Abstract
Context
The oral minimal model is a widely accepted noninvasive tool to quantify both β-cell responsiveness and insulin sensitivity (SI) from glucose, C-peptide, and insulin concentrations during a 3-hour 9-point oral glucose tolerance test (OGTT).
Objective
Here, we aimed to validate a 2-hour 7-point protocol against the 3-hour OGTT and to test how variation in early sampling frequency impacts estimates of β-cell responsiveness and SI.
Methods
We conducted a secondary analysis on 15 lean youth with stage 1 type 1 diabetes (T1D; ≥ 2 islet autoantibodies with no dysglycemia) who underwent a 3-hour 9-point OGTT. The oral minimal model was used to quantitate β-cell responsiveness (φtotal) and insulin sensitivity (SI), allowing assessment of β-cell function by the disposition index (DI = φtotal × SI). Seven- and 5-point 2-hour OGTT protocols were tested against the 3-hour 9-point gold standard to determine agreement between estimates of φtotal and its dynamic and static components, SI, and DI across different sampling strategies.
Results
The 2-hour estimates for the disposition index exhibited a strong correlation with 3-hour measures (r = 0.975; P < .001) with similar results for β-cell responsiveness and SI (r = 0.997 and r = 0.982; P < .001, respectively). The agreement of the 3 estimates between the 7-point 2-hour and 9-point 3-hour protocols fell within the 95% CI on the Bland-Altman grid with a median difference of 16.9% (−35.3 to 32.5), 0.2% (−0.6 to 1.3), and 14.9% (−1.4 to 28.3) for DI, φtotal, and SI. Conversely, the 5-point protocol did not provide reliable estimates of φ dynamic and static components.
Conclusion
The 2-hour 7-point OGTT is reliable in individuals with stage 1 T1D for assessment of β-cell responsiveness, SI, and DI. Incorporation of these analyses into current 2-hour diabetes staging and monitoring OGTTs offers the potential to more accurately quantify risk of progression in the early stages of T1D.
Keywords: prediabetes, type 1 diabetes, islet autoimmunity, oral minimal model, insulin sensitivity, β-cell function
The pathologic process that causes type 1 diabetes (T1D) begins long before the disease becomes clinically apparent, progressing through predictable and identifiable stages (1). The rate of progression, however, is variable. Epidemiologic data have yielded population estimates for progression risk based on the presence and pattern of diabetes autoantibodies, combined with results obtained from a standard 2-hour oral glucose tolerance test (OGTT). While younger age is in general associated with faster disease progression, one cannot yet determine on an individual basis how rapidly someone will develop clinical diabetes requiring insulin therapy (stage 3 T1D) once autoantibodies are present. Current risk assessment methods rely heavily on measures of β-cell function. Reduced insulin sensitivity (SI) is also a prominent feature of T1D (2), but it is more difficult to measure. Understanding the full metabolic picture of diabetes in its early stages requires simultaneous measurement of both processes. A more complete knowledge of the functional impairments that underlie progression to stage 3 diabetes is necessary to identify those at risk of rapid progression and to optimize prevention strategies.
The 3-hour oral minimal model is a powerful noninvasive research tool to assess both β-cell responsiveness and SI from glucose, C-peptide, and insulin concentrations during the OGTT (3). It uses 2 different models, one relying on glucose and C-peptide levels to estimate insulin secretion (β-cell responsiveness), and one relying on glucose and insulin concentrations to estimate SI. We previously reported using this method to study children and adolescents with stage 1 T1D (≥ 2 diabetes autoantibodies and normal glucose tolerance) who were part of the National Institutes of Health–funded TrialNet Pathway to Prevention study (4). Even at this early stage of diabetes, we were able to identify defects in β-cell responsiveness, SI, and insulin clearance. Such data may help inform novel investigational paths aimed at targeting both β-cell responsiveness and sensitivity for future diabetes-prevention strategies. A major downside of this method is that it involves a 3-hour, 9-sample OGTT protocol (5), which may be too long to be practical in young children (3, 6-8). Although extended OGTTs provide detailed information regarding the metabolic phenotype of at-risk individuals, the attrition of participants in OGTT-based screening programs approaches approximately 60% over time even in those with stage 1 T1D (9). Strategies to minimize the burden of the metabolic assessment are expected to affect the success of screening program in at-risk populations.
The test length and sampling frequency are dependent on the specific characteristics of the oral minimal model methodology used for describing the dynamics of glucose and insulin metabolism. A short sampling minimal model protocol (2-hour vs 3-hour) was found to be able to accurately identify model indices in a population of healthy adults with normal glucose tolerance (10). However, there are no available data on the validity of a shorter sampling protocol in the youth with normal glucose tolerance that represent the target population of most T1D screening programs. This young group would benefit from a shorter protocol with respect to the study length and the sampling frequency.
Herein we present a secondary analysis conducted on a cohort of youth with stage 1 T1D to test the accuracy of a reduced 2-hour protocol with a 7- or 5-sample OGTT against the gold-standard 3-hour 9 sample OGTT. If a 2-hour OGTT with a 5- or 7-sample protocol yields reliable minimal model results, it would be feasible to incorporate this into the usual OGTT screening and monitoring that is performed in children in the early stages of T1D, thus providing substantial additional information with minimal extra effort or cost.
Materials and Methods
Participants
The study took place at the University of Minnesota (UMN) and the Indiana University School of Medicine (IU). Participants were relatives of individuals with T1D who were screened for diabetes autoantibodies as part of the Type 1 Diabetes TrialNet Pathway to Prevention study, as previously described (4). Those with 2 or more autoantibodies were invited, per TrialNet protocol, to have a staging OGTT and then to be followed every 6 months with monitoring OGTTs. The TrialNet OGTT protocol specifies that individuals be excluded who have previous or current use of medications for the control of hyperglycemia or diabetes, who are currently using immunosuppressive or immunomodulatory therapies (including systemic steroids), or who have known severe active diseases and/or diseases that are likely to limit life expectancy or lead to the use of chronic immunosuppressive or immunomodulatory therapies. Individuals younger than 18 years who were scheduled for routine TrialNet OGTTs were invited to participate in the study. They would ordinarily have had a 2-hour OGTT performed, and they and their parents consented/assented to extend the OGTT to 3 hours and add extra blood samples. The study protocol was approved by the human subjects committee at each site. After obtaining the OGTT results, participants were included in the present data analysis if they had normal fasting and 2-hour glucose levels and no glucose level greater than or equal to 200 mg/dL during the OGTT.
Fourteen healthy control individuals were selected from the Yale Pediatric Diabetes Obesity Clinic repository, among the children without obesity and dysglycemia. A detailed description of this cohort is provided elsewhere (4).
We previously published the 3-hour oral minimal model estimates of insulin secretion, SI, and disposition index (DI) in 14 individuals from the TrialNet cohort compared to normal controls (4). The present report represents a secondary analysis of the data, and includes an additional participant for whom there was no suitable control in the original study, as the secondary analysis did not require matched controls.
Oral Glucose Tolerance Test
Participants underwent testing after an overnight fast. Blood samples were drawn through a peripheral intravenous line. Flavored dextrose in a dose of 1.75 g per kilogram of body weight (up to a maximum of 75 g) was given orally with time 0 set as immediately before the onset of beverage consumption. Blood samples were obtained at −10, 0, 10, 20, 30, 60, 90, 120, 150, and 180 minutes for plasma measurement of glucose, insulin, and C-peptide. Baseline values were computed as the average of −10 and 0 minutes samples.
Insulin was measured by radioimmunoassay (Linco assay, MO, USA) at the UMN and IU sites and Millipore (MA, USA - RRID:AB_2801577) at the Yale site. The assays had less than 1% cross-reactivity with C-peptide and proinsulin. The intra-assay coefficient of variation was 1.5% for a control sample with a mean insulin of 164.4 ± 2.44 μU/mL, 1.4% for a control sample with mean insulin of 92.2 ± 1.32 μU/mL, and 2.3% for a control sample with a mean insulin of 12.3 ± 0.28 μU/mL. The interassay precision coefficient of variation was 2.5% for a control sample with a mean insulin of 165.5 ± 4.15 μU/mL, 2.1% for a control sample with mean insulin of 93.0 ± 1.94 μU/mL, and 4.4% for a control sample with a mean insulin of 12.4 ± 0.54 μU/mL (11). The Millipore assay displayed similar performance as per the manufacturer booklet (RRID:AB_2801577).
C-peptide levels were determined with an immunoenzymometric assay from Diagnostic Product (Tosoh Bioscience, CA, USA) at the UMN and IU sites and Millipore (MA,USA - RRID:AB_2891151) at Yale. The interassay coefficient of variation for the C-peptide for the Diagnostic Product assay was 2.0% in a control sample with a mean C-peptide of 5.23 ± 0.11 ng/mL, and 2.8% in a control sample with a mean C-peptide of 1.71 ± 0.05 ng/mL. The interassay precision coefficient of variation was 3.1% in a control sample with a mean C-peptide of 5.06 ± 0.16 ng/mL, and 2.6% in a control sample with a mean C-peptide of 1.65 ± 0.04 ng/mL (11). The Millipore assay displayed similar performance as per the manufacturer booklet (RRID:AB_2891151).
Oral Glucose Tolerance Test-derived Measures of β-Cell Responsiveness, Insulin Sensitivity, and β-Cell Function
The oral minimal model method (3) was used to estimate SI, β-cell responsiveness (φtotal), and β-cell function with respect to the prevailing SI (DI). The 3-hour model has been previously validated in adults, children, and adolescents against model-independent measurements using multiple-tracer meal protocols and euglycemic and hyperglycemic clamps (5, 12). The physiology underpinnings of the oral minimal model are outlined in Fig. 1 (7, 13).
Figure 1.
Outline of the physiology of the oral glucose models. Continuous lines represent fluxes, dotted lines regulatory signals. Panel A demonstrates the C-peptide minimal model of β-cell responsiveness, or insulin secretion in response to glucose. The figure shows the dynamic component, which results in immediate release of stored insulin vesicles in response to the rate of glucose change, and the static component, which represents a later response to increased glucose levels involving the production of a new pool of insulin. Panel B represents insulin sensitivity, including uptake of glucose by the liver and skeletal muscle and adipose tissue, and suppression of hepatic glucose production, as measured by the glucose-insulin minimal model. The relation between glucose and C-peptide levels over time in the static (red circles) and dynamic (blue circles) components of β-cell responsiveness are shown in panel C, the time curves below the figure, while the relation between glucose and insulin levels is displayed in panel D. Panel E describes the hyperbolic relationship between insulin secretion and sensitivity and suggests the role of insulin clearance in a 3-component frame. The resulting disposition index (DI)—namely the ability to dispose of a glucose load at given levels of insulin secretion capacity and insulin sensitivity—results from the relative changes of these components: For example, a transient reduction of insulin sensitivity can be compensated for by reduced insulin clearance and/or by increased insulin secretion to prevent dysglycemia. The failure of one or more components of such a model results in deterioration from normal glycemia (green curve) to dysglycemia (yellow curve) and diabetes (red curve).
Insulin sensitivity
SI was quantified from the OGTT plasma glucose and insulin concentrations using the oral glucose minimal model (3), which measures the overall effect of insulin on stimulating glucose disposal and inhibiting hepatic glucose production. A lower number represents increased insulin resistance. β-Cell responsiveness (φtotal) was quantified from the OGTT plasma glucose and C-peptide concentrations using the oral C-peptide minimal model. The model assumes that glucose-stimulated insulin secretion is made up of two components: a dynamic component, representing the secretion of readily-releasable insulin that is stimulated by the rate of increase in glucose concentration (φdynamic), and a static component, which measures new insulin production in response to a given increment in glucose above basal concentrations (φstatic). These two components do not exactly mirror the classic intravenous GTT first- and second-phase insulin secretion parameters because they partially overlap. From φdynamic and φstatic, a total β-cell responsiveness index (φtotal) is derived, which measures the overall ability of the β-cell to respond to a glucose stimulus (3, 14, 15). A higher number suggests greater insulin secretion in response to a rise in glucose. The DI is calculated as the product of φ total and SI and represents the appropriateness of the β-cell response relative to the level of insulin sensitivity.
Insulin sensitivity
Insulin clearance (CL) was estimated as the ratio of the area under the curve (AUC) of insulin secretion rate (ISR) over the plasma insulin AUC (ISRAUC/IAUC) (16, 17), The ISR was calculated by C-peptide deconvolution (14, 18).
Islet autoantibody assays
The radiobinding assays for autoantibodies to insulin (19), glutamic acid decarboxylase (20), protein tyrosine phosphatase (20), and zinc transporter 8 (21) were performed at the Barbara Davis Center for Diabetes, University of Colorado, from samples collected at UMN and IU using previously published assay methods. Islet autoantibodies from the Yale cohort were performed at the Yale Center for Clinical Investigation core laboratory facility, which uses the same methodology described earlier.
Statistical Analysis
The primary outcome of the study was the comparison of DI estimated using a 3-hour 9-point OGTT vs the same index using a 2-hour 7-point test. Secondary analyses included the comparison of individual components of DI (φtotal, φstatic, φdynamic, and SI) as computed from a 3-hour vs a 2-hour test. For the 2-hour test, an additional analysis was conducted to quantify the effect of reduced frequency of sampling (traditional every 30 minute 5-point OGTT sampling) vs 7-point sampling by removing the 10-minute and 20-minute time points from the computational procedure. For the 2-hour model computational process, we assumed a spontaneous return to baseline of glucose and insulin after 4 hours for the estimate of SI regardless of the actual 3-hour values. This approach has been previously adopted (22).
The association between the indices computed with the 2-hour vs 3-hour protocols, as well as between the 7-sample and 5-sample models, was quantified with a linear regression analysis. Nonnormally distributed variables were naturally log-transformed before the analysis. Continuous variables were examined for departure from normality in distribution using histograms and descriptive indicators such as skewness and kurtosis.
The Bland-Altman plot is a method for analyzing agreement between 2 different measurements with a 95% CI (±1.96 SD). The x-axis shows the actual mean of the 2-hour and 3-hour index values for each individual, while the y-axis shows the difference between the 2 paired measurements plotted against their mean. A wider point distribution spread along the y-axis corresponds to greater “bias,” that is, lower agreement. If 95% of the differences on the y-axis fall between −1.96 and +1.96 SDs of the mean, the differences are considered to be in good agreement.
An exploratory analysis was conducted to describe the distribution of both the 3-hour and the 2-hour model-derived indices (DI, φtotal and SI) and islet autoantibody prevalence. Analysis of variance tests were used to compare model-derived indices among those without islet autoimmunity, with 2 to 3 antibodies and with more than 3 antibodies. Similarly, the AUC for C-peptide (AUCc-peptide) and glucose (AUCc-peptide), and the Index60 were calculated for the 3 groups. Index60 was computed as 0.3695 (log fasting C-peptide) + 0.0165 × (60-minute glucose) − 0.3644 × (60-minute C-peptide) as previously described (23). The analysis was meant to provide the variability of each index within the study population with respect to a clinically relevant outcome, as the number of islet autoantibodies is known to be associated with the risk of progression to stage 3 T1D (24).
Results are shown as mean ± SD or median (25th-75th percentile) as appropriate. The oral minimal model was numerically identified by nonlinear least squares, as implemented in SAAM II v.2.3 (The Epsilon Group 2012-2013). Analyses were performed using STATA.13 software (StataCorp. 2013. Stata Statistical Software: Release 13. StataCorp LP) and Prism 8.0 (GraphPad Software).
Results
Participants
Baseline subject characteristics are presented in Table 1. Of the 15 cases included in this analysis, 4 were positive for 2 antibodies, and 11 had 3 or more autoantibodies. Nine individuals were positive for anti-insulin antibodies. Only one participant was diagnosed with diabetes during 12 months of follow-up after the study OGTT. OGTT glucose, insulin, and C-peptide levels have been previously presented for 14 of the participants (4). Briefly, despite normal glucose tolerance, they had modestly but statistically significantly higher mid-OGTT glucose excursion compared to controls and lower C-peptide excursion. The one additional patient added to the present report followed a similar pattern (Fig. 2). The 14 antibody-negative healthy controls did not differ from the cases with respect to metabolic and anthropometric characteristics, except for a lower 1-hour glucose level (P = .034).
Table 1.
Participant characteristics
Cases (n = 15) | Controls (n = 14) | |
---|---|---|
Age, y | 11 (8.5 to 14.9) | 11.5 (10.4 to 14.9) |
BMI | 19.2 (16.5 to 23.8) | 19.6 (18.4 to 20.4) |
z score BMI | 0.75 (−0.07 to 1.25) | 0.80 (−0.07 to 0.90) |
Sex (F) n (%) | 7 (47) | 6 (43) |
Fasting glucose, mg/dL | 92 (89 to 100) | 88 (84 to 89.5) |
1-h glucose, mg/dL | 151 (125 to 175) | 129.5 (103 to 132) |
2-h glucose, mg/dL | 111 (93 to 130) | 102.5 (85 to 110) |
Fasting insulin, µU/mL | 13 (8 to 26) | 14 (10.5 to 17.0) |
Fasting C-peptide, pmol/L | 244 (150 to 390) | 682.5 (508 to 790) |
Continuous data are reported as median (25th-75th percentile).
Abbreviation: BMI, body mass index.
Figure 2.
A, Glucose; B, C-peptide; and C, insulin profiles during the 3-hour oral glucose tolerance test of cases (n = 15) and controls (n = 14).
β-Cell Function and its Components in Stage 1 Type 1 Diabetes: 2-Hour vs 3-Hour Protocols
All comparisons of the 2- and 3-hour protocols showed high levels of correlation. Similarly, Bland-Altman bias assessment showed good agreement between the 2 minimal model approaches (Fig. 3). Specifically, for the assessment of DI, excellent correlation was seen between the 2-hour 7 point and the 3-hour 9-point protocols (r = 0.975; P < .001; Fig. 3A). The agreement of DI between the 2 protocols is displayed in Fig. 3B, with all the participants falling within 1.96 SD from the mean difference and exhibiting a median difference of 16.9% (−35.3 to 32.5) between the 2-hour and 3-hour estimates and a Bland-Altman bias −0.163 ± 0.245). The estimates of SI between the 2-hour and the 3-hour protocols were also highly correlated (r = 0.982; P < .001; Fig. 3C), with a median difference of 14.9% (−1.4 to 28.3) between the 2-hour and the 3-hour protocols and a mean Bland-Altman bias of −0.141 ± 0.297 (Fig. 3D). β-cell responsiveness (φtotal) was highly correlated (Fig. 3E), with minimal variation seen between the 2-hour and the 3-hour OGTT (r = 0.997; P < .001) and a median difference of 0.2% (−0.6 to 1.3) with a Bland-Altman bias of −0.020 ± 0.078 (Fig. 3F). The φdynamic and the φstatic components of φtotal exhibited similar agreement between the 2-hour and 3-hour protocols (r = 0.992 and r = 0.994, respectively; both P < .001). CL estimates from the 2- and 3-hour protocols exhibited a high level of agreement (r = 0.998; P < .001; Bland-Altman bias was −4.09 ± 6.44; Fig. 3G and 3H) with a median difference between 2-hour and 3-hour protocols of 8.7% (−14.5 to 13.0).
Figure 3.
Linear regression analysis (left side) and Bland-Altman agreement plots (right side) of mean (x-axis) and difference (y-axis) for 2-hour vs 3-hour estimates of the A and B, disposition index; C and D, insulin sensitivity; E and F, β-cell responsivity; and G and H, insulin clearance.
Healthy controls
Similar results were obtained for the 14 autoantibody-negative normal control individuals. DI, SI, φtotal and CL during the 3-hour vs 2-hour protocols demonstrated a high level of correlation: DI: r = 0.72; P < .001; SI: r = 0.98; P = .010; φtotal: r = 0.89; P < .001; and CL: r = 0.88; P < .001. The median differences between the 2-hour and the 3-hour protocols were −7.8% (−32 to 23); −11.2% (−28.1 to 9.4); −5.3% (−11.0 to 5.4) for DI, SI, and φtotal respectively, and the median difference for CL was 7.2% (−12.5 to 13.5).
Two-Hour Minimal Model Protocol: 5- vs 7-Point Blood Sampling
The standard oral C-peptide minimal model estimate of β-cell function incorporates early measures of glucose, insulin, and C-peptide, with levels obtained at 10 and 20 minutes. These early levels have not been historically included in OGTTs performed in large T1D clinical networks, where the first blood sample after time 0 is generally at 30 minutes. Herein we tested the effect of omitting the 10 and 20 minute samples on the 2-hour OGTT minimal model estimates. The secretion model illustrated in Fig. 4A displays an estimate of the ISR during the test. It shows that the absence of 10- and 20-minute time points resulted in a delay of the estimated insulin secretion peak. The absence of 10- and 20-minute time points in the minimal model assessment led to a significant underestimate of total insulin secretion following the oral load when quantified as AUCISR (2-hour AUCISR 17 395 ± 737.2 vs 16 819 ± 788 pmol × min/L for the 7-point and 5-point OGTT, respectively [P = .048]). This occurred primarily in the first 60 minutes of the OGTT.
Figure 4.
A, The insulin secretion rate based on oral minimal model C-peptide and glucose measures during the 5-point (red line) and the 7-point (blue line) 2-hour protocols. Linear regression analysis for 5-point 2-hour vs 7-point 2-hour estimates of the B, φtotal; C,φdynamic; and D, φstatic; E, insulin sensitivity; and F, disposition index.
In contrast to the excellent correlation and agreement seen between the 3-hour 9-point OGTT and the 2-hour 7-point OGTT minimal model approaches, the absence of the 10- and 20-minute time points in the 2-hour 5-point minimal model assessment resulted in reduced precision of estimated βcell responsiveness (φtotalr = 0.032; P = .909; φstaticr = 0.057; P = 0.840, and φdynamicr = 0.105; P = .905; Fig. 4B-4D), with a low agreement demonstrated by Bland-Altman analysis (bias 31.6 ± 56.5 for φtotal). In contrast to estimates of insulin secretion, estimated SI (r = 0.998; P < .001; Fig. 4E) was not negatively affected by the absence of the 10- and 20-minute time points, and there was limited effect on the overall DI estimate (r = 0.897; P < .001; Fig. 4F) and on CL (r = 0.975; P < .001).
Healthy controls
The agreement between the 5-point and 7-point protocols in the diabetes autoantibody-negative control cohort was high for SI (r = 0.99; P < .001), φtotal (r = 0.97; P < 0.001), and DI (r = 0.98; P < .001) as well as for the static component of insulin secretion (φstatic, r = 0.979; P < .001) and CL (r = 0.945; P < .001). However, the reduced 5-point protocol failed to reproduce the dynamic component (φdynamicr = 0.014; P = .959) in normal control participants. Thus, with the exception of the dynamic component of insulin secretion, oral minimal model analysis using data from either the traditional 5-point OGTT or the 7-point OGTT mostly performed equally well in normal controls, highlighting the ability of the 7-point study to identify early differences in individuals whose β-cells are under autoimmune attack as demonstrated by 2 or more diabetes autoantibodies.
β-Cell Function and Islet Autoantibody Number
This was an exploratory analysis because of the small number of participants. While β-cell responsiveness (φtotal) did not differ between the control group and the 4 individuals with 2 to 3 islet autoantibodies, it was approximately 60% lower in the group of 11 individuals with 3 or more autoantibodies (Fig. 5A; P = .015). Conversely, approximately 50% lower SI was observed in the 2 to 3 islet autoantibodies group with respect to the healthy controls (P = .015), with even lower SI in the group with more than 3 antibodies (P = .011; Fig. 5B). The DI (Fig. 5C) confirmed this pattern with a DI that was about 80% lower in both autoantibody positive groups compared to healthy controls (P < .048). Other commonly used indices of β-cell function (AUC C-peptide), metabolic control (AUC glucose), and T1D risk for progression (Index60) did not differentiate the 3 groups at this stage, as shown in Fig. 5D to 5F. Similar results were obtained with the 2-hour 7-point OGTT protocol.
Figure 5.
A, β-Cell responsiveness (φtotal); B, insulin sensitivity (SI); C, disposition index (DI); D, area under the curve (AUC)-C-peptide; E, AUC glucose; and F, Index60 for the healthy controls (blue), those with 2 to 3 antibodies (red) and those with 4 to 5 antibodies (green). Data are presented as median, interquartile range (25th-75th percentile).
Discussion
Oral minimal model predictions of rapidly declining β-cell function are particularly relevant in the pediatric population because diabetes risk is increased. In addition, the model incorporates changes both in insulin secretion and SI, the latter of which is highly variable in youth and may play an important role in T1D progression. Recently, we used the 3-hour OGTT minimal model in 14 youth from the TrialNet study who were in the earliest stage of diabetes, stage 1, to report novel observations (4). Compared to healthy controls, these children and adolescents had a lower DI due to simultaneous reduction of total body SI and insulin secretion in response to oral glucose. They also had lower baseline and postload CL, and higher hepatic insulin resistance. Incorporating these minimal model-derived metabolic parameters into methods estimating the rate of progression to insulin-requiring clinical diabetes might substantially improve disease prediction efforts. However, the 3-hour OGTT test is inconvenient for patients and staff, and would require a change in current monitoring methods. In this data reanalysis, we now show for the first time that compared to the 3-hour test, a shorter 2-hour 7-point OGTT allows an equivalent assessment of β-cell responsivity, DI, and SI in children and adolescents with early T1D. We showed similar agreement between the 2-hour and 3-hour protocols in a cohort of autoantibody-negative healthy controls.
When comparing the 2- and 3-hour OGTT minimal model tests, we observed a difference of approximately 20% for the DI, which is not clinically relevant, as we previously demonstrated that changes in glucose tolerance status and time to glucose peak during the OGTT are associated with a 50% DI variation in youth with obesity (25, 26), while the median DI difference between children with stage 1 T1D and their peers without islet autoantibodies is greater than 80% (4). Similarly, for the sensitivity index and β-cell responsiveness, the observed differences are not expected to be clinically meaningful as the median approximately 15% difference for SI between the 2-hour and 3-hour protocols is substantially lower than the 30% to 60% changes in SI described during transition from normal to impaired glucose tolerance in youth using the oral minimal model or other surrogate estimates (27-31). Notably, the observed differences have been shown to become even less relevant if the minimal model methodology is used consistently in longitudinal cohorts, such as in the context of clinical trials testing the effect of medications targeting one of the components of β-cell function, including insulin sensitivity and β-cell function (25, 32).
In the context of the reduced 2-hour OGTT, we tested whether including intermediate 10- and 20-minute time points was necessary, or if values obtained at the traditional 0-, 30-, 60-, 90-, and 120-minute times would suffice for a 2-hour oral minimal model assessment in individuals with stage 1 T1D. The absence of the 10- and 20-minute time points negatively affected the accuracy of the estimates of the early components of insulin secretion and overall β-cell responsiveness. In contrast, SI and CL estimates did not change when the early time points were omitted. The DI was only modestly affected by reduced sampling, suggesting that the small but significant changes in early insulin secretion detected by the 2-hour 7-point method were masked by unchanged estimates of SI. Thus, in participants with stage 1 T1D, the standard 5-point OGTT failed to discern defects in β-cell responsiveness. Identification of these earliest changes may prove to be important in diabetes prediction and prevention trials. Interestingly, the standard 5-point OGTT performed better in normal control participants, highlighting the role of insulin secretory defects in the first 30 minutes following a glucose load as one of the first signs of β-cell dysfunction.
In a novel exploratory analysis, we assessed the relation between the number of autoantibodies and minimal-model indices. Clinically meaningful changes in these indices ranged from 50% to 80% when healthy controls were compared to those carrying either 2 to 3 islet autoantibodies or 4 or more autoantibodies. Even in this small cohort, the DI (mainly driven by lower SI) was able to differentiate those with 2 to 3 islet autoantibodies from healthy controls, while commonly used indices of β-cell function (AUC 3-hour C-peptide), glycemic control (AUC 3-hour glucose), and risk of diabetes progression (Index60), were not able to distinguish the 2 groups. These analyses were limited by low participant numbers, but the fact that differences were seen in this small cohort suggests that the DI may be one of the earliest indicators of β-cell stress.
T1D is continuum that progresses through identifiable stages, starting with the presence of 2 or more diabetes-related autoantibodies (1). An OGTT is then required for staging. In stage 1 disease, the OGTT is normal. Positive antibodies serve as a marker that an immune attack on pancreatic β cells has begun, but β-cell mass is still sufficient to maintain normal glucose levels. In stage 2, glucose tolerance has started to deteriorate but has not yet reached the glucose thresholds used to define diabetes. These individuals are clinically asymptomatic despite OGTT abnormalities. Stage 3 is what used to be considered the onset of T1D, with fasting glucose levels greater than or equal to 126 mg/dL (7.0 mmol/L) and/or 2-hour OGTT glucose levels greater than or equal to 200 mg/dL (11.1 mmol/L) and classic symptoms of diabetes.
While the progression from seroconversion to a stage 3 diabetes diagnosis is much faster in youth than in adults, the individual time interval between the appearance of autoantibodies and progression to clinical stage 3 diabetes is quite variable—it may be less than 1 year or may be 15 or more years (24). C-peptide levels tend to drift downward slowly until about 6 months before the onset of stage 3 diabetes, at which point there is a change in slope and the rate of C-peptide decline sharply accelerates (33). Unfortunately, because it happens over a relatively short time interval, this change in slope can usually be detected only retrospectively. Current methodology is not able to prospectively identify individuals who are approaching this phase of rapid disease progression.
Identifying clinically asymptomatic individuals who are in the early stages of T1D, and especially those who are progressing more rapidly or headed toward rapid progression, is important both to reduce the risk that the disease will subsequently present as life-threatening ketoacidosis, and to identify the most likely candidates for participation in efforts attempting slow disease progression. In addition to screening through the TrialNet network, which screens approximately 15 000 relatives of people with T1D per year, several groups have established screening programs for the general population. In particular, the new JDRF-funded T1Detect program is expected to greatly expand T1D screening in the US population. Individuals with early diabetes may be eligible for intervention studies or treatments designed to slow or halt β-cell loss. For example, TrialNet demonstrated that the drug teplizumab can delay progression from stage 2 to stage 3 diabetes in individuals aged 8 years and older (34), and the Food and Drug Administration has approved the use of this drug for that population. The potential benefit of slowing progression to insulin-requiring stage 3 disease is clear, particularly in children. But there are also risks associated with these therapies, most of which involve immune system manipulation. The risk-benefit ratio of early T1D treatments depends in part on how rapidly a patient is likely to progress to overt clinical disease. Identifying those who are at greatest risk for rapid progression is challenging, even within high-risk populations.
Research studies such as TrialNet have sought methods to better quantify risk. The Index60 and the Diabetes Prevention Trial Risk Score (DPTRS) are examples of models that use OGTT glucose and C-peptide levels to better define rapidly deteriorating β-cell function (35, 36). Index60 was developed from a proportional-hazards model and combines the log fasting C-peptide level with the 60-minute glucose and 60-minute C-peptide levels from OGTTs. The DPTRS was derived from a proportional-hazards model that includes log fasting C-peptide, log body mass index (BMI), age, and the C-peptide sum and glucose sum from 30, 60, 90, and 120 minutes. These models assess declining β-cell function without considering SI or CL, providing a limited picture of the overall metabolic state.
The gold-standard techniques for quantifying insulin secretion and SI are hyperglycemic and euglycemic insulin clamps, respectively. However, these are labor intensive, difficult to apply in clinical practice or in large epidemiological studies, and challenging to perform in children. The intravenous minimal model method, first published in 1979 (37), employs a mathematical model to independently measure insulin secretion and action and to estimate the β-cell function as the result of insulin secretion in the context of whole-body SI (3). The test requires a skilled research center and is expensive and personnel intensive due to the frequency and large number of required blood samples. The oral minimal model uses data from a 3-hour OGTT and has been successfully validated in adults, children, and adolescents using multiple-tracer meal protocols and euglycemic and hyperglycemic clamps (5, 12). It is relatively inexpensive and does not require a large number of samples or a lot of blood, but the length of the test is a challenge for children and for busy research centers.
We have now shown that similar findings can be obtained from a 2-hour OGTT minimal model assessment in youth with stage 1 T1D. By simply adding 10- and 20-minute blood samples to the usual every 30-minute sampling, the oral minimal model can be routinely integrated into the standard 2-hour diabetes staging OGTT. For this study, we did not obtain a blood sample at 15 minutes. Potentially, a single 15-minute sample could replace the 10- and 20-minute samples, but this would need to be tested.
The major strength of this study is the availability of a well-defined pediatric population with stage 1 T1D for simultaneous comparison of a 2-hour OGTT minimal model with the standard 3-hour study. The major weakness is the lack of longitudinal assessment and the correlation with clinically relevant variables such as pubertal development or BMI changes. The limited sample size did not allow any inference about the role of ethnic background on the reliability of the model-based estimates. However, the oral minimal model has been previously validated in pediatric and adult multiethnic cohorts (5, 8, 26), thus suggesting that our findings could be extended to different ethnic groups. The estimate of SI through the minimal model may be affected by the length of the OGTT as it relies on the assumption that glucose returns to baseline at the end of the test. Such a limit can be easily circumvented by assuming a delayed return to baseline (eg, after 4 or 6 hours) for those with impaired glucose tolerance or diabetes undergoing a 2-hour test without affecting the final estimates (8, 22, 38). Conversely, β-cell responsiveness is not expected to be affected by changes in glucose tolerance status (3). SI could theoretically be affected by the presence of anti-insulin antibodies; the present study did not have a sufficient number of participants to test this hypothesis.
In summary, with T1D autoantibody screening of the general population becoming increasingly common, it is more critical than ever that methods be in place to help determine the best candidates for intervention, particularly in children as they have the most rapid rate of progression. We have shown that the 2-hour 7-point OGTT is sufficient in patients with stage 1 T1D for accurate minimal model assessment of β-cell responsiveness, SI, and DI. These measures offer the potential to more accurately quantify risk of progression in the early stages of T1D. The ability to obtain these data from a 2-hour OGTT makes it possible to incorporate these analyses into current research on OGTT staging and monitoring programs. Longitudinal studies are necessary to determine the long-term significance of minimal model findings and how they might contribute to overall risk quantification. These data will help inform novel investigational approaches to diabetes prevention, and may eventually affect clinical screening guidelines.
Acknowledgments
We acknowledge TrialNet for providing the infrastructure for this ancillary study and for analyzing the C-peptide levels.
Abbreviations
- AUC
area under the curve
- BMI
body mass index
- CL
insulin clearance
- DI
disposition index
- DPTRS
diabetes prevention trial risk score
- ISR
insulin secretion rate
- IU
Indiana University
- OGTT
oral glucose tolerance test
- Phi (φdynamic)
dynamic component of insulin secretion refers to early secretion of pre-formed insulin from a readily-releasable vesicle pool stimulated by the rate of increase in glucose concentration
- Phi (φstatic)
the static component of insulin secretion consists of new insulin production in response to an increase in glucose above basal concentrations
- Phi (φtotal)
β-cell responsiveness—the overall ability of the β-cell of respond to a glucose stimulus
- SI
insulin sensitivity
- T1D
type 1 diabetes
- UMN
University of Minnesota
Contributor Information
Alfonso Galderisi, Department of Woman and Child's Health, University of Padova, 35128 Padua, Italy.
Carmella Evans-Molina, Center for Diabetes and Metabolic Diseases, Indiana University, Indianapolis, Indiana 46202, USA.
Mariangela Martino, Department of Woman and Child's Health, University of Padova, 35128 Padua, Italy.
Sonia Caprio, Department of Pediatrics, Yale University, New Haven, Connecticut 06520, USA.
Claudio Cobelli, Department of Woman and Child's Health, University of Padova, 35128 Padua, Italy.
Antoinette Moran, Department of Pediatrics, University of Minnesota, Minneapolis, Minnesota 55454, USA.
Financial Support
This work was supported by the National Institutes of Health U01-DK085476 (TrialNet) and CTSA program UL1-TR002494 (to A.M.); the National Institutes of Health NU01-DK085505 (TrialNet), R01-DK-093954, UC4-DK-127786, R21-DK119800, R01DK127308-01, and P30DK097512, VA Merit Award I01BX001733, and JDRF Strategic Research Agreement (to C.E.M.); Institute for Pediatric Research (IRP) (to A.G.); and European Commission HORIZON2020, FORGETDIABETES-FET-EU951933 (to A.G. and C.C.).
Disclosures
The authors have nothing to disclose.
Data Availability
Some or all data sets generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on reasonable request. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.
References
- 1. Insel R, Dunne J, Atkinson M, et al. Staging presymptomatic type 1 diabetes: a scientific statement of JDRF, the Endocrine Society, and the American Diabetes Association. Diabetes Care. 2015;38(10):1964‐1974. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Donga E, Dekkers O, Corssmit E, Romijn J. Insulin resistance in patients with type 1 diabetes assessed by glucose clamp studies: systematic review and meta-analysis. Eur J Endocrinol. 2015;173(1):101‐109. [DOI] [PubMed] [Google Scholar]
- 3. Cobelli C, Dalla Man C, Toffolo G, Basu R, Vella A, Rizza R. The oral minimal model method. Diabetes. 2014;63(4):1203‐1213. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Galderisi A, Moran A, Evans-Molina C, et al. Early impairment of insulin sensitivity, β-cell responsiveness, and insulin clearance in youth with stage 1 type 1 diabetes. J Clin Endocrinol Metab. 2021;106(9):2660‐2669. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Sunehag A, Dalla Man C, Toffolo G, Haymond M, Bier D, Cobelli C. Beta-cell function and insulin sensitivity in adolescents from an OGTT. Obesity (Silver Spring). 2009;17(2):233‐239. [DOI] [PubMed] [Google Scholar]
- 6. Breda E, Cavaghan MK, Toffolo G, Polonsky KS, Cobelli C. Oral glucose tolerance test minimal model indexes of beta-cell function and insulin sensitivity. Diabetes. 2001;50(1):150‐158. [DOI] [PubMed] [Google Scholar]
- 7. Dalla Man C, Caumo A, Basu R, Rizza R, Toffolo G, Cobelli C. Minimal model estimation of glucose absorption and insulin sensitivity from oral test: validation with a tracer method. Am J Physiol Endocrinol Metab. 2004;287(4):E637‐E643. [DOI] [PubMed] [Google Scholar]
- 8. Dalla Man C, Yarasheski KE, Caumo A, et al. Insulin sensitivity by oral glucose minimal models: validation against clamp. Am J Physiol Endocrinol Metab. 2005;289(6):E954‐E959. [DOI] [PubMed] [Google Scholar]
- 9. Driscoll K, Tamura R, Johnson S, et al. ; TEDDY Study Group . Adherence to oral glucose tolerance testing in children in stage 1 of type 1 diabetes: the TEDDY study. Pediatr Diabetes. 2021;22(2):360‐368. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Dalla Man C, Campioni M, Polonsky KS, et al. Two-hour seven-sample oral glucose tolerance test and meal protocol: minimal model assessment of beta-cell responsivity and insulin sensitivity in nondiabetic individuals. Diabetes. 2005;54(11):3265‐3273. [DOI] [PubMed] [Google Scholar]
- 11. Baidal D, Warnock M, Xu P, et al. Oral glucose tolerance test measures of first-phase insulin response and their predictive ability for type 1 diabetes. J Clin Endocrinol Metab. 2022;107(8):e3273‐e3280. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Cali AM, Dalla Man C, Cobelli C, et al. Primary defects in beta-cell function further exacerbated by worsening of insulin resistance mark the development of impaired glucose tolerance in obese adolescents. Diabetes Care. 2009;32(3):456‐461. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Pedersen M, Dalla Man C, Cobelli C. Multiscale modeling of insulin secretion. IEEE Trans Bio-med Eng. 2011;58(10):3020‐3023. [DOI] [PubMed] [Google Scholar]
- 14. Van Cauter E, Mestrez F, Sturis J, Polonsky KS. Estimation of insulin secretion rates from C-peptide levels. Comparison of individual and standard kinetic parameters for C-peptide clearance. Diabetes. 1992;41(3):368‐377. [DOI] [PubMed] [Google Scholar]
- 15. Dalla Man C, Micheletto F, Sathananthan A, Rizza RA, Vella A, Cobelli C. A model of GLP-1 action on insulin secretion in nondiabetic subjects. Am J Physiol Endocrinol Metab. 2010;298(6):E1115‐E1121. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Piccinini F, Bergman R. The measurement of insulin clearance. Diabetes Care. 2020;43(9):2296‐2302. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Trico D, Galderisi A, Mari A, Santoro N, Caprio S. One-hour post-load plasma glucose predicts progression to prediabetes in a multi-ethnic cohort of obese youths. Diabetes Obes Metab. 2019;21(5):1191‐1198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Polonsky KS, Given BD, Hirsch L, et al. Quantitative study of insulin secretion and clearance in normal and obese subjects. J Clin Invest. 1988;81(2):435‐441. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Yu L, Robles D, Abiru N, et al. Early expression of antiinsulin autoantibodies of humans and the NOD mouse: evidence for early determination of subsequent diabetes. Proc Natl Acad Sci USA. 2000;97(4):1701‐1706. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Bonifacio E, Yu L, Williams AK, et al. Harmonization of glutamic acid decarboxylase and islet antigen-2 autoantibody assays for National Institute of Diabetes and Digestive and Kidney Diseases Consortia. J Clin Endocrinol Metab. 2010;95(7):3360‐3367. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Yu L, Boulware D, Beam C, et al. Zinc transporter-8 autoantibodies improve prediction of type 1 diabetes in relatives positive for the standard biochemical autoantibodies. Diabetes Care. 2012;35(6):1213‐1218. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Pacini G, Bergman R. MINMOD: a computer program to calculate insulin sensitivity and pancreatic responsivity from the frequently sampled intravenous glucose tolerance test. Comput Methods Programs Biomed. 1986;23(2):113‐122. [DOI] [PubMed] [Google Scholar]
- 23. Nathan B, Redondo M, Ismail H, et al. Index60 identifies individuals at appreciable risk for stage 3 among an autoantibody-positive population with normal 2-hour glucose levels: implications for current staging criteria of type 1 diabetes. Diabetes Care. 2022;45(2):311‐318. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Ziegler A, Rewers M, Simell O, et al. Seroconversion to multiple islet autoantibodies and risk of progression to diabetes in children. JAMA. 2013;309(23):2473‐2249. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Galderisi A, Tricò D, Dalla Man C, et al. Metabolic and genetic determinants of glucose shape after oral challenge in obese youths: a longitudinal study. J Clin Endocrinol Metab. 2020;105(2):534‐542. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Cropano C, Santoro N, Groop L, et al. The rs7903146 variant in the TCF7L2 gene increases the risk of prediabetes/type 2 diabetes in obese adolescents by impairing β-cell function and hepatic insulin sensitivity. Diabetes Care. 2017;40(8):1082‐1089. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Giannini C, Weiss R, Cali A, et al. Evidence for early defects in insulin sensitivity and secretion before the onset of glucose dysregulation in obese youths: a longitudinal study. Diabetes. 2012;61(3):606‐614. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Saad R, Gungor N, Arslanian S. Progression from normal glucose tolerance to type 2 diabetes in a young girl: longitudinal changes in insulin sensitivity and secretion assessed by the clamp technique and surrogate estimates. Pediatr Diabetes. 2005;6(2):95‐99. [DOI] [PubMed] [Google Scholar]
- 29. Michaliszyn SF, Mari A, Lee S, et al. β-Cell function, incretin effect, and incretin hormones in obese youth along the span of glucose tolerance from normal to prediabetes to type 2 diabetes. Diabetes. 2014;63(11):3846‐3855. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Arslanian S, Suprasongsin C. Differences in the in vivo insulin secretion and sensitivity of healthy black versus white adolescents. J Pediatr. 1996;129(3):440‐443. [DOI] [PubMed] [Google Scholar]
- 31. Dalla Man C, Piccinini F, Basu R, Basu A, Rizza RA, Cobelli C. Modeling hepatic insulin sensitivity during a meal: validation against the euglycemic hyperinsulinemic clamp. Am J Physiol Endocrinol Metab. 2013;304(8):E819‐E825. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Yadav Y, Dunagan K, Khot R, et al. Inhibition of 11β-hydroxysteroid dehydrogenase-1 with AZD4017 in patients with nonalcoholic steatohepatitis or nonalcoholic fatty liver disease: a randomized, double-blind, placebo-controlled, phase II study. Diabetes Obes Metab. 2022;24(5):881‐890. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Bogun M, Bundy B, Goland R, Greenbaum C. C-peptide levels in subjects followed longitudinally before and after type 1 diabetes diagnosis in TrialNet. Diabetes Care. 2020;43(8):1836‐1842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Herold K, Bundy B, Long S, et al. An anti-CD3 antibody, teplizumab, in relatives at risk for type 1 diabetes. N Engl J Med. 2019;381(7):603‐613. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Nathan B, Boulware D, Geyer S, et al. Dysglycemia and Index60 as prediagnostic end points for type 1 diabetes prevention trials. Diabetes Care. 2017;40(11):1494‐1499. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Sosenko JM, Skyler JS, Herold KC, Palmer JP;Type 1 Diabetes TrialNet and Diabetes Prevention Trial–Type 1 Study Groups . The metabolic progression to type 1 diabetes as indicated by serial oral glucose tolerance testing in the Diabetes Prevention Trial–Type 1. Diabetes. 2012;61(6):1331‐1337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Bergman RN, Ider YZ, Bowden CR, Cobelli C. Quantitative estimation of insulin sensitivity. Am J Physiol. 1979;236(6):E667‐E677. [DOI] [PubMed] [Google Scholar]
- 38. Dalla Man C, Camilleri M, Cobelli C. A system model of oral glucose absorption: validation on gold standard data. IEEE Trans Biomed Eng. 2006;53(12):2472‐2478. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Some or all data sets generated during and/or analyzed during the present study are not publicly available but are available from the corresponding author on reasonable request. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.