Genomics|HuGENet|Reviews|HLA-DQ and Type 1 Diabetes

Contributors: J.S. Brach, R.Z. Ezzedine, M.J. Forlenza, I.M. Gathuru, J.L. Gibbon, M.A. Hagan, L.L. Lee, T.L. Loucks, E.K. Luedecking, C.E. McFarland, C. Pollice, J.R. Rager, S.A. Riddler, S.E. Riechman, N.J. Schmidt, D.L. Stricklin, T.M. Truitt, E.N. Vergis, R.P. Wildman, R.G. Willenkin, C.S. Winters.

Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, Pittsburgh, PA 15261.

Acknowledgments: This research was supported by NIH Grants RO1-DK42316, RO1-DK44590 and RO1-DK49588.

Address Correspondence to:

Janice S. Dorman, Ph.D.
A 548 Crabtree Hall Department of Epidemiology
Graduate School of Public Health
University of Pittsburgh
Pittsburgh, PA 15261
Phone: 412-383-1286
Fax: 412-383-1022
Email: Jansdorman@aol.com

Type 1 diabetes is one of the most frequent chronic diseases in children. Although it is an autoimmune disorder, the etiology of type 1 diabetes remains unclear. However, there is considerable evidence that both genetic and environmental factors are major determinants. In terms of genetic markers, type 1 diabetes is primarily determined by genes in the HLA region of chromosome 6. The HLA-DQ locus appears to be the best single marker of susceptibility, particularly among Caucasians. However, recent genome screens have identified at least 15 additional loci that may also contribute to disease risk. Although type 1 diabetes is likely a polygenic disorder, epidemiologic patterns of type 1 diabetes, including seasonal and temporal changes in incidence, suggest that environmental factors are involved. With the exception of a possible role for viruses and infant nutrition, the environmental factors that initiate or precipitate the onset of type 1 diabetes have not been established. This HuGE Review will focus on HLA-DQ associations, as well as their interactions with other genetic and environmental risk factors, in the development of type 1 diabetes. The Review was prepared by the students enrolled in Epidemiology 2600, “Introduction to Molecular Epidemiology” course, Department of Epidemiology, Graduate School of Public Health, University of Pittsburgh, under the direction of Dr. Janice S. Dorman and Dr. Clareann H. Bunker.

Key words: HLA-DQA1, HLA-DQB1, type 1 diabetes, genetic markers, risk factors, epidemiology

The Human Leukocyte Antigen (HLA) complex is located on the short arm of chromosome 6 at p21.3 (1-4). It encompasses approximately 3500 kb of DNA, and contains at least 150 genes. It is the primary region of susceptibility for type 1 diabetes, as well as other autoimmune disorders. Recent genome screens have identified the class II sub-region (i.e., HLA-DR, DQ, DP loci) as IDDM1.

The DQ locus, which is the focus of this review, consists of two tightly linked genes (i.e., DQA1 and DQB1) that encode alpha and beta glycoproteins, respectively. These molecules combine non-covalently to form functional a-b heterodimers. DQA1 and DQB1 genes are highly polymorphic. Allelic variation is observed primarily in the second exon, which corresponds to the peptide-binding cleft. HLA-DQ and other class II molecules present extracellular antigens to helper T-cells and stimulate immune response. They have a restricted tissue distribution and are located mainly on macrophages, B cells and activated T cells.

Transcription of DQA1 and DQB1 is complex and involves cis- and trans-acting factors. Critical upstream regulatory sequences have been reported for DQA1 and DQB1. Variation in promoter sequences affects gene expression, and may also be involved in the pathogenesis of autoimmune disorders. In addition, post transcription activities appear to influence disease risk. For example, functional DQab heterodimers can be formed from the non-covalent association of products of DQA1 and DQB1 genes in cis (5). Alternatively, the combined a and b glycoproteins can represent molecules encoded by genes in trans. Hybrid DQ molecules with DR or DP glycoproteins have also been observed.

There are eight major allelic variants of the HLA-DQA1 gene (Table 1). The data presented are based on the WHO DiaMond Molecular Epidemiology Project (6). The populations included represent those with similar race-specific incidence rates. All recruited cases were diagnosed between 1979 - 1990, and identified from type 1 incidence registries established according to the WHO DiaMond Project protocol (7). Registration criteria included: 1) diagnosed by a physician with type 1 diabetes, 2) placed on daily insulin injection before the 15th birthday, and 3) residency in the area of registration at the time of the first insulin administration. The selected cases were age 18 years, on average, and approximately half were males.

Unrelated non-diabetic controls were identified for each participating center using a standardized computer program that performed a simple random sample stratified by age and sex. This method permitted the identification of representative controls who were at-risk during the time the registry was established. They were recruited from consecutive hospital admissions for injuries / surgeries or the local population. Because of ethical concerns, all controls were required to be age 15 or older at the time of evaluation. As a group, they were somewhat older than the cases (age 22 years, on average). Approximately half were males. The target sample sizes for the case and control groups were 100.

Published frequencies from the 12^th International Histocompatibility Workshop and Conference have also been included in Table 1 (8). These represent pooled estimates from approximately 20 Caucasian populations worldwide. Cases were defined according to National Diabetes Data Group criteria for type 1 diabetes, and were not restricted by age at onset. Proportions diagnosed before age 15 years, between ages 15 and 30 years, and over age 30 years were 66%, 23% and 11%, respectively. Approximately half were males. Controls were drawn from the same ethnic and geographic populations as the cases. However, eligibility and selection criteria were not defined, nor were they standardized across populations.

In addition to the information provided in Table 1, there is a plethora of case-control studies regarding HLA-DQ associations and type 1 diabetes for Caucasian and other ethnic groups, many of which are not based on epidemiologic methods. Cases were generally selected from convenience samples (i.e., hospital clinics or medical practices). This can lead to an overrepresentation of multiplex families, as well as the inclusion of type 2 diabetics who use insulin in their treatment regime. Control frequencies were often determined from blood donors or other groups that may not truly reflect the population at-risk. These investigations also varied in terms of the approaches utilized for HLA molecular typing. Because of these inconsistencies, the results have not been included in Table 1. However, the reader is referred to the references 9 - 18 for additional information regarding HLA-DQ and type 1 diabetes in non-Caucasian populations (9-18).

As illustrated in Table 1, the frequency of the DQA1 alleles varied across ethnic groups. Significant differences were observed between type 1 diabetic and non-diabetic individuals from the same population. DQA1 alleles coding for arginine at position 52 (Arg-52) have been associated with susceptibility to type 1 diabetes (19-21). Although there are four DQA1*Arg-52 alleles, only DQA1*0301 has been shown to have a significant independent effect on type 1 diabetes risk. In some populations, DQA1*0501 is also associated with the disease. However, in other areas, this allele is neutral. Other than DQA1*0102, no DQA1 variant is an independent protective marker for type 1 diabetes.

Nineteen major DQB1 alleles have been identified. Their frequency also varied by ethnicity, as shown in Table 2 for the WHO DiaMond Molecular Epidemiology Project (6) and the 12^th International Histocompatibility Workshop and Conference (8). Significant differences between type 1 diabetics and healthy unrelated controls were observed. DNA sequences coding for an amino acid other than aspartic acid in position 57 (non-Asp-57) have been associated with type 1 diabetes in all ethnic groups (19-22) except the Japanese (22,23). In particular, DQB1*0302 and DQB1*0201, which are in linkage disequilibrium with DR4 and DR3, respectively, have been consistent susceptibility markers for type 1 diabetes. Other non-Asp-57 alleles (i.e., DQB1*0501, *0502, *0604, *0605) were generally neutral. DQB1*0602, which codes for aspartic acid in position 57, was significantly less prevalent among cases compared to controls.

Some researchers have suggested that DQB1 polymorphisms may be more important than those in the DQA1 gene for determining peptide binding specificity (24,25). However, variation in both genes contribute to susceptibility to type 1 diabetes. In particular, the DQA1*0501-DQB1*0201 and DQA1*0301-DQB1*0302 haplotypes confer the highest type 1 diabetes risk. In combination, their effect is even stronger than that observed for individuals homozygous for DQA1*0501-DQB1*0201 or DQA1*0301-DQB1*0302, suggesting that heterodimers formed from gene products in trans (i.e., DQA1*0501 and DQB1*0302) may be particularly diabetogenic (5). Other high risk DQ haploypes for type 1 diabetes include DQA1*0301-DQB1*0201 among African Americans, DQA1*0301-DQB1*0303 in the Japanese and DQA1*0301-DQB1*0401 in the Chinese (26). The DQA1*0102-DQB1*0602 haplotype is protective, and associated with a reduced risk for type 1 diabetes in most populations.

In the U.S., the prevalence of type 1 diabetes approximates 2 per 1000 for children less than age 20 years (27). This rate is higher than those reported for other childhood chronic disorders, such as cystic fibrosis, juvenile arthritis, etc. The onset of type 1 diabetes can occur at any age, but it is usually diagnosed during childhood and adolescence, with a peak incidence around the time of puberty. This pattern has been reported for most populations throughout the world.

Type 1 diabetes incidence rates for males and females are similar, although a female preponderance has been noted in low-risk populations, such as the Japanese (28). An excess risk for males has been observed in some areas where the overall incidence is high, such as in Finland. There is also a notable seasonal variation in the incidence of type 1 diabetes in most countries. Lower rates have been reported for the late spring and summer, but rates are higher in the winter for populations in both the Northern and Southern hemispheres (29).

Temporal trends in incidence have recently been observed. A significant increase in type 1 diabetes incidence has been reported by many population-based registries in Northern and Central Europe (30,31), as well as those in Asian and Western Pacific populations (32,33). In addition, epidemics of type 1 diabetes, such as the one that occurred in the Virgin Islands during the mid 1980s (34), have been observed.

Type 1 diabetes also exhibits dramatic geographic and ethnic variation. The highest incidence rates in the world (> 35/100,000 per year) have been reported for Finland and Sardinia, Italy (35). The lowest incidence rates are observed in the Asian countries (< 3/100,000 per year) including Japan, China and Korea. The Native American, Cuban, Chilean and Mexican populations also have extremely low rates of type 1 diabetes. In most other Caucasian populations in Europe and the Americas, incidence rates are moderate (~ 10-20/100,000 per year).

Ethnic variations in risk for populations residing in the same geographic area have been observed for type 1 diabetes (35). African Americans and Hispanics generally have lower incidence rates than Caucasians living in the same community. Ethnic variations were also apparent in China, an extremely low risk country represented by over 50 ethnic groups. In 1998, Yang et al. reported annual incidence rates of 0.3/100,000 per year versus 1.8/100,000 per year, respectively for the Zhuang and Mongols (36). Similar findings were observed for Finland, an extremely high-risk country, where the rates ranged from 4 to 245/100,000 per year (37). Unlike China, the Finnish population is genetically homogenous. Reasons for these geographic and ethnic differences in type 1 diabetes incidence are currently being investigated, and appear to reflect differences in both genetic and environmental risk factors (6).

Viruses have been implicated in the etiology of type 1 diabetes for the past several decades (38). They are thought to act as initiators, accelerators or precipitators of the disease, and may function by direct or indirect mechanisms. Viruses may attack and destroy the beta cells of the pancreas and directly cause diabetes, with or without autoimmunity (39). Alternatively, viruses may initiate or potentiate a autoimmune response against beta cells through molecular mimicry or ‘bystander’ autoimmune activation, the latter of which results from the induction of inflammatory cytokines after infection.

Many epidemiologic investigations have supported the involvement of Coxsackie virus B (CVB) in the etiology of type 1 diabetes (40). The most recent studies were based on molecular analyses, and revealed positive associations between the presence of enteroviral mRNA and the development of beta cell autoimmunity (41), and type 1 diabetes (42,43). Sequence homology between a highly conserved non-structural CVB4 protein (P2C) and glutamic acid decarboxylase (GAD), a potential diabetes autoantigen, has been reported (44). Several investigations indicated that antibodies to P2C and GAD cross-reacted (43,44). However, P2C antibodies were also observed among healthy GAD negative controls (45), suggesting that the response was not specific for type 1 diabetes.

Other viruses have also been associated with type 1 diabetes. Finnish investigators observed an increase in the incidence of type 1 diabetes two to four years after a mumps epidemic (46). Although the type and variant of the virus is clearly important, the age at exposure may also influence disease risk. Recent studies have shown that exposure to enteroviruses in utero increases the risk of developing the disease (47,48). Moreover, 10-20% of children with congenital rubella, particularly those who carry high risk HLA alleles, develop autoimmune type 1 diabetes (49). Thus, early viral exposure appears to be particularly diabetogenic.

Ecologic analyses have shown positive correlations between type 1 diabetes incidence and average milk consumption / breast feeding rates across populations (50). In addition, many case-control studies demonstrated weak positive associations between exposure to milk at an early age (< 3 months) and type 1 diabetes (OR from meta analysis = 1.4, 95% CI = 1.2 – 1.6) (51). Although breast feeding itself may be protective, it was hypothesized that the observed effect may indirectly reflect exposure to dietary proteins when the infant’s gut is not completely developed, and still permeable to antigenic peptides (51).

In 1992, Karjalainen et al. demonstrated that virtually all Finnish children with newly diagnosed diabetes had elevated levels of IgG antibodies to the whey protein, bovine serum albumin (BSA) (53). Most of the antibodies were specific to the 17 amino acid peptide ABBOS. The ABBOS peptide and p69, an islet cell surface protein, are similar in sequence. It was, therefore, suggested that early exposure to cow’s milk triggers an immune response that may lead to beta cell autoimmunity because of molecular mimicry. However, the negative T cell proliferation studies in response to cow’s milk antigen led one to question the molecular mimicry hypothesis (54). In addition, conflicting evidence came from recent reports of similar cellular responses to b casein among type 1 diabetic and non-diabetic subjects (55). Concern has also been raised by a short-term natural history study that showed no association between infant feeding patterns and the development of beta cell autoimmunity (56). Thus, the role of infant diet and type 1 diabetes is far from clear.

One of the major issues in molecular epidemiology relates to the definition of genetic susceptibility. For type 1 diabetes, relevant questions include: What loci, alleles and / or haplotypes should be considered as high risk? How should they be determined? Are there protective and/or neutral alleles or haplotypes that should also be evaluated? These questions are not at all trival; and for type 1 diabetes, they have been debated among epidemiologists, geneticists, immunologists and clinicians for many years. Despite numerous discussions, there is no uniform agreement on the ‘best’ definition of genetic susceptibility for type 1 diabetes. For purposes of this HuGE Review, we will outline the rationale for the strategy proposed for the WHO DiaMond Molecular Epidemiology Project (6,57). This approach considered biological significance, cost-effectiveness, and the statistical properties of the data.

Although the class II region has been identified as IDDM1, the strong linkage disequilibrium between DRB1 and DQB1 has made it difficult to assess the contribution of HLA-DQ independent of DR. This issue has been addressed by examining case-control differences in linkage disequilibrium for haplotypes containing high risk DQ alleles (i.e., DQA1*Arg-52, DQB1*non-Asp-57), but low risk DR alleles (i.e., not DR3 or DR4). Type 1 diabetics had a significantly greater frequency of high risk DQ alleles in low risk DR haplotypes than controls, suggesting that DQ better defined susceptibility haplotypes than DR (58). Thus, the HLA-DQ locus has been considered to be the strongest single genetic marker for type 1 diabetes, particularly among Caucasians (59) .

As indicated, polymorphisms in the DQA1 and DQB1 genes appear to be of biological importance, and are likely involved in the etiology of the disease. However, analyses of allelic associations within ethnic groups revealed that not all DQA1*Arg-52 and DQB1*non-Asp-57 alleles increased type 1 diabetes risk. Moreover, in the Asian populations, DQB1*Asp-57 alleles were significantly more common among type 1 diabetes cases compared to controls. Thus, defining type 1 diabetes susceptibility alleles by DQA1*Arg-52 and DQB1*non-Asp-57 was not accurate.

For the WHO DiaMond Molecular Epidemiology Project, genetic susceptibility was, therefore, defined for each population, and was based on the DQA1 and DQB1 alleles / haplotypes that were statistically significantly increased among cases compared to controls (Tables 1 and 2). All other alleles were considered neutral / protective for that population. This method accounted, in part, for contributions of specific DRB1 alleles, which were not directly evaluated, but known to be in linkage disequilibrium with high risk DQA1-DQB1 haplotypes. It also permitted a more accurate classification of disease susceptibility genes than approaches previously employed. It was, therefore, established DQA1 and DQB1 alleles would be the minimum analysis required for the WHO DiaMond Molecular Epidemiology Project. However, DRB1 typing was also recommended, particularly in non-Caucasian populations.

Type 1 diabetes associations with DQA1-DQB1 genotypes are presented in Table 3 for the WHO DiaMond Molecular Epidemiology Project (6). As illustrated, comparisons of odds ratios revealed statistically significant dose-response relationships (i.e., larger ORs for homozygote susceptibles (SS) than for heterozygotes (SP) relative to homozygote non-susceptibiles (PP)). Although these data present useful information regarding the strength of disease associations, cumulative incidence rates (i.e., absolute risks estimates) for specific genotypes can be more meaningful from a clinical and public health perspective. They represent the actual risk of developing the disease during a specific time period, and are determined directly from prospective studies. However, they also can be estimated from population-based case-control studies conducted in areas where the overall incidence of type 1 diabetes is known (60).

This approach was employed for the WHO DiaMond Molecular Epidemiology Project (Table 3) (6). Caucasians and African Americans with the SS genotype had, approximately, a 3% chance of developing type 1 diabetes by age 30 years. Interestingly, this risk was similar to the rate observed for siblings of affected individuals, who are typically considered to be at high risk for developing the disease. Thus, using molecular DQA1 and DQB1 typing, one can identify a sub-group of individuals in the general Caucasian population who have a similar risk for type 1 diabetes as first degree relatives. These markers were much less predictive in the Asian groups, where the disease is rare.

Population attributable fractions (PAF) for type 1 diabetes also provide important information regarding potential public health implications for disease prevention strategies, as they reflect the proportion of the total population incidence that can be attributed to genetic susceptibility. Table 3 includes PAFs for Caucasians, African Americans and Asians from the WHO DiaMond Molecular Epidemiology Project. PAFs were lower for the Asian compared to Caucasian or African American groups. This indicates that the DQ genotypes, as defined for the study, are better genetic markers in areas where the overall disease incidence is higher. Therefore, disease interventions in genetically susceptible individuals would likely have a greater impact among Caucasians or African Americans compared to Asians.

Although recent studies have concluded that DQA1-DQB1 haplotypes are the primary markers of susceptibility for type 1 diabetes, their effect can be modified by DRB1. Studies investigating this issue have examined differences in DRB1*04 alleles among DQA1*0301-DQB1*0302 positive cases and controls. In Caucasians, the DRB1*0401-DQA1*0301-DQB1*0302 haplotype has been shown to be increased in frequency among type 1 diabetic cases compared to controls (61). However, in combination with DQA1*0301-DQB1*0301, DRB1*0401 was negatively associated with the disease. Thus, it is unlikely that DRB1*0401 confers an independent risk for type 1 diabetes.

Studies of the DRB1*0404-DQA1*0301-DQB1*0302 haplotype have yielded conflicting results. Positive (61,62) and negative (63-65) associations have been observed. These discrepancies may be related to ethnic differences in linkage disequilibrium with B locus alleles, such as B39, which was significantly increased in DRB1*0404-DQB1*0302 cases compared to controls from Estonia, Latvia and Russia (64). DRB1*0405-DQA1*0301-DQB1*0302 appears to be significantly more common among type 1 diabetic cases compared to controls in groups such as Mexican Americans (18). Variations in the peptide-binding motifs, which are reflected by the DRB1*04 polymorphisms, may be the primary determinants of risk differences associated with the DRB1*04-DQA1*0301-DQB1*0302 haplotype (66).

In Caucasians, it has been demonstrated that the insulin gene region (INS), located on chromosome 11p15.5, contains the second major susceptibility locus for type 1 diabetes (i.e., IDDM2) (67). Positive associations have been observed with a non-transcribed minisatellite region (VNTR) in the 5’ flanking region. There are two common alleles; the shorter class I allele predisposes to type 1 diabetes, while the longer class III allele appears protective. The biological plausibility of these associations may relate to the expression of insulin mRNA in the thymus (68). Class III alleles generate higher levels of insulin mRNA than class 1 alleles. These differences may contribute to a better immune tolerance for class III positive individuals by increasing the likelihood of negative selection for auto-reactive T cell clones.

The effect of INS appears to vary by ethnicity. Undlien et al. found that class I INS alleles were significantly associated with type 1 diabetes in Caucasians, borderline significant in Tanzanian Blacks, and non-significant in Japanese (69). In contrast, Kawaguchi et al. found a significant positive association between INS class I alleles and type 1 diabetes in the Japanese (70). Methodological differences, such as heterogenous case and control groups, and variations in allele frequencies in the general populations, may be responsible for the inconsistencies in the literature.

Interaction between the INS and HLA DR-DQ has also been explored. Several groups observed as association between the INS class I alleles and type 1 diabetes, but only in the presence of low / moderate DQ genotypes (71,72). However, other investigators observed no difference in risk for INS class I alleles in analysis stratified by HLA-DQ (73,74). There is also a report of an association with the INS gene in the presence of DR4 (75). These contradictory findings suggest that further investigation of the INS gene and type 1 diabetes is warranted.

Most investigations of cow's milk exposure as a potential environmental trigger for type 1 diabetes were based on controls from the general population, many of whom were not genetically susceptible. Given the importance of allele-specific differences in immune response, failure to control for high risk HLA genotypes may obscure the effect of environmental risk factors. One of the first studies to address this issue was conducted in Denver, where it was observed that exposure to cow’s milk was associated with type 1 diabetes among individuals with high, not low risk DQB1 genotypes (76). Individuals who were homozygous for DQB1*non-Asp-57 alleles and exposed to cow’s milk at an early age had a markedly increased risk (OR = 11.3, p < 0.05) compared to non-susceptible, unexposed individuals. This was greater than the estimated relative risk reported for non-susceptible but exposed persons (OR=3.7 p > 0.05). Subsequent reports revealed similar findings (77,78). However, it remains to be determined whether the effect of breast feeding is independent of, or modified by, high risk HLA genotypes.

Prior to the development of molecular techniques, serologic typing for HLA-DR and DQ was the standard for the field. This method required living B cells obtained from peripheral blood samples that were subsequently tested against known antisera. However, serological evaluations were limited by unsuccessful or inaccurate results in approximately 11% of samples due to technical problems, lack of specific typing reagents and crossreactivity between alleles (79). These issues were resolved by the implementation of molecular typing methods. Compared with serological approaches, molecular tests revealed discrepancies, which were, in part, allele specific (80,81). However, direct sequencing confirmed the results of the molecular studies. Reproducibility of molecular DQB1 typing methods has also been demonstrated (82).

Several techniques have been developed for HLA-DQ molecular typing based on genomic DNA, which can be easily isolated from a variety of sources (i.e., lymphocytes, dried blood spots, buccal brushes, etc.). These include restriction fragment length polymorphism (RFLP) analysis or polymerase chain reaction (PCR)-based methods, the latter of which is most often utilized at the present time (81). Resolution with PCR methods depends on the amplification region (defined by the primers), as well as the number and specificity of the oligonucleotide probes used for hybridization. The WHO DiaMond Molecular Epidemiology Project employed sequence specific oligonucleotide (SSO) probes for DQA1 and DQB1 molecular typing (6), as utilized for the National Marrow Donor Program (81),

At present, there is no cure for type 1 diabetes, and lifelong insulin therapy is the only available treatment. However, several large clinical trials have begun to evaluate a variety of primary and secondary interventions in family members. Eligible persons are identified from families of affected probands using one of two general strategies. The first involves screening for high risk HLA-DQ alleles. The Trial to Reduce Type 1 Diabetes in the Genetically At-Risk (TRIGR) screens newborns for DQB1*0302, DQB1*0201, DQB1*0602/3 and DQB1*0301 (83). Although this approach does not permit exact genotyping, it can be used to identify individuals at high risk (i.e., positive for DQB1*0302 and / or DQB1*0201) and low risk (i.e., positive for DQB1*0602/3 and / or DQB1*0301) for purposes of inclusion and exclusion, respectively. Those at high risk are eligible for randomization. In TRIGR, the intervention group receives a casein hydrolysate formula, and / or breast milk for the first six months of life. In contrast, the control group receives a typical cow’s milk formula.

The second screening approach in families is based on an initial evaluation of beta cell autoantibodies (i.e., ICA, GAD, IA2, IAA). Although these antibodies are rarely observed among first degree relatives (~2-5%), and not all antibody positive individuals develop the disease, those who are positive for multiple autoantibodies appear to be at very high risk (84,85). Some studies have estimated the positive predictive value associated with two and three autoantibodies, respectively, as 65% and >90%. Thus, current investigations are adopting a two-stage screening strategy, with an initial test for one or two autoantibodies (86,88). Those who test positive are subsequently evaluated for additional immunologic, metabolic or genetic tests. Because the development of beta cell autoantibodies is sequential, family members must be screened at regular follow-up intervals to detect the appearance of multiple autoantibodies.

Although most intervention trials are based on first degree relatives, about 90% of those who develop the disease have a negative family history. Thus, for interventions to have a public health impact, they must be based on the general population. Unfortunately, the genetic screening strategies described above are not as effective in the general population as they are in families of affected probands (89). For example, in Finland, sensitivity estimates high risk alleles for siblings and the general population were 38% and 24%, respectively. Corresponding figures for specificity were 86% and 39%, respectively. However, one clinical trial based on screening in the general population has recently been initiated.

The Diabetes Prediction and Prevention Project (DIPP) is designed to determine whether it is possible to delay the clinical manifestations of type 1 diabetes with nasal insulin by at least three years in high risk Finnish newborns (90). This is the only intervention targeted toward the general population, and the screening strategy empolyed by DIPP is based on that developed for TRIGR. Those at high genetic risk are randomized. The treatment group receives daily intranasal insulin at a dose of 1IU / kg. The control group is observed, and receives no intervention. Recruitment is expected to be complete by 2002, with final results available in 2005.

The Diabetes Autoimmunity Study in the Young (DAISY) is a natural history study, which is also based on newborn screening for high risk HLA class II alleles (91). This investigation involves families of varied ethnic backgrounds (Hispanic, Non-Hispanic white, Black and Asian). The susceptibility alleles required for inclusion were defined as DRB1*03, DRB1*04 and DQB1*0302; and DRB1*15/16 (DR2) was considered as an exclusion criteria. Since the study is in its early phases, the sensitivity and specificity of the screening have yet to be determined.

In summary, HLA-DQ screening for type 1 diabetes is now being conducted in high risk families and the general population for intervention trials and natural history studies. Thus, there is a critical need to reconsider the risks, benefits, ethics, legal and social issues regarding genetic and / or autoantibody testing for type 1 diabetes. In addition, population-based risk factor specific incidence rates are urgently needed for all ethnic groups. Translating research findings for prediction and prevention outside a research environment also requires genetic counseling and genetic education programs for type 1 diabetes family members, as well as health care professionals. During the next millennium, these issues should be among the top priorities for funding for type 1 diabetes.