A Randomized Controlled Trial of Dapagliflozin Plus Once-Weekly Exenatide Versus Placebo in Individuals with Obesity and Without Diabetes: Metabolic Effects and Markers Associated with Bodyweight Loss

The World Health Organization statistics for 2016 showed that ~ 39% (1.9 billion) of the global adult population was overweight; of these, ~ 13% (650 million) were obese [1]. Between 1980 and 2016, the prevalence of obesity worldwide more than doubled.

Overweight and obesity are associated with significant comorbidities, such as impaired glucose tolerance, type 2 diabetes mellitus (T2DM), hypertension, and dyslipidemia [2, 3]. Individuals with obesity often have multiple comorbidities [3], placing a substantial health-related burden on both the individual and the healthcare system.

Weight loss among overweight or obese individuals ameliorates impaired glucose tolerance, hypertension, and dyslipidemia [4‐6]. However, bodyweight loss is often not maintained over the long term [7]. Subsequent bodyweight gain results in these conditions returning to baseline levels or worse [4‐6], emphasizing the importance of bodyweight loss maintenance when managing these patients.

Bodyweight loss with single-use pharmacotherapies is generally modest and is often limited by physiologic counterregulatory mechanisms. Combination pharmacotherapies, on the other hand, may achieve greater bodyweight loss, particularly if they reduce counterregulation, thus providing additive or even synergistic effects. In addition, they may allow reduced doses of the individual agents, thereby minimizing side effect risk [8].

Clinical trials of some glucose-lowering medications have demonstrated bodyweight loss among patients with T2DM, especially those investigating selective sodium–glucose cotransporter 2 (SGLT2) inhibitors (e.g., dapagliflozin and empagliflozin) and glucagon-like peptide-1 receptor agonists (GLP-1RAs; e.g., exenatide and liraglutide). Importantly, the bodyweight loss achieved with exenatide and dapagliflozin is maintained over the long term, with trials demonstrating sustained reductions from baseline for over 6 years with exenatide and for up to 4 years with dapagliflozin [9‐11].

The mechanisms by which GLP-1RAs (appetite suppression) and SGLT2 inhibitors (urinary caloric loss secondary to glucosuria) induce bodyweight loss are different and complementary [12‐14]. The recent DURATION-8 study demonstrated that mean bodyweight loss with the combination of exenatide once weekly (QW) and dapagliflozin once daily (QD) among patients with T2DM was greater than that with exenatide and dapagliflozin alone [15].

The current study, which investigated the combination of exenatide QW and dapagliflozin QD on bodyweight loss among individuals with obesity and without diabetes, found significantly greater mean bodyweight loss, systolic blood pressure reduction, and glycemia reduction with active treatment versus placebo during the initial 24-week double-blind phase, which was maintained at 52 weeks in those continuing active treatment during the open-label extension phase [16, 17]. In addition, individuals initially randomized to receive placebo who switched to active treatment during the open-label extension phase achieved similar bodyweight loss at 52 weeks to those remaining on active treatment throughout the study. However, trajectories of bodyweight change varied substantially between individuals [16, 17].

This post hoc analysis aimed to investigate the metabolic effects of dapagliflozin plus exenatide QW on substrate use and selected hormone levels and to conduct exploratory hypothesis-generating analyses on associations between baseline variables and bodyweight loss over time that might account for the observed interindividual variability.

For individuals with overweight and obesity, achieving and maintaining bodyweight loss present many challenges, including physiologic counterregulatory mechanisms of restoring bodyweight [23], hunger-induced food cravings [24], and ongoing sedentary lifestyles [1]. When administered in combination, the complementary mechanisms of action of an SGLT2 inhibitor (increasing urinary caloric loss) and a GLP-1RA (suppressing appetite) may represent a novel treatment approach for obesity.

Counterregulatory mechanisms in response to SGLT2 inhibitor-induced urinary caloric loss have been described, including compensatory increases in appetite/caloric intake [25, 26]. Therefore, the mechanisms of action of GLP-1RAs, including centrally mediated appetite suppression and possibly also delayed gastric emptying [13], may attenuate dapagliflozin-induced counterregulatory mechanisms resulting in improved bodyweight loss.

Meta-analyses of studies among patients with T2DM evaluating exenatide QW and dapagliflozin administered separately estimate bodyweight loss versus placebo at 24 weeks to range from 0.8 [27] to 1.6 kg [28, 29] and 1.6 [30] to 2.2 kg [27, 31], respectively. Assuming an additive effect with combined administration, bodyweight loss versus placebo at 24 weeks with DAPA + ExQW would be expected to be in the approximate range of 2.4–3.8 kg. Although it is not possible to make direct comparisons with other studies or to confirm additive bodyweight loss because of the lack of single-agent comparator arms, in the current study among participants without diabetes bodyweight loss versus placebo at 24 weeks was 4.1 kg. Although the current study did not assess caloric intake or satiety, the magnitude of bodyweight loss supports the hypothesis that the compensatory increase in caloric intake following SGLT2 inhibition may be attenuated to a degree by the combination with a GLP-1RA. Further support that at least additive weight loss can be achieved by combining an SGLT2 inhibitor with agents that increase satiety comes from a recent study of combined therapy with canagliflozin and phentermine among individuals with obesity and without diabetes. In this study, which, unlike the current study, employed a structured diet and exercise program, combined canagliflozin and phentermine achieved a 6.9% reduction in bodyweight at 28 weeks versus placebo compared with reductions of 1.9 and 4.1% with single administration of canagliflozin or phentermine, respectively [32].

Among patients with T2DM [25], estimated caloric intake is increased by an average of 10–15% during chronic treatment with an SGLT2 inhibitor; this offsets about 9 kg of an expected glucosuria-induced bodyweight loss of 11 kg over 2 years [33]. From the magnitude of bodyweight loss observed in the current study, it might be inferred that food intake was reduced by the addition of exenatide QW, thus counteracting the dapagliflozin-induced increase in food intake. However, the mechanisms involved have not been elucidated, and it is not clear whether the degree of bodyweight loss with DAPA + ExQW achieved among participants with obesity and without diabetes is similar to that achieved among patients with T2DM.

In the current analysis, we first explored possible metabolic adaptations during DAPA + ExQW treatment. These included changes in substrate use (FFAs, glycerol, beta-OH-butyrate, and glucose), insulin and glucagon during OGTTs, and insulin secretion and sensitivity. Change from baseline at 24 weeks in 2-h FFAs was significantly increased versus placebo (Table 1). Although no placebo comparisons were available at 52 weeks, DAPA + ExQW significantly increased 2-h FFAs and significantly decreased fasting glycerol and 2-h beta-OH-butyrate over 52 weeks. Fasting beta-OH-butyrate remained unchanged. The changes in 2-h FFAs and fasting glycerol were similar to those observed among individuals without diabetes receiving the SGLT2 inhibitor empagliflozin [34], supporting the hypothesis that SGLT2-induced glucosuria favors a shift towards greater utilization of lipids for energy production [34, 35]. However, the lack of increase in glycerol during the OGTT could suggest that the observed increase in FFAs during the OGTT might be due to reduced FFA uptake into various tissues rather than increased lipolysis in adipose tissue. Given that other work has identified that raised serum FFAs are associated with the subsequent development of insulin resistance [36], our finding of raised postprandial FFAs is of potential concern. However, no detrimental effects on insulin sensitivity have been found in clinical studies of SGLT2 inhibitors. In the current study, change in insulin sensitivity at 24 weeks with DAPA + ExQW was not significantly different from placebo, and among patients with T2DM, dapagliflozin therapy has been reported to improve insulin sensitivity [37]. Of note, in other studies among individuals with overweight/obesity receiving exenatide alone, no changes in postprandial lipolysis [38], enhanced insulin-mediated suppression of lipolysis and serum FFA levels [39], or suppression of FFA levels during the OGTT [40] were observed, indicating that the increase in 2-h FFAs observed in the current study was driven by the dapagliflozin component. Moreover, neither fasting beta-OH-butyrate nor 2-h beta-OH-butyrate was elevated by DAPA + ExQW, in contrast to elevations observed among individuals without diabetes (mild elevations) and in those with diabetes (greater elevations) treated with empagliflozin alone [34]. Indeed, 2-h beta-OH-butyrate was significantly reduced with DAPA + ExQW. While the finding of reduced 2-h beta-OH-butyrate with DAPA + ExQW may be reassuring with respect to infrequent reports of euglycemic ketoacidosis with SGLT2 inhibitor use [41], the recent hypothesis that increased production of ketone bodies with SGLT2 inhibitor treatment may underpin the cardioprotective effect observed in the EMPA-REG OUTCOME trial should also be considered [42, 43]. The lack of a beta-OH-butyrate increase upon coadministration of dapagliflozin with exenatide QW may be explained by glucagon levels remaining unchanged in the current study. This was likely due to the exenatide component inhibiting the expected increase in glucagon when dapagliflozin [44] is administered alone.

The attenuation of the expected increase in glucagon secretion and associated endogenous glucose production with dapagliflozin [44] by coadministration with exenatide QW, together with the significant reductions in glucose (fasting, 120-min OGTT, and AUC) and insulin (fasting, 120-min OGTT, and AUC) levels over 52 weeks (Table 1), suggest that DAPA + ExQW may have a role in diabetes prevention among individuals with obesity. Indeed, we have previously reported a reduction in prediabetes prevalence at 24 and 52 weeks in this population [16, 17]. However, in these participants without T2DM, we did not observe clear changes in insulin secretion (as measured by the IGI) or insulin sensitivity (as measured by the MI-UGE).

Our analysis exploring baseline associations with bodyweight loss at 24 weeks over 52 weeks demonstrated significant findings regarding SNP rs10010131, BMI, and the IGI. However, we did not find significant associations with markers of GLP-1–related pathways (amylin, GLP-1, and IL-6).

The rs10010131 SNP resides in an intron of the wolframin gene (WFS1). This gene encodes a ubiquitously expressed endoplasmic reticulum membrane glycoprotein that participates in the regulation of cellular calcium homeostasis [45]. WFS1 is mutated in Wolfram syndrome, a rare autosomal recessive disorder characterized by diabetes mellitus, optic atrophy, diabetes insipidus, and deafness [46]. In both humans and mice, wolframin deficiency results in loss of pancreatic β cells, possibly as a consequence of an enhanced endoplasmic reticulum stress response leading to increased β-cell apoptosis [47‐49]. In the context of diabetes risk in humans, the G allele of the rs10010131 SNP variant of WFS1 is associated with impaired GLP-1–induced insulin secretion and increased diabetes risk [50‐52], and the A allele of rs10010131 is protective against developing T2DM [53]. In the current study, we found that, for every copy of the A allele in an additive association model, an additional 2.4 kg of bodyweight loss occurred. While some degree of bodyweight loss occurred during the first 24 weeks for placebo-treated participants with two copies of the A allele, which was similar to that achieved among participants with two G alleles receiving DAPA + ExQW, this change was not significantly different from baseline. Thus, it is likely that bodyweight loss with DAPA + ExQW was accentuated among participants with two A alleles and attenuated in those with two G alleles. However, the treatment-by-rs10010131-by-week-24 interaction term was not significant in this analysis, which would rather favor independent effects of the rs10010131 A allele and DAPA + ExQW treatment on bodyweight loss. In a separate investigation, another SNP associated with diabetes risk, rs7903146 in TCF7L2, attenuated bodyweight loss in response to lifestyle intervention among individuals at risk for T2DM [54]. We evaluated rs7903146 in the current study but found no association with bodyweight loss. To the best of our knowledge, this is the first study to describe an association with rs10010131 and bodyweight loss. Further research is required in larger independent samples to confirm/validate whether the rs10010131 variant influences bodyweight loss independently or in response to SGLT2 inhibitors, GLP-1RAs, or combined therapy.

Baseline BMI does not appear to influence the degree of bodyweight loss among patients with T2DM receiving dapagliflozin or exenatide administered alone. However, Ferrannini and colleagues [25] identified that, among patients with T2DM treated with empagliflozin, the estimated compensatory increased caloric intake was inversely correlated with baseline BMI (partial r = −0.34; P < 0.01). Thus, if these findings are equivalent in SGLT2 inhibitor-treated individuals without diabetes, the greater bodyweight loss among individuals with lower baseline BMI receiving DAPA + ExQW might be explained by the exenatide component attenuating the greater degree of dapagliflozin-associated compensatory increase in caloric intake among individuals with lower baseline BMI. This hypothesis is supported by the significant treatment-by-BMI-by-week-24 interaction term in this analysis, suggesting that the treatment effect was influenced by baseline BMI. However, more DAPA + ExQW-treated participants with higher baseline bodyweight had the rs10010131 G/G genotype than placebo-treated participants with higher baseline bodyweight, which may have influenced these findings. Alternatively, the association with lower baseline BMI and greater bodyweight loss observed in the current study may be independent of DAPA + ExQW treatment, as lower baseline BMI is associated with greater bodyweight loss following dietary [55] and surgical [56] interventions.

Individuals with a higher IGI (i.e., more insulin secretion per increase in glucose) showed a poorer bodyweight response with DAPA + ExQW, and the treatment-by-IGI-by-week-24 interaction term was significant in this analysis, suggesting that the treatment effect was influenced by baseline IGI. Although baseline IGI was significantly lower in the DAPA + ExQW versus placebo group, the MMRM analysis of change in bodyweight containing terms for both treatment and baseline IGI adjusted for these differences. This finding is consistent with reports that post-challenge hyperinsulinemia [57], as well as fasting hyperinsulinemia [58], predicts later bodyweight gain. However, other researchers have not replicated this observation [59]. Hyperinsulinemia may not only be a consequence of obesity-associated insulin resistance but may also promote obesity itself.

This study has limitations. First, measures of satiety, appetite, and caloric intake before and during DAPA + ExQW treatment were not undertaken. Second, the small sample size and the presence of missing data for some variables limited the power to detect associations. However, the analytic strategy employing modern variable selection techniques minimized the impact of these limitations. Third, as in any exploratory study without independent validation, statistical overfitting was a potential issue. Consequently, these findings cannot be generalized to future populations undergoing treatment with DAPA + ExQW and should be regarded as hypothesis-generating in nature. Fourth, placebo comparisons were only available from weeks 0–24. While the switch to active treatment at 24 weeks among those who initially received placebo allowed internal validation of the effect of time-invariant covariates, such as the SNP variant rs10010131, which showed a similar pattern of change post-switch as for initial active treatment, this was not the case for time-varying covariates. Because time-varying covariates (e.g., BMI) change over time, the new covariate “baseline” value at 24 weeks among those who switched from placebo to active treatment will vary nonrandomly, making assessments of changes in bodyweight beyond 24 weeks by the baseline value at week 0 unreliable in this group. Fifth, while patients were observed over 52 weeks, obesity is usually a life-long disorder, and longer-term studies are needed to determine the durability of response with DAPA + ExQW as potential attenuation of dapagliflozin-induced compensatory mechanisms by exenatide QW may change over time. Finally, the study design lacked monotherapy arms, so firm conclusions cannot be drawn about potential additive effects of dapagliflozin and exenatide on bodyweight loss.

Metabolic effects following dual therapy with DAPA + ExQW included less suppression of circulating FFAs compared with placebo during the OGTT, suggesting compensatory lipid mobilization for energy production when glucose availability was reduced because of glucosuria. The increase in glucagon secretion previously reported with dapagliflozin alone did not occur with the dual therapy, suggesting that the GLP1 receptor agonist component had an opposite effect to the SGLT2 inhibitor, thus preventing a glucagon rise. Baseline variables associated with greater bodyweight loss during DAPA + ExQW therapy included the A allele of the SNP variant rs10010131, lower baseline adiposity (BMI), and lower glucose-stimulated insulin secretion at baseline. Further research in larger populations with varying baseline demographics and genetic characteristics is required to validate these associations and their potential predictive impact for identification of treatment responders and nonresponders. Such work may also support future efforts in precision medicine within the obesity field.