Abstract
Introduction
This study examined how insulin resistance and impaired insulin secretion are associated with hyperglucagonemia during oral glucose tolerance tests (OGTT) in type 2 diabetes mellitus (T2DM).
Research design and methods
A retrospective analysis included 247 patients with T2DM treated at Tianjin Medical University General Hospital from October 2022 to March 2025. All underwent a 75 g OGTT, with blood samples collected at 0, 0.5, 1, 2, and 3 hours for glucose, insulin, C-peptide, and glucagon measurement. Insulin resistance was assessed via homeostasis model assessment of insulin resistance (HomaIR) and C-peptide immunoreactivity insulin resistance (CPRIR), while insulin secretion was evaluated using C-peptide area under the curve (AUCcp), homeostasis model assessment of beta cell function (HomaB), first-phase and second-phase insulin secretion during OGTT (first PH and second PH). Generalized linear models and mediation analyses examined associations between glucagon levels and above indices. Model fitness and robustness were evaluated via residual diagnostics, Cook’s distance, and bootstrapped CIs.
Results
Glucagon levels during the OGTT were significantly elevated in patients with severe insulin resistance (HomaIR Q3 and CPRIR Q1), even after adjusting for confounders. In contrast, groups with the poorest insulin secretion (AUCcp Q1, HomaB Q1, first PH Q1, and second PH Q1) did not show elevated glucagon levels compared with those with better secretion. Mediation analysis confirmed that neither AUCcp nor HomaB mediated the relationship between insulin resistance and glucagon levels. Residual diagnostics demonstrated a satisfactory model fit. Furthermore, sensitivity analyses, both by excluding influential points identified via Cook’s distance and by applying bootstrapped CIs, yielded consistent results, thereby affirming the robustness of the model.
Conclusions
In T2DM, impaired glucagon suppression during OGTT is associated with insulin resistance closely, highlighting insulin resistance as a key factor in alpha-cell dysfunction.
Keywords: Glucagon; Insulin Resistance; Insulin Secretion; Diabetes Mellitus, Type 2
WHAT IS ALREADY KNOWN ON THIS TOPIC
In vivo, insulin modulates glucagon secretion from pancreatic alpha cells via paracrine signaling mechanisms. Previous studies have indicated that alpha cell dysfunction in type 2 diabetes mellitus (T2DM) may stem from reduced insulin secretion by neighboring beta cells, as well as insulin resistance within the alpha cells themselves.
WHAT THIS STUDY ADDS
Insulin resistance (assessed via homeostasis model assessment of insulin resistance and C-peptide immunoreactivity insulin resistance) was strongly associated with elevated glucagon levels, whereas insulin secretion (measured by C-peptide area under the curve, homeostasis model assessment of beta cell function, first PH, and second PH) showed no significant correlation. Mediation analysis further confirmed that insulin resistance, rather than impaired insulin secretion, was associated with glucagon secretion dysregulation in T2DM.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
In T2DM, impaired glucagon suppression during oral glucose tolerance test is linked to insulin resistance but not to reduced insulin secretion, highlighting insulin resistance as a key factor associated with alpha-cell dysfunction. These findings suggest that therapeutic strategies targeting insulin resistance may be more closely associated with improved glucagon regulation than those focusing solely on insulin secretion.
Insulin resistance and impaired insulin secretion are two key factors contributing to poor glycemic control in type 2 diabetes mellitus (T2DM). Regarding insulin secretion, in vivo studies show that insulin modulates glucagon secretion from pancreatic alpha cells via paracrine signaling mechanisms.1 2 Normally, glucagon and insulin are secreted in a coordinated, pulsatile manner, but this relationship is disrupted in diabetes.3 4 Raskin et al demonstrated that intravenous administration of physiological insulin resulted in a smaller reduction in glucagon levels in patients with diabetes compared with nondiabetic individuals.5 Unger et al further found that even supraphysiological insulin doses fail to promptly restore alpha cell responsiveness to hyperglycemia, suggesting that insulin deficiency alone may not fully explain alpha cell dysfunction in diabetes.6 In addition to insulin deficiency, insulin resistance, which is commonly observed in the liver, muscle, and adipose tissue in T2DM, may also affect pancreatic alpha cells. Alpha cell insulin resistance is supported by a study demonstrating that alpha cell-specific insulin receptor knockout mice exhibit elevated glucagon levels both during fasting and on stimulation.1 This finding underscores the significant role of alpha cell insulin resistance in the dysregulation of glucagon.7 Despite extensive research on insulin secretion, the mechanisms governing glucagon secretion, particularly in diabetes, remain incompletely understood. Therefore, this study aimed to investigate the relationships between glucagon dynamics during an oral glucose tolerance test (OGTT), insulin resistance, and insulin secretion in individuals with T2DM, using generalized linear model (GLM) and mediation analysis.
Methodologically, the hyperinsulinemic-euglycemic clamp and hyperglycemic clamp remain the gold standard for directly assessing insulin resistance and secretion8; however, due to their complexity and invasiveness, surrogate indices derived from the OGTT are widely used in clinical and epidemiological research. This study employed OGTT-derived calculated indices, homeostasis model assessment of insulin resistance (HomaIR) and C-peptide immunoreactivity insulin resistance (CPRIR), to reflect insulin resistance, and used homeostasis model assessment of beta-cell function (HomaB), the area under the curve for C-peptide (AUCcp), as well as first-phase insulin secretion (first PH), second-phase insulin secretion (second PH) to reflect insulin secretion.
Research design and methods
Design and population
This study involved patients diagnosed with T2DM who received care at the Department of Endocrinology and Metabolism, Tianjin Medical University General Hospital, between October 2022 and March 2025. The exclusion criteria were as follows: (1) individuals under the age of 18 or those who are pregnant; (2) comorbidities that may influence glucose metabolism, including pancreatitis, previous pancreatic surgery, thyroid dysfunction, Cushing’s syndrome, severe gastrointestinal diseases, cancer, or hepatic or renal insufficiency.
All participants fasted for 8 hours prior to undergoing a 75 g OGTT. On the day of the OGTT, patients were instructed to suspend all hypoglycemic medications to eliminate any potential acute pharmacological interference. Blood samples were collected at baseline (0 hours), and at intervals of 0.5 hours, 1 hour, 2 hours, and 3 hours post-glucose ingestion in order to measure levels of blood glucose, insulin, C-peptide, and glucagon. Clinical data were systematically retrieved from the hospital’s electronic medical record system, encompassing demographic characteristics (gender, age), anthropometric measurements (height, weight), clinical variables (duration of diabetes, family history of diabetes, presence of fatty liver disease), and medication use. Special attention was given to the use of hypoglycemic drugs within 1 week before the OGTT, including glucagon-like peptide-1 receptor agonists, dipeptidyl peptidase-4 inhibitors, sodium-glucose co-transporter-2 inhibitors, insulin, insulin secretagogues, insulin sensitizers, and α-glucosidase inhibitors.
Biochemical measurement
Glucose levels were measured using the hexokinase method, while insulin and C-peptide levels were primarily assessed via chemiluminescence immunoassay. A glucagon stabilizer was added to the glucagon collection vessel to prevent glucagon degradation (patent number: ZL201810066819.6). We employed dual monoclonal antibodies that target the N and C terminals of glucagon and determined it by chemiluminescence immunoassay using a Fully Automatic Chemiluminescence analyzer (HomoG 100, China).
Statistical analysis and calculations
Continuous variables with normal distribution were reported as mean and SD or SE, non-normally distributed variables were summarized as median (IQR). Categorical variables were presented as counts (percentages). Other data presentation methods are shown in the footnotes of the figures or tables. Differences in continuous variables across groups were analyzed using the Kruskal-Wallis H test; categorical variables were compared using the χ2 test. In the GLM analysis, we applied natural logarithmic transformations to the relevant variables and selected an identity link function. To evaluate the suitability of the model, we conducted residual diagnostics and analyzed Cook’s distance. To assess the robustness of our findings, a bootstrapped CI analysis was performed. Additionally, mediation analysis was carried out to determine whether insulin secretion mediated the relationship between insulin resistance and glucagon secretion. In addressing the use of hypoglycemic drugs, we constructed categorical variables based on their mechanisms of action and incorporated them into the model. Statistical analyses were performed using SPSS V.23 (IBM) and R software (V.4.1.1). Graphs were created with GraphPad Prism V.8.0.2.
The mathematical formulas used in this section are presented below:
CPRIR=60/(CP0h *Glu0h)9 10
HomaIR=1.5+Glu0h*CP0h*333/280011
HomaB=0.27*CP0h*333/(Glu0h−3.5)12
1st PH=1283+1.829*insulin0.5h−138.7*Glu0.5h+3.772*insulin0h13
2nd PH=287+0.4164*insulin0.5h−26.07*Glu0.5h+0.9226*insulin0h13
The AUC for blood glucose, C-peptide, and glucagon was calculated using the trapezoidal method. In the above formula, the unit of blood glucose is mmol/L, the unit of C-peptide is ng/mL, and the unit of insulin is pmol/L.
Results
General characteristics of the study population
This study enrolled a total of 247 patients diagnosed with T2DM, consisting of 57.89% males and 42.11% females. The median age was 54.00 years, the median body mass index (BMI) was 26.12 kg/m², and the median duration of diabetes was 72.00 months. Baseline demographic and clinical characteristics, including gender, age, BMI, disease duration, blood glucose levels, prevalence of fatty liver disease, and usage of hypoglycemic drugs, are summarized in table 1.
Table 1. Characteristics of patients.
| Patient (n=247) | |
|---|---|
| Sex, n (%) | Male, 143 (57.89) Female, 104 (42.11) |
| Age, years | 54.00 (42.00, 63.00) |
| BMI, kg/m2 | 26.12 (23.34, 29.38) |
| Duration of diabetes, months | 72.00 (6.00, 120.00) |
| Glu0h, mmol/L | 6.61 (5.64, 7.80) |
| Glu2h, mmol/L | 16.33±4.07 |
| Family history of diabetes, n (%) | 88 (35.63) |
| Fatty liver, n (%) | 181 (73.28) |
| GLP-1R agonist/DPP-4 inhibitor, n (%) | 73 (29.55) |
| SGLT-2 inhibitor, n (%) | 78 (31.58) |
| Insulin, n (%) | 125 (50.61) |
| Insulin secretagogue, n (%) | 30 (12.15) |
| Insulin sensitizer, n (%) | 119 (48.18) |
| α-glucosidase inhibitors, n (%) | 96 (38.87) |
BMI, body mass index; DPP-4, dipeptidyl peptidase-4; GLP-1R, glucagon-like peptide-1 receptor; Glu0h, fasting glucose; Glu2h, glucose level at 2 hours in OGTT; SGLT-2, sodium-glucose co-transporter-2.
Association of glucagon levels during OGTT with insulin resistance and insulin secretion
Insulin resistance was assessed using the CPRIR and HomaIR calculated based on fasting blood glucose and C-peptide levels during OGTT. Both indices were divided into tertiles. CPRIR Q1 indicated the highest insulin resistance and Q3 the lowest, while for HomaIR, Q3 reflected the greatest resistance and Q1 the mildest. Baseline characteristics are summarized in online supplemental Table 1. Significant differences in the area under the glucagon curve during OGTT (AUCgcg) were found across CPRIR and HomaIR tertiles.
Insulin secretion was assessed using the area under the C-peptide curve and HomaB, both also categorized into tertiles. AUCcp Q3 and HomaB Q3 indicated the highest insulin secretion, and Q1 the lowest. Baseline data are shown in online supplemental Table 2. AUCgcg varied significantly across AUCcp tertiles but not among HomaB tertiles.
Glucagon levels across different groups were visualized using line graphs, as presented in figure 1. Glucagon levels increased progressively from CPRIR Q3 to Q1 and from HomaIR Q1 to Q3, indicating that greater insulin resistance is associated with higher glucagon levels and reduced suppression of glucagon secretion. Notably, the AUCcp Q1 and HomaB Q1 groups—both showing the poorest insulin secretion—had lower glucagon levels than groups with better insulin secretion. This indicates that in T2DM, glucagon secretion does not rise as insulin secretion declines.
Figure 1. Glucagon levels in OGTT across insulin sensitivity and resistance groups. Note: The data in the graph are presented as mean±SE. AUCcp, area under the C-peptide curve in oral glucose tolerance test (OGTT); CPRIR, C-peptide immunoreactivity insulin resistance; HomaB, homeostasis model assessment of beta-cell function; HomaIR, homeostasis model assessment for insulin resistance.
The GLM was used to analyze glucagon levels across different CPRIR and HomaIR groups. In model 1, with no adjustment for confounding factors, the CPRIR Q3 and Q2 groups exhibited significantly lower glucagon levels than the CPRIR Q1 group. Additionally, AUCgcg, GCG2h, and GCG3h were lower in CPRIR Q3 than in CPRIR Q2. These results remained consistent in models 2 and 3 after adjusting for age, BMI, fatty liver, and medication. Similar patterns were observed in the HomaIR-stratified groups. In the HomaIR Q3 group, glucagon levels were significantly higher than those in both the HomaIR Q1 and Q2 groups. Moreover, AUCgcg, GCG2h, and GCG3h were lower in HomaIR Q2 than in HomaIR Q1. These differences persisted even after adjustment for confounding variables (as shown in table 2). To exclude potential interference from exogenous insulin use on the analytical results, we performed a stratified analysis based on insulin use to compare AUCgcg across different groups. The results remained consistent with the primary analysis and are shown in online supplemental Tables 3 and 4. These findings suggest that in T2DM, glucagon secretion during OGTT gradually increases as insulin resistance worsens.
Table 2. Comparison of glucagon levels across different CPRIR and HomaIR groups.
| GCG0h，pmol/L | GCG0.5h，pmol/L | GCG1h，pmol/L | CCG2h，pmol/L | GCG3h，pmol/L | AUCgcg，pmol/L*hour | |
|---|---|---|---|---|---|---|
| Model 1 | ||||||
| CPRIR Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| CPRIR Q2 | −0.25 (−0.40 to −0.10)* | −0.26 (−0.41 to −0.11)* | −0.28 (−0.45 to −0.12)* | −0.22 (−0.40 to −0.03)† | −0.18 (−0.36 to 0.004) | −0.23 (−0.38 to −0.09)* |
| CPRIR Q3 | −0.41 (−0.55 to −0.26)‡ § | −0.39 (−0.54 to −0.24)‡ | −0.42 (−0.58 to −0.25)‡ | −0.38 (−0.56 to −0.19)‡ | −0.40 (−0.58 to −0.21)‡ § | −0.39 (−0.54 to −0.24)‡ § |
| HomaIR Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| HomaIR Q2 | 0.16 (0.01 to 0.31)† | 0.13 (−0.02 to 0.29) | 0.13 (−0.04 to 0.29) | 0.16 (−0.02 to 0.35) | 0.21 (0.03 to 0.40)† | 0.15 (0.003 to 0.30)† |
| HomaIR Q3 | 0.40 (0.25 to 0.55)‡ ¶ | 0.38 (0.23 to 0.53)‡ ¶ | 0.41 (0.25 to 0.58)‡ ¶ | 0.37 (0.18 to 0.55)‡ § | 0.39 (0.21 to 0.57)‡ | 0.38 (0.23 to 0.53)‡ ¶ |
| Model 2 | ||||||
| CPRIR Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| CPRIR Q2 | −0.28 (−0.45 to −0.12)* | −0.24 (−0.40 to −0.07)* | −0.29 (−0.47 to −0.11)* | −0.29 (−0.49 to −0.09)* | −0.23 (−0.42 to −0.05)† | −0.26 (−0.42 to −0.10)* |
| CPRIR Q3 | −0.41 (−0.59 to −0.23)‡ | −0.31 (−0.49 to −0.13)* | −0.39 (−0.58 to −0.20)‡ | −0.48 (−0.70 to −0.27)‡ § | −0.50 (−0.70, −0.30)‡ ¶ | −0.40 (−0.58 to −0.23)‡ § |
| HomaIR Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| HomaIR Q2 | 0.13 (−0.02 to 0.29) | 0.09 (−0.07 to 0.24) | 0.10 (−0.07 to 0.27) | 0.21 (0.02 to 0.39)† | 0.28 (0.10 to 0.45)* | 0.15 (−0.001 to 0.30) |
| HomaIR Q3 | 0.40 (0.22 to 0.58)‡ ¶ | 0.29 (0.11 to 0.47)* § | 0.38 (0.19 to 0.57)‡ ¶ | 0.46 (0.24 to 0.67)‡ § | 0.47 (0.28 to 0.67)‡ § | 0.39 (0.22 to 0.56)‡ ¶ |
| Model 3 | ||||||
| CPRIR Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| CPRIR Q2 | −0.28 (−0.44 to −0.11)* | −0.23 (−0.39 to −0.06)* | −0.29 (−0.47 to −0.12)* | −0.29 (−0.49 to −0.09)* | −0.22 (−0.40 to −0.03)† | −0.26 (−0.42 to −0.10)* |
| CPRIR Q3 | −0.41 (−0.60 to −0.22)‡ | −0.31 (−0.50 to −0.13)* | −0.42 (−0.62 to −0.22)‡ | −0.50 (−0.73 to −0.28)‡ § | −0.48 (−0.68 to −0.27)‡ ¶ | −0.42 (−0.60 to −0.24)‡ § |
| HomaIR Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| HomaIR Q2 | 0.14 (−0.02 to 0.30) | 0.09 (−0.07 to 0.25) | 0.13 (−0.04 to 0.30) | 0.22 (0.03 to 0.41)† | 0.27 (0.09 to 0.45)* | 0.16 (0.01 to 0.32)† |
| HomaIR Q3 | 0.40 (0.21 to 0.59)‡ ¶ | 0.29 (0.10 to 0.47)* § | 0.41 (0.21 to 0.61)‡ ¶ | 0.48 (0.25 to 0.70)‡ § | 0.45 (0.25 to 0.66)‡ | 0.40 (0.22 to 0.58)‡ ¶ |
The dependent variables were log-transformed prior to analysis. Results are presented as mean difference (95% Wald CI).
Model 1: Unadjusted for confounding factors.
Model 2: Adjusted for age, BMI, disease duration, fatty liver and insulin use.
Model 3: Model 2+AUCglu+GLP-1R agonist/DPP-4 inhibitor+SGLT-2 inhibitor+insulin+insulin secretagogue+insulin sensitizer+α-glucosidase inhibitor.
p<0.01 compared with Q1 group.
p<0.05 compared with Q1 group.
p<0.001 compared with Q1 group.
p<0.05 compared with Q2 group.
p<0.01 compared with Q2 group.
AUCgcg, area under the glucagon curve in OGTT; AUCglu, area under the glucose curve in OGTT; BMI, body mass index; CPRIR, C-peptide immunoreactivity insulin resistance; DPP-4, dipeptidyl peptidase-4; GCG0h, fasting glucagon; GCG1h, glucagon level at 1h in OGTT; GCG2h, glucagon level at 2h in OGTT; GCG3h, glucagon level at 3h in OGTT; GCG0.5h, glucagon level at 0.5h in OGTT; GLP-1R, glucagon-like peptide-1 receptor; HomaIR, homeostasis model assessment for insulin resistance; SGLT-2, sodium-glucose co-transporter-2.
A comparative analysis using a GLM was performed to assess glucagon levels across different groups categorized by AUCcp and HomaB. The results indicated that, regardless of adjustments for confounding factors, the Q3 group—characterized by enhanced insulin secretion response—did not display the lowest glucagon levels. In fact, both the Q2 and Q3 groups exhibited higher glucagon levels compared with the Q1 group. These findings are summarized in table 3. The same trend was observed irrespective of insulin use, that is, higher insulin secretion groups consistently failed to demonstrate lower glucagon levels than their lower-secretion counterparts (online supplemental Tables 3 and 4). We also calculated the first-phase and second-phase insulin secretion (first PH and second PH) during the OGTT in patients not receiving insulin therapy and analyzed their association with glucagon levels. In unadjusted analyses, only the first PH Q3 group had significantly lower GCG3h than the Q1 group, with no differences among the three second PH groups. In model 3, only the first PH Q3 group showed lower GCG0.5h than the Q1 group, and the second PH groups again did not differ from one another. Results are shown in online supplemental Table 5. Subsequently, we adjusted for the CPRIR and reanalyzed the association between insulin secretion indices and glucagon levels. The results similarly demonstrated no increase in glucagon levels alongside the decline in insulin secretion (online supplemental Table 6). The corresponding Akaike information criterion, Bayesian information criterion, and pseudo-R2 values for the analysis models mentioned above are provided in online supplemental Table 7.
Table 3. Comparison of glucagon levels across different AUCcp and HomaB groups.
| GCG0h, pmol/L | GCG0.5h, pmol/L | GCG1h, pmol/L | CCG2h, pmol/L | GCG3h, pmol/L | AUCgcg, pmol/L*hour | |
|---|---|---|---|---|---|---|
| Model 1 | ||||||
| AUCcp Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| AUCcp Q2 | 0.25 (0.10 to 0.41)* | 0.25 (0.09 to 0.40)* | 0.23 (0.06 to 0.41)* | 0.20 (0.01 to 0.39)† | 0.19 (0.01 to 0.38)† | 0.23 (0.08 to 0.38)* |
| AUCcp Q3 | 0.26 (0.11 to 0.41)* | 0.25 (0.10 to 0.41)* | 0.13 (−0.04 to 0.30) | 0.06 (−0.13 to 0.25) | 0.05 (−0.14 to 0.23) | 0.15 (−0.004 to 0.31) |
| HomaB Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| HomaB Q2 | 0.15 (−0.01 to 0.30) | 0.18 (0.02 to 0.34)† | 0.19 (0.02 to 0.37)† | 0.07 (−0.12 to 0.26) | −0.01 (−0.20 to 0.18) | 0.13 (−0.02 to 0.29) |
| HomaB Q3 | 0.22 (0.07 to 0.38)* | 0.20 (0.04 to 0.36)† | 0.16 (−0.01 to 0.34) | 0.02 (−0.17 to 0.21) | −0.05 (−0.24 to 0.14) | 0.12 (−0.04 to 0.27) |
| Model 2 | ||||||
| AUCcp Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| AUCcp Q2 | 0.26 (0.10 to 0.43)* | 0.23 (0.07 to 0.39)* | 0.24 (0.06 to 0.41)* | 0.30 (0.10 to 0.50)* | 0.28 (0.10 to 0.46)* | 0.27 (0.11, 0.43)* |
| AUCcp Q3 | 0.19 (−0.01 to 0.39) | 0.18 (−0.02 to 0.37) | 0.08 (−0.12 to 0.29) | 0.15 (−0.08 to 0.39) | 0.11 (−0.10 to 0.33) | 0.15 (−0.04 to 0.34) |
| HomaB Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| HomaB Q2 | 0.10 (−0.07 to 0.27) | 0.10 (−0.07 to 0.26) | 0.12 (−0.06 to 0.30) | 0.04 (−0.16 to 0.25) | −0.06 (−0.25 to 0.13) | 0.08 (−0.08 to 0.24) |
| HomaB Q3 | 0.11 (−0.09 to 0.31) | 0.02 (−0.17 to 0.22) | 0.07 (−0.15 to 0.28) | −0.01 (−0.25 to 0.23) | −0.13 (−0.35 to 0.09) | 0.03 (−0.16 to 0.22) |
| Model 3 | ||||||
| AUCcp Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| AUCcp Q2 | 0.26 (0.08 to 0.42)* | 0.23 (0.07 to 0.40)* | 0.23 (0.05 to 0.40)† | 0.29 (0.10 to 0.49)* | 0.28 (0.09 to 0.46)* | 0.26 (0.10 to 0.42)* |
| AUCcp Q3 | 0.22 (0.02 to 0.42)† | 0.25 (0.05 to 0.45)† | 0.13 (−0.08 to 0.35) | 0.21 (−0.03 to 0.45)† | 0.17 (−0.05 to 0.40) | 0.20 (0.01 to 0.39)† |
| HomaB Q1 | Reference | Reference | Reference | Reference | Reference | Reference |
| HomaB Q2 | 0.14 (−0.04 to 0.31) | 0.11 (−0.06 to 0.29) | 0.11 (−0.07 to 0.30) | 0.04 (−0.17 to 0.26) | −0.05 (−0.25 to 0.15) | 0.09 (−0.08 to 0.26) |
| HomaB Q3 | 0.19 (−0.04 to 0.41) | 0.07 (−0.14 to 0.29) | 0.08 (−0.16 to 0.32) | 0.01 (−0.25 to 0.28) | −0.09 (−0.34 to 0.16) | 0.07 (−0.15 to 0.28) |
Note: The dependent variables were log−transformed prior to analysis. Results are presented as mean difference (95% Wald CI).
Model 1: Unadjusted for confounding factors.
Model 2: Adjusted for age, BMI, disease duration, fatty liver and insulin use.
Model 3: Model 2+AUCglu+GLP−1R agonist/DPP-4 inhibitor+SGLT−2 inhibitor+insulin+insulin secretagogue+insulin sensitizer+α−glucosidase inhibitor.
p<0.01 compared with Q1 group.
p<0.05 compared with Q1 group.
AUCcp, area under the C−peptide curve in OGTT; AUCgcg, area under the glucagon curve in OGTT; AUCglu, area under the glucose curve in OGTT; BMI, body mass index; DPP−4, dipeptidyl peptidase−4; GCG0h, fasting glucagon; GCG1h, glucagon level at 1 hour in OGTT; GCG2h, glucagon level at 2 hours in OGTT; GCG3h, glucagon level at 3 hours in OGTT; GCG0.5h, glucagon level at 0.5 hours in OGTT; GLP−1R, glucagon−like peptide-1 receptor; HomaB, homeostasis model assessment of beta−cell function; OGTT, oral glucose tolerance test; SGLT−2, sodium−glucose co-transporter-2.
To evaluate the fitness of GLM, we then performed model diagnostics, consisting of residual distribution assessment and Cook’s distance analysis. The residual diagnostic plots suggested a satisfactory model fit (online supplemental figures 1 and 2), whereas Cook’s distance plots identified several influential observations (online supplemental figures 3 and 4). On removing these influential points and repeating the analysis, the results remained consistent (online supplemental Table 8). Furthermore, to evaluate the robustness of the findings, the bootstrapped CI method was employed. The results demonstrated no significant changes (online supplemental Table 9).
Mediation analysis of insulin secretion in the relationship between insulin resistance and glucagon secretion
A mediation analysis was conducted to determine whether insulin secretion mediates the relationship between insulin resistance and glucagon secretion. In this analysis, AUCgcg served as the outcome variable, HomaIR and CPRIR as independent variables, and AUCcp and HomaB as potential mediators. When HomaIR or CPRIR was used as the independent variable, AUCcp did not exhibit a negative mediating effect in the unadjusted model. This result remained unchanged after controlling for gender, age, BMI, AUCglu, and medication (figure 2a,b). Similarly, HomaB showed no mediating role in any of the models examined (figure 2c,d). The adjusted mediation results are also presented in online supplemental Table 10. The accompanying path diagram that visualizes these relationships, along with the standardized beta coefficients, is provided in online supplemental figure 5. These findings corroborate the earlier GLM results: in patients with T2DM, glucagon secretion during OGTT becomes less suppressed as insulin resistance increases, but does not progressively rise as insulin secretion declines.
Figure 2. Medicating effects of AUCcp and HomaB in the relationships between HomaIR, CPRIR and AUCgcg. Note: The variables were log-transformed prior to analysis. HomaIR is the independent variable and AUCcp is the mediating variable; (b) CPRIR is the independent variable and AUCcp is the mediating variable; (c) HomaIR is the independent variable and HomaB is the mediating variable; (d) CPRIR is the independent variable and HomaB is the mediating variable model 1: unadjusted for confounding factors; model 2: adjusted for gender, age, BMI, disease duration, AUCglu, and fatty liver; model 3: model 2+Use of GLP-1R agonists/DPP-4 inhibitors+SGLT-2 inhibitors+insulin+insulin secretagogues+insulin sensitizers+α-glucosidase inhibitors. ACME, average causal mediation effect; ADE, average direct effect; AUCcp, area under the C-peptide curve in oral glucose tolerance test (OGTT); AUCgcg, area under the glucagon curve in OGTT; AUCglu, area under the glucose curve in OGTT; BMI, body mass index; CPRIR, C-peptide immunoreactivity insulin resistance; DPP-4, dipeptidyl peptidase-4; GLP-1R, glucagon-like peptide-1 receptor; HomaB, homeostasis model assessment of beta; HomaIR, homeostasis model assessment for insulin resistance; SGLT-2, sodium-glucose co-transporter-2.
Discussion
This study examined how insulin resistance and insulin secretion relate to glucagon secretion during an OGTT in T2DM patients. Our findings reveal that increased insulin resistance correlates with diminished suppression of glucagon. Conversely, decreased insulin secretion is not accompanied by a rise in glucagon levels.
Glucagon functions as a physiological antagonist of insulin, increasing blood glucose levels through the inhibition of glycolysis and the promotion of gluconeogenesis and glycogenolysis.14 15 Its secretion is regulated by nutrients, hormones (autocrine, paracrine, endocrine), and nerves.16,18 Paracrine signals include insulin from nearby beta cells and somatostatin from delta cells.18,20 Alpha cell dysfunction in T2DM may stem from reduced insulin secretion by neighboring beta cells. A pancreatic perfusion study showed insulin suppresses glucagon secretion in alloxan-induced diabetic dogs.21 When about 75% of beta cell mass is lost, glucose-induced glucagon suppression in mice disappears and cannot be restored by exogenous insulin injections. This suggests that glucagon inhibition depends on local insulin release, not systemic insulin levels.22 Alpha cells may exhibit insulin resistance in diabetic conditions.7 23 High expression of insulin receptors has been detected on the surface of alpha cells, and genetic deletion of the insulin receptor has been shown to elevate both blood glucose and glucagon levels.24 25 In mice with alpha cell-specific insulin receptor knockout, plasma glucagon levels rise postprandially, and these same animals demonstrate impaired glucagon secretion during episodes of hypoglycemia.1 Additionally, the alpha cell line cultured under chronic high glucose and palmitate exposure demonstrates abnormalities in the insulin signaling pathway.26 27
The hyperinsulinemic-euglycemic clamp technique remains the gold standard for assessing insulin resistance; however, it is labor-intensive and time-consuming.8 For broader applicability, a series of straightforward surrogate indices for insulin sensitivity and resistance are commonly employed. The indices include those derived from fasting steady-state measurements and indices obtained from dynamic procedures like the OGTT.28 Previous evidence suggests that HomaIR correlates well with insulin sensitivity as assessed by the hyperinsulinemic-euglycemic clamp and CPRIR was correlated strongly with glucose infusion rate measured during hyperinsulinemic-euglycemic glucose clamps,9 10 29 so HomaIR and CPRIR were adopted in this study. Given that nearly half of our patients use exogenous insulin, and Li et al showed a strong correlation between fasting C-peptide-based modified HomaIR and insulin-based traditional HomaIR in diabetes,11 we used C-peptide—not insulin—to calculate HomaIR.
Tsuchiyama’s study demonstrated that glucagon secretion in response to arginine stimulation is positively correlated with HomaIR in patients with T2DM.23 Similarly, Henkel et al’s research identified a significant positive association between postprandial glucagon levels and HomaIR.30 In this study, we observed a positive correlation between HomaIR and OGTT-derived glucagon levels. Consistently, glucagon levels in T2DM progressively increased as CPRIR declined. Other OGTT-derived formulas for estimating insulin sensitivity, such as the Stumvoll index and the Gutt index, are also widely used in clinical practice. However, since their calculations rely entirely on insulin levels, it remains unclear whether C-peptide could be substituted for insulin in these formulas, so those indices did not adopt here.1331,33 However, it should be noted that HomaIR and CPRIR are not indicators of insulin sensitivity specifically in pancreatic alpha cells, and no established clinical measures currently exist to directly assess alpha-cell insulin sensitivity. Therefore, based on the present findings, we can only infer potential associations rather than conclude that alpha-cell insulin resistance exists in patients with T2DM, nor can we establish a causal relationship between hyperglucagonemia and insulin resistance.
Although insulin has a physiological effect of inhibiting glucagon, some previous studies have not observed the expected negative correlation between the two or between C-peptide levels and glucagon. Kobayashi assessed pancreatic beta cell function using HomaB and the insulinogenic index derived from OGTT results, and reported no negative correlation between glucagon level and insulin secretion.34 Hosokawa also found fasting glucagon levels rise with C-peptide in T2DM.35 In this study, AUCcp and HomaB are utilized as preliminary indicators of insulin secretion status. Our findings reveal that glucagon levels do not decrease in tandem with elevated AUCcp and HomaB values. Given that the hyperglycemic clamp, while regarded as the “gold standard” for assessing insulin secretion, is operationally complex, we further utilized OGTT-derived indices (first PH and second PH) developed by Stumvoll, which correlate strongly with clamp-measured first-phase and second-phase insulin secretion, to more precisely evaluate dynamic beta cell function.8 13 36 Since these indices depend on endogenous insulin measurements and some of our study population was subject to interference from exogenous insulin, we restricted the analysis to patients who were not using insulin therapy. Our findings indicate that glucagon levels did not decrease with the increase in both first PH and second PH, which is consistent with the aforementioned results.
We conducted mediation analysis to assess whether changes in insulin secretion affected the relationship between insulin resistance and glucagon levels. Results showed that AUCcp and HomaB did not mediate this association, indicating that reduced insulin secretion in T2DM did not strengthen the positive association between insulin resistance and elevated glucagon. Tsuchiyama’s study identified distinct mechanisms underlying elevated glucagon levels in T1DM and T2DM, with insulin deficiency and alpha cell insulin resistance driving glucagon secretion dysregulation in each type respectively.23 These results support our finding that glucagon increases with insulin resistance but is not associated with reduced insulin secretion in patients with T2DM. Clinical studies demonstrate that glucagon levels do not increase proportionally with declining insulin secretion,23 28 while basic research indicates that impaired insulin secretion enhances glucagon release.37 38 This apparent contradiction may be explained by several factors: (1) Basic studies frequently employ models of severe beta cell damage, whereas most patients with T2DM maintain partial beta cell function, and in those individuals, concurrent insulin resistance may compensate for the reduced effects of insulin secretion on glucagon regulation; (2) In T2DM, pancreatic alpha cell insulin resistance attenuates insulin’s direct inhibitory action on glucagon secretion; (3) Glucagon secretion is modulated by a complex interplay of multiple factors, which may confound its relationship with insulin secretion; and (4) The relationship between insulin resistance and insulin secretion is a dynamic process; however, the parameters used in this study to assess insulin resistance and secretion were derived from cross-sectional, single-time-point measurements, which cannot fully capture such temporal changes. This discrepancy between cross-sectional data and dynamic processes may partially explain the unexpected finding that glucagon levels did not show a corresponding increase as insulin secretion declined. Therefore, further longitudinal clinical and basic studies are needed to clarify the complex interactions among glucagon secretion, insulin secretion, and insulin resistance in T2DM.
Our study identifies a link between insulin resistance and elevated glucagon levels in T2DM. This observation raises the possibility that improving insulin sensitivity may influence glucagon regulation, although the underlying mechanisms require further investigation using more specific methodologies. Future studies using direct measures of alpha-cell function are needed to determine whether enhancing insulin sensitivity in alpha cells could have therapeutic implications.
Our study has several limitations. First, the cross-sectional design of the study precludes causal inference. Second, potential confounding from both gastrointestinal hormones and hypoglycemic medications cannot be excluded. Third, insulin sensitivity was assessed using surrogate indices derived from the OGTT rather than the hyperinsulinemic-euglycemic clamp technique. Nevertheless, our findings provide clinical insights into glucagon secretion dysregulation in T2DM, adding new content to the existing framework.
Conclusions
In patients with T2DM, as insulin resistance increases, the inhibition of glucagon secretion during OGTT weakens; however, increased glucagon levels are not associated with a decrease in insulin secretion. In those with severe insulin resistance, inhibiting glucagon may bring greater improvements in glycemic control.
Supplementary material
Footnotes
Funding: This work was supported by Tianjin Key Medical Discipline (Specialty) Construction Project (TJYXZDXK-3-002C), Tianjin Medical University Clinical Special Disease Research Center-Neuroendocrine Tumor Clinical Special Disease Research Center (N/A), Tianjin Major Science and Technology Special Projects and Programs (grant number 24ZXYXSY00070), Tianjin Medical University General Hospital Sailing Program (grant number 22TSC020), National Natural Science Foundation of China (82200963), and Baotou Municipal Health Commission (2023wsjkkj09).
Provenance and peer review: Not commissioned; externally peer reviewed.
Patient consent for publication: Not applicable.
Ethics approval: The study protocol was approved by the Institutional Review Board of Tianjin Medical University General Hospital (IRB2024-YX-139-01). Given that fully anonymized retrospective data were used in this research, the requirement for obtaining individual informed consent from participants was waived.
Data availability statement
Data are available on reasonable request.
References
- 1.Kawamori D, Kurpad AJ, Hu J, et al. Insulin signaling in alpha cells modulates glucagon secretion in vivo. Cell Metab. 2009;9:350–61. doi: 10.1016/j.cmet.2009.02.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Maruyama H, Tominaga M, Bolli G, et al. The alpha cell response to glucose change during perfusion of anti-insulin serum in pancreas isolated from normal rats. Diabetologia. 1985;28:836–40. doi: 10.1007/BF00291074. [DOI] [PubMed] [Google Scholar]
- 3.Meier JJ, Kjems LL, Veldhuis JD, et al. Postprandial suppression of glucagon secretion depends on intact pulsatile insulin secretion: further evidence for the intraislet insulin hypothesis. Diabetes. 2006;55:1051–6. doi: 10.2337/diabetes.55.04.06.db05-1449. [DOI] [PubMed] [Google Scholar]
- 4.Menge BA, Grüber L, Jørgensen SM, et al. Loss of inverse relationship between pulsatile insulin and glucagon secretion in patients with type 2 diabetes. Diabetes. 2011;60:2160–8. doi: 10.2337/db11-0251. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Raskin P, Fujita Y, Unger RH. Effect of insulin-glucose infusions on plasma glucagon levels in fasting diabetics and nondiabetics. J Clin Invest. 1975;56:1132–8. doi: 10.1172/JCI108188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Unger RH, Madison LL, Muller WA. Abnormal alpha cell function in diabetics response to insulin. Diabetes. 1972;21:301–7. doi: 10.2337/diab.21.5.301. [DOI] [PubMed] [Google Scholar]
- 7.Honzawa N, Fujimoto K, Kitamura T. Cell Autonomous Dysfunction and Insulin Resistance in Pancreatic α Cells. Int J Mol Sci. 2019;20:3699. doi: 10.3390/ijms20153699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.DeFronzo RA, Tobin JD, Andres R. Glucose clamp technique: a method for quantifying insulin secretion and resistance. Am J Physiol. 1979;237:E214–23. doi: 10.1152/ajpendo.1979.237.3.E214. [DOI] [PubMed] [Google Scholar]
- 9.Ohkura T, Shiochi H, Fujioka Y, et al. 20/(fasting C-peptide × fasting plasma glucose) is a simple and effective index of insulin resistance in patients with type 2 diabetes mellitus: a preliminary report. Cardiovasc Diabetol. 2013;12:21. doi: 10.1186/1475-2840-12-21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Okura T, Nakamura R, Fujioka Y, et al. CPR-IR is an insulin resistance index that is minimally affected by hepatic insulin clearance-A preliminary research. PLoS One. 2018;13:e0197663. doi: 10.1371/journal.pone.0197663. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Li X, Zhou ZG, Qi HY, et al. Using Fasting C-Peptide to Replace Insulin in the Modified HOMA Formula for Assessing Insulin Resistance and Pancreatic β-Cell Function (in Chinese) Journal of Central South University (Medical Sciences) 2004:419–23. [PubMed] [Google Scholar]
- 12.Zhu J, Han J, Liu L, et al. Clinical expert consensus on the assessment and protection of pancreatic islet β-cell function in type 2 diabetes mellitus. Diabetes Res Clin Pract. 2023;197:110568. doi: 10.1016/j.diabres.2023.110568. [DOI] [PubMed] [Google Scholar]
- 13.Stumvoll M, Mitrakou A, Pimenta W, et al. Use of the oral glucose tolerance test to assess insulin release and insulin sensitivity. Diabetes Care. 2000;23:295–301. doi: 10.2337/diacare.23.3.295. [DOI] [PubMed] [Google Scholar]
- 14.Jiang G, Zhang BB. Glucagon and regulation of glucose metabolism. Am J Physiol Endocrinol Metab. 2003;284:E671–8. doi: 10.1152/ajpendo.00492.2002. [DOI] [PubMed] [Google Scholar]
- 15.Wewer Albrechtsen NJ, Kuhre RE, Pedersen J, et al. The biology of glucagon and the consequences of hyperglucagonemia. Biomark Med. 2016;10:1141–51. doi: 10.2217/bmm-2016-0090. [DOI] [PubMed] [Google Scholar]
- 16.Quesada I, Tudurí E, Ripoll C, et al. Physiology of the pancreatic alpha-cell and glucagon secretion: role in glucose homeostasis and diabetes. J Endocrinol. 2008;199:5–19. doi: 10.1677/JOE-08-0290. [DOI] [PubMed] [Google Scholar]
- 17.Andersen DB, Holst JJ. Peptides in the regulation of glucagon secretion. Peptides. 2022;148 doi: 10.1016/j.peptides.2021.170683. [DOI] [PubMed] [Google Scholar]
- 18.Gromada J, Franklin I, Wollheim CB. Alpha-cells of the endocrine pancreas: 35 years of research but the enigma remains. Endocr Rev. 2007;28:84–116. doi: 10.1210/er.2006-0007. [DOI] [PubMed] [Google Scholar]
- 19.Hauge-Evans AC, King AJ, Carmignac D, et al. Somatostatin secreted by islet delta-cells fulfills multiple roles as a paracrine regulator of islet function. Diabetes. 2009;58:403–11. doi: 10.2337/db08-0792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Unger RH, Orci L. Paracrinology of islets and the paracrinopathy of diabetes. Proc Natl Acad Sci U S A. 2010;107:16009–12. doi: 10.1073/pnas.1006639107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Braaten JT, Faloona GR, Unger RH. The effect of insulin on the alpha-cell response to hyperglycemia in long-standing alloxan diabetes. J Clin Invest. 1974;53:1017–21. doi: 10.1172/JCI107638. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Meier JJ, Ueberberg S, Korbas S, et al. Diminished glucagon suppression after β-cell reduction is due to impaired α-cell function rather than an expansion of α-cell mass. Am J Physiol Endocrinol Metab. 2011;300:E717–23. doi: 10.1152/ajpendo.00315.2010. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Tsuchiyama N, Takamura T, Ando H, et al. Possible role of alpha-cell insulin resistance in exaggerated glucagon responses to arginine in type 2 diabetes. Diabetes Care. 2007;30:2583–7. doi: 10.2337/dc07-0066. [DOI] [PubMed] [Google Scholar]
- 24.Elliott AD, Ustione A, Piston DW. Somatostatin and insulin mediate glucose-inhibited glucagon secretion in the pancreatic α-cell by lowering cAMP. Am J Physiol Endocrinol Metab. 2015;308:E130–43. doi: 10.1152/ajpendo.00344.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Patel YC, Amherdt M, Orci L. Quantitative electron microscopic autoradiography of insulin, glucagon, and somatostatin binding sites on islets. Science. 1982;217:1155–6. doi: 10.1126/science.6126003. [DOI] [PubMed] [Google Scholar]
- 26.Shen X, Li H, Pan L, et al. Glucotoxicity and α Cell Dysfunction: Involvement of the PI3K/Akt Pathway in Glucose-Induced Insulin Resistance in Rat Islets and Clonal αTC1-6 Cells. Endocr Res. 2012;37:12–24. doi: 10.3109/07435800.2011.610855. [DOI] [PubMed] [Google Scholar]
- 27.Filippello A, Scamporrino A, Di Mauro S, et al. Direct Effects of D-Chiro-Inositol on Insulin Signaling and Glucagon Secretion of Pancreatic Alpha Cells. Biomolecules. 2020;10:1404. doi: 10.3390/biom10101404. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Muniyappa R, Lee S, Chen H, et al. Current approaches for assessing insulin sensitivity and resistance in vivo: advantages, limitations, and appropriate usage. Am J Physiol Endocrinol Metab. 2008;294:E15–26. doi: 10.1152/ajpendo.00645.2007. [DOI] [PubMed] [Google Scholar]
- 29.Katsuki A, Sumida Y, Gabazza EC, et al. Homeostasis model assessment is a reliable indicator of insulin resistance during follow-up of patients with type 2 diabetes. Diabetes Care. 2001;24:362–5. doi: 10.2337/diacare.24.2.362. [DOI] [PubMed] [Google Scholar]
- 30.Henkel E, Menschikowski M, Koehler C, et al. Impact of glucagon response on postprandial hyperglycemia in men with impaired glucose tolerance and type 2 diabetes mellitus. Metab Clin Exp. 2005;54:1168–73. doi: 10.1016/j.metabol.2005.03.024. [DOI] [PubMed] [Google Scholar]
- 31.Stumvoll M, Van Haeften T, Fritsche A, et al. Oral glucose tolerance test indexes for insulin sensitivity and secretion based on various availabilities of sampling times. Diabetes Care. 2001;24:796–7. doi: 10.2337/diacare.24.4.796. [DOI] [PubMed] [Google Scholar]
- 32.Hulman A, Simmons RK, Brunner EJ, et al. Trajectories of glycaemia, insulin sensitivity and insulin secretion in South Asian and white individuals before diagnosis of type 2 diabetes: a longitudinal analysis from the Whitehall II cohort study. Diabetologia. 2017;60:1252–60. doi: 10.1007/s00125-017-4275-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Gutt M, Davis CL, Spitzer SB, et al. Validation of the insulin sensitivity index (ISI(0,120)): comparison with other measures. Diabetes Res Clin Pract. 2000;47:177–84. doi: 10.1016/s0168-8227(99)00116-3. [DOI] [PubMed] [Google Scholar]
- 34.Kobayashi M, Satoh H, Matsuo T, et al. Plasma glucagon levels measured by sandwich ELISA are correlated with impaired glucose tolerance in type 2 diabetes. Endocr J. 2020;67:903–22. doi: 10.1507/endocrj.EJ20-0079. [DOI] [PubMed] [Google Scholar]
- 35.Hosokawa Y, Kozawa J, Nishizawa H, et al. Positive correlation between fasting plasma glucagon and serum C-peptide in Japanese patients with diabetes. Heliyon. 2019;5:e01715. doi: 10.1016/j.heliyon.2019.e01715. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Lorenzo C, Williams K, Haffner SM. Insulin secretion based on the late oral glucose tolerance test period and incident diabetes: the San Antonio Heart Study. Diabet Med. 2012;29:e151–8. doi: 10.1111/j.1464-5491.2012.03660.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.He S, Wang D, Lu Y, et al. Increasing glucagon secretion could antagonize the action of exogenous insulin for glycemic control in streptozocin-induced diabetic rhesus monkeys. Exp Biol Med (Maywood) 2013;238:385–91. doi: 10.1177/1535370213477974. [DOI] [PubMed] [Google Scholar]
- 38.Kawamori D, Kulkarni RN. Insulin modulation of glucagon secretion: the role of insulin and other factors in the regulation of glucagon secretion. Islets. 2009;1:276–9. doi: 10.4161/isl.1.3.9967. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
Data are available on reasonable request.