Obesity, Islet Cell Autoimmunity, and Cardiovascular Risk Factors in Youth at Onset of Type 1 Autoimmune Diabetes

Abstract

Context:

The current increase in childhood type 1 diabetes (T1D) and obesity has led to two conflicting hypotheses and conflicting reports regarding the effects of overweight on initiation and spreading of islet cell autoimmunity vs earlier clinical manifestation of preexisting autoimmune β-cell damage driven by excess weight.

Objective:

The objective of the study was to address the question of whether the degree of β-cell autoimmunity and age are related to overweight at diabetes onset in a large cohort of T1D youth.

Design:

This was a prospective cross-sectional study of youth with autoimmune T1D consecutively recruited at diabetes onset.

Setting:

The study was conducted at a regional academic pediatric diabetes center.

Patients:

Two hundred sixty-three consecutive children younger than 19 years at onset of T1D participated in the study.

Main Outcome Measures:

Relationships between body mass index and central obesity (waist circumference and waist to height ratio) and antigen spreading (islet cell autoantibody number), age, and cardiovascular (CVD) risk factors examined at onset and/or 3 months after the diagnosis were measured.

Results:

There were no significant associations between number of autoantibodies with measures of adiposity. Age relationships revealed that a greater proportion of those with central obesity (21%) were in the youngest age group (0–4 y) compared with those without central obesity (6%) (P = .001). Patients with central obesity had increased CVD risk factors and higher onset C-peptide levels (P < .05).

Conclusions:

No evidence was found to support the concept that obesity accelerates progression of autoantibody spreading once autoimmunity, marked by standard islet cell autoantibody assays, is present. Central obesity was present in almost one-third of the subjects and was associated with early CVD risk markers already at onset.

Contemporaneous with the worldwide increasing prevalence of pediatric obesity (1, 2), the prevalence of overweight in children at the onset of type 1 diabetes (T1D) increased significantly in our population from 12.6% (1979–1989) to 43.9% (1999–2002) (3, 4), confirmed in other cohorts (5), and paralleling the rising incidence of T1D (6, 7), especially in the youngest age group (0–5 y) (8, 9). Based on these data, we speculated a decade ago that weight excess may accelerate T1D onset, contributing to its increased incidence (10). Individuals with T1D were documented to be heavier and taller during childhood, suggesting a link between β-cell demand and T1D risk (11). The accelerator hypothesis posits that obesity-driven insulin resistance (IR) in genetically predisposed individuals, leads to β-cell autoimmunity, and accelerated diabetes (12). In contrast, we speculate that obesity-driven IR accelerates the onset of clinical hyperglycemia at an early stage of established autoimmune β-cell damage, with possible additional acceleration of antigen spreading (13).

Smaller retrospective studies testing the accelerator hypothesis reported contradictory and controversial results (11, 14, 15). Some, but not all, described associations between body mass index (BMI) and earlier age at T1D onset and a few suggested associations of IR with younger age only in those with decreased insulin secretion or low C-peptide levels (11, 15).

This prospective cross-sectional analysis of a large unselected new-onset pediatric T1D population evaluates relationships between adiposity and the degree of β-cell autoimmunity and/or age.

Materials and Methods

Subjects

All patients age younger than 19 years, diagnosed with clinical T1D from January 2004 to December 2006 at Children's Hospital of Pittsburgh were recruited at onset. Inclusion criteria were as follows: 1) informed consent, 2) diagnosis of diabetes requiring insulin, 3) insulin treatment at hospital discharge, and 4) available research laboratory results for three or more β-cell autoantibodies (AAs), including islet cell autoantibodies (ICAs), glutamic acid decarboxylase (GAD) (65 kDa isoform), insulin antibody (IA)-2AA, and insulin autoantibody (IAA). Cases with clinical maturity-onset diabetes of the young, secondary diabetes, and without AA were excluded.

Demographic and clinical data

Clinical data at onset and follow-up at 2–3 months were obtained from hospital and research records including gender, race, age, height, weight, and blood pressure (BP). Waist circumference (WC) was obtained at follow-up only. BMI percentiles and BMI Z scores were calculated using the Centers for Disease Control and Prevention 2000 growth data with overweight defined as BMI of the 85th percentile or greater to less than the 95th percentile for age and gender and obesity as BMI of the 95th percentile or greater. Waist to height ratio (WHtR) was used to determine central obesity (WHtR ≥ 0.5) (16). Age, gender, and ethnicity-specific WC percentiles were calculated based on the nationally representative tables (17). The BP percentiles were adjusted by height, age, and gender.

Assays

Blood samples obtained at diagnosis and/or follow-up (within 100 d) were assayed for glycated hemoglobin (HbA1c), lipids, postprandial serum C-peptide levels, human leukocyte antigen (HLA) typing, and autoantibodies. IAA was measured only in samples obtained within 1 week of diagnosis (n = 113). IA-2AA, GAD, 65 kDa isoform (GAD-65) AA, IAA, and ICA using group 0 human pancreas were measured as previously described (10). Sensitivities and specificities have been consistently 80%–100% for ICA, GAD, and IA-2 autoantibodies with 60% sensitivity and 93% specificity for IAA.

Subjects were instructed to eat a typical meal 1–2 hours before their blood draw. Postprandial C-peptide levels were determined using a human C-peptide RIA kit (Linco Research) with a lower detection limit of 0.1 ng/mL (0.0333 nmol/L). Inter- and intraassay coefficients of variation were 4.7% and 4.6%, respectively. Serum lipid levels were measured in the Nutrition Laboratory (3, 4), and low-density lipoprotein (LDL) cholesterol levels were estimated using Friedewald's equation. HLA typing was determined for the presence of the DQ2 and/or DQ8 haplotypes (18).

Statistical analysis

Comparisons of continuous variables between subgroups were assessed using a t test and ANOVA procedures. Comparison of proportions for categorical variables between subgroups used χ² test procedures. Associations between continuous variables used Pearson correlations. Nonparametric alternatives were used when distributional assumptions were not met. Linear regression modeling assessed the relationship between age and individual anthropometric measures of adiposity. Covariates including demographics, HLA, number of AAs, C-peptide, and HbA1c were included in the modeling. Logistic regression modeling assessed the associations of the presence of AAs with various adiposity measures. Cumulative logit models assessed relationships between adiposity and the number of AAs. Multivariate models included covariates of age and/or HbA1c, depending on the adiposity measure. Analyses were performed using SAS 9.3 (SAS Institute).

Results

Subjects

Of the 351 subjects recruited, the 295 who had three or more AAs measured were older (9.9 ± 4.1 vs 7.8 ± 4.6 y, P < .001), with fewer aged 0–4 years (14% vs 32%, P = .002), more males (59% vs 45%, P = .04), and less centrally obese (WHtR 0.5 ± 0.1 vs 0.6 ± 0.1, P = .001) compared with the 56 excluded. There were no significant differences in C-peptide levels at onset [median (25th percentile, 75th percentile) 0.6 (0.2, 0.9) vs 0.6 (0.2, 0.8) ng/mL, P = .75] or at follow-up [1.4 (0.7, 2.6) vs 2.4 (1.1, 4.4) ng/mL, P = .4], HbA1c at onset (11.7 ± 2.5 vs 11.6 ± 2.6%, P = .75) or at follow-up (7.4 ± 0.9 vs 7.6 ± 1.1%, P = .11), BMI percentile at follow-up (74.5 ± 22.3 vs 77.7 ± 22.5, P = .36), WC (67.0 ± 14.9 vs 70.6 ± 21.6 cm, P = .45), WC greater than the 75th percentile (67% vs 55%, P = .41) or height Z score (0.3 ± 1.0 vs 0.1 ± 1.1, P = .21), respectively. There were also no significant differences in the frequency of high-risk DQ2 and/or DQ8 HLA alleles DQ2 (28% vs 24%), DQ8 (35% vs 44%), DQ2/DQ8 (18% vs 15%), XX (19% vs 18%, P = .77). Most exclusions were attributed to insufficient blood volume (probably explaining the age difference) or the sample drawn out of window.

AAs were present in 263 of 295 (89%) with three or more in most (17.9% with 1, 34.6% with 2, 47.5% with ≥3). The most common AAs were ICA and IA2, followed by IAA and GAD-65. Characteristics of the study population are presented in Table 1. There were no differences in any adiposity characteristics from a control group of 43 low-HLA-risk siblings with similar age and gender (data not shown).

Table 1.

Demographic and Clinical Characteristics of Study Group (n = 263)

Characteristics	Value
Age, y	9.7 ± 4.1
Age group, y, n
0–4	14% (37)
5–9	37% (96)
10–14	40% (106)
15–18	9% (24)
Gender (% male), n	59% (155)
Race, white/black/other, %	94/5/1
BMI percentile at onset	51.3 ± 34.5
BMI Z score at onset	−0.0 ± 1.5
BMI percentile at 3 mo	73.2 ± 22.4a
BMI Z score at 3 mo	0.8 ± 0.9a
Waist class at 3 mo, % >75th percentile, n	67% (129)
Height Z score at 3 mo	0.4 ± 1.0b
WHtR at 3 mo	0.5 ± 0.1c
Centrally obese, %, n	30% (57)
C-peptide at onset, ng/mL	0.6 [0.2, 0.9]d
C-peptide at 3 mo, ng/mL	1.4 [0.7, 2.5]e
HbA1c at onset, %	11.7 ± 2.5
HbA1c at 3 mo, %	7.4 ± 0.8
HLA DQ2 and DQ8, %, n
XX	17% (41)
DQ2	30% (72)
DQ8	36% (89)
DQ2/DQ8	17% (42)
ICA h, % positive, n	84% (222)
GAD-65, % positive, n	56% (148)
IA2, % positive, n	73% (193)
IAA, % positive, n	64% (68)

Data are mean ± SD unless otherwise noted. C-peptide is presented as median [interquartile range].

^aBased on 247 subjects.

^bBased on 253 subjects.

^cBased on 192 subjects.

^dBased on 244 subjects.

^eBased on 219 subjects.

Relationships between islet autoantibodies and measures of adiposity

The assessment of associations between the number of positive AAs and measures of adiposity, using cumulative logistic regression modeling, revealed no significant relationships between any measure of adiposity and number of AAs, irrespective of the classification. Adjustment by age and/or HbA1c did not reveal significant associations between adiposity and number of AA.

Role of central obesity

Central obesity was present in 30% (n = 57) who had a significantly higher BMI, prevalence of overweight/obesity (80%), and WC of the 75th percentile or greater (91%) than the noncentral obese group at follow-up (Table 2). The centrally obese group included a significantly greater proportion of younger (0–4 y) (21% vs 6%) and older (15–18 y) (16% vs 7%) children vs the noncentrally obese group and had slightly but significantly higher C-peptide levels at onset but not at follow-up (Table 2). Components of the metabolic syndrome at follow-up, ie, systolic BP (SBP) and triglycerides (TGs), were significantly higher and high-density lipoprotein (HDL) significantly lower than in the noncentrally obese group, with LDL marginally higher (P = .063). There were no differences in ethnicity, HLA type, or number and type of AA (Table 2).

Table 2.

Demographics and Measurements by Different Obesity Criteria: Centrally Obese Versus Noncentrally Obese and BMI > 85th Percentile and BMI ≤ 85th Percentile

	Central Obese (n = 57)	Noncentral Obese (n = 135)	P Value	BMI > 85th Percentile (n = 98)	BMI ≤ 85th Percentile (n = 149)	P Value
Age, y	10.0 ± 4.6	9.9 ± 3.4	.82	9.8 ± 3.7	9.7 ± 4.0	.88
Age group, y			.001			.73
0–4	21% (12)	6% (8)		11% (11)	13% (19)
5–9	25% (14)	44% (60)		42% (41)	36% (53)
10–14	39% (22)	43% (58)		38% (37)	44% (65)
15–18	16% (9)	7% (9)		9% (9)	8% (12)
Gender, % male, n	56% (32)	55% (74)	.87	58% (57)	58% (87)	.97
Race, white/black/other, %	91/7/2	93/5/2	0.87	89/9/2	96/3/1	.05
BMI percentile at 3 mo	90.1 ± 14.2	66.6 ± 20.6	<.001	93.7 ± 4.5	59.7 ± 18.9	<.001
BMI Z score at 3 mo	1.6 ± 0.8a	0.5 ± 0.7	<.001	1.7 ± 0.5	0.3 ± 0.6	<.001
Height Z score	0.6 ± 0.9	0.3 ± 1.1	.16	0.6 ± 0.9	0.2 ± 1.0	.001
C-peptide at onset, ng/mL	0.8 [0.6, 1.1]	0.5 [0.2, 0.7]	<.001	0.7 [0.4, 1.0]	0.5 [0.2, 0.7]b	.000 01
C-peptide at 3 mo, ng/mL	0.9 [0.3, 3.1]	1.42 [0.8, 2.4]	.48	1.7 [0.6, 3.0]	1.2 [0.7, 2.3]b	.48
HbA1c at onset	11.8 ± 2.38 (55)	11.8 ± 2.4 (133)	.96	11.7 ± 2.5 (93)	11.7 ± 2.4 (147)	.82
HbA1c at 3 mo	7.4 ± 0.7a	7.3 ± 0.8 (128)	.45	7.2 ± 0.8 (86)	7.4 ± 0.8 (142)	.06
Multiple (two or more) antibodies	86% (49)	81% (109)	.39	84% (82)	83% (124)	.93
+ICA (H), %, n	89% (51)	84% (113)	.30	88% (86)	85% (126)	.48
+GAD, %, n	58% (33)	59% (80)	.86	55% (54)	59% (88)	.54
+IA2, %, n	75% (43)	75% (101)	.93	71% (70)	75% (112)	.51
+IAA, %, n	74% (14)	58% (28)	.24	63% (25)	66% (38)	.76
Positive antibodies, n			.67			.88
1+, %, n	14% (8)	19% (26)		16% (16)	17% (25)
2+, %, n	39% (22)	33% (45)		38% (37)	33% (49)
3+, %, n	33% (19)	37% (50)		36% (35)	40% (59)
4+, %, n	14% (8)	10% (14)		10% (10)	11% (16)
Positive antibodies, n	2.5 ± 0.9 (57)	2.4 ± 0.9 (135)	.54	2.4 ± 0.9 (98)	2.4 ± 0.9 (149)	.70
Percentile for SBP at 3 mo, n	76.2 ± 20.4 (45)	61.1 ± 27.9 (118)	.001	72.6 ± 22.0 (73)	60.4 ± 27.7 (130)	.001
Percentile for DBP at 3 mo, n	63.5 ± 22.9 (45)	57.6 ± 20.0 (117)	.11	59.6 ± 19.3 (72)	58.6 ± 21.5 (130)	.73
HDL for 3 mo, mg/dL, n	43.2 ± 8.5 (47)	50.7 ± 10.9 (131)	<.001	44.6 ± 10.4 (78)	51.3 ± 11.1 (122)	<.001
TGs at 3 mo, mg/dL, n	143.7 ± 92.5 (48)	114.8 ± 55.4 ( 131)	.046	131.2 ± 82.1 (79)	114.2 ± 53.6 (123)	.10
Cholesterol for 3 mo, mg/dL, n	158.4 ± 26.9 (48)	154.3 ± 25.2 (131)	.35	155.5 ± 28.5 (79)	154.4 ± 23.4 (123)	.77
LDL for 3 mo, mg/dL, n	88.0 ± 24.4 (47)	80.7 ± 22.2 (131)	.063	84.2 ± 24.7 (78)	80.0 ± 21.2 (122)	.19

Abbreviation: DBP, diastolic blood pressure. Data are mean ± SD unless otherwise noted. C-peptide is presented as median [interquartile range].

^aBased on 54 subjects.

^bBased on 149 subjects.

Overweight/obesity was more common in blacks and was associated with significantly higher WC percentiles, height, C-peptide levels at onset, and SBP and lower HDL but not with age, prevalence of AA, HbA1c, or LDL (Table 2).

Relationships between age and adiposity

Age was not a determinant of BMI, WC percentile, or WHtR, even when controlling for potential confounders of HbA1c, C-peptide levels, and two or more positive AAs (P > .05).

Univariate linear regression showed positive associations between age and GAD-65AA (R² = 6%, P < .001), onset C-peptide levels (R² = 13%, P < .001), follow-up C-peptide levels (R² = 9%, P < .001), baseline HbA1c (R² = 8%, P < .001) and follow-up HbA1c (R² = 10%, P < .001). Age was negatively associated with height percentile and Z score when adjusted for HbA1c, C-peptide, and presence of two or more positive AAs. No significant associations were found between age and AA number or HLA type.

Discussion

The term, double diabetes, was conceived as the coexistence of autoimmunity, insulin deficiency, and insulin resistance (13), with the latter postulated to play a role in both the progression of insulin deficiency and timing of presentation of T1D in addition to causing obesity-related morbidities. The pathogenic concept has resulted in many contradictory reports and controversy around the accelerator hypothesis (11). However, there is increasing evidence that IR increases the risk of cardiovascular complications in T1D adults (19, 20).

We were unable to confirm the hypothesis that obesity accelerates the clinical presentation of T1D at early phases of β-cell damage, as reflected by fewer diabetes-related AAs and higher β-cell reserve (C-peptide levels). Multiple AAs are suggested to indicate aggressive progression of β-cell destruction. We found no associations between any measure of adiposity with either the number of AAs or their type. Our evaluation did not support the postulate that obesity is more likely in young T1D patients due to acceleration of the disease (13). There was no relationship between any measure of adiposity at follow-up and age, even after adjustment for HbA1c and C-peptide levels, assuming that this recovered status reflected the patients' prediabetes adiposity.

These data, in a large, well-characterized cohort, with research measurements of islet AAs and HLA typing, do not support the accelerator hypothesis once AAs are detected but neither prove nor disprove obesity being an initiator of the β-cell damage (13). Of concern, a significant number of children with T1D had central obesity assessed by the WHtR, a proposed marker of central adiposity and IR. This centrally obese group included a significantly greater proportion of younger children than those without central obesity. They had higher C-peptide levels at diagnosis, suggesting slightly more residual β-cell function. This implies that abdominal obesity-related IR may accelerate the age of onset of clinical diabetes during the course of autoimmune β-cell damage. As documented in children with type 2 diabetes, the centrally obese group had significantly increased cardiometabolic risk factors already at T1D diagnosis, raising the concern that these become superimposed on the known T1D associated increased CVD risk.

In summary, we found no evidence that obesity accelerates the autoimmunity or presentation of T1D, once autoantibodies are present. This is consistent with recent (11, 14), but in contrast to earlier publications (11), possibly due to the different proportions of central obesity among the populations. The prevalence of central obesity in the youngest age group suggests that abdominal adiposity may play a role in accelerating the age of T1D onset. The lack of a relationship of AA number with age-adjusted WC, the best measure of IR in nondiabetic children, does not support a role for IR in AA spreading. Prospective studies are needed to assess a potential role of IR before AAs are produced. The major concern raised is that the centrally obese T1D children demonstrate cardiovascular risk markers already at diagnosis and therefore may be at even greater risk for cardiovascular morbidity than imposed by T1D per se.