Role of Insulin Resistance in Predicting Progression to Type 1 Diabetes

The DPT-1 screened 103,390 first- and second-degree nondiabetic relatives of individuals in whom type 1 diabetes had been diagnosed before the age of 45 years. To have been eligible for screening, an individual must have been a first-degree relative of a patient with type 1 diabetes and aged between 3 and 45 years or a second-degree relative aged between 3 and 20 years. Staging of 3,480 relatives who were found to be ICA-positive was performed to quantify the projected 5-year risk of diabetes.

Staging consisted of ICA confirmation, HLA-DQ typing, determination of insulin antibodies, and glucose tolerance testing (intravenous glucose tolerance test [IVGTT] and oral glucose tolerance test [OGTT]). Relatives with HLA-DQA 1*0102/DQB1*0602 were excluded from the DPT-1. Of 808 subjects, those considered to have a >50% 5-year risk were eligible for entry into the parenteral insulin trial (n = 339) if either the FPIR was <1st or 10th percentile (depending on age and relation to the proband) on two occasions and/or there were abnormalities (other than diabetes) on an OGTT. If none of these metabolic abnormalities were present, the FPIR was above threshold (if ≥10 percentile for siblings, offspring, and second-degree relatives or if ≥1st percentile for parents), testing for insulin autoantibody (IAA) was positive, and the 5-year risk was considered to be 25–50%, participants (n = 372) were entered into the oral insulin trial. The subjects were randomly assigned to receive antigen intervention or matched placebo/observation. Neither intervention showed a statistically significant difference between the treatment arms in terms of progression to type 1 diabetes. To study the natural history of this disease, we report here the results of 186 subjects from the oral insulin trial (moderate-risk group) and 170 subjects from the parenteral trial (high-risk group) who were randomly assigned to the placebo/observation arms of the studies.

All subjects were seen every 6 months, at which point an OGTT was administered to assess glycemic status. The dose of oral glucose was 1.75 g/kg (maximum, 75 g of carbohydrate). Blood samples were obtained for C-peptide measurements in the fasting state and then 30, 60, 90, and 120 min later. The peak C-peptide was the maximum point of all measurements.

An IVGTT was performed after an overnight fast at baseline, annually thereafter for moderate-risk subjects or biannually thereafter for high-risk subjects, and at study end for all subjects. Samples were obtained through a temporary indwelling intravenous catheter. Measurements of fasting glucose, fasting insulin, FPIR, and HOMA-IR came from the IVGTT sample.

Impaired glucose tolerance by American Diabetes Association criteria (1997) was defined as fasting plasma glucose <126 mg/dl and 2-h plasma glucose between 140 and 199 mg/dl.

Diabetes was diagnosed according to American Diabetes Association criteria: fasting glucose ≥126 mg/dl, 2-h glucose ≥200 mg/dl with confirmation by either an elevated fasting or 2-h glucose level at a follow-up visit, or random plasma glucose ≥200 mg/dl accompanied by symptoms of polyuria, polydipsia, and/or weight loss.

Plasma glucose was measured by the glucose oxidize method. Glucose tolerance tests were performed after an overnight fast. FPIR was calculated as the sum of the serum insulin concentrations at 1 and 3 min after intravenous injection of glucose. HOMA-IR was calculated as the fasting insulin (milliunits per liter) × fasting glucose (millimoles per liter)/22.5 from the mean of fasting insulin at 10 and 4 min and fasting glucose at 4 min before each IVGTT performed.

Cytoplasmic ICAs were determined on frozen sections of human pancreas. Samples were considered positive at ≥10 Juvenile Diabetes Foundation units.

IAAs were measured by a competitive liquid-phase radioassay. As noted in the DPT-1 protocol, the initial entry criterion for IAA was a level ≥80 nU/ml, which was subsequently changed to a level ≥39 nU/ml (17).

GAD and ICA512 autoantibody levels were measured simultaneously in a combined GAD and ICA512 autoantibody radioassay. The assay was performed in 96-well filtration plates with autoantibody-bound [³H]GAD65 and [³⁵S]ICA512 precipitated with protein A–Sepharose. The cut points were set at indexes of 0.032 (mean ± 2 SD for GAD antibodies) and 0.049 (mean ± 6 SD for ICA512 antibodies), the 99th percentiles of 198 normal control subjects.

For HLA-DQ typing, DNA was extracted from the buffy coats of peripheral blood leukocytes, and HLA-DQA1 and DQB1 alleles were amplified by PCR with the use of sequence-specific probes. In the analysis of this article, a high-risk HLA genotype was defined as having one of the following haplotype combinations: DQA1*0301-DQB1*0302, DQA1*0501-DQB1*0201, DQA1*04-DQB1*0201, or DQA1*0301-DQB*0201.

Variables (ICA titers, IAA titers, peak C-peptide, fasting glucose, fasting insulin, FPIR, HOMA-IR, and the FPIR–to–HOMA-IR ratio) that were not normally distributed were log transformed for analysis. For clarity of presentation, the results in Table 1 are expressed as back-transformed mean ± SD or median (interquartile range) values. Categorical variables were compared by Pearson's χ² test. The continuous variables were tested by t test for the differences in means or by Wilcoxon's rank-sum test for the differences in order. The association of HOMA-IR and FPIR–to–HOMA-IR ratio with time to type 1 diabetes was adjusted for potential confounders using the multivariate Cox proportional hazards model. The β coefficient parameters indicated the direction of the association between the predictor and response variables, and hazard ratios (HRs) were obtained. A paired t test was used to compare the relative change of HOMA-IR and the FPIR–to–HOMA-IR ratio. Correlation coefficients were obtained by Spearman's nonparametric correlation method. Forward selection using a Cox proportional hazards model was applied to select the best risk factors for type 1 diabetes. Tests of significance were two tailed. P < 0.05 was considered to be statistically significant. Statistical analyses were performed with SAS (version 9; SAS Institute, Cary, NC). The analysis was performed separately for the subjects in the moderate-risk group and in the high-risk group.

Table 1 shows the clinical and metabolic characteristics of subjects by risk group. By design, subjects in the moderate-risk group who were ICA⁺ and IAA⁺ with an FPIR above the threshold (n = 186) had a projected 25−50% risk for diabetes over 5 years, and subjects in the high-risk group who were ICA⁺ and had a below- threshold FPIR or impaired glucose tolerance (n = 170) had a projected >50% risk for diabetes over 5 years. In both groups, those who progressed to clinical diabetes were younger than those who did not (moderate-risk group P = 0.035; high-risk group P = 0.002). Progressors and nonprogressors did not differ in sex, relationship to proband, race, or proportion of subjects with a high-risk HLA. In terms of immunologic factors, the progressors presented higher IAA and ICA titers in both risk groups (moderate-risk group P = 0.001; high-risk group P = 0.002). Those progressors in the high-risk group had a higher proportion of GAD65 antibodies present than nonprogressors (P = 0.012), an observation not found in the moderate-risk group (P = 0.674). This finding suggests that the number of antibodies, even among those with loss of FPIR and/or glucose intolerance, is an important predictor of type 1 diabetes.

At baseline, we did not detect a difference in HOMA-IR, but the ratio of FPIR to HOMA-IR was significantly different in both risk groups, comparing progressors to nonprogressors (moderate-risk group P < 0.001; high-risk group P = 0.010). In the high-risk group, FPIR, peak C-peptide, and proportion of subjects with impaired glucose tolerance were significantly different, comparing progressors with nonprogressors at P = 0.004, P < 0.001, and P = 0.003, respectively.

Before the end of the study, 53 subjects in the moderate-risk group and 70 subjects in the high-risk group developed type 1 diabetes. The median follow-up was 4.3 years for moderate-risk subjects and 3.7 years for high-risk subjects.

In Table 2, the time to type 1 diabetes was significantly associated with HOMA-IR (P = 0.038) and the FPIR–to–HOMA-IR ratio (P < 0.001) at baseline in a univariate model in the moderate-risk group. In the high-risk group, there was a significant association between the FPIR–to–HOMA-IR ratio and time to type 1 diabetes (P = 0.002) and a marginally significant association between HOMA-IR and time to type 1 diabetes (P = 0.058). Time to type 1 diabetes was also shown to be associated with age, ICA titers, IAA titers, ICA512 positivity, A1C, and fasting insulin in the moderate-risk group and with age, ICA titers, IAA titers, ICA512 positivity, impaired glucose tolerance, A1C, and peak C-peptide in the high-risk group.

From the multivariate analysis results (Table 3), progression to type 1 diabetes was strongly associated with HOMA-IR and FPIR in both risk groups (model 1). In addition, the baseline ratio of FPIR to HOMA-IR was independently associated with progression to type 1 diabetes (model 2). The HRs of HOMA-IR were 2.70 (95% CI 1.45–5.06) in the moderate-risk group and 1.83 (1.19–2.82) in the high-risk group. This indicates that increased insulin resistance increases the risk for type 1 diabetes. The HRs for the FPIR–to–HOMA-IR ratio were 0.32 (0.18–0.57) in the moderate-risk group and 0.56 (0.40–0.78) in the high-risk group. When the ratio of insulin secretion to insulin resistance decreased, the risk of type 1 diabetes increased (online appendix Fig. 1 [available at http://dx.doi.org/10.2337/dc06-2389]).

For the progressors, we compared relative changes from the baseline to the last IVGTT test before diabetes onset. In both groups, the average insulin resistance increased during the study course; meanwhile, the ratio of FPIR to HOMA-IR decreased. In the moderate-risk group, the average ± SD HOMA-IR of the progressors was 5.06 ± 4.25 from the last IVGTT test, compared with 3.79 ± 2.69 at baseline (P = 0.07), and the FPIR–to–HOMA-IR ratio was 26.35 ± 19.84 versus 44.28 ± 19.20 with P < 0.001 using a paired t test. In the high-risk group, the HOMA-IR of the progressors was 3.08 ± 1.61 from the last IVGTT test, compared with 2.60 ± 1.34 at baseline (P < 0.001), and the FPIR–to–HOMA-IR ratio was 18.37 ± 12.09 versus 28.91 ± 14.59 with P < 0.001.

HOMA-IR and fasting insulin were significantly and highly correlated in both risk groups (r = 0.98, P < 0.001). HOMA-IR was significantly correlated with fasting glucose as might be expected (moderate-risk group r = 0.51, P < 0.001; high-risk group r = 0.59, P < 0.001). The correlations between HOMA-IR and FPIR was 0.56 (P < 0.001) in the moderate-risk group and 0.37 (P < 0.001) in the high-risk group. In the moderate-risk group, the correlations between A1C, peak C-peptide, and HOMA-IR were significant with r = 0.27 (P < 0.001) and r = 0.50 (P < 0.001), respectively. In the high-risk group, the correlations between A1C, peak C-peptide, impaired glucose tolerance, and HOMA-IR were significant with r = 0.32 (P < 0.001), r = 0.30 (P < 0.001), and r = 0.28 (P = 0.002), respectively. Because of these correlations, we added fasting insulin and fasting glucose into a multivariate Cox proportional hazards model together with HOMA-IR, FPIR, and the FPIR–to–HOMA-IR ratio to test their independent contribution to type 1 diabetes risk. A forward selection model showed the same risk factors as model 2 (Table 3). Thus, the combination of FPIR and HOMA-IR, measured as the FPIR–to–HOMA-IR ratio, was a better indicator than FPIR and HOMA-IR themselves (or the components, fasting insulin and glucose), suggesting that there was an interaction effect of FPIR and HOMA-IR.

Differing degrees of insulin sensitivity can be accommodated by altering insulin secretion to maintain normal blood glucose concentrations for those who have normal β-cell function (18). Defects in insulin secretion resulting from β-cell death and dysfunction characterize type 1 diabetes. However, the state of glucose tolerance is determined by the physiological relationship between insulin secretion and insulin resistance, which can also influence the progression to clinical disease. The regulatory mechanism governing hepatic glucose output and insulin secretion suggests that increasing insulin resistance generally results in increased insulin secretion (19,20). A 1996 study showed that normoglycemic ICA⁺ first-degree relatives of type 1 diabetic patients became more insulin resistant and had a decrease in intravenous glucose tolerance after 2 weeks of nicotinamide treatment and that these effects were associated with the expected increase in insulin secretion (21). It is postulated that the increased incidence of clinical type 1 diabetes during puberty and the anecdotal reports of a precedent illness could be attributed to the influence of relative insulin resistance during these time periods. Insulin resistance during the development of type 1 diabetes may be related to counter-regulatory hormones such as those associated with puberty (22,23).

Three longitudinal studies have been conducted to evaluate the risk of insulin resistance downregulating insulin action and leading to glucose abnormalities in the progression of subclinical type 1 diabetes. In the Melbourne Pre-Diabetes Family Study, insulin resistance measured by HOMA-IR in relation to secretion assessed by FPIR in IAA⁺ individuals accounted for increased risk for progression to clinical diabetes. Similar results were observed in the Childhood Diabetes in Finland Study. In contrast, using estimates of insulin resistance derived from a modification of the Bergman minimal model from the frequently sampled IVGTT, the Seattle Family Study showed that such measurements of insulin resistance did not improve the prediction of clinical disease. In the DPT-1, we followed two different populations with a 25–50% or >50% projected risk of developing type 1 diabetes within 5 years. In this study, we found a significant association between HOMA-IR, the ratio of FPIR to HOMA-IR, and the onset of diabetes in both risk groups after adjustments for other risk factors. In addition, the subjects who developed type 1 diabetes in both groups showed an increased HOMA-IR and decreased FPIR–to–HOMA-IR ratio at the time of diagnosis, relative to their baseline values. These findings confirmed earlier studies (13,14) using nearly the same methodology for measuring insulin resistance in a larger well-defined multi-institutional cohort accessioned across the U.S. and Canada. The one study that did not show the same association between insulin resistance and progression to type 1 diabetes used a different methodology, which might suggest an explanation for its findings.

We note, however, that Wallace et al. (24) suggested that insulin resistance measured by HOMA-IR, as was done in this study, may underestimate actual insulin resistance values. Despite this possible underestimation of actual resistance, HOMA-IR and the ratio of FPIR–to–HOMA-IR are shown to be useful in predicting the risk of type 1 diabetes in this longitudinal study. They confirm the importance of the ratio of insulin secretion after an IVGTT to the product of the fasting insulin and glucose levels. The findings also support the prognostic significance of the baseline A1C values in both risk groups.

In summary, using a multivariate Cox proportional hazards model in our studies, we have found clear evidence of an association between insulin resistance and progression to type 1 diabetes. The FPIR–to–HOMA-IR ratio, which is a measure of the combination of insulin resistance and insulin secretion, has also been found to be a better risk predictor than either measure used separately. With a better understanding of the metabolic mechanism of glucose production and uptake for this disease, effective intervention therapies may be developed based on targeting both insulin secretion and insulin sensitivity in the preclinical period of type 1 diabetes.

This work was sponsored by cooperative agreements with the Division of Diabetes, Endocrinology and Metabolic Diseases of the National Institute of Diabetes and Digestive and Kidney Diseases, the National Institute of Allergy and Infectious Diseases, the National Institute of Child Health and Human Development, the National Center for Research Resources, the American Diabetes Association, and the Juvenile Diabetes Research Foundation. Supplies were provided by Eli Lilly, Bayer, Becton Dickinson, International Technidyne, LifeScan, Mead Johnson Nutritionals Division of Bristol-Myers Squibb, Medisense Division of Abbott Laboratories, MiniMed, and Roche Diagnostics.