), families will be instructed regarding a healthy diet for their child and arrange­ments will be made for subsequent follow‑up visits. The subjects will visit the research centre at the age of3, 6, 9, 12, 18 and 24 months, and at 3, 4, 5, 6 and 10 years of age, or when clinical diabe­tes develops. The presence or absence of diabetes will be determined according to the criteria outlined below (see Outcome Assessment, Section d.h.c.). Central findings at each visit (e.g. weight, height) will be recorded on the case reportforms.

Blood for serology will be drawn after the application of an analgesic ointment on the venipuncture site at the above mentioned visits and stored centrally. In addition, the 6 month serum sample will be available for use in the assess­ment of dietary compliance. All serum samples will be aliquoted and stored at -70 C°. Local measure­ment of plasma glucose and glycated hemoglobin takes place at 12, 18, 24 months, and 3, 4, 5, 6 and 10 years. The specimens for plasma glucose are preferentially taken 1-2 hours postprandially.

A summary and timetable of the various events are given in Appendix VII.

d.h.c. Outcome Assessment

d.i.Methodology to be Applied

d.i.a. Methodology for Genetic Screening

The method used for HLA-DQ typing has been developed for screening relevant DQB1 and DQA1 alleles (73,74). EDTA treated cord blood is collected and a blood spot prepared on filter papers. Alternatively capillary blood is taken directly on the filter paper. A 2 mm diameter disc is punched into microtitre plate wells where the polymorphic gene region is amplified using biotinylated primers. The amplification product is subsequently transferred to streptavidin coated microtitre plates where it is bound by the solid phase streptavidin. After denaturation the hybridisation reaction is performed with a mixture of lanthanide (Europium, Samarium or Terbium) labelled sequence specific oligonucleotide probes and specific hybridisation signals detected by time-resolved fluorometry after repeated washes and the addition of enhancement solution. The filter papers are sent by air mail or special delivery from the U.S. and Canadian centers to the tissue typing laboratory of Professor Massimo Trucco, University of Pittsburgh, Pa., USA, and from the various European and other centres to the tissue typing laboratory of Dr. Jorma Ilonen, Universityof Turku, Finland. According to experiences from the second pilot study of the project (Finland, Sweden, Estonia and Hungary), the result of the genetic screening will be available within 1 week for the decision concerning final inclusion or exclusion. The results will be sent by e-mail to the central project office and the hospital in question.

d.i.b. Methodology for Humoral Markers of b-Cell Autoimmunity

We will use islet cell antibodies (ICA), insulin autoantibodies (IAA), antibodies to the 65 kD isoform of glutamic acid decarboxylase (GADA) and antibodies to the protein tyrosine phosphatase related IA-2 molecule (IA-2A) as humoral markers of b-cell autoimmunity. All autoantibody analyses will be performed in the laboratory of Professor Mikael Knip, University of Tampere, Finland.This laboratory has a long experience and acknowledged expertise in the analysis of diabetes-associated autoantibodies.The laboratory will pay particular attention to elaborate quality control of the assays used for the detection of autoantibodies.All samples will be recoded before the antibody assays and the analyses will be performed on a blinded basis.Samples from the same individual will, however, be recorded in such a way that they can be run in the same assay. These will be analyzed in two assay rounds, the first of which comprise samples from birth up to the age of 2 or 3 years, whereas the other assay includes samples from the age of 2 or 3 years up to the age of6 or 7 years.Possible assay drift over time will be monitored by analyzing blindly with each assay three standards (low, medium and high antibody levels) once a month.An assay drift exceeding 10% is considered unacceptable.The laboratory is presently in the process of collecting optimal standards with sufficient volume for a period of 15 years.

ICA are analyzed with a standard immunofluorescence assay performed on sections of frozen human pancreas from a blood group O donor (75). Fluorescein-conjugated anti-human IgG (Sigma, St Louis, MO, USA) is used to detect ICA. All initially ICA-positive samples are retested to confirm antibody positivity.End-point dilution titers are identified and the results are expressed in Juvenile Diabetes Foundation (JDF) units relative to an international reference standard (76). The detection limit is 2.5 JDF units.The sensitivity of this ICA assay was 100%, the specificity 98%, the validity 98% and the consistency 98% in the most recent international standardization round.

Insulin autoantibodies (IAA) are measured with a microassay method, modified from that described by Williams et al. (77). In this method endogenous insulin is not removed before the assay. Immune complexes are precipitated with Protein A Sepharose (Pharmacia Biotech, Uppsala, Sweden) after incubation for 72 h of the serum sample (5 µl/well) with mono 125I-(TyrA14)-labeled human insulin (Amersham, Little Chalfont, Bucks, UK) in the presence or absence of an excess of unlabeled insulin. The volume of the incubation reaction is doubled by adding the reaction buffer (TBT; 50 mM Tris, pH 8.0, 1 % (v/v) Tween 20). After thorough washing with the reaction buffer the samples are transferred from the deep well plates to microtitration plates, scintillation liquid is added and the bound activity measured with a liquid scintillation counter (1450 MicroBeta Trilux; EG&G Wallac, Turku, Finland). The specific binding is expressed as relative units (RU) based on a standard curve run on each plate using the MultiCalc™ software program (EG&G Wallac). The standard curve is constructed from nine serial dilutions of a serum from a patient with a high IAA titer and a serum of an IAA-negative subject. The negative human serum is considered as the lowest point of the standard curve. Due to the competition a proportion of the samples with no detectable IAA give a cpm below the lowest standard(<1 RU). All such samples are assigned a value of 1 RU in the statistical evaluation. The cut-off limit for IAA positivity is set at the 99th centile (1.56 RU) in 371 non-diabetic Finnish subjects. Samples with an initial IAA level exceeding the 95th centile are reanalyzed to verify the antibody status.The disease sensitivity of this assay was 35 % and the disease specificity 100 % based on 140 samples included in the Multiple Autoantibody Workshop (78).

An immunoprecipitation radioligand assay is used to detect GADA (79,80). The recombinant plasmid pGEM3 encoding the whole 65 kD form of the GAD protein (585 amino acids) is multiplied in the Escherichia coli JM 109 by standard techniques.The GAD65 protein is produced by in vitro transcription and translation of the purified plasmid using the TNT Coupled Reticulocyte Lysate System (Promega, Madison, WI, USA) in the presence of35S-methionine (Amersham). Unincor-porated 35S-methionine is removed by gel chromatography on a NAP-5 column (Pharmacia Biotech).Sera (2 µl) are incubated overnight at + 4 °C with approximately 20,000 cpm human GAD65 in a total volume of 50 µl TBST.To isolate the immune complexes, 10 µl Protein A-Sepharose® CL-4B (Pharmacia Biotech) is added on the following day. A scintillation counter is used to count the amount of precipitated immune complexes. The results are expressed in relative units (RU) based on a standard curve run on each plate using a commercial software program (MultiCalcÔ, EG&G Wallac). The cut-off limit for antibody positivity is set at the 99th percentile in 373 non-diabetic children and adolescents, i.e. 5.36 relative units (RU).All samples with an initial GADA level between the 97.5th and 99.5th percentiles are reanalyzed to verify the antibody status.This assay had a disease sensitivity of 69% and a specificity of 100% based on 140 samples included in the 1995 Multiple Autoantibody Workshop (78).

A radiobinding assay is also used to analyze the IA-2A (81). The recombinant plasmid pSP64poly(A) encoding the intracellular portion of the full lentgh IA-2 protein, including amino acids 605-979 is transformed in Escherichia coli JM 109 cells and then purified by standard techniques.The radioactive IA-2 protein is produced with the TNT Coupled Reticulocyte Lysate System (Promega) by in vitro transcription and translation of the purified plasmid in the presence of 35S-methionine. Sera are incubated overnight at +4 °C with 10,000 cpm of labeled IA-2 protein. Protein-A Sepharose® (Pharmacia) is used to isolate the immune complexes on the following day. After thorough washing the radioactivity of the samples is measured by a liquid scintillation counter (1450 Microbeta® Trilux,EG&G Wallac). The results are expressed in RU based on a standard curve constructed from the dilution of a pool of strongly positive samples and a pool of negative samples. The standard curve is run on each plate. A subject is considered IA-2A positive, if the serum antibody levels are equal to or exceed 0.43 RU, which represents the 99th percentile in 374 non-diabetic Finnish children and adolescents. Samples with an initial IA-2A level between the 97.5th and 99.5th percentiles are reanalyzed to verify the antibody status.The disease sensitivity of this assay was 62% and the specificity 97% based on 140 samples included in the 1995 Multiple Autoantibody Workshop (78).

d.i.c. Methodology for the Analysis of Antibodies to Cow's Milk Proteins

After washing, 100 ml alkaline-phosphatase-conjugated monospecific rabbit affinity purified F(ab’)2 antihuman IgG, IgA and IgM antisera (dilutions between1:600 to 1:1200) (Dako A/S Glostrup, Denmark) were added and the plates were incubated for 60 minutes at 37 C.After washing, 100 ml of p-nitrophenylphosphate substrate(2 mg/ml in diethanolamine buffer, pH 10.0, (both from Medix Biochemica, Helsinki, Finland) were added.The reaction was stopped after 30 minutes with 100 ml 1 M NaOH.The end product was measured at 405 nm in a semi-automatic photometer (Titertek Multiscan®, Elflab Inc., Helsinki, Finland).The mean value of the two absorbancies for the wells coated withblocking solution was subtracted from the mean value for the three absorbancies in the antigen-coated wells.This result was compared by point to point analysys by the computerized photometer with that for 2-fold serial dilutions of a standard serum with a very high titre of CM antibodies. The value for standard serum was given as 100%.If the absorbance of the sample was not on the linear part of the standard curve, the sample was diluted further.The levels of antibodies were expressed as percentages of the standard.

d.i.c.b. Antibodies to human and bovine insulin

Development of humoral immune response to insulin is detected by EIA. The microtiter plates (Labsystems) are coated with 1 mg of bovine or human insulin (Sigma) per well. Human serum albumin is used as a blocking agent. The serum samples are diluted 1:20 for the measurement of IgG-class antibodies to insulin. Alkaline phosphatase conjugated polyclonal rabbit anti-human IgG-antibodies (Jackson ImmunoResearch) are used as the second antibody. The reaction is developed by using p-nitrophenyl phosphate as a substrate. The specificity of insulin-binding antibodies is checked by inhibition experiments using bovine or human insulin at the concentrations of 1, 10, 100 and 1000 mg/ml in liquid-phase as an inhibitor. The assays are performed by Dr. Outi Vaarala, Department of Biochemistry, National Public Health Institute, Helsinki, Finland.

d.j.Contamination

It is extremely important to ensure that subjects who are randomised to receive the test formula are not exposed to CMP during the intervention period. The risk is especially high in the immediate postpartum period when a newborn infant may be given the wrong formula for supplementation in the hospital nursery. To prevent such contamination, infants who are in the study will be clearly identified to all nursery personnel. Any formula supple­mentation for these infants will only be with the allocated study formula.

d.k. Data Base

A specific database programme has been developed for the storage and quality control of the information collected within the project (described in Appendix VIII). The main goal has been to make an application, which is easy to use and that would store the information so that the data processing would be as versatile as possible at the analysis stage. The software has been developed to meet the needs of the project. The software will be continuously improved along with the progress of the project.

d.l.Statistical Methods

Two different analytic methods will be used to determine whether early avoidance of CMP decreases the cumulative incidence of diabetes-associated autoantibodies and/or clinical diabetes. First, survival curves will be calculated using a life table method and compared using the Mantel-Haenszel chi-square method. They will be used to analyse the time until diabetes occurs in the two groups. Secondly, the proportion of diabetic subjects in each group after 6 and 10 years will be compared using the Mantel-Haenszel chi-square or Fisher exact test.These same analytic methods will be used to determine whether the intervention prevented or delayed the appearance of significant titres of ICA antibodies or other relevant autoantibodies (which are associated with subsequent Type 1 diabetes). For this analysis, an event will be defined as the first time point when a significant titre is detected even if the titre subsequently falls.

Finally, the acceptability of the intervention will be determined on the basis of clinical measures of compliance and the dropout rate during the intervention period. This, as well as morbidity, will be described as proportions in both groups and will be compared using Pearson's chi square test and Student's t tests. The height and weight in both groups at 1, 2, 3, 4, 5, 6 and 10 years will be compared usingStudent's t test.

The application of statistical methods is supervised by Dr Paul Knekt, Ph.D., Department of Health and Disability, National Public Health Institute, Helsinki, Finland.

d.m. Interim Analysis

An interim analysis will be performed annually in the study proper by the Monitoring Committee, beginning 2 years after the last subject is recruited, in order to ensure that an unexpec­tedly large protective effect of the intervention will be detected early. The data will be analysed to determine if the P value for the difference in the rate of autoantibody development or manifest diabetes between the two groups is < 0.01. The follow-up will be discontinued, the code opened and the results published if any of the analyses reaches the above level of statistical significance.

d.n. Ethical Issues

Ethical approval will be obtained at each study centre. Written informed consent will be asked from the parents.The ethical issues are described in more detail elsewhere in this application.

d.o.Project planning and time table

2. year
- Recruitment and genetic screening are continue
- Clinical follow up of the children and collection of the follow-up samples are continued
- Dietary advice and assessment continue
- Monitoring of the implementation of the dietary intervention continues
- Studies on T-cell responses in fresh blood samples are continued
-Investi­ga­tors meeting in the middle of this period, review of items as listed above for the firstyear.
- Investi­ga­tors meeting at the end of this period for the mid-term review procedure, review of items as listed above for the first year

d.p. Role of co-workers in the project

The study group members represent a versatile and high-standard expertise in the fields of pediatrics, pediatric diabetology, infant nutrition, pediatric gastroenterology, neonatology, obstetrics, nutrition science, genetics, immunology and epidemiology. Several members have published extensively on the prediction of type 1 diabetes and autoimmune markers of the disease.

Several members are internationally recognized experts in their fields related to the research on type 1 diabetes, particularly on the etiology, pathogenesis and prevention of the disease. As examples may be mentioned in the following studies:

Here, please, study group members, mention the most important pieces of your work, which you consider relevant (a few sentences, and the references).

Success in the execution of a demanding research plan, like this, depends besides in general favourable conditions (the intelligence and acumen of the investigators, material resources etc.) also on the ability to collaborate. Particularly in projects like this, requiring many years for completion it is of great importance to know one´s closest collaborators also from this point of view. The undersigned P.I. has for a long time collaborated with the large majority of the present study group members, in the case ofFinnish investigators e.g. in the DiMe-Study and the two pilot studies in the present TRIGR-project, and our collaboration has been characterized by mutual respect and joy of discovering new things together. Therefore I consider that our research team also from this point of view has good chances to succeed in the demanding task. The undersigned has as well had collaboration with the foreign scientists over many years, and I trust the confidence between us is mutual.

d.q. Mechanistic studies outside the present core protocol

d.r. Previous application to the NIH

We submitted a grant application to the NIH in January 1995 on the same topic, but it was not funded. In the present application we have taken notice of the received criticism as much as possible.