Introduction

Type 1 and type 2 diabetes are not completely distinctive clinically or aetiologically, and may overlap considerably, e.g. acute presentation or insulin requirement may be present in type 2 diabetes, and autoimmune phenomena may arise in combination with insulin resistance [13]. The ‘accelerator hypothesis’, first proposed in 2001, postulates a shared basis for type 1 and type 2 diabetes, with predisposition, insulin resistance and autoimmunity leading to beta cell insufficiency [4]. However, it has long been known that enhanced weight gain in infancy may be associated with a higher risk of diabetes in children [5]. A possible explanation is that enhanced insulin secretion and hyperinsulinism as a result of increased demands favour harmful effects on the beta cells at a critical period in early life. This assumption is supported by in vitro studies and by experiments in diabetes-prone animals [6]. It has also been demonstrated that the implication of early weight gain as a possible risk factor for accelerated type 1 diabetes manifestation early in life may vary between different ethnic groups [79]. Based on these findings, the aim of this study was to explore the relationships of body weight, height and BMI with onset of type 1 diabetes in a large cohort of Caucasoid patients in whom diabetes became manifest between 1990 and 2003, and to investigate both time courses and age-related effects to test the accelerator hypothesis.

Subjects and methods

Subjects and study design

Informed consent was obtained from all participants, and the investigations were carried out in accordance with the principles of the Declaration of Helsinki as revised in 2000 (available at: http://www.wma.net/e/policy/b3.htm, last accessed in August 2005). We analysed anonymised data on 9,248 patients of European extraction (4,360 girls, 4,888 boys) aged 0–20 years. The data had been obtained during the first 2 months after diagnosis, excluding the first week, and were collected from 1990 to 2003 in 116 clinics in Germany and Austria by means of the registry of the Working Group on Pediatric Diabetology in Ulm, Germany. Their database, Diabetes Register for Prospective Follow-Up and Survey Version (DPV), was established for quality control and scientific surveys [10]. Patients were classified into four groups according to age at diagnosis (0–4.9 years, 5–9.9 years, 10–14.9 years and 15–20 years). Height, weight and BMI data were converted into age- and sex-adjusted SD scores (SDSs) using current reference data on 17,147 boys and 17,275 girls, which were obtained from multicenter surveys carried out over a comparable time-span throughout Germany [11, 12]. To test our hypothesis that a higher BMI SDS is associated with an earlier onset of type 1 diabetes, we applied a calculation method proposed by Cole [13]. The curves for each variable (e.g. BMI) were derived using the LMS method Box–Cox power transformation, which adjusts the distribution of the parameters for skewness and allows individual data to be expressed as an SDS or a z score [13]. Subjects were diagnosed according to the criteria set out by the World Health Organization, based on clinical symptoms, hyperglycaemia, ketonuria, ketoacidosis and insulin requirement [14].

Statistical analysis

Statistical analysis was performed using SAS/STAT version 8.2 (SAS Institute, Cary, NC, USA) to calculate means, SEM, SDS, the multi-sample Kruskal–Wallis test, the Wilcoxon signed rank test and multivariate analysis based on a generalised linear model. A p value of less than 0.05 was considered statistically significant after adjustment for multiple comparisons by the Bonferroni stepdown correction.

Results

Age distribution of patients

In total, 1,822 patients (19.7%) were diagnosed between 0 and 4.9 years of age, 3,132 children (33.9%) between 5 and 9.9 years of age, 3,598 patients (38.9%) between 10 and 14.9 years of age, and 696 patients (7.5%) between 15 and 20 years of age.

Weight and BMI of patients at diabetes manifestation

Mean BMI values for the patients and age-matched controls are given in Table 1. The results indicate that, in all age groups, diabetic patients had a significantly higher mean BMI than the controls. Using the values for the patient cohort as a whole, weight SDS was +0.33 and BMI SDS +0.34, indicating that both weight and BMI were significantly higher in patients at diagnosis than in controls (p<0.00001). Mean BMI SDS at diagnosis was +0.51 in the 0–4.9-years age group, +0.35 in the 5–9.9-years age group, +0.28 in the 10–14.9-years age group (p<0.00001), and +0.18 in the 15–20-years age group (p<0.001). Higher weight SDSs compared with age-matched reference data were found in almost all age groups (p<0.05 to p<0.00001). Both males and females in all age groups had higher BMI SDSs than controls. In children under 5 years of age (boys and girls), BMI SDS was consistently found to be elevated (p<0.00001). Weight SDS and BMI SDS at diabetes manifestation were both higher in boys aged 10–14.9 years than in age-matched girls (p<0.001).

Table 1 Mean BMIs of 9,248 patients with type 1 diabetes, categorised into four groups according to age at diagnosis, and 34,422 age-matched controls [11, 12]

Between manifestation years 1990 and 2003 there was a significant rise in the BMI SDS of both sexes (Fig. 1), especially in children aged 5–9.9 years (p<0.0001), those aged 10–14 years (p<0.001), and in the cohort as a whole (p<0.0001, slope 0.0164, 95% CI 0.0102–0.0227). In the study group, weight SDS increased from +0.06 to +0.38 over the observation period, with the increase being particularly marked in children aged 5–14.9 years (p<0.001), and BMI SDS increased from +0.01 to +0.39 (p<0.0001). In the cohort as a whole, there was a significant negative correlation between the patients’ BMI SDS and age at diagnosis, with a mean annual decrease in BMI SDS of −0.0248 (95% CI −0.0294 to −0.0202, p<0.0001) observed.

Fig. 1
figure 1

BMI SDSs of the different age groups (symbols) and the entire cohort (dashed line) at diagnosis of type 1 diabetes over a 10-year study period. German reference data [11, 12] representing an SDS of ±0 are indicated by a dark horizontal line. Filled circles, children aged 0–4.9 years at diabetes onset; empty circles, 5–9.9 years; empty triangles, adolescents aged 10–14.9 years; filled triangles, 15–20 years. ***p<0.001, ****p<0.0001 vs reference data

Height of patients at diabetes manifestation

In the entire cohort, height SDS was +0.15, indicating that patients were significantly taller than controls (p<0.0001). In the three different age groups between 0 years and 14.9 years of age, height SDS was significantly higher than the reference data, for the entire group and for both sexes (p<0.01 to p<0.00001). Height SDS at diabetes onset was significantly higher in boys aged 10–14.9 years than in girls of the same age (p<0.05). There was a decrease in height SDS from early childhood to adulthood, ranging from +0.24 in group I (0–4.9 years) down to −0.07 in group IV (15–20 years), for the entire cohort and for both sexes. Therefore, height SDS was not significantly different from the reference values in patients aged 15–20 years. Between manifestation years 1990 and 2003, no significant increase in height SDS was observed for any of the cohorts studied.

Discussion

Our data, gathered from 9,248 children and adolescents with newly diagnosed type 1 diabetes, demonstrate a clear association between a higher BMI SDS and a younger age at diabetes manifestation in both girls and boys. Infants diagnosed with diabetes under 5 years of age presented with higher weight SDSs and BMI SDSs than older patients. These results suggest that an increase in BMI SDS could be a risk factor for the onset of type 1 diabetes early in life. This is in line with other studies that have focussed on anthropometric data in children prior to diabetes onset [5, 7, 15, 16]. Excess weight gain in infancy has been shown to be associated with an increased risk of early diabetes manifestation in susceptible individuals [17]. Furthermore, an Austrian study showed a positive trend between the BMI of the background population and diabetes incidence rates in children [18]. As in our study, BMI SDSs have previously been observed to be inversely related to age at diagnosis [19]. Increased body fat or the crossing of a certain threshold weight centile during a particular time-span seem to be crucial for an early manifestation of type 1 diabetes in children [20]. However, a recent study [9] that investigated diabetic children who were either from the UK or originally from South Asia did not confirm the accelerator hypothesis, possibly due to the small number of participants enrolled.

In our study, both the weight SDS and the BMI SDS of the cohort increased significantly between 1990 and 2003. In Germany, 20% of children and adolescents are overweight [21], as are 17% of diabetic patients [22], in whom body weight is influenced by metabolic control and concomitant autoimmune disorders [23].

An additional finding in our study was that height SDSs at diabetes onset were significantly elevated in younger age groups but not in adolescents. This finding is most likely the result of an increased growth velocity associated with a higher BMI, as obese children frequently exhibit accelerated growth [7, 15].

A shift to an early onset of type 1 diabetes, rather than a true increase in the incidence of the disease, can be seen [24, 25]. Only 7.5% of our patients were in the 15–20-years age category, which may be the result of lower incidence rates or incomplete retrieval [26]. Interestingly, the absolute increase in the incidence of type 1 diabetes is reported to be comparable in infants, children and adolescents [27].

The accelerator hypothesis [4] proposes that type 1 and type 2 diabetes form part of a spectrum of disturbances associated with a single metabolic disorder, based on genetics, insulin resistance, hypoinsulinaemia and autoimmune phenomena. Further studies are necessary to clarify whether age at diagnosis of type 1 diabetes is indeed a function of body fat mass.

Conclusion

A higher BMI was associated with a younger age of diabetes manifestation in our cohort of 9,248 patients from Germany and Austria. Increased weight gain in childhood could therefore be an additional factor for the early manifestation of type 1 diabetes through metabolic and immunological disturbances. Clinical trials aimed at the prevention of overfeeding and the control of weight gain in children may not only prevent type 2 diabetes in later life, but may also contribute to a delayed manifestation of type 1 diabetes.