In this descriptive prospective study, we analysed data from children developing diabetes in Skåne, the most southern part of Sweden. Of 48,000 children born from September 2000 to August 2004, 60 had developed diabetes as of December 2006. These children were all born during the screening years of the DiPiS study, a population-based prospective investigation of diabetes in children [14, 28, 29], allowing us to select a total of 155 matched control children.
The aim of the DiPiS is to determine the positive predictive value for type 1 diabetes of genetic risk combined with islet cell autoantibody markers and to identify factors during and after pregnancy that may trigger the disease . Cord blood from more than 35,000 children was genotyped for HLA and analysed for islet autoantibodies. Participating parents filled in questionnaires when the child was 2 months of age . Children with increased risk of type 1 diabetes, according to a scoring based upon HLA genotype, cord blood islet autoantibodies or other risk factors such as high birthweight, infections during pregnancy and first-grade relative with diabetes, are followed annually from 2 years of age with blood samples for islet autoantibodies and questionnaires. Children developing two to three islet autoantibodies are followed every 3 months by a paediatrician, including performance of blood tests for islet autoantibodies, glucose and HbA1c. The Lund University Ethics Committee approved the DiPiS study.
Type 1 diabetic children
The 60 children (35 boys, 25 girls) developing diabetes before the end of December 2006 (Fig. 1) were between 9 months and 5 years and 10 months old (median 2.75, range 0.75–5.84 years) at diagnosis. Seven boys and four girls developed diabetes before 18 months of age. A cord blood sample for the DiPiS study was obtained at delivery from 50 of the children. Gestational age ranged from 30 to 42 weeks, with 3/60 children born prematurely (one at 30 weeks, two at 36 weeks). An additional three children were born to mothers with gestational diabetes, but none of the children had a mother with type 1 diabetes. Growth curves were available for 58/60 children.
We randomly selected 155 controls, all born to non-diabetic mothers, but matched for sex, gestational week and date of birth (year and month and as near date as possible), from the 25,378 children participating in the DiPiS. Among the control children, 51 were also matched for HLA.
Swedish children are regularly examined by Child Health Services (Barnavårdscentraler) before starting school. At these visits, weight and height of the child are recorded, and a growth chart completed. These growth charts, which contain birthweight, birth length, parental height and growth in infancy compared with standard Swedish normograms , were obtained with parental consent. Information on both parents’ heights was available for 95% of the children.
The genotypes HLA-DQA1 and HLA-DQB1 were determined by PCR and sequence-specific probes as previously described [14, 31, 32]. HLA genotypes were available from 55 diabetic children, either from cord blood or from samples obtained at the diagnosis of diabetes. HLA-DQA1*-B1* genotypes were ranked according to the risk designation criteria used to identify newborns at risk of type 1 diabetes in the DiPiS as described previously . A ‘risk code’ was established as an arbitrary numeric description of the risk designation (Table 1).
Birthweight and birth length SDS were calculated as described . Height and weight SDS after birth were calculated from the Swedish reference population . MPH was calculated according to the following previously described formula : SDS MPH = (height SDS of father + height SDS of mother)/1.61. BMI SDS was calculated according to Karlberg et al. . Difference from MPH was calculated as SDS birth length or height minus SDS MPH.
Parametric tests were used. If, in tests of between-group differences, unequal variances were observed, the number of degrees of freedom were corrected according to Sattherwaite .
In post hoc tests following ANOVA, correction for multiple testing was performed with Dunnett’s test. Using multiple regression analysis we tested whether gestational age, diabetes, MPH or HLA high-risk genotypes were associated with birth length SDS, birthweight SDS and gain in height SDS. In the regression analyses, HLA risk code genotypes 3 and 4 were combined and used as a dummy variable, and compared with all other genotypes (Table 1).
We considered p values less than 0.05 to be significant. In children developing diabetes before 18 months of age, values of height and BMI were not included at diabetes onset and thereafter.