Diabetes may feature impaired insulin kinetics, which could be aggravated by altered hepatic metabolism and glycemic control. Thus, we examined insulin clearance and its possible determinants in individuals with recent-onset diabetes.
Participants of the German Diabetes Study (GDS) with type 1 diabetes (T1D) (n = 306), type 2 diabetes (T2D) (n = 489), or normal glucose tolerance (control [CON]) (n = 167) underwent hyperinsulinemic-euglycemic clamps for assessment of whole-body insulin sensitivity (M value) and insulin clearance (ICCLAMP). Insulin clearance rates were further calculated during intravenous glucose tolerance tests (ICIVGTT) and mixed-meal tests (ICMMT). Hepatocellular lipid content (HCL) was quantified with 1H-MRS.
Both T1D and T2D groups had lower ICCLAMP (0.12 ± 0.07 and 0.21 ± 0.06 vs. 0.28 ± 0.14 arbitrary units [a.u.], respectively, all P < 0.05) and ICMMT (0.71 ± 0.35 and 0.99 ± 0.33 vs. 1.20 ± 0.36 a.u., all P < 0.05) than CON. In T1D, ICCLAMP, ICIVGTT, and ICMMT correlated negatively with HbA1c (all P < 0.05). M value correlated positively with ICIVGTT in CON and T2D (r = 0.199 and r = 0.178, P < 0.05) and with ICMMT in CON (r = 0.176, P < 0.05). HCL negatively associated with ICIVGTT and ICMMT in T2D (r = −0.005 and r = −0.037) and CON (r = −0.127 and r = −0.058, all P < 0.05). In line, T2D or CON subjects with steatosis featured lower ICMMT than those without steatosis (both P < 0.05).
Insulin clearance is reduced in both T1D and T2D within the first year after diagnosis but correlates negatively with liver lipid content rather in T2D. Moreover, insulin clearance differently associates with glycemic control and insulin sensitivity in each diabetes type, which may suggest specific mechanisms affecting insulin kinetics.
Introduction
Insulin clearance describes the uptake and degradation of insulin in all insulin-sensitive tissues and thereby contributes to the regulation of circulating insulin concentration. The balance among insulin secretion, sensitivity, and clearance prevents inadequate insulinemia and its metabolic consequences (1). The liver is the main site for insulin clearance and accounts for degrading up to 80% of endogenously secreted and 50% of intravenously infused insulin during the first portal passage (2). Extrahepatic insulin clearance occurs mainly in skeletal muscle and kidneys (3,4).
Some studies showed that insulin clearance is blunted in young individuals with incipient stages of type 1 diabetes (T1D) when estimated from oral glucose tolerance tests (5) and in individuals with long-standing T2D when derived from clamp tests (6). In contrast, findings of other studies indicated no difference in insulin clearance between individuals with T2D and those with normal glucose tolerance (7–9). Of note, insulin concentrations vary broadly in recent-onset T2D, which may relate to different degrees of insulin resistance and β-cell dysfunction (10), but other factors such as insulin clearance could also contribute to disturbed insulinemia. Insulin clearance can be also decreased in obesity and nonalcoholic fatty liver disease (NAFLD) (11). Impaired insulin clearance appears to be associated with increased liver fat content, possibly explaining the hyperinsulinemia observed in advanced liver disease (9). Although NAFLD mainly results from augmented lipid availability from hypercaloric nutrition and adipose tissue lipolysis leading to subsequent lipotoxic hepatic insulin resistance (12,13), it has been hypothesized that reduced insulin clearance could also contribute to hyperinsulinemia and thereby accelerate NAFLD (14).
The use of direct arteriovenous methods for measuring insulin secretion and clearance in vivo is limited in humans due to ethical considerations as well as methodological issues, as blood sampling is required from the portal vein and hepatic artery and vein with simultaneous assessment of hepatic blood flow. To surmount these limitations, investigators have used indirect approaches based on kinetic modeling as an accurate alternative to calculate insulin clearance during oral (e.g., mixed-meal tests [MMT]) and intravenous (frequently sampled intravenous glucose tolerance test [IVGTT]) glucose challenges (15). However, insulin clearance as assessed under dynamic conditions of oral (ICMMT) and intravenous (ICIVGTT) glucose loading as well as hyperinsulinemic clamps (ICCLAMP) has not yet been compared in cohorts with newly diagnosed diabetes.
Despite the common feature of hyperglycemia and impaired β-cell function, T1D and T2D differ in their pathophysiology and other metabolic abnormalities. We therefore hypothesized that also insulin clearance may be distinctly modulated due to specific metabolic features of T1D and T2D likely affecting insulin clearance such as fatty liver disease and insulin resistance.
Hence, in this study we aimed to assess insulin clearance by distinct methods and their possible associations with insulin sensitivity and hepatocellular lipids (HCL) in individuals with recent-onset T1D or T2D or with normal glucose tolerance. We hypothesized that insulin clearance is impaired in recent-onset diabetes and relates to higher hepatic lipid content.
Research Design and Methods
Study Population
The study included 489 individuals with T2D and 306 with T1D with a known duration of diabetes of <12 months as well as 167 with normal glucose tolerance (control [CON]), all recruited from the multicentric prospective German Diabetes Study (GDS) (10,16). Diagnosis of diabetes was based on the criteria of the American Diabetes Association (17). All participants provide written informed consent prior to enrollment. The study is registered at ClinicalTrials.gov (clinical trial reg. no. NCT01055093), is approved by the ethics committee of the Medical Faculty of Heinrich Heine University (reference number 4508), and complies with the Declaration of Helsinki (16). The exclusion criteria have previously been described in detail (16) and comprise, among others, pregnancy, HbA1c >75 mmol/mol (9.0%), severe acute or chronic diseases including severe liver, kidney, or inflammatory diseases, or a history of cancer.
All participants underwent detailed clinical examination and blood chemistry analyses in the Clinical Research Center of the German Diabetes Center (DDZ) (16). In line with previous studies oral glucose-lowering medication was discontinued for at least 3 days before the metabolic tests (16,18). Short-acting insulin was discontinued on the morning prior to metabolic tests and long-acting insulin for at least 12 h.
Laboratory Tests
Routine laboratory parameters and diabetes-related autoantibodies were measured systematically for every study participant as previously described (16). The estimated glomerular filtration rate was computed with the Cockcroft-Gault formula.
Metabolic Tests
All participants underwent a modified Botnia clamp test, which consists of an IVGTT followed by a hyperinsulinemic-euglycemic clamp test (16) and an MMT (18) on two separate days. The IVGTT was started with administration of a 30% glucose infusion bolus (1 mg/kg body wt) followed by repeated blood sampling for 60 min. Total C-peptide secretion, as a measure of ββ-cell function, was determined from the incremental area under the curve (AUC) during IVGTT. Then human insulin was given as a primed infusion [10 mU ⋅ (kg body wt)−1 ⋅ min−1] for 10 min continued by constant infusion [1.5 mU ⋅ kg (body wt)−1 ⋅ min−1] (Insuman Rapid; Sanofi, Frankfurt, Germany). Blood glucose concentrations were maintained at 90 mg/dL with a variable 20% glucose infusion. During the clamp test, whole-body insulin sensitivity (M value) was assessed from mean glucose infusion rates during steady state with glucose space correction (16). A primed-continuous infusion of 98% enriched D-[6,6-2H2]glucose was initiated at min 0 and continued until the end of the clamp test for assessment of endogenous glucose production (EGP) and hepatic insulin sensitivity from % EGP suppression during clamp steady state (19).
On another day, the MMT was performed with a standardized liquid meal (BOOST; Nestlé HealthCare Nutrition) and blood samples were obtained before (−1 min) and 15, 30, 60, 90, 120, and 180 min after meal ingestion for measurement of blood glucose as well as plasma insulin, C-peptide, and incretins (18).
On another day, the glucagon stimulation test was performed as previously described (16). Blood glucose, plasma insulin, and C-peptide levels were measured before (0 min) and C-peptide was also measured after (6 min) an intravenous injection of 1 mg glucagon (GlucaGen; Novo Nordisk, Mainz, Germany) (16). The difference in C-peptide concentrations between 0 and 6 min reveals the glucagon-stimulated C-peptide secretion capacity (ΔC-peptide).
Calculation of Insulin Clearance
Using validated formulas (20,21), quantification of insulin clearance was based on measurements of the C-peptide–to–insulin ratios during oral (ICMMT) or intravenous (ICIVGTT) glucose challenges or during clamp tests (ICCLAMP),
where Vins is the volume of distribution of insulin, equaling 141 mL/kg;Of note, this formula includes consideration not of the currently obsolete AUC(C-peptide) / AUC(ins) but, rather, the AUC of the ratio C-peptide(t)/insulin(t) at each time point (t), where 180 is the last time point and 0 is the initial, fasting time point of IVGTT or MMT or clamp test (21).
Whole-body insulin clearance (ICCLAMP) was calculated during clamp steady state (ISS) (22),
where II is the insulin infusion rate (mU × (kg body wt)−1 ⋅ min−1), EI is the residual endogenous insulin secretion rate, and ISS is clamp steady-state plasma mean insulin levels measured from 200 to 240 min (mU ⋅ min−1).The EI during the clamp test was calculated as previously described (20):
MRI and MRS
HCL was measured with MRI and 1H-MRS, on a 3.0T magnet (Achieva X-series; Philips Health Care, Best, the Netherlands) (23). For quantification of HCL a fast localizer MRI was performed and the volume of interest (25 × 25 × 25 mm3) for 1H-MRS was placed in the posterior region of the liver with avoidance of major vessels and the gallbladder. After automatic shimming, 1H spectra were acquired using stimulated echo acquisition mode sequence (repetition time/echo time/mixing time = 4,000/10/16 ms) (24). Both water-suppressed and nonsuppressed 1H spectra were taken within the identical voxel, and the water signal at 4.7 ppm in non-water-suppressed MRS served as internal reference for lipid quantification as previously described (23,24).
Statistics
Numeric variables are presented as mean ± SD. Data were also summarized in percentages for categorical variables. Categorical variables were compared with χ2 or Fisher exact test. For the continuous parameters with normal distribution, two-tailed t test was used. Logarithmic transformations were performed for the nonnormally distributed parameters. Comparisons among the three groups were performed with ANOVA. Pearson or Spearman correlations were used for numerical variables. In regression models, data were adjusted for possible confounders. Correlations were considered strong at r ≥ 0.600, moderate at 0.200 ≤ r < 0.600, and weak at r < 0.200. A two-tailed P value <0.05 was considered to indicate a statistically significant difference or correlation. Statistical analyses were performed with SAS (version 9.4; SAS Institute, Cary, NC). Figures were drawn with GraphPad Prism (version 9; GraphPad Software, San Diego, CA).
Data and Resource Availability
The data sets used for the current study are available from the GDS (principal investigator: M.R.) on reasonable request.
Results
Participants’ Characteristics
Anthropometric and clinical data of the participants are listed in Table 1. Both diabetes groups presented with comparable glycemic control (HbA1c) and known diabetes duration. As expected, there was more frequent use of antihypertensive drugs, lipid-lowering drugs, and metformin among the T2D group and of insulin among the T1D group (Supplementary Table 1).
Insulinemia, Insulin Sensitivity, and Insulin Clearance
Compared with CON, T2D featured almost doubled fasting C-peptide and insulin concentrations (both P < 0.05) (Table 1 and Fig. 1A). Whole-body insulin sensitivity (M value) was higher in CON than in T1D and T2D (all P < 0.05) (Table 1). Hepatic insulin sensitivity (% EGP suppression) did not differ among CON, T1D, and T2D (100 ± 26%, 95 ± 45%, and 92 ± 18%, respectively; P = 0.56).
ICCLAMP was higher in CON than in T1D and T2D (0.28 ± 0.14 vs. 0.12 ± 0.07 and 0.21 ± 0.06 arbitrary units [a.u.], respectively; all P < 0.05). ICIVGTT was lowest in T1D (Fig. 1C). ICMMT was higher in CON than T1D and T2D (1.20 ± 0.36 vs. 0.71 ± 0.35 and 0.99 ± 0.33 a.u.; all P < 0.05) (Fig. 1D).
Measurements of insulin clearance derived from IVGTT and MMT showed excellent agreement (r = 0.812; P < 0.001) (Supplementary Fig. 1A). The distinct measurements of insulin clearance showed moderate and strong correlations in each group, individually (Supplementary Fig. 1B–D). The highest correlation coefficients were observed for CON. The presence of steatosis did not affect the strength of these correlations in T2D but showed relevant differences between T1D with and T1D without steatosis (Supplementary Fig. 2).
Within all groups, ICCLAMP was correlated with C-peptide secretion during the IVGTT (incremental AUC, CON r = 0.198, T1D r = 0.313, T2D r = 0.132; all P < 0.05) as well as with ΔC-peptide from the glucagon stimulation test in the diabetes groups (T1D r = 0.578, T2D r = 0.211; all P < 0.05).
Liver Lipid Content and Hepatic Steatosis
HCL was lower in CON and T1D than in T2D (2.7 ± 4.4%, 1.6 ± 4.0%, and 9.1 ± 8.4%, respectively; P < 0.05) (Table 1). Hepatic steatosis, as diagnosed at HCL values >5.56% (25), was present in 16% of CON, 7% of T1D, and 53% of T2D. M value was lower in all groups with steatosis (Fig. 2A). Those with steatosis and with T2D or without diabetes had lower ICMMT than those without steatosis (Fig. 2D). However, the presence of steatosis did not affect ICCLAMP or ICIVGTT in T2D or CON (Fig. 2B and C). Steatosis did not modify any measure of insulin clearance in T1D (Fig. 2B–D). Across all groups, BMI was negatively correlated with ICIVGTT (r = −0.121; P < 0.05).
Correlation Analyses
Across the whole cohort, measures of insulin clearance (ICMMT, ICIVGTT, and ICCLAMP) did not correlate with circulating nonesterified fatty acids, triglycerides, HCL, or M value. Only ICIVGTT negatively associated with HbA1c in all groups combined (r = 0.092; P < 0.05).
Considering each group individually, ICCLAMP did not correlate with HCL, glycemic control, or M value in CON (Supplementary Fig. 3A–C). ICIVGTT correlated positively with whole-body insulin sensitivity (r = 0.199; P < 0.05) (Supplementary Fig. 3F) and negatively with HCL (r = −0.127; P < 0.05) (Supplementary Fig. 3D). Similarly, ICMMT correlated positively with whole-body insulin sensitivity (r = 0.047; P < 0.05) (Supplementary Fig. 3I) and negatively with HCL (r = −0.058; P < 0.05) (Supplementary Fig. 3G) in CON.
In T1D, ICCLAMP, ICIVGTT, and ICMMT were negatively correlated with HbA1c (all P < 0.05) (Fig. 3B–F).
In T2D, ICCLAMP correlated negatively with HbA1c (r = −0.004; P < 0.05) (Fig. 3B) but not with HCL or M value (Supplementary Fig. 3A and C). On the other hand, ICIVGTT correlated positively with M value (r = −0.178; P < 0.05) (Supplementary Fig. 3F) and negatively with HCL (r = −0.005; P < 0.05) (Fig. 3D). ICMMT correlated negatively with HCL (r = −0.037; P < 0.05) (Fig. 3E).
Conclusions
The results of this study demonstrate that insulin clearance is already reduced in those with recently diagnosed T1D and T2D in comparison with individuals without diabetes. Furthermore, insulin clearance as assessed with insulin and C-peptide excursions during MMT is reduced in individuals with steatosis in the presence or absence of T2D compared with those without steatosis. Insulin clearance correlates with whole-body insulin sensitivity in individuals with normal glucose tolerance and T2D but not T1D. On the other hand, certain measures of insulin clearance are associated with impaired glycemic control in individuals with T1D, who also feature lower hepatic insulin extraction in comparison with individuals with T2D or normal glucose tolerance. These findings suggest diabetes-type specific alterations of insulin clearance.
The current study included use of different methods to assess insulin clearance. Insulin clearance was assessed from steady-state insulin infusion rates and insulin concentrations during the hyperinsulinemic clamp, which also takes into account extrahepatic clearance (26). Further validated methods allowed for estimating insulin clearance during IVGTT and MMT from the C-peptide–to–insulin ratios at different time points, based on the assumption that, unlike insulin, C-peptide does not undergo hepatic metabolism (27). Nevertheless, contribution of peripheral insulin clearance can be expected to decrease proportionally to the degree of insulin resistance (21,28). Taken together, ICCLAMP, ICIVGTT, and ICMMT showed very good agreement across all examined groups.
The findings of lower insulin clearance in individuals with T1D than in those with T2D and without diabetes are in agreement with the findings of one study in youth with T1D, who were compared with glucose tolerant subjects (5). Young people with T1D may have their highest insulin clearance in mid-adolescence, which declines with age and insulin resistance (29). In the current study, ∼91% of individuals with T1D still had detectable C-peptide concentrations (>0.1 ng/dL), while previous study populations mainly comprised T1D cohorts with undetectable C-peptide levels, markedly longer mean diabetes duration (∼15 years), and worse glycemic control (30).
In addition, other factors such as presence of components of the metabolic syndrome or impaired liver function could also affect insulin clearance. In the current study, hepatic steatosis affected insulin clearance as measured with MMT in T2D and CON, whereas HbA1c negatively associated with insulin clearance more consistently in recent-onset T1D, suggesting a major role for chronic hyperglycemia in modulating hepatic insulin kinetics. Still, ICCLAMP negatively correlated with HbA1c also in T2D, which suggests that hyperglycemia, as a common feature of both diabetes types, may be a main driver of insulin clearance regulation. Furthermore, it should be noted that individuals with T2D were more obese and more insulin resistant, conditions that are independently associated with reduced insulin clearance (11) and contribute to the observed differences between groups. Interestingly, the prevalence of NAFLD in the T1D group was lower compared with 8.6% reported in a recent meta-analysis (31). This could be due to the use of different methods of quantifying liver lipids but more likely results from the features of the current study population. Indeed, this meta-analysis did not address disease duration, glycemic control, or lifestyle habits, which can affect liver fat content and prevalence of NAFLD, as previously discussed (23). Of note, all included participants had a known disease duration of maximally 1 year, near-normoglycemic control, and a presumably healthier lifestyle and/or better compliance, which probably account for the lower prevalence of steatosis in comparisons with some other studies in humans with diabetes (25,31).
Interestingly, lower insulin clearance was observed in T1D and T2D despite the short known disease duration. So far, studies on insulin clearance rates in T2D yielded conflicting results, with reporting of lower (32), unchanged (7), or even higher (33) insulin clearance in comparisons with metabolically healthy individuals. Such differences may be due to the use of different formulas and different tests (e.g., hyperinsulinemic clamp vs. MMT or oral glucose tolerance test) and the possible underestimation of insulin clearance rates, by ∼15% in individuals with glucose tolerant and by 5% in those with T2D (11).
We further found a reduction in insulin clearance specifically in individuals with or without T2D in the presence of increased liver lipid content. Authors of previous analyses hypothesized a causal link between liver lipid content and whole-body, but not hepatic, insulin clearance (3). In obese Japanese individuals without diabetes, insulin clearance was also decreased in the presence of liver steatosis, independently of BMI (34). In humans without diabetes increased liver lipid content seems to be associated with impaired insulin clearance and hepatic insulin resistance (35). Bril et al. (36) compared data from individuals with histologically proven steatosis, nonalcoholic steatohepatitis, and cirrhosis suggesting that even a mild increase in liver lipid content associates with decreased insulin clearance as derived from the clamp test. The present finding of an exclusive decrease in insulin clearance in the presence of steatosis in the CON and T2D groups may be due to the groups’ higher degree of insulinemia and insulin resistance. Nevertheless, it remains to be established whether hyperinsulinemia and reduced insulin clearance cause peripheral insulin resistance or vice versa, as hyperinsulinemia may cause steatosis due to its adipogenic effect (19). Prolonged portal hyperinsulinemia can lead to insulin receptor downregulation with subsequently reduced clearance (20). In this context, chronic hyperinsulinemia appears to be both cause and consequence of impaired insulin clearance as part of a vicious cycle. Defective insulin clearance leads to hyperinsulinemia, while chronic hyperinsulinemia favors de novo lipogenesis in liver, further affecting its ability to degrade insulin (11).
In the present studies the three measures of insulin clearance exhibit differences in the association with glycemic control and insulin sensitivity despite their overall good agreement. This may derive from differences in hepatic exposure to insulin in the three tests. IVGTT and MMT have high portal exposure related to EGP, while during the clamp high insulin concentrations are present mainly in hepatic artery due to the exogenous intravenous application. In T2D, ICCLAMP only correlated with glycemic control, ICIVGTT only with insulin sensitivity, and ICMMT only with HCL, suggesting the operation of different mechanisms depending on the route of glucose administration and degree of insulinemia. In this context, a recent study concluded that decreases in hepatic insulin clearance in T2D may reflect a compensatory response to the inadequate insulin secretion to overcome insulin resistance and maintain normoglycemia (20).
It was previously shown that insulin sensitivity and insulin clearance decline gradually from normal to impaired glucose tolerance to T2D and might therefore predict progression of glucose intolerance (37). Piccinini et al. (38) reported that lower hepatic insulin clearance is associated with higher risk for T2D. For this group, the current study showed that certain measures of insulin clearance correlated positively with insulin sensitivity but negatively with liver lipid contents, both of which are key determinants of the pathogenesis of T2D (12). Thus, one might hypothesize that a decrease in insulin clearance along with an increase in peripheral insulin concentrations could serve as a tool for assessing diabetes risk (37) for the subtypes of prediabetes and high risk of early complications.
The current study benefits from the relatively large size of the comprehensively phenotyped study population with a well-defined short diabetes duration, which reduces the impact of chronic diabetes-related metabolic alterations such as—but not only—hyperglycemia and dyslipidemia. This study also has some limitations. First, all participants are of European descent, which does not allow for generalization to other ethnicities. Indeed, decreased peripheral rather than hepatic insulin clearance may be more common in Japanese individuals with T2D (39) and African Americans may have higher insulin concentrations but lower insulin clearance (6,40). Second, direct arteriovenous methods were not applied for ethics reasons and the interpretation of insulin clearance from the C-peptide–to–insulin ratio is limited in humans with T1D due to exogenous insulin application. The low prevalence of steatosis in T1D may render the interpretation of the analysis restrictive for this specific group. Also, different glucose-lowering treatments may differently affect measures of insulin clearance (4). Although the study participants were not treatment naïve, glucose-lowering treatment was discontinued for at least 3 days prior to the study (16).
In conclusion, in this study we found that insulin clearance 1) is already reduced in humans with recent-onset T1D and T2D, 2) is further reduced in humans in the presence of hepatic steatosis independently of diabetes, and 3) correlates with liver lipid content rather in T2D and with impaired glycemic control rather in T1D. These differences in insulin clearance suggest specific alterations of insulin kinetics.
Clinical trial reg. no. NCT01055093, clinicaltrials.gov
This article contains supplementary material online at https://doi.org/10.2337/figshare.24156732.
O.-P.Z. and S.A. contributed equally to this work.
A full list of members of the GDS Group can be found in the appendix.
Article Information
Acknowledgments. The authors thank the staff of DDZ for their excellent support.
Funding. This study was supported in part by DDZ. The GDS was initiated and financed by DDZ, which is funded by the Ministry of Culture and Science of the State of North Rhine-Westphalia and the German Federal Ministry of Health (BMG) and by grants of the Federal Ministry for Research (BMBF) to the German Center for Diabetes Research (DZD) (DZD Grant 2016) and the Schmutzler-Stiftung. The research of O.P.Z is supported by grants from the European Foundation for the Study of Diabetes (Rising Star Fellowship Programme) and Deutsche Diabetes Gesellschaft (DDG) (Adam-Heller prize) and the Multi-Omics Data Science (MODS) initiative (grant PROFILNRW-2020-107-A). The research of M.R. is further supported by grants from the German Research Foundation (DFG) (GRK 2576) and the German European Community (HORIZON-HLTH-2022-STAYHLTH-02-01: Panel A) to the INTERCEPT-T2D consortium.
The sole responsibility for the content of this publication lies with the authors. The funding sources had no role in study design, data collection, data analysis, data interpretation, or writing of the manuscript.
Duality of Interest. M.R. received fees for lectures and/or advisory boards from AstraZeneca, Boehringer Ingelheim, Eli Lilly, Novo Nordisk, and Target RWE and investigator-initiated research support from Boehringer Ingelheim, Nutricia/Danone, and Sanofi. O.-P.Z. and R.W. declare lecture fees from Sanofi. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. O.-P.Z. and S.A. drafted the manuscript and researched data. O.-P.Z., S.A., Y.Ka., and K.B. performed the metabolic tests and researched data. Y.Ku. and V.S.-H. performed the metabolic imaging assessments. P.B. performed the statistical analyses. A.G., J.S., R.W., V.B., and M.R. contributed to the discussion and reviewed and edited the manuscript. All authors read, critically reviewed, and approved the final manuscript. M.R. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Prior Presentation. Parts of this work were presented at the 79th Scientific Sessions of the American Diabetes Association, San Francisco, CA, 7–11 June 2019, and at 55th Annual Meeting of the European Association for the Study of Diabetes, 16–20 September 2019, Barcelona, Spain.
Appendix
The GDS group consists of M. Roden (speaker), H. Al-Hasani, B. Belgardt, G. Bönhof, G. Geerling, C. Herder, A. Icks, K. Jandeleit-Dahm, J. Kotzka, O. Kuß, E. Lammert, W. Rathmann, S. Schlesinger, V. Schrauwen-Hinderling, J. Szendroedi, S. Trenkamp, and R. Wagner and their co-workers who contributed to the design and conduct of the GDS.