Weight Loss–Independent Effect of Liraglutide on Insulin Sensitivity in Individuals With Obesity and Prediabetes

Glucagon-like peptide 1 (GLP-1) receptor agonists decrease hemoglobin A_1c, cause weight loss, and reduce incidence of major adverse cardiovascular events (1). Although GLP-1 receptor (GLP-1R) agonists are known to decrease fasting and postprandial glucose levels, their effects on other metabolic hormones, including insulin, glucagon, and endogenous incretins, are less clear and vary by population and study design (2–9). Prior studies are also complicated by the confounding effects of weight loss, which modifies insulin resistance and metabolic hormones (10). Dipeptidyl peptidase 4 (DPP-4) inhibitors increase endogenous GLP-1 as well as other incretins such as gastric inhibitory peptide (GIP) and nonincretin peptides (1). As compared with GLP-1R agonists, DPP-4 inhibitors are less effective at reducing glucose levels and do not cause weight loss or reduce incidence of cardiovascular events (1). Thus, it is unclear which metabolic effects of GLP-1R agonist treatment are due to specific receptor activation and which are due to weight loss. In addition, the question remains if there are differences in the metabolic effects of GLP-1R activation by endogenous GLP-1 versus targeted pharmacologic therapy.

To address the question of weight loss–dependent versus–independent effects of GLP-1R agonist treatment as well as the effects of endogenous versus pharmacologic GLP-1R activation, we performed a three-arm, parallel, randomized study of the GLP-1R agonist liraglutide, the DPP-4 inhibitor sitagliptin, and a hypocaloric diet. We enrolled individuals with obesity and prediabetes and performed mixed-meal tolerance tests at baseline, after 2 weeks of therapy to evaluate short-term effects before significant weight loss, and after 14 weeks of therapy to evaluate long-term effects. We measured fasting and postprandial glucose and metabolic hormone levels. We additionally assessed GLP-1R–dependent effects by performing a crossover study of the GLP-1R antagonist exendin(9-39) and placebo at both 2 and 14 weeks.

Details of the protocol and the primary outcomes have been previously published (11). We now report prespecified secondary measures. In brief, men and women aged 18–65 years with obesity (BMI ≥30 kg/m²) and prediabetes, according to the American Diabetes Association criteria, were enrolled in this trial (ClinicalTrials.gov identifier NCT03101930). Participants underwent a baseline study day for anthropometric measurements and ingested a standardized mixed meal of 712 calories (16 oz [473 mL] Glucerna 1.5 Cal, Abbott Nutrition, Columbus OH) over 15 min. Participants were then randomized in a 2:1:1 ratio, stratified by race, to treatment with liraglutide 1.8 mg/day (Novo Nordisk), sitagliptin 100 mg/day (Merck and Co, Inc.), or a hypocaloric diet. Treatment with liraglutide or sitagliptin was double-blind and placebo controlled, whereas treatment with the hypocaloric diet was unblinded.

Participants underwent study days at 2 weeks to measure short-term effects of treatment and at 14 weeks to measure sustained chronic effects (Supplementary Fig. 1A). For the first 74 participants, we conducted a two-by-two crossover study of the GLP-1R antagonist exendin(9-39) (Clinalfa; Bachem Distribution Services, Weil am Rhein, Germany; investigational new drug no. 122,217) and placebo to assess the contribution of GLP-1R activation to any observed effects. These participants completed two study days at 2 weeks and two study days at 14 weeks, each separated by 48 h, during which exendin(9-39) or placebo was administered in randomized order (Supplementary Fig. 1B). exendin(9-39) was given as an intravenous bolus of 7,500 pmol/kg over 1 min followed by continuous infusion of 350 pmol/kg/min for the remainder of the study. Infusion with placebo or exendin(9-39) was double-blind. Two hours after the infusion, participants were given the mixed-meal tolerance test. The last 14 individuals studied did not participate in the crossover study with exendin(9-39), due to lack of drug availability, and only underwent two study days, one after 2 weeks and one after 14 weeks. Placebo vehicle was infused during each of these study days.

A YSI glucose analyzer (YSI Life Sciences, Yellow Springs, OH) was used to measure plasma glucose level immediately after collection. All remaining samples were stored at −80°C in aliquots until the time of assay. Plasma insulin samples were collected in tubes containing aprotinin, and insulin was measured by radioimmunoassay (EMD Millipore, Billerica, MA). The assay cross-reacts with 38% intact proinsulin but not with C-peptide (≤0.01%). Samples for GLP-1, C-peptide, GIP, and glucagon were collected in tubes containing aprotinin plus DPP-4 inhibitor. Active or intact GLP-1 [GLP-1(7-36) and GLP-1(7-37)], C-peptide, active GIP, and glucagon were measured via multiplex immunoassay (Meso Scale Discovery; Meso Scale Diagnostics, Gaithersburg, MD). The cross-reactivity of the GLP-1 assay for liraglutide is 0.03%. The cross-reactivity of glucagon for glicentin is 30%.

We previously reported patients’ baseline characteristics, change in weight, fasting blood glucose and insulin levels, and HOMA for insulin resistance (HOMA-IR), and these measures were analyzed as described (11). The prespecified secondary end points reported here include measures of glucose, insulin, and GLP-1 during the mixed-meal test. Additional end points include mixed-meal measures of C-peptide, glucagon, and GIP. The area under the curve (AUC) of the end point was calculated using a linear trapezoidal method, and the peak was defined as the maximal value observed for each participant. All analyses were by assigned groups. Because sex was distributed similarly among treatment arms in this randomized trial, we did not adjust for sex in the statistical analysis of the data.

Changes in outcomes from baseline to 2 and 14 weeks within each treatment group were examined using the Wilcoxon signed-rank test. To evaluate treatment effects between groups, we fit separate, multivariable, generalized least squares linear regression models. Treatment (liraglutide, sitagliptin, or diet), infusion (vehicle or exendin(9-39)), time (2 or 14 weeks), baseline measurement as well as all the two-way and three-way interactions among treatment, time, and infusion were included as independent variables. A compound symmetry structure for within-subject correlation was used. For each model, we examined the residual plots to confirm that model assumptions were not violated. Inferences on the contrasts of interest were conducted using Wald statistic. All the analyses were performed using the statistical software R, version 4.1.0.

All data will be made available upon reasonable request to the corresponding author. No resources were generated during this study.

Data from 88 individuals (liraglutide, n = 44; sitagliptin, n = 22; hypocaloric diet, n = 22) who completed the study are presented, and participant characteristics were reported previously (Supplementary Table 1) (11). Baseline characteristics were similar among treatment arms. Fasting measures of glucose, insulin, C-peptide, GLP-1, GIP, and glucagon were similar at baseline (Table 1). Calculated measures of HOMA-IR, the updated HOMA model (HOMA2), the Matsuda index of insulin sensitivity, insulinogenic index, disposition index, and AUC of insulin to glucose at 30 min (InsAUC30/GluAUC30) were also similar at baseline.

As reported previously, liraglutide caused weight loss at 14 weeks and diet caused significant weight loss at 2 and 14 weeks, whereas sitagliptin did not (Table 2). Liraglutide significantly decreased fasting glucose levels at both 2 and 14 weeks. This was accompanied by a decrease in fasting insulin level at 2 weeks, and a decrease in glucagon level at both 2 and 14 weeks. There was no effect of liraglutide on fasting C-peptide level. Liraglutide significantly improved calculated measures of insulin resistance, with a decrease in HOMA-IR at 2 and 14 weeks, and a decrease in HOMA2 at 2 weeks. We also measured the incretins GLP-1 and GIP. Because of cross-reactivity of liraglutide with the GLP-1 assay, we do not report concentrations of fasting GLP-1 during liraglutide treatment. Liraglutide had no effect on fasting GIP level.

The hypocaloric diet did not change fasting glucose levels (Table 2). Diet decreased fasting insulin levels at 2 and 14 weeks and fasting C-peptide and glucagon levels at 2 weeks but not at 14 weeks. This was accompanied by a decrease in HOMA-IR and HOMA2 at 2 and 14 weeks. There was no effect of diet on fasting GLP-1 or GIP values. By contrast, sitagliptin did not change fasting glucose, insulin, C-peptide, or glucagon levels at 2 or 14 weeks, and did not change insulin sensitivity. As expected, sitagliptin increased fasting levels of both GLP-1 and GIP.

Liraglutide decreased peak postprandial glucose and the AUC for glucose at 2 and 14 weeks (Table 2 and Fig. 1A). This was accompanied by a decrease in peak insulin and C-peptide levels at both 2 and 14 weeks, and a decrease in insulin AUC at 2 weeks and C-peptide AUC at 14 weeks. Insulin and C-peptide levels normalized for glucose showed similar trends (Fig. 1B and C). To further estimate insulin sensitivity and secretion using data from the mixed-meal tolerance test, we calculated additional indices using postprandial measures. Liraglutide increased the Matsuda index of insulin sensitivity at 2 and 14 weeks. There was no effect of liraglutide on the disposition index, the insulinogenic index, or the InsAUC30/GluAUC30. Liraglutide decreased peak glucagon level at 2 weeks but did not significantly change the glucagon AUC (Fig. 2A). To evaluate the effect of liraglutide on GLP-1 after the mixed-meal tolerance test, we subtracted the time 0 level of GLP-1 (which reflects both endogenous GLP-1 and cross-reactive [0.03%] liraglutide). Given that liraglutide drug level is expected to be at steady state and stable throughout the study day, we quantified endogenous GLP-1 release in response to the mixed meal using this time 0–subtracted measure. There was a small increase in peak postprandial GLP-1 at 2 weeks but no other significant effects of liraglutide treatment on endogenous GLP-1 response to the mixed meal (Fig. 2B). Similarly, liraglutide had no effect on GIP levels after a meal (Fig. 2C).

Similar to liraglutide, sitagliptin decreased peak postprandial glucose level and glucose AUC at 2 and 14 weeks (Table 2 and Fig. 1A). There was no effect of sitagliptin treatment on postprandial insulin or C-peptide levels (Fig. 1B and C) with the exception of a decrease in the C-peptide/glucose peak at 14 weeks. There was also no effect of sitagliptin treatment on postprandial measures of insulin sensitivity or insulin secretion. Sitagliptin significantly decreased the postprandial glucagon peak and AUC (Fig. 2A) and increased GLP-1 and GIP peaks and AUC at 2 and 14 weeks (Fig. 2B and C).

Despite significant weight loss, the hypocaloric diet did not decrease peak glucose level or glucose AUC after a mixed meal (Table 2 and Fig. 1A). There was no effect of hypocaloric diet–induced weight loss on postprandial insulin or C-peptide levels (Fig. 1B and C); no effect on Matsuda index, disposition index, insulinogenic index, or InsAUC30/GluAUC30; and no effect on incretin levels (Fig. 2B and C). There was a small decrease in glucagon AUC at 2 weeks but no effect on peak glucagon (Fig. 2A).

As expected and reported previously (11), GLP-1R antagonism with exendin(9-39) increased fasting glucose levels in all treatment groups (Table 3). Exendin also increased peak postprandial glucose level and glucose AUC in all treatment groups (Fig. 3A). In liraglutide-treated individuals, exendin(9-39) increased postprandial insulin and C-peptide levels (both peak and AUC), but there was no significant effect when normalized for glucose (Fig. 3B and Supplementary Fig. 2A). This was accompanied by a significant decrease in the Matsuda index of insulin sensitivity by exendin(9-39) in liraglutide-treated individuals (Table 3). exendin(9-39) also increased postprandial GLP-1 peak and AUC in liraglutide-treated individuals (Fig. 4A) and increased fasting and postprandial glucagon levels (Fig. 4B). Of note, there was a significant interactive effect of liraglutide and exendin on glucagon concentrations after a mixed-meal test (Table 3).

In contrast, in sitagliptin-treated individuals, exendin had no effect on postprandial insulin, C-peptide, or glucagon levels (Table 3). There was a significant interactive effect of sitagliptin and exendin on GLP-1 concentrations after the mixed meal (Fig. 4A). In hypocaloric diet–treated individuals, exendin increased peak insulin level and insulin AUC at 14 weeks, and peak glucagon level and glucagon AUC at 2 weeks. There was no effect of exendin on GLP-1 levels in the hypocaloric diet–treated individuals. There was also no effect of exendin on GIP levels in any treatment arm.

To understand the GLP-1R–specific effects of a long-acting GLP-1R agonist, we compared the metabolic effects of liraglutide with the effects of weight loss induced by hypocaloric intake and with the effects of increasing endogenous GLP-1 by inhibiting DPP-4 with sitagliptin. We further assessed the contribution of GLP-1R activation to the metabolic effects of each treatment, using the GLP-1R antagonist exendin(9-39). This study differed from prior studies in that the effect of GLP-1R antagonism was measured both after short-term treatment (2 weeks) and after 14-week treatment associated with significant weight loss.

We found that liraglutide increased insulin sensitivity as measured by HOMA-IR, HOMA2, and Matsuda index even after 2 weeks of therapy, before there was significant weight loss. The effect of liraglutide on HOMA-IR (and on HOMA2 at 2 weeks) was comparable to the effect of diet-induced weight loss, whereas HOMA-IR and HOMA2 were unchanged during sitagliptin treatment. Similarly, decreased fasting insulin levels, peak postprandial insulin to glucose ratio and AUC (2 weeks), as well as peak C-peptide to glucose ratio in the liraglutide group were all consistent with increased insulin sensitivity. Stimulation of insulin secretion by GLP-1 and GLP-1R agonists is glucose dependent and it is also possible that decreased glucose concentrations during liraglutide treatment may have contributed to decreased insulin secretion (12,13). Several investigators have reported that long-acting GLP-1R agonists increase β-cell function as measured by HOMA2-B (14) and increase insulin sensitivity as measured by clamp (15–17) or mixed-meal test (18), and potential molecular mechanisms of GLP-1 action on insulin sensitivity, including decreasing oxidative stress and inflammation, have been reviewed (19). Decreased appetite during liraglutide treatment (20) could also contribute to decreased fasting insulin level at 2 weeks, even prior to weight loss (21,22). Notably, although fasting calculations of insulin sensitivity, such as HOMA-IR and HOMA2, predominantly reflect hepatic insulin resistance (23), measures incorporating postprandial values, such as the Matsuda index of insulin sensitivity, additionally reflect peripheral insulin resistance (24). The contrasting effect of liraglutide on the Matsuda index with the lack of effect of sitagliptin or a hypocaloric diet suggests differences in tissue insulin sensitivity after these treatments. This is consistent with results of a preclinical study in rats that found liraglutide improves insulin sensitivity in the liver and in muscles in insulin-resistant states (25), and requires further investigation in humans.

GLP-1R agonists and DPP-4 inhibitors differ mechanistically in their effect on endogenous GLP-1 and other incretins. Thus, whereas endogenous postprandial GLP-1 and GIP concentrations were increased during sitagliptin treatment, endogenous incretins were not increased during liraglutide treatment. Despite a significant effect of DPP-4 inhibition on increasing endogenous GIP and GLP-1 concentrations, sitagliptin treatment did not result in weight loss. This finding is relevant to recent trials of the combined long-acting agonists of GLP-1 and GIP, which induce dramatic weight loss (26), and suggests that coactivation of the GIP receptor alone may not account for the enhanced effect of the combined agonist tirzepatide on weight loss compared with GLP-1R agonists alone, as recently reviewed by Nauck and D’Alessio (27). Alternatively, pharmacological GIP agonists and endogenous GIP may differ in their effects at the GIP receptor in vivo. Although tirzepatide and GIP had similar signaling properties at the GIP receptor in vitro and in mice (28), in vivo pharmacodynamics in humans may drive differences in response. Finally, DPP-4 inhibition prevents the degradation of a number of additional peptides (29), which may diminish or counteract the effects of increasing endogenous GIP and account for differences in clinical outcomes between DPP-4 inhibition and dual GLP-1/GIP agonists.

Liraglutide and sitagliptin also differed in their effects on postprandial glucagon concentrations. In this study, liraglutide did not have a significant sustained effect on postprandial glucagon level, similar to results in multiple prior reports (5,7,30). This suggests decreased glucagon level did not contribute to the profound effect of liraglutide on postprandial glucose that persisted at 14 weeks. Sitagliptin significantly decreased postprandial glucagon levels at both 2 and 14 weeks, as has been found in prior studies of DPP-4 inhibition (31–33). Mechanistically, GLP-1 may suppress glucagon both indirectly, including via paracrine effects on the α cell by increasing insulin and somatostatin, as well as directly via a small number of GLP-1Rs on α-cells, as reviewed by Müller et al. (34). The differences seen in glucagon responses after increased endogenous GLP-1 signaling with DPP-4 inhibition versus GLP-1R agonist effects may be due to allosteric modulation by the latter.

Contrasting effects of exendin(9-39) on postprandial GLP-1 and glucagon concentrations in liraglutide-treated and sitagliptin-treated groups also highlight different consequences of GLP-1R activation by the long-acting pharmacologic GLP-1R agonist versus endogenous GLP-1. Thus, exendin(9-39) increased postprandial active GLP-1 to a significantly greater effect in the sitagliptin-treated group compared with in the liraglutide-treated group. This suggests that decreased degradation of endogenous active GLP-1 in the presence of DPP-4 inhibition exerts greater feedback inhibition of L-cell secretion than does the long-acting analog. Conversely, concurrent treatment with exendin(9-39) increased fasting and postprandial glucagon (peak and AUC) levels compared with placebo in the liraglutide-treated group but not in the sitagliptin-treated group. Interestingly exendin(9-39) did not increase fasting or postprandial GIP values in any treatment group, suggesting a lack of feedback inhibition on K cell secretion. A lack of effect of GLP-1R inhibition on GIP concentrations after a meal in individuals with diabetes and/or obesity has been reported in most prior studies (35–46), although Nauck et al. (47) reported an increase in GIP during exendin(9-39) in patients with diabetes and Salehi et al. (48) reported the same in patients undergoing bariatric surgery.

Pharmacokinetic differences in the duration of GLP-1 agonism may have contributed to differences between the effect of liraglutide and of sitagliptin. Sitagliptin acts as a competitive inhibitor of DPP-4 and, thus, effects may be short-acting. The GLP-1R agonist liraglutide and endogenous GLP-1, levels of which were increased during sitagliptin treatment, also differ in their effects on GLP-1R signaling. Although liraglutide and GLP-1 activate Gαs proteins to promote the production of cAMP with equal potency, liraglutide is more potent than GLP-1 in activating phosphorylation of extracellular regulated kinases 1 and 2 (pERK1/2), leading to biased agonism (49). Liraglutide induces greater receptor internalization than GLP-1 and liraglutide-stimulated receptors recycle more slowly. Thus, liraglutide stimulates pERK1/2 in the cytosol as well as at the membrane.

Strengths of this study include differentiation of effects of pharmacologic GLP-1R activation by liraglutide versus endogenous GLP-1R activation by sitagliptin versus caloric restriction–induced weight loss without drug treatment. In addition, evaluation of measures at two time points, including at 2 weeks before weight loss in the liraglutide group, as well as at 14 weeks in the chronic state, allows better appreciation of weight loss–independent mechanisms. Limitations of the study include cross-reactivity of the selected GLP-1 assay with liraglutide at 0.03%, which prevented us from quantifying endogenous GLP-1 levels in the fasting state in liraglutide-treated individuals. Assays measuring active GLP-1 also do not quantify any fraction bound to albumin. The glucagon assay we performed has 30% cross-reactivity with glicentin, another proglucagon-derived peptide that increases after a mixed meal. Thus, our measured glucagon levels may be high; however, liraglutide treatment does not affect glicentin concentrations (30). Furthermore, we selected the 350 pmol/kg/min dose of exendin(9-39), which is consistent with that used in multiple prior studies (47,50); however, a 2021 review by Gasbjerg et al. (51) suggests that an infusion rate of 400 pmol/kg/min is optimal. It is possible that liraglutide’s effects on change in gut motility could drive some of the glucose-lowering effects independently of the measured metabolic hormones and this was not captured in our study. We studied obese individuals with prediabetes rather than diabetes to avoid confounding effects of antidiabetic therapy. Nevertheless, the effects of each treatment appear congruent with observations from studies in individuals with type 2 diabetes.

In conclusion, the GLP-1R agonist liraglutide rapidly improves insulin sensitivity and fasting and postprandial blood glucose levels as early as 2 weeks and prior to weight loss, while decreasing insulin and C-peptide levels, and decreasing fasting glucagon levels in individuals with obesity and prediabetes. These effects are not recapitulated by diet-induced weight loss or increased endogenous GLP-1, and are reversed by GLP-1R antagonism with exendin(9-39). Future studies investigating the molecular mechanisms of GLP-1R activation by these agonists, including on pathways important in insulin resistance such as modulation of reactive oxygen species and inflammation, is crucial as newer GLP-1R agonists and coagonists rapidly enter the market.

Acknowledgments. The authors acknowledge contributions from study nurse Patricia Wright; study dietitian Dianna Olson; research assistants Sara E. Howard, Bradley Perkins, and Eric C. Olson; and laboratory manager Anthony Dematteo, Vanderbilt University Medical Center. The graphical abstract was created with BioRender.com.

Funding. Research reported in this article was supported by the American Heart Association (grant 17SFRN33520017 to M.M., H.N., D.M., J.K.D., C.Y., H.J.S., J.M.L., and N.J.B.); the National Center for Advancing Translational Sciences (grant 5UL1TR002243), and the National Institute of Diabetes and Digestive and Kidney Diseases (grant T32DK007061 to M.M.). This work used the core(s) of the Vanderbilt Diabetes Research and Training Center, funded by grant DK020593 from the National Institute of Diabetes and Digestive and Kidney Diseases. Novo Nordisk provided liraglutide and matching placebo.

Duality of Interest. J.M.L. has served on the advisory board for Mineralys. N.J.B. serves on the scientific advisory board for Alnylam Pharmaceuticals, is a consultant for Pharvaris Gmbh and eBioStar Tech, and owns equity in AbbVie and Johnson & Johnson Pharmaceuticals. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.M. researched data and wrote the manuscript. H.N. and C.Y. performed statistical analysis. D.M. designed the database and managed study days. J.K.D. contributed to study design. J.L.G. researched data. H.J.S. designed and managed the diet intervention and assessments, researched data, and edited the manuscript. K.N. contributed to study design, obtained funding, and edited the manuscript. J.M.L. researched data and edited the manuscript. N.J.B. designed the study, obtained funding, researched data, managed the team, and edited the manuscript. N.J.B. 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 study were presented in abstract form at the ENDO 2022 annual meeting, Atlanta, GA, 11–14 June 2022.