Changing Prevalence of Overweight Children and Adolescents at Onset of Insulin-Treated Diabetes

All patients treated with insulin at the onset of diabetes were admitted to Children’s Hospital of Pittsburgh. As part of an ongoing study, all 130 black patients diagnosed between 1 January 1979 and 31 December 1998 were matched by sex, age (within 1 year), and year of diagnosis (within 1 year) to an equal number of white patients. The criteria for inclusion in this study were as follows: 1) being assessed to have diabetes and require insulin by a physician, 2) <19 years of age at diagnosis, and 3) treated with insulin therapy at the time of hospital discharge. All cases of secondary diabetes and type 2 diabetes, diagnosed on the basis of clinical criteria, were excluded. The initial management of new patients with diabetes, which did not change over the time period of this study, involved starting insulin therapy in all patients with a presumptive diagnosis of type 1 diabetes, irrespective of obesity, if they had ketosis and/or severe hyperglycemia and/or marked weight loss. If obese patients improved dramatically or became euglycemic on small insulin doses while on diet, insulin was stopped within a few days and they were treated with diet and oral hypoglycemic agents. These patients and those treated without insulin at onset were not included in this study.

Data were obtained from review of medical records and included sex, race, date and age at onset, height, weight (measured in all children at diagnosis), family history of diabetes, bicarbonate level, presence of ketosis, and diabetic ketoacidosis. BMI was expressed as body weight in kilograms divided by the square of height in meters (kg/m²) (19), and overweight was defined as BMI ≥85th percentile for age and sex. Period I was defined as 1 January 1979 to 31 December 1989 and period II as 1 January 1990 to 31 December 1998.

Blood samples were obtained within 1 week of diagnosis for insulin autoantibodies (IAA) or within 3 months for autoantibodies other than IAA. Blood was stored frozen at 20°C before antibody testing. Antibody positivity was defined as having at least one of the four conventional autoantibodies present, either islet cell antibody (ICA) measured on human pancreas (n = 184), IAA (n = 95), GAD65 autoantibodies (n = 175), or insulinoma-associated protein 2 (IA-2) autoantibodies (n = 184).

ICAs were detected by a modification of an immunoperoxidase method on human blood group 0 (H) fresh frozen pancreas (20). This assay is sensitive to Juvenile Diabetes Foundation (JDF) units using the JDF serum as standard. The specificity of this assay was 77–100%, and sensitivity varied between 88 and 99% in the JDF proficiency workshops conducted by the University of Florida in Gainesville from 1991 to 1996.

The insulin autoantibody assay was performed only on those subjects with serum available within 5 days of diagnosis. The radioimmunoassay used I¹²⁵-mono-iodinated insulin obtained from New England Nuclear and protein A separation using the assay described by Williams et al. (21). This assay detected 16% of subsets of type 1 diabetes and was 100% specific in the Diabetes Autoantibody Standardization Program workshop in 2000, which was average for participating laboratories. The intra-assay coefficient of variation was 8% and the interassay coefficients of variation (CVs) were 8.9, 6.5, and 21.6%, respectively, for low, medium, and high controls.

GAD65 autoantibodies were detected by a radiobinding assay using ³⁵S-[Met]-labeled recombinant human GAD65 produced in vitro with the TNT reticulocyte R transcription/translation kit as described by Grubin et al. (22). The GAD65 construct used for this study was donated by Dr. Åke Lernmark, whereas the IA-2 construct (ICA512bdc) was provided by Dr. George Eisenbarth. The results were expressed as an index (index = sample cpm − negative control cpm/positive control cpm − negative control cpm) as previously reported (23). Of the results from the proficiency workshops organized by the University of Florida in Gainesville (1995, 1996, and 1997) and the Diabetes Autoantibody Standardization Program workshop (2000) organized by the World Health Organization, the data were 76–100% sensitivity, 90–100% specificity (100% specificity three times), and 100% validity for GAD autoantibodies; and 48–78.5% sensitivity, 98–100% specificity, 87.5% validity, and 91.6% consistency in the 1996 and 2000 workshops for IA-2 autoantibodies.

Student’s t test was used to compare continuous variables and the χ² test to compare proportions. When the expected values were <5, Fisher’s exact test was used. P values <0.05 were deemed statistically significant. Logistic regression analysis was used to determine which variables were associated with presence of overweight. Variables introduced into the regression analysis included sex, race, age at diagnosis, time period, presence of at least one autoantibody to the islet cells, and presence of ketosis. Computations were done using the SPSS statistical software package (SPSS, Chicago, IL).

Complete data including blood for antibody testing were available for 185 (71%) of the 260 subjects, of whom 96 were black and 89 were white. Their characteristics are presented in Table 1. Of the 96 black subjects, 38% (n = 36) were diagnosed during period I and 62% (n = 60) during period II. Similarly, 40% (n = 35) of the white subjects presented during period I, whereas 60% (n = 54) were diagnosed during period II (P = 0.79), which is consistent with the increased number of newly diagnosed patients in the Children’s Hospital of Pittsburgh Registry from 50–60 per year in the early 1980s to 120–150 per year at the end of the 1990s.

The prevalence of being overweight through the 20-year period was 27.5%. There was no significant difference by sex (females vs. males, 26.7 vs. 28.5%, P = 0.91). However, blacks had a higher prevalence of overweight than whites (42 vs. 11%, P = 0.0001), and adolescents (aged ≥11 years) had a higher prevalence of overweight than children aged <11 years (40 vs. 16%, P = 0.0002). Among those with autoimmunity, the prevalence of being overweight was 16%, but it was less than in the group with no evidence of autoimmunity (66%, P = 0.00001).

When comparing the two periods, the prevalence of being overweight increased threefold, from 12.6% in the decade of the 1980s to 36.8% in the decade of the 1990s (P = 0.0003) (Fig. 1). This increase was seen in both males (16–37.5%, P = 0.036) and females (8.5–36.4%, P = 0.002). In whites, the prevalence rose five times, from 2.9 to 16.6% (P = 0.04), whereas in blacks it more than doubled from 22 to 55% (P = 0.001). In younger children (aged <11 years at onset), the prevalence of being overweight increased from 7.3 to 22.2% (P = 0.04), whereas in the group ≥11 years of age, it increased from 20 to 50% (P = 0.006).

In the group with islet autoantibodies (n = 144), who by definition have type 1a diabetes, the prevalence of obesity increased almost five times, from 5.1% in the 1980s to 24.4% in the 1990s (P = 0.002). The increase was statistically significant among blacks (7.4 vs. 36.8%, P = 0.006), males (6.8 vs. 28.2%, P = 0.02), females (3.4 vs. 21.2%, P = 0.04), and the adolescent group (5 vs. 33.3%, P = 0.02). In the group without autoantibodies, the prevalence of being overweight was higher and also increased from 46.1 to 75% without reaching statistical significances (P = 0.07), which was likely due to the small numbers (n = 41).

Table 2 shows the biochemical data at presentation and frequency of autoantibodies by race and period. No differences were seen in the severity of presentation between periods in either blacks or whites. In period I, 39% of the group presented in diabetic ketoacidosis compared with 35% in period II (P = 0.99). There were also no significant differences in prevalence of ketosis (81 vs. 73%, P = 0.2), mean bicarbonate level (16.7 vs. 17.5 meq/l, P = 0.3), or venous pH (7.29 vs. 7.27, P = 0.3). There was also no statistically significant difference in severity of presentation (presence of diabetic ketoacidosis) between the overweight patients with and without evidence of autoimmunity (29 vs. 15%, P = 0.37).

In the logistic regression, being black (odds ratio 5.84; 95% CI 4.94–6.74), older (2.52; 1.61–3.38), diagnosed during the 1990s (5.95; 4.97–6.97), and not having islet antibodies (7.39; 6.51–8.27) were significantly associated with being overweight.

These data demonstrate that being overweight is increasingly prevalent in children and adolescents with new-onset insulin-treated diabetes similar to the increasing prevalence of obesity observed in the general population (7,8). This increase was seen irrespective of sex, race, and age at onset, but it was more striking among whites and those with islet autoantibodies. The most dramatic increase was seen in those with autoimmune-associated type 1 diabetes, with a fivefold increase in the prevalence of being overweight. In the children without these autoantibodies, which is usually associated with type 2 diabetes and described in studies of “atypical diabetes” (24–26), the initially high prevalence of overweight in period I also almost doubled. This weight excess was obvious despite the frequent weight loss during the prodromal phase of the disease and was unlikely to have accounted for differences between the two periods as the frequency of ketosis and diabetic ketoacidosis were not different.

These observations provide further evidence that there may be a link between the prevalence of childhood weight excess and the rising incidence of childhood type 1 diabetes. The incidence of insulin-treated diabetes in adolescents in Allegheny County, Pennsylvania, increased more than threefold from the early 1980s to 1994 (15), which is concurrent with an increase in incidence of type 1 diabetes in children reported around the world (27). Our data from the Children’s Hospital of Pittsburgh Registry, which has previously been shown to be representative of the Allegheny County population, reveal an apparent continued increase in these numbers. This requires validation by an update of the population-based registry. The BMI data demonstrate an increase in the prevalence of excess weight in children diagnosed with diabetes, as is also seen in the general population and is contemporaneous with the worldwide increased incidence of type 1 diabetes (27).

Our data support a potential mechanistic link between excess weight gain and increasing frequency and earlier onset of type 1a diabetes with the age decreasing to include most of early childhood (28). Obesity-associated insulin resistance increases insulin secretory demands on β-cells and may accelerate autoimmune damage in these metabolically upregulated cells by increasing antigen presentation (29). Also, the clinical presentation of autoimmune-mediated diabetes can probably occur earlier in the face of relative insulin deficiency, which otherwise may not become clinically apparent until later in life in susceptible individuals. These concepts are included in the recently described “accelerator hypothesis” (5). The theory of obesity-induced acceleration of type 1 diabetes is supported by the reports of an inverse correlation between age at onset and BMI at diagnosis in a group of children <16 years of age (6) as well as the studies that have shown that children who develop type 1 diabetes were heavier in their pre-diabetes infancy and childhood periods than control children (2–4).

In summary, this study shows that the prevalence of being overweight at onset of insulin-treated diabetes has tripled in children and adolescents from the 1980s to the 1990s, which follows the trend seen in the general population. The rate of this increase is most dramatic among those with type 1 diabetes. This leads to the speculation that excess weight gain may be an accelerating factor for onset of the disease and that it may have contributed to the increase in incidence seen in some populations. This implication of mixed ethiopathogenesis may impact classification, treatment, and prevention of childhood diabetes. Moreover, these data raise a cautionary note in the general assumption that all overweight children with diabetes have type 2 diabetes. Instead, children with type 1 diabetes are increasingly likely to present with a phenotype that is similar to that of type 2 diabetes. Thus, both autoimmunity and insulin resistance may play a role in the precipitation of clinical disease.

This work was supported in part by a Resident Research Grant from the American Academy of Pediatrics (to I. L.), by National Institutes of Health Grants T32-DK-07729 (to I.L.), R01-DK-24021-20 (to D.B.), R01-DK-53456, and R01-DK-56200 (to M.P.), and by the American Diabetes Association (Career Development Award to M.P.). The authors also acknowledge the General Clinical Research Center (NIH 5M01RR00084), the Renziehausen Trust Fund, and the Cochrane-Webber Trust Fund.

We thank Angela Darnley, Jamie Gay, Susan Pietropaolo, Karen Riley, and the general clinical research center nurses for their assistance.