High Familial Risk and Genetic Susceptibility in Early Onset Childhood Diabetes

The Bart’s-Oxford prospective family study has recruited parents and siblings of patients with type 1 diabetes diagnosed before age 15 years from within the Oxford Regional Health Authority area in England since 1985 (12,13). The overall levels of case ascertainment estimated from independent survey of general practitioners in the region are >95%, and 89% of eligible families have participated in the prospective study. This report is based on the study population as of 1 March 2000, by which date 1,299 families had been recruited. Details of diabetes in all first-degree relatives were collected by interview. Venous blood samples were taken from all consenting family members and analyzed for islet cell antibodies (ICAs) and antibodies to GAD and protein tyrosine phosphatase IA-2. Insulin autoantibodies (IAAs) were also measured in samples from nondiabetic family members (14). Families were followed by annual visit or telephone call, and the diabetes status of every family member was ascertained on each occasion. The study cohort comprised the families of 298 children diagnosed before age 5 years, 442 diagnosed between the ages of 5 and 10 years, and 559 diagnosed between ages 10 and 15 years and included 3,878 first-degree relatives (1,430 siblings and 2,448 parents). A total of 5 mothers had a history of gestational diabetes, 21 parents were classified as having type 2 diabetes, and 3 parents had insufficient data available for classification. These 29 individuals were therefore excluded, and analysis was based on the remaining 2,419 parents. Subject characteristics are given in Table 1. The median duration of follow-up for unaffected siblings and parents was 6.8 and 5.5 years, respectively.

Serum samples collected at study entry were available from 1,169 of the 1,430 nondiabetic siblings included in the analysis (median age 11 years) and 2,130 of the 2,328 nondiabetic parents (1,165 mothers and 965 fathers, median age 38 years). Of these, 19 siblings and 12 parents developed type 1 diabetes during follow-up.

DNA samples for HLA class II typing were available from 573 of the diabetic probands in this cohort. This included 157 diagnosed before 5 years of age (86 male and 71 female), 201 diagnosed between 5 and 9 years of age (102 male and 99 female), and 215 diagnosed between 10 and 14 years of age (115 male and 100 female).

ICAs and antibodies to GAD and the protein tyrosine phosphatase IA-2 were measured in all samples. As IAAs are generally the first antibody marker to appear and may be found in isolation in young children, these were tested in samples from all children aged <10 years (15). Any samples in which levels of one or more of the other markers were greater than or equal to the 97.5th percentile of 2,860 schoolchildren control subjects were also tested for IAA.

ICAs were measured in undiluted sera by indirect immunofluorescence as previously described (16). End point titers of test samples were converted to Juvenile Diabetes Foundation (JDF) units by comparison with a standard curve of log₂ JDF units versus log₂ of end point titer of the standard sera. The threshold of detection was 4 JDF units. The interassay coefficient of variation was 10% at 13 JDF units and 4.3% at 80 JDF units. The assay achieved 78.4% sensitivity with 98% specificity in the First Immunology of Diabetes Society (IDS) Combined Antibody Workshop (17).

Antibodies to in vitro–translated [³⁵S]GAD₆₅ and [³⁵S]PTP-IA-2_ic were measured by immunoassay as previously described (18). The interassay coefficient of variation of the GAD antibody assay was 17% for samples that had 1.5 units and 9% for samples with 17 units of antibody. The interassay coefficient of variation of the IA-2 antibody assay was 15% for samples with 1 unit and 21% for samples with 9 units of antibody. The GAD antibody assay achieved 91% sensitivity with 99% specificity and the IA-2 antibody assay 74.4% sensitivity with 99% specificity in the First IDS Combined Antibody Workshop (17).

Antibodies to ¹²⁵I-labeled insulin were measured as previously described, using a format similar to that used to measure antibodies to GAD and IA-2 (19). The two-stage assay involved initial screening for insulin binding followed by a competition assay in samples with insulin binding above the 95th percentile of the schoolchildren. Immune complexes were isolated with protein A Sepharose (Pharmacia Biotech, Uppsala, Sweden) and expressed in arbitrary units derived from a standard curve. The assay achieved 58% sensitivity with 99% specificity on the samples included in the First IDS Combined Antibody Workshop (17).

HLA genotyping was carried out on DNA from blood or mouth swab samples. Details of DNA extraction methods and HLA class II analysis have been previously published (20). Briefly, mouth swab extractions were carried out using a guanadinium chloride/phenol chloroform method. DNA was extracted from blood using a salting-out method (21). DNA quantitation was achieved using OD₂₆₀ readings for blood DNA and picoGreen analysis for mouth swab DNA. Low yield DNA samples from mouth swabs underwent whole genome amplification by primer extension preamplification. HLA analysis was carried out by polymerase chain reaction (PCR) using sequence-specific primers (22). PCR products were separated on 2% agarose containing 0.002% ethidium bromide and visualized on a dual-intensity gel documentation system (Ultra Violet Products, Cambridge, U.K.).

Individuals were defined as having type 1 diabetes if they fulfilled one or more of the following criteria: permanent insulin treatment within 6 months of diagnosis, presence of one (or more) islet autoantibodies at study entry, or diagnosis before age 30 years with permanent insulin treatment. Patients were classified as having type 2 diabetes if they were antibody negative and diet or tablet treated or if they were insulin treated and antibody negative in samples collected within 6 months of diagnosis.

Life table analysis was used to estimate cumulative risk of development of diabetes in siblings and parents (23). The end of follow-up for each family member was defined as the age at diagnosis of diabetes or the age at last contact. Survival experience in sibings and parents subdivided according to age at diagnosis of the proband was compared using the Wilcoxon statistic. The prevalence of antibody markers in subgroups of relatives was compared using the χ² test. Haplotypes in diabetic probands were subdivided according to age at onset, and differences in genotype according to age at onset were analyzed by χ² analysis.

At the time of analysis, 53 (3.7%) of 1,430 siblings and 106 (4.4%) of 2,419 parents had developed type 1 diabetes by median age 16.5 years (range 25–40) and age 44 years (range 1–94), respectively. Of the siblings, 27 (1.9%) were diagnosed before study entry and 26 (1.8%) during follow-up. Of the parents, 91 (3.7%) were diagnosed before study entry and 15 (0.6%) during follow-up. The median age at diagnosis of the siblings was 6.7 years (range 0.5–17) and of the parents was 21 years (range 1–56). In siblings, the overall risk of diabetes by age 20 was 4.3% (95% CI 3.1–5.4). In parents, the overall risk by age 40 was 4.2% (3.4–5.0). The risk in fathers was 5.7% (4.3–7.0) compared with 2.8% (1.9–3.8) in mothers (P = 0.009).

Figure 1 shows the survival curves for diabetes development in siblings and parents subdivided according to age at diagnosis of the proband. Siblings of a proband diagnosed before age 5 years had a cumulative risk of diabetes by age 20 years of 11.7% (95% CI 6.7–16.8), compared with 3.6% (1.8–5.4) for diagnosis from 5–9 years and 2.3% (1.1–3.5) for 10–14 years (P < 0.0001). In parents, the cumulative risk by age 40 years was 5.9% (3.9–7.9) if the proband was diagnosed before age 5 years, compared with 3.7% (2.3–5.0) for diagnosis from 5 to 9 years and 3.7% (2.5–4.8) for 10–14 years (P = 0.04). In fathers, the cumulative risks of diabetes by age 40 years in these three groups were 8.5% (5.1–11.9), 5.0% (2.8–7.1), and 4.5% (2.7–6.4), respectively (P = 0.05), and in mothers were 3.3% (1.1–5.4), 2.4% (1.0–3.9) and 2.8% (1.4–4.2), respectively (P = 0.7).

Of 1,169 unaffected siblings tested at study entry, 159 had at least one islet autoantibody with levels above the 97.5th percentile of schoolchildren control subjects. The relation between the number of antibodies and age at onset of the proband is shown in Table 2. Detection of multiple antibodies in siblings was strongly associated with the age at onset of the proband. Two or more markers greater than or equal to the 97.5th percentile were found in 13 of 178 (7.3%) siblings of probands diagnosed before 5 years of age, 8 of 368 (2.2%) where the proband was 5–9 years of age, and 15 of 623 (2.4%) where the proband was 10–14 years of age (P = 0.002). Of those who remained nondiabetic at the time of analysis, two or more markers were detected in 7 of 168 (4.2%) siblings of probands diagnosed before 5 years of age, 7 of 366 (1.9%) where the proband was 5–9 years of age, and 8 of 615 (1.3%) where the proband was 10–14 years of age (P = 0.06). Of 2,130 unaffected parents tested, 228 had at least one antibody with levels greater than or equal to the 97.5th percentile of control subjects. A total of 189 had one marker, and 39 had two or more antibody markers above this level (Table 2). There was no association between the number of antibodies detected in the parent and the age at onset of the proband.

The frequency of the highest risk genotype decreased with increasing age at onset. The details are outlined in Table 3. Of 157 children diagnosed before the age of 5 years, 52% were heterozygous for HLA DRB1*03-DQA1*0301-DQB1*02/DRB1*04-DQA1*0501-DQB1*0302, compared with 32% of those diagnosed at 5–9 years and 33% of those diagnosed at 10–14 years (P < 0.001). The frequency of the other HLA class II genotypes was similar in the three age groups.

This study has demonstrated that the risk of diabetes in relatives is strongly influenced by the age at diagnosis of the proband. Siblings of children diagnosed before the age of 5 years had a three- to fivefold greater cumulative risk of diabetes by age 20 years than siblings of children diagnosed between 5 and 15 years of age, and their parents were also more likely to develop diabetes. Previous studies have produced conflicting results. An analysis of 194 families of individuals with type 1 diabetes in Wisconsin found that risk was highest in siblings of those diagnosed before the age of 10 years (24), as did a clinic-based study of 493 families in Minnesota (9). In contrast, the Pittsburgh family study (25) found that risk was not increased in siblings of children diagnosed before age 5 years. The EURODIAB study (1) compared the prevalence of diabetes in relatives by age at diagnosis of 3,960 children with newly diagnosed type 1 diabetes in 16 European countries and found an increased prevalence of diabetes in the parents of children diagnosed before age 5 years, particularly fathers, but no difference in the prevalence of diabetes in their siblings. However, the EURODIAB study analyzed family history at the time of diagnosis, whereas our study included up to 15 years of follow-up, during which the number of affected siblings almost doubled.

We extended this observation by examination of islet autoantibodies in unaffected relatives, thus identifying those at the highest risk of subsequent progression to diabetes. This showed a similar trend, with 4.2% of siblings of children aged <5 years at diagnosis having two or more islet autoantibodies, compared with 1.9 and 1.3% of siblings of those diagnosed at ages 5–9 and 10–14 years, respectively (P = 0.06). Siblings of children diagnosed before age 5 years therefore have a higher prevalence of both overt and preclinical type 1 diabetes than older children.

These findings in siblings should be interpreted with some caution because affected family members tend to develop type 1 diabetes at around the same age (25). Early onset in the proband might therefore be a marker of early onset in relatives, rather than a marker of increased risk over a lifetime. This possibility cannot be excluded but would not be consistent with the increased prevalence of autoantibodies we observed in siblings of early onset cases. Furthermore, age-clustering cannot explain differences in parental risk that persisted up to age 40 years. We therefore interpret our results as indicating an overall increase in risk of diabetes in the relatives of early onset patients, rather than early harvesting of susceptible individuals.

Our observation that risk to parents is associated with age of disease onset in the proband confirms several previous studies (8,11,26) and once again shows that fathers of children with diabetes are more often affected than mothers. In addition, as in the EURODIAB study (11), the effect of early onset in the child on parental risk was largely confined to fathers, whereas maternal risk varied little between onset age groups.

Analysis of the HLA class II genes DRB1, DQA1, and DQB1 in 573 probands from this cohort showed a significant increase in the number of HLA DRB1*03-DQA1*0301-DQB1*02/DRB1*04-DQA1*0501-DQB1*0302 heterozygotes in those diagnosed under the age of 5 years compared with older age groups. This confirms previous reports that young age at onset is associated with a greater genetic load (4) and correlates well with the results of a Finnish study (2) in which the same age groups were analyzed. The high proportion of DRB1 *03/*04 heterozygotes indicates that both parents carry at least one susceptibility haplotype for type 1 diabetes in many families of those diagnosed under the age of 5 years. In this situation, the siblings have a three in four chance of having at least one susceptibility haplotype for type 1 diabetes in common with the proband and at least a one in four chance of being heterozygous for DRB1 *03/*04. Their predicted risk of diabetes would therefore be greater than that of siblings of a proband with only one high-risk allele. The difference in sibling risk observed in our study could therefore be explained in terms of transmission of high-risk HLA alleles, although a role for non-HLA genes cannot be excluded.

Our study was performed in a population with an increasing incidence of childhood diabetes, particularly marked in the very young. The cumulative risk of diabetes by the age of 5 years in the Oxford region has doubled since the study began in 1985 (13). A rapid increase in a genetically stable population implies that environmental agents encountered in early life must play an important role in disease pathogenesis. Early onset diabetes is characterized by high genetic susceptibility, and we confirmed a high prevalence of the highest risk genotype HLA DRB1*03-DQA1*0301-DQB1*02/DRB1*04-DQA1*0501-DQB1*0302 in those diagnosed under the age of 5 years. Susceptibility alleles are therefore over-represented in these families, and we have shown that this is associated with considerably increased risk of progression to diabetes. Taken together, these observations suggest that a changing environment is interacting with the highest risk HLA susceptibility alleles and that each new case in early childhood represents another high-risk family. There is an urgent need to understand the forces that are driving this harmful interaction.

The Bart’s-Oxford study is funded by Diabetes U.K. and the Wellcome Trust.

We thank the physicians and pediatricians in the Oxford Region as well as the families taking part in the study for their continuing support. We also thank the project administrators, Isabel Wilson and Suzanne Weeks, and the BOX study fieldworkers for their help. We are grateful to Alistair Williams, Alistair Norcross, William Moore, Rachel Dix, Kirsty Hunter, and Rhona Langley for technical assistance.