Obesity is a growing epidemic, and current medical therapies have proven inadequate. Endogenous satiety hormones provide an attractive target for the development of drugs that aim to cause effective weight loss with minimal side effects. Both glucagon and GLP-1 reduce appetite and cause weight loss. Additionally, glucagon increases energy expenditure. We hypothesized that the combination of both peptides, administered at doses that are individually subanorectic, would reduce appetite, while GLP-1 would protect against the hyperglycemic effect of glucagon. In this double-blind crossover study, subanorectic doses of each peptide alone, both peptides in combination, or placebo was infused into 13 human volunteers for 120 min. An ad libitum meal was provided after 90 min, and calorie intake determined. Resting energy expenditure was measured by indirect calorimetry at baseline and during infusion. Glucagon or GLP-1, given individually at subanorectic doses, did not significantly reduce food intake. Coinfusion at the same doses led to a significant reduction in food intake of 13%. Furthermore, the addition of GLP-1 protected against glucagon-induced hyperglycemia, and an increase in energy expenditure of 53 kcal/day was seen on coinfusion. These observations support the concept of GLP-1 and glucagon dual agonism as a possible treatment for obesity and diabetes.
Glucagon is a counterregulatory hormone, secreted at high levels during hypoglycemia and fasting. It promotes glycogenolysis and gluconeogenesis, as well as hepatic fatty acid β-oxidation and ketogenesis (1). The glucagon receptor is expressed in a broad range of tissues, including the liver, kidney, adipose tissue, pancreas, heart, brain, gastrointestinal tract, and adrenal glands (2). Its effects may therefore be more widespread than just control of glucose metabolism, and it is increasingly recognized that glucagon plays a key role in general energy homeostasis. Glucagon has been shown to potently increase satiety and acutely reduce food intake in humans (3). Additionally, glucagon has the ability to significantly increase energy expenditure during infusion in man (4,5) and has been reported to promote nonshivering thermogenesis in brown adipose tissue in rodents (6). Appetite inhibition classically results in defense of body weight by a reduction of energy expenditure (7,8). The increased energy expenditure in association with anorexia induced by glucagon thus potentially enhances its usefulness as an antiobesity therapy.
The prohormone for glucagon, proglucagon, is processed to GLP-1 in the gut. GLP-1 is secreted postprandially in response to direct stimulation of mucosal L cells by nutrients within the gut lumen and indirectly via neuronal pathways within the enteric nervous system. GLP-1 binds to the G-protein–coupled GLP-1 receptor found in pancreatic islet cells as well as brain, heart, and lung tissue (9) and exerts an incretin effect, stimulating glucose-dependent insulin release by β-cells (10). Acute intravenous injection of GLP-1 has also been shown to also reduce appetite and calorie intake (11), an effect that has been observed in lean, obese, and type 2 diabetic volunteers. As a result, GLP-1 is capable of achieving a modest reduction in body weight (12). GLP-1 also causes nausea and delayed gastric emptying, which limits the dose that can be used clinically (13). At present, GLP-1 analogs, such as exenatide and liraglutide, are licensed for the treatment of type 2 diabetes, improving glucose control and resulting in mild weight loss (14).
Obesity is a growing global epidemic. By 2015, projections suggest that 4 billion adults will be overweight and >700 million will be obese (15). It is therefore clear that new strategies are urgently needed to tackle obesity. Both glucagon and GLP-1 are apparently involved in physiological regulation of appetite and are consequently attractive targets for the development of drugs for weight loss. GLP-1 analogs produce only a small weight loss in diabetic subjects or obese patients (14). Moreover, glucagon would be expected to cause hyperglycemia, an undesired effect, especially in patients with diabetes. Dual administration of glucagon and GLP-1, or analogs thereof, could provide additional benefit over GLP-1 analogs alone. We have previously shown that the proglucagon derivative and gut hormone oxyntomodulin (OXM), which is an agonist at both the glucagon and GLP-1 receptors, is able to reduce body weight and increase energy expenditure without causing hyperglycemia in man (16). Interestingly, the anorectic effect of OXM is abolished in Glp1r−/− mice (17), suggesting that OXM exerts its effect on food intake via the GLP-1 receptor. However, it is a relatively weak agonist for the GLP-1 receptor, being less potent by two orders of magnitude compared with GLP-1 (18). This calls into question the mechanism of OXM’s anorectic effect, as OXM is secreted postprandially at concentrations that are in the same order as GLP-1 itself (19). We propose that this unexpectedly strong inhibition of appetite by OXM might be due to its combined action on both the glucagon and GLP-1 receptor. Others have demonstrated that dual agonism at both the GLP-1 and glucagon receptors augments appetite reduction and weight loss in rodents and improves glucose homeostasis in animal models of diet-induced obesity and diabetes (20,21). This approach could combine the appetite-suppressive effects of GLP-1 and glucagon with the energy expenditure–increasing effects of glucagon. Recent work by our group has also demonstrated that the combination of glucagon and GLP-1 does indeed increase energy expenditure and that the hyperglycemic effects of glucagon are counterbalanced by the action of GLP-1 in humans (5).
We hypothesized that the combination of glucagon and GLP-1 might enhance the reduction of food intake over that observed when the respective hormones are given alone. The current study was therefore designed to demonstrate the acute effects of intravenous infusion of GLP-1 and glucagon when given at low doses, both alone and in combination, on food intake (22), glucose homeostasis, and energy expenditure in humans. This approach was taken, instead of using OXM, as it gave us the flexibility to determine the subanorectic doses of GLP-1 and glucagon individually.
Research Design and Methods
This study was reviewed and approved by the West London Research Ethics Committee (10/H0707/80) and carried out according to the principles of Good Clinical Practice and the Declaration of Helsinki.
Sixteen nondiabetic, overweight volunteers with a mean BMI of 27 kg/m2 (range 24–32.9) were recruited by advertisement. All participants underwent health screening including medical history, physical examination, biochemical and hematological testing, and 12-lead electrocardiogram. Any abnormal eating behavior was assessed using the Dutch Eating Behavior Questionnaire (23) and the SCOFF questionnaire (24). Female volunteers were premenopausal, with regular menstrual cycles, and were not taking hormonal contraceptives. Smokers were excluded. Informed and written consent was obtained. Of the initial 16 recruited, 3 were excluded from final analysis, prior to unblinding, due to abnormal eating behavior not picked up by the standard questionnaires. One participant ate less than the minimum required 300 kcal at the acclimatization visit, one participant did not finish eating within the allotted time (20 min), and the third demonstrated progressive aversion to his chosen meal throughout the course of the study as demonstrated by visual analog scores (VAS). Participants’ demographics are described in Table 1.
An initial dose-finding phase was undertaken in order to establish a subanorectic dose of glucagon. We used a known subanorectic dose of GLP-1 (11,25). After the dose-finding phase, participants attended for five study visits. The first visit was an unblinded acclimatization visit, during which participants were infused with placebo alone (Gelofusine; B. Braun, Crawley, U.K.) in order to become familiar with study protocol. The four subsequent studies were conducted in a double-blind, four-way crossover, randomized, controlled manner at least 2 days apart. Infusions consisted of 1) placebo (Gelofusine), 2) GLP-17–36 amide (0.4 pmol/kg/min; Clinalfa Basic, Bachem, Switzerland), 3) glucagon (2.8 pmol/kg/min; Novo Nordisk, Crawley, U.K.), or 4) combined GLP-1 and glucagon at the above doses. Gelofusine was used as the vehicle for hormone infusions in order to minimize adsorption of peptides to infusion lines and syringes (26).
Volunteers attended the clinical research facility at 0830 h, having fasted from 2200 h the night before, and refrained from alcohol and strenuous exercise for the preceding 24 h. The study room was kept at a consistent temperature of 21°C, which was centrally controlled.
The study commenced at −60 min with the placement of two venous cannulae: one for blood sampling and one for intravenous hormone infusion. After cannulation, volunteers were encouraged to relax and were seated in reclining chairs. They were permitted to watch television or listen to music. After 30 min (−30 min), they were placed under an indirect calorimeter canopy (Gas Exchange Monitor; GEMNutrition, Daresbury, U.K.). Prior to measurement of energy expenditure, the calorimeter was calibrated with “zero” (0.000% O2, 0.000% CO2) and “span” (20% O2, 1.125% CO2) gases (BOC Gases, Surrey, U.K.). Indirect calorimetry measurement was performed as previously described (5) for 30 min, allowing time for initial stabilization of readings, with the last 10 min of measurements used for analysis. Resting energy expenditure (REE), respiratory quotient (RQ), and carbohydrate and fat oxidation rates were calculated from the VO2 and VCO2 measured at 1-min intervals, and adjusted for urinary nitrogen excretion (27,28). After 30 min of calorimetry, the canopy was removed, and infusion of hormones was commenced (0 min). The infusion was initially ramped in order to rapidly achieve a steady-state plasma concentration of hormone. Ramping was carried out at four times the nominal infusion rate for 5 min and then twice the nominal infusion rate for a further 5 min and then reduced to the nominal rate for the remaining 120 min. At 40 min, further 30-min measurements of REE and substrate oxidation rates were made, still in the fasting phase. At 90 min, an ad libitum meal of known specific calorific value was served (spaghetti bolognese 188 kcal/100 g, chicken tikka masala 178 kcal/100 g, or macaroni cheese 194 kcal/100 g; Sainsbury’s Supermarkets Ltd, London, U.K.). Participants had tasted their chosen meal during the acclimatization visit, and all deemed it to be palatable. The same meal was served at all five visits. Participants were allowed 20 min to eat and instructed to eat until comfortably full and then stop. The hormone infusion continued for 120 min in total and was terminated 10 min after the end of the meal. Participants remained in the study room for 60 min after termination of the infusion. At 180 min, the cannulae were removed, and the participants emptied their bladders. Urinary volume was measured in order to calculate urinary nitrogen excretion for estimation of protein oxidation. The participants were then discharged home.
During the study, pulse and blood pressure were measured at −60, −30, 0, 40, 70, 90, 120, 150, and 180 min. At these times, blood samples were also taken for measurement of glucose, insulin, glucagon, and active GLP-1. Glucose and insulin levels were measured by the Department of Chemical Pathology, Imperial College Healthcare National Health Service Trust (coefficient of variation [CV] <5% and <10%, respectively, across the working range). Samples for active GLP-1 and glucagon were collected in lithium heparin tubes containing 1,000 kallikrein inhibitor units. Active GLP-1 and total ghrelin were measured using commercially available ELISA kits (Millipore, Livingston, U.K.) according to the manufacturer’s instructions (CV <7% and 8%, respectively), as was acylated ghrelin (BioVendor, Brno, Czech Republic) (CV <7%). Glucagon and total peptide YY (PYY) were assayed according to established immunoassay protocols by in-house radioimmunoassay (10,29) (CV <10% and <15%, respectively). At each of the above time points, a VAS was completed by the participant for assessment of nausea and satiety.
Statistical analysis was carried out using GraphPad Prism 5.0d (GraphPad Software, San Diego, CA). Two-way repeated-measures ANOVA with Bonferroni post hoc test was used to compare differences in glucose, insulin, PYY levels, blood pressure, pulse, and VAS. One-way repeated-measures ANOVA with Newman-Keuls and Bonferroni post hoc tests was used to compare food intake, change in REE, and substrate oxidation rates between groups. Area under the curve (AUC) was calculated using the trapezoidal rule, and differences between treatment groups were compared using one-way repeated-measures ANOVA with Tukey post hoc test. Paired Student t test was used to compare ghrelin levels at baseline and during infusion. Data are reported mean ± SEM unless otherwise stated.
Baseline plasma active GLP-1 levels at −30 min were 4–5 pmol/L. In the experimental groups receiving GLP-1 alone or GLP-1 with glucagon, levels rose to 11–16 pmol/L at 70 min postinfusion (Fig. 1A). Although active GLP-1 plasma levels appeared to be lower during the combination infusion compared with the GLP-1–only infusion, AUCGLP-1 was not significantly different during the infusion period (0–120 min [Supplementary Fig. 1A]). Mean plasma glucagon levels at −30 min were 14–19 pmol/L, rising to 147–173 pmol/L at 70 min in those groups receiving glucagon infusion (Fig. 1B).
Plasma glucose and serum insulin responses to placebo, glucagon, GLP-1, or combination infusions are shown in Fig. 2. In the placebo group, glucose and insulin remained constant during infusion and, as expected, rose in response to the meal served at 90 min. Glucagon infusion caused a rise in glucose from 4.8 ± 0.08 mmol/L to a peak of 6.5 ± 0.3 mmol/L at 40 min, with a corresponding rise in insulin to 31.2 ± 3.8 mU/L. GLP-1 infusion reduced plasma glucose during infusion from 4.9 ± 0.1 mmol/L to 4.1 ± 0.3 mmol/L at 40 min, with serum insulin levels similar to that observed in the placebo arm. After glucagon and GLP-1 coadministration, AUCglucose was similar to that seen with placebo and significantly lower than with glucagon alone (Fig. 2C). There was a significant increase in insulin during the glucagon/GLP-1 coinfusion of greater magnitude than seen with GLP-1 or glucagon alone (Fig. 2D). Postmeal, where calorie intake differed between treatment groups, glucose and insulin levels were not significantly different between groups.
As expected, glucagon alone and GLP-1 alone, at the doses given, did not significantly reduce food intake. However, glucagon and GLP-1 coinfusion significantly reduced food intake by 13% at the study meal compared with placebo (P < 0.05), which was also a significantly greater reduction than seen during infusion of glucagon alone (P < 0.05) or GLP-1 alone (P < 0.05) (mean energy intake at study meal: 1,086 ± 110.1 kcal [placebo], 1,086 ± 96.9 kcal [glucagon], 1,052 ± 81.3 kcal [GLP-1], and 879 ± 94.2 kcal [combined glucagon plus GLP-1]) (Fig. 3).
Neither the palatability of the buffet meal nor other satiety-related VAS responses were altered significantly by any of the infusions (Fig. 4A, C, and D). The nausea score significantly increased postmeal (120 min) during the combined infusion of glucagon and GLP-1 (Fig. 4B). Three participants reported mild nausea after the combined infusion, and two participants vomited after glucagon infusion. In all cases, this occurred postmeal between 120 and 160 min.
There were no significant differences in baseline REE between groups: 1,336 ± 65.8 kcal/day (placebo), 1,314 ± 53.0 kcal/day (glucagon), 1,330 ± 71.9 kcal/day (GLP-1), and 1,341 ± 56.6 kcal/day (combined glucagon plus GLP-1); P = 0.7275. The mean within-subject CV was 4.1 ± 1.3%. After infusion, there was a trend toward higher REE in response to glucagon alone and glucagon/GLP-1 coadministration by a mean of 66.8 and 52.5 kcal/day, respectively (Fig. 5A).
RQ values and carbohydrate oxidation rates at baseline were similar in all treatment groups, and RQ did not change after GLP-1 infusion. A significant increase in RQ and carbohydrate oxidation was observed with both the glucagon and combination infusions (Fig. 5B and C). Glucagon alone and in combination with GLP-1 significantly reduced fat oxidation rates (Fig. 5D). Protein oxidation rate was calculated over the entire study period for each infusion, and none of the treatment arms were significantly different from placebo (data not shown). There were no significant changes in pulse or systolic or diastolic blood pressure (Supplementary Fig. 2) with any of the treatment groups.
Infusion of GLP-1 or glucagon alone did not affect total or acylated ghrelin levels. However, coinfusion led to a significant fall in both total (P < 0.05) and acylated (P < 0.05) ghrelin (Fig. 6A and B). Plasma PYY levels were unaffected by GLP-1 and glucagon individually or in combination (Fig. 6C).
This study shows that dual infusion of GLP-1 and glucagon reduces food intake significantly, whereas the same low doses of glucagon and GLP-1, when administered separately, do not exert a similar anorectic effect.
The dose of glucagon used in this study (2.8 pmol/kg/min) was established in a dose-finding study to be subanorectic. It is higher than the dose used previously by Geary et al. (3) (0.84 pmol/kg/min), which was demonstrated to reduce food intake. Our intention was to examine the effect of raised preprandial levels of glucagon and GLP-1 on food intake in a fasting state. In contrast, the study by Geary et al. examined the effect of elevated postprandial levels of glucagon after consumption of 500 g of tomato soup. The soup would be expected to stimulate anorectic gut hormone secretion (e.g., PYY, GLP-1) and suppress ghrelin secretion, explaining the differences in glucagon doses.
Three subjects receiving the combination infusion experienced nausea. This nausea only became apparent postprandially, with no significant nausea occurring during the 90 min of infusion before the meal was served. The nausea is therefore unlikely to explain the reduction in food intake that was noted. GLP-1–based therapies for diabetes can cause nausea (14), as can glucagon (30). However, the doses used in this study were far smaller than those administered in these clinical situations. The postprandial nausea seen in this study may instead be accounted for by a delay in gastric emptying triggered by glucagon and GLP-1 (31,32).
Despite the significant reduction in food intake with coinfusion, no differences were observed in perceived hunger and satiety scores. Palatability of the meal was reduced with coinfusion, although this difference did not reach statistical significance. This discrepancy between perceived appetite and energy intake highlights the importance of measuring energy intake in an ad libitum meal, a more robust end point than VAS, which can be affected by other factors such as age, sex, and physical activity (33).
Multiple pathways are responsible for the food intake reduction observed with both hormones. The hypothalamus expresses both glucagon and GLP-1 receptors (34,35), and intracerebroventricular glucagon and GLP-1 are both capable of reducing food intake (36,37), suggesting that peripheral glucagon and GLP-1 could exert a direct effect on the hypothalamus after crossing the incomplete blood-brain barrier at the median eminence. A second mechanism might be via activation of vagal afferents to the brainstem, as vagotomy attenuates the anorectic effect of glucagon and GLP-1 after peripheral administration (38,39). A third mechanism for food intake reduction might be via indirect effects on other gut hormones. Coadministration of GLP-1 (0.8 pmol/kg/min) and glucagon (14 pmol/kg/min) causes a significant reduction in circulating levels of the orexigenic hormone ghrelin (5). The current study corroborates these findings at a lower dose of GLP-1 (0.4 pmol/kg/min) and a far lower dose of glucagon (2.8 pmol/kg/min). In our study, neither GLP-1 nor glucagon, alone or in combination, affected plasma PYY levels. Näslund et al. (40) demonstrated a small inhibitory effect of GLP-1 on PYY secretion, where an infusion of 0.75 pmol/kg/min reduced postprandial PYY levels by 4–5 pmol/L. In contrast, our study examined fasting PYY levels and used a smaller dose of GLP-1. Thus, it appears that GLP-1 and glucagon, at the doses used here, can modulate ghrelin secretion but not PYY.
The hyperglycemic effect of glucagon is undesirable in patients with diabetes or impaired glucose tolerance, albeit the greatest element results from a one-off stimulation of glycogenolysis, which would be expected to decline with time. Coadministration with GLP-1 attenuates this hyperglycemia, consequent on enhanced insulin release and glucose disposal (5). Both GLP-1 and glucagon act directly on the β-cell to release insulin and, in addition, the hyperglycemia itself is a stimulus for insulin release (41). GLP-1’s insulinotropic effects are dependent on the prevailing glucose level (10), which accounts for the relatively small insulinotropic response observed during GLP-1 infusion alone, as glucose levels tend to decline after the start of the infusion (Fig. 2A). The insulinotropic response with the coinfusion is much larger in amplitude owing to the triple effect of GLP-1, glucagon, and hyperglycemia. The insulin level during the coinfusion is sustained even when glucose returns to ≤5 mmol/L at 70 min (Fig. 2A and B) because glucagon continues to exert an insulinotropic effect independent of the prevailing glucose level (42). Therefore, the addition of GLP-1 to glucagon in the doses used for our coinfusion is able to neutralize the undesirable hyperglycemic effect of glucagon alone.
Moreover, the reduction in food intake seen with the combination infusion is likely to contribute to the attenuated postprandial glycemic response. The postmeal glucose response to GLP-1 alone is attenuated compared with placebo. However, the rise in insulin is delayed with infusion of GLP-1, suggesting that this is not an incretin effect. This phenomenon may be related to delayed gastric emptying with GLP-1. Analysis of the glucose and insulin response to the meal is complex, as the subjects ate different amounts. Further studies examining the effect of glucagon and GLP-1 combination on the glucose and insulin response to a standardized calorie load are warranted in order to formally assess the effects on carbohydrate tolerance, particularly in diabetic patients who may have compromised β-cell reserve.
Consistent with our previous study, we demonstrated an increase in REE of ∼50 kcal/day in both the glucagon alone and combination infusion groups, although this did not reach statistical significance. We also found a rise in RQ, rise in carbohydrate oxidation rate and fall in fat oxidation rate with glucagon alone and combination infusion, likely to be related to the relative substrate availabilities of glucose versus free fatty acid (5). The fact the rise in REE did not reach statistical significance was not unexpected in this current study, since the dose of glucagon used was a fifth that of our previous study (5). We speculate that the chronic sustained effects of this small increase in REE would have an important impact on body weight in the long-term when combined with the food intake reduction. The mechanism behind the increase in REE mediated by glucagon remains unclear. This phenomenon could be mediated by increased thermogenesis in brown adipose tissue (6) and/or by futile substrate cycling (43). These effects may be direct, via tissue glucagon receptor (e.g., in brown adipose tissue), or indirect, via an increase in catecholamines (44).
We also found a rise in RQ, a rise in carbohydrate oxidation rate, and a fall in fat oxidation rate with glucagon alone consistent with our previous study and likely to be related to the relative substrate availabilities of glucose versus free fatty acids (5). In contrast, GLP-1 alone caused a small reduction in carbohydrate oxidation and a small increase in fat oxidation consistent with previous studies (45). Interestingly, coinfusion caused an increase in RQ, increase in carbohydrate oxidation, and decrease in fat oxidation with magnitudes approximately double those seen with glucagon alone. This phenomenon is consistent with our observation of a similar increase in RQ, increase in carbohydrate oxidation, and decrease in fat oxidation with the combination compared with glucagon alone, although there was no significant statistical difference at the higher doses used in our previous study (5). It is possible that the combination of GLP-1 with glucagon may increase carbohydrate oxidation through a combined stimulation of glycolysis and glycogenolysis, and this requires further study.
In conclusion, this study reports that coadministration of glucagon and GLP-1, at doses which are individually subanorectic, significantly reduces food intake in humans. Furthermore, the coadministration of GLP-1 ameliorates the hyperglycemia of glucagon. These data are consistent with findings seen with acute infusion of OXM (16), suggesting that the anorectic and energy expenditure effects of OXM can be explained by costimulation of both the GLP-1 and glucagon receptors. These observations provide support for the further development of GLP-1/glucagon receptor coagonists as a therapeutic approach for obesity. This study has only examined the acute effects of GLP-1/glucagon coagonism, and further chronic studies need to be performed in humans to establish a therapeutically useful anorectic effect without exerting nausea as well as maintaining euglycemia. Establishing these effects will be the key to therapeutic exploitation.
Acknowledgments. The authors acknowledge the staff of the National Institute for Health Research/Wellcome Trust Clinical Research Facility at Imperial College Healthcare National Health Service Trust, without whom this study would not have been possible. The authors also thank the staff of the Pharmacy Department at Imperial College Healthcare National Health Service Trust for their assistance with peptide preparation.
Funding. This study was supported by the Imperial College Academic Health Sciences Centre/Biomedical Research Centre (BRC). Investigative Medicine is funded by the Medical Research Council (MRC), the Biotechnology and Biological Sciences Research Council (BBSRC), the National Institute for Health Research (NIHR), an Integrative Mammalian Biology Capacity Building award, an FP7-HEALTH-2009-241592 EuroCHIP grant, and the NIHR Imperial Biomedical Research Centre within the Academic Health Science Centre. R.C.T. is a recipient of an MRC Clinical Research Training fellowship. B.J., G.T., and J.K. are NIHR Academic Clinical Fellows. K.A.M. and J.Ce. are recipients of Wellcome Trust Clinical Research Training fellowships. C.T.L., N.P., and M.H. are NIHR Academic Foundation Year 2 trainees. E.S.C. is supported by a BBSRC Diet and Health Research Industry Club grant. T.M.T. is supported by grants from the MRC. S.R.B. is supported by an NIHR Senior Investigator Award and the MRC.
Duality of Interest. No potential conflicts of interest relevant to this article were reported.
Author Contributions. J.Ce. and R.C.T. conducted the study, wrote the manuscript, and reviewed and commented on the manuscript. B.J., G.T., J.K., K.A.M., C.T.L., N.P., M.H., and E.S.C. conducted the study and reviewed and commented on the manuscript. J.M., J.Cu., M.A.G., and K.M. executed and advised on the hormone immunoassays and reviewed and commented on the manuscript. T.M.T. and S.R.B. designed the study and reviewed and commented on the manuscript. All authors gave approval of the final version to be published. S.R.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.