Bariatric Surgery Alters the Postprandial Recovery From Hypoglycemia, Mediated by Cholinergic Signal Free

Bariatric surgery has seen an increasing rise in clinical use over the past two decades because it is the most effective intervention for weight loss and corrects many of the comorbidities of obesity such as diabetes. The surgical rearrangement of the gastrointestinal (GI) tract with the two most common weight loss surgeries, Roux-en-Y gastric bypass (GB) and sleeve gastrectomy (SG), causes rapid passage of nutrients into the small bowel and increases the rate of nutrient appearance into the circulation (1–3).

Rearrangement of GI tract anatomy due to GB or SG leads to a larger prandial glucose excursion with higher and earlier peak levels of glucose and lower nadir glucose along with enhanced insulin secretion, resetting the balance between glucose appearance and clearance after this procedure (2,4–8). The larger glucose excursion caused by GB or SG can lead to prandial hypoglycemia, which often is asymptomatic (9,10), indicative of a reset in counterregulatory glucose response to hypoglycemia (11,12). Prandial hyperglucagonemia and hyperinsulinemia in post-GB and -SG subjects without diabetes (1,6,9,13) raise the possibility that increased parasympathetic nervous system (PNS) (the neural component of enteroinsular axis) in the fed state plays a role in glucose metabolism after these surgeries. The role of autonomic nervous system (ANS), both PNS and sympathetic nervous system activity, has also been implicated in hepatic glucose production (14–16) in addition to islet cell secretory response to oral glucose or meal ingestion (17–21).

Of bariatric surgery patients 1 in 10 develop a late complication of hypoglycemia (22,23) and 1 in 150 suffer from severe hypoglycemia requiring emergency room visit or hospitalization (22). The first line of treatment for acute hypoglycemia after GB or SG is oral carbohydrate mixed with fat/protein. Yet, the glucose counterregulatory response (endogenous glucose production [EGP]) to insulin-induced hypoglycemia after meal ingestion is reduced after GB or SG in comparison with nonsurgical subjects (24). While these observations hint toward diminished hepatic sensitivity to glucagon after GB or SG, hepatic glucose production can also be altered by cholinergic signal. In the perfused rat liver, parasympathetic hepatic nerve stimulation diminishes glucagon-stimulated glucose release (25). Further, preclinical studies in dogs have shown that liver glucose uptake is increased by augmented glucose load and the route by which glucose is administered (portal vein vs. hepatic artery) (26,27). In vivo studies in dogs as well as ex vivo studies of isolated perfused rat liver have also demonstrated that the routed glucose delivery in hepatic portal versus peripheral venous increases parasympathetic over sympathetic nervous system signaling and enhances net hepatic glucose uptake (16,28). Therefore, it is plausible that faster nutrient emptying after GB and SG due to rerouted GI tract anatomy reduces the prandial EGP by promoting a larger hepatic glucose load, a negative arterial-portal venous glucose gradient across the liver (i.e., portal vein greater than hepatic artery), as well as a greater PNS contribution to prandial glucose kinetics.

The goal of the present study was to examine the role of the PNS blockade by atropine infusion in the counterregulatory hormonal response and glucose kinetics involved with recovery from insulin-induced hypoglycemia provoked by meal ingestion in subjects with history of GB or SG versus healthy control subjects. We hypothesized that PNS blockade increases prandial EGP response to hypoglycemia in subjects with bariatric surgery compared with a matched control group.

Nine subjects without diabetes with GB and 7 with SG and 5 healthy control subjects without prior GI surgery (CN), were recruited consecutively in order of presentation to the advertisement or to the clinic. All subjects had stable body weight for at least 3 months prior to the studies. The control subjects had no personal or family history of diabetes and had a normal oral glucose tolerance test before enrollment. The surgical subjects regardless of history of hypoglycemia were recruited at least 2 years after surgery and had HbA_1c <6%. All subjects were free of diabetes and GI disease and had normal renal and hepatic function. The Institutional Review Board of The University of Texas at San Antonio approved the protocol, and all participants provided written informed consent before participating in any experiments. The study is registered with ClinicalTrials.gov (NCT02823665).

All studies were performed at the Bartter Clinical Research Unit at Audie L. Murphy Memorial Veterans’ Hospital in the morning after a 10-h overnight fast. Participants were instructed to maintain normal carbohydrate ingestion for 3 days before each visit and not to engage in excessive physical activity. Body composition was assessed with DEXA. Intravenous catheters were placed in each antecubital or hand for withdrawal of blood and infusion of study drugs. The arm or hand used for blood sampling was continuously warmed with a heating pad.

Each subject was studied twice, separated by at least 1 week. After withdrawal of fasting blood samples at −120 min, a primed-continuous infusion of [6,6-²H₂]glucose (28 µmol/kg prime and 0.28 µmol/kg/min constant) (6) was initiated and continued for 2 h, when the rate was reduced to one-half in anticipation of the insulin-mediated reduction in hepatic glucose output for the remainder of the study. At 0 min, a primed-continuous infusion of recombinant human insulin (Humulin, 100 units/mL) diluted in isotonic saline mixed with 2 mL of the subject’s blood was started and continued at 120 mU/min/m² for the duration of the study. Blood was sampled frequently, and a variable infusion of 20% glucose (1% enriched with [6,6-²H₂]glucose) was infused to maintain glucose at a target of 55–60 mg/dL (∼3 mmol/L). At 60 min subjects received either atropine sulfate (American Regent, Shirley, NY) (10 µg/kg lean body mass [LBM]/2 min followed by a constant infusion of 10 µg/kg LBM/h) for the remainder of the study (29) or saline. At 120 min subjects consumed a 140-mL liquid mixed meal containing 33 g whey protein, 12.7 g corn oil, and 15 g glucose mixed with 0.2 g [U-¹³C]glucose over 10 min. Plasma was separated within 60 min and stored at −80°C until assay. Each subject’s heart rate (HR) and systolic and diastolic blood pressure (BP) were monitored throughout the studies, and values were averaged over every 5–10 min.

Blood samples were collected in heparin for measurement of insulin and glucose and in aprotinin/heparin/EDTA for assay of other hormones (9,13). Plasma glucose was determined with GM9 Glucose Analyser (Analox Instruments, Stourbridge, U.K.). Insulin (DIAsource, Neuve, Belgium), and C-peptide and glucagon (Millipore, Billerica, MA) were measured by commercial radioimmunoassay. The Millipore glucagon radioimmunoassay kit has a cross-reactivity of <2% with oxyntomodulin and glicentin with a sensitivity of ∼10 pmol/L (30). GLP-1, GIP, and PP were measured with commercial Multiplex ELISA (Millipore) according to the manufacturer specifications. Tracer enrichment was measured with gas chromatography–mass spectrometry as previously described (31).

Basal fasting plasma glucose and hormone concentrations represent the mean of three samples drawn before studies, and the premeal values reflect the average of the samples drawn before the test meal (110–120 min). We calculated insulin secretion rates (ISRs) from C-peptide concentrations using deconvolution with population estimates of C-peptide (32). With use of the trapezoidal rule, the incremental area under the concentration curve (AUC) of islet cell and GI hormones was calculated from 0–180 and 0–60 min after meal for evaluation of the total and early responses. Rates of total glucose appearance (R_a), glucose disappearance (R_d), systemic appearance of ingested glucose (R_aO), and endogenous glucose production (EGP) were derived from plasma [6,6-²H₂]glucose and [U-¹³C]glucose enrichments as previously described with the Steele equation (6,24), and cumulative values for 1 and 3 h from meal ingestion were estimated. Peripheral insulin action was calculated as metabolic clearance of glucose, measured as R_d/plasma glucose, expressed per insulin levels at baseline, pre-meal (110–120 min), and post-meal (120–300 min) (31). The relative change in the parameters of interest from saline to atropine was calculated as the percentage of the absolute change in comparison with the saline value.

Data are presented as mean ± SEM. The effect of atropine versus saline infusion during the clamp studies and the group effect (GB, SG, and CN), as well as their interaction with regard to experimental outcomes, were analyzed with repeated-measures ANOVA; Tukey honestly significant difference was used for post hoc analysis to compare differences among the three groups. The parameters of interest at baseline as well as the relative changes in parameters between the two studies were compared with χ² or ANOVA based on prespecified comparisons among the groups (surgical subjects vs. CN and GB vs. SG). Statistical analyses were performed with SPSS 28 (SPSS, Chicago, IL).

The subjects in the three groups had similar BMI, waist circumference, body fat mass and LBM, age, sex, and HbA_1c (Table 1). Surgical subjects were similar regarding preoperative BMI, weight loss, and time since surgery (Table 1).

Parts of glucose, insulin, and glucagon data during saline studies were previously published as part of studies that focused on the glucose counterregulatory response to hypoglycemia and insulin clearance (24,32). The effects of PNS signals on glucose kinetics and hormonal response are highlighted in this article.

Fasting plasma glucose values did not differ among the three groups or between the two studies (Fig. 1A). For all experiments, plasma glucose declined to the glycemic target (∼55–60 mg/dL) in 30 min with infusion of insulin and were kept at the target level before meal ingestion (Fig. 1A). The glucose infusion rates (GIR) needed to achieve the glycemic target before meal (110–120 min) did not differ across the groups or between the two studies (Fig. 1B).

After meal ingestion, to maintain glycemia at the target level, we reduced the GIR in three groups (Fig. 1B). Despite a complete discontinuation of glucose infusion in subjects, in four GB, one SG, and one CN, glycemia increased to higher than the target concentration but remained below fasting level (Fig. 1A). Mean ± SEM coefficients of variation of plasma glucose for each study from 60 to 300 min were 17 ± 2%, 11 ± 1%, and 9 ± 2% for GB, SG, and CN, respectively, and did not differ among the groups significantly (P = 0.183). The coefficients of variation of glucose for 60–300 min from studies with and without atropine in each subject were 5 ± 1%, 5 ± 1%, and 2 ± 1% for GB, SG, and CN, respectively, demonstrating similarity of glucose levels among the two study conditions in each subject. Atropine infusion did not affect prandial GIR in GB, but it increased the need to lower GIR in anticipation of ingested glucose appearance in SG and robustly decreased the need to lower the GIR in CN (P < 0.05 for interaction) (Fig. 1B).

Fasting and steady-state plasma insulin concentrations and ISR were similar among the three groups and between two studies (Fig. 2A). In response to exogenous insulin administration, endogenous ISR decreased similarly in the three groups (Fig. 2A). PNS blockade had no significant effect on endogenous ISR suppression (Fig. 2A).

In parallel with glycemic response, meal ingestion increased insulin secretory response in the first hour in GB but with minimal change in SG and did not change prandial ISR in CN (Fig. 2A). However, PNS blockade diminished prandial ISR across all groups despite similar glycemia between the two studies (P < 0.05); this effect was more evident in surgical subjects, whose prandial ISR was raised during control studies with saline (Fig. 2A).

Basal PP did not differ among three groups or between two studies and increased in response to insulin-induced hypoglycemia in all groups (Fig. 2B). Atropine infusion reduced PP to the lowest level detectable by the assay before a meal in all three groups, indicative of an adequate PNS blockade in all groups (Fig. 2B).

Meal ingestion also augmented PP values during saline study in all three groups (Fig. 2B). During atropine study, PP levels remained suppressed and unchanged from premeal values as intended (Fig. 2B).

Fasting glucagon values and the rise in response to insulin-induced hypoglycemia before meal ingestion were similar among the groups and between two studies (Fig. 2C).

GB and SG subjects had larger post-meal glucagon responses compared with CN (P < 0.05) (Fig. 2C). Atropine infusion suppressed early glucagon response (AUC_{Glucagon 1 h)} in all three groups (P < 0.01) (Fig. 2C). However, reduction in AUC_{Glucagon 3 h} by atropine was only observed in CN (mean ± SEM relative change over the 3 h: −8 ± 25%, −26 ± 16%, and 97 ± 43% for GB, SG, and CN, respectively; P < 0.05).

Following an overnight fast, steady-state conditions prevail, and the rates of total glucose utilization (R_d) equal the rates of EGP and were similar among three groups and between the studies with and without atropine infusion (Table 2).

As expected, hyperinsulinemia suppressed EGP within the first 60 min in all three groups (Fig. 3A), but then extension of hypoglycemia beyond the first 60 min increased EGP in the 60- to 120-min period (Fig. 3A). PNS blockade had no significant effect on premeal EGP values (Fig. 3A).

Prandial R_a did not differ among the groups and was not significantly affected by atropine infusion studies (Table 2). As reported previously (24), R_aO was greater in surgical subjects compared with CN during the saline studies (Fig. 1C and Table 2). Atropine infusion markedly diminished R_aO in CN but had no influence in ingested glucose appearance into circulation in GB or SG (mean ± SEM R_aO decremental change over the 3 h from saline to atropine studies: −10 ± 7%, 5 ± 15%, and 31 ± 15% for GB, SG, and CN, respectively; P < 0.05) (Fig. 1C). In comparisons with saline studies, EGP over a 3-h postprandial period was increased by ∼10% in SG and decreased by ∼40% in CN without any significant changes in GB during atropine infusion (P < 0.05) (Fig. 3A and Table 2).

The rates of metabolic clearance of glucose (R_d/glucose) per corresponding insulin concentrations, representing peripheral insulin action, were similar for basal, steady-state period (110–120 min), and prandial condition (120–300 min) among the three groups and between the two studies (Table 2).

Basal GLP-1 and GIP levels were similar among the three groups and two studies (data not shown). Prandial GLP-1 increased in both surgical and nonsurgical groups, with the largest early response occurring in the surgical subjects in comparison with CN (P < 0.05) (Fig. 4A). Compared with CN, GB and SG subjects had a larger AUC_{GIP 1 h} (P < 0.05) (Fig. 4B, inset), but AUC_{GIP 3 h} were similar among the groups (Fig. 4B). Atropine infusion decreased GLP-1 secretion, particularly in the first hour (P < 0.05) (Fig. 4A). Total GIP was not affected by atropine infusion, given a biphasic GIP response to PNS blockade (Fig. 4B), i.e., a reduction in the first hour of meal ingestion followed by an increase in the second phase of meal study, in SG and CN subjects.

Basal HR and BP values and their response to hypoglycemia before meal ingestion were similar among the groups and between the two studies. Meal ingestion increased mean ± SEM HR above premeal values by 10 ± 4%, 6 ± 2%, and −2 ± 1.5% for GB, SG, and CN, respectively, during saline studies (P < 0.01). Meal ingestion did not have any additional effect on systolic BP, but diastolic BP diminished further in all three groups (P < 0.05). HRs that were elevated by atropine infusion remained unchanged after meal ingestion (Supplementary Table 1).

In this study we examined the effect of atropine infusion on physiological responses to insulin-induced hypoglycemia during meal ingestion in subjects without diabetes with and without bariatric surgery. To this end, hyperinsulinemic clamp was used to maintain glucose levels at subfasting levels and minimize the effects of glucose and insulin on the outcomes of interest. Our major findings are that following mixed-meal ingestion, acute PNS blockade during insulin-induced hypoglycemia 1) slows the rates of systemic appearance of ingested glucose in CN without any significant influence in GB or SG; 2) affects the meal-induced EGP enhancement differentially among the groups, with EGP reduced by 40% in CN, increased by 13% in SG, and unchanged in GB; 3) reduces early prandial α-cell secretion in both surgical and CN subjects but eradicates overall post-meal glucagon secretion only in CN; 4) diminishes the meal-induced β-cell secretion particularly in surgical subjects, an effect that is glucose independent; and 5) reduces postprandial GLP-1 secretion in both surgical and CN groups but has no effect on total GIP response. We also have found that consistent with the findings from previous study in dogs (33), the effect of atropine infusion on counterregulatory responses to hypoglycemia before meal ingestion is negligible. To our knowledge, this is the first study to investigate the contribution of the PNS in meal-based recovery response to insulin-induced hypoglycemia in humans with and without prior history of bariatric surgery. Our findings suggest a differential regulatory effect of cholinergic signal on glucose metabolism in the absorptive state between surgical versus nonsurgical subjects and between patients with GB and SG.

We used atropine infusion that led to achieving maximum HRs and complete suppression of pancreatic polypeptide in all three groups, reflecting suppression of the vagal input to pancreas (34). A test meal containing 15 g glucose mixed with protein and fat was used to replicate the recommended treatment for acute hypoglycemia in patients with bariatric surgery (4). Also, the macronutrient composition of our test meal (high protein and lower glucose) allowed us to magnify the prandial α-cell secretory response without hindering our ability to maintain glucose concentrations below fasting values. We studied GB and SG subjects with distinct alterations in GI tract anatomy to examine potential differences between the two surgical groups.

Atropine infusion during an oral glucose tolerance test in patients with prandial reactive hypoglycemia with intact GI anatomy increases the nadir glucose concentration and decreases the peak glucose level by 30–40 mg/dL along with diminished insulin response by >50% (35). In absence of hypoglycemia, atropine reduces the prandial insulin response in humans (20,21) or animals (17,19) with minimal effect on the glycemic response. These observations indicate that glucose metabolism in the prandial state in subjects with normal GI anatomy is affected by PNS signaling; however, whether the PNS glucoregulatory effects are mediated by a direct islet cell effect or changes in GI function or glucose flux is largely unknown. Here, we demonstrate that cholinergic signals contributed to prandial insulin secretion during fixed hypoglycemia in all three groups but to a larger extent in surgical subjects compared with CN. Considering that glucose concentrations were maintained under fasting values and were similar between the two studies, we infer from these data that insulinotropic effect of PNS in the fed condition is independent of systemic glucose concentrations; thus, even at glucose concentrations below fasting values, vagal activation can add to prandial hyperinsulinemia after GB or SG.

In vivo studies in animals (36,37) or humans without GI surgeries (38) have demonstrated that 70–80% of the glucagon response to hypoglycemia is accounted for by ANS activity. A role for cholinergic input was also previously shown in subjects with truncal vagotomy whose hypoglycemia-induced glucagon secretion was 50% lower than that in subjects with a normal innervated pancreas (39). Consistent with previous work, in our experiments, atropine infusion diminished early post-meal glucagon secretion in all groups, but the α-cell effect of PNS blockade was much more evident in nonsurgical subjects whose meal-induced glucagon secretion was almost abolished. The differential effect of atropine infusion on prandial glucagon response among surgical and nonsurgical groups in our studies, however, raises the question of whether this is the result of PNS blockade on the islet cells or an indirect effect due to changes in GI function reflected in R_aO and consequently the amount of nutrient reaching the pancreatic islets to stimulate glucagon secretion.

Atropine infusion in nonsurgical subjects, lean or obese, delays gastric emptying of both liquid and solid food (40). In our study, PNS blockade robustly reduced R_aO in CN but had minimal effect in GB and SG, suggesting that the GI effects of PNS in surgical subjects are trivial. However, there was no association between the atropine-induced changes in R_aO and changes in glucagon response; thus, the differential effect of PNS blockade on glucagon response cannot be explained only by cholinergic GI function.

The liver response to hypoglycemia has been attributed to glucose sensing at the liver and the brain, as well as to efferent impulses from the brain to the liver (41). In nonsurgical subjects, neural signals play a key role in EGP response to hypoglycemia (14,42). The EGP response to insulin-induced hypoglycemia is blunted after hepatic denervation in pigs (15) or humans (42), indicating an important role of ANS.

Activation of the PNS promotes glucose uptake by the liver (16). Interrupting the vagal signal in our study had no effect on EGP response to hypoglycemia prior to meal ingestion. This finding is consistent with previous report from a previous preclinical study in that hormonal or EGP response to hypoglycemia was similar with and without vagal blockade by atropine infusion (33). In the prandial state, although atropine infusion increased EGP in SG-treated subjects, it eradicated EGP in CN and had minimal effect in GB subjects (Fig. 3A). This conclusion was also supported by the rate of glucose infusion required to maintain glucose concentration at the target during atropine study being significantly higher in CN and lower in SG subject despite equivalent glucose utilization rates among the surgical and CN subjects. The obvious question that arises from these findings is how they fit with the differences in GI function among the surgical versus CN or between GB and SG subjects. Our data do not directly address this question, but they do not rule out involvement of humoral factors, such as glucagon, not kept at fixed levels, or variations in hepatic glucose load (26,27), on the observed glucose kinetic responses. However, contribution of diminished prandial glucagon to the lower EGP in nonsurgical subjects during atropine infusion is unlikely, since the effect of PNS blockade on EGP is instant (Fig. 3A) when glucagon level is not reduced (Fig. 2C), indicative of a direct effect of PNS signals on liver glucose metabolism. Finally, a differential effect of GB versus SG on prandial EGP responses with and without atropine cannot be explained by variations in hepatic glucose load that are likely to be similar between the surgical groups and among the atropine and control studies (Fig. 1C).

Prandial glucose metabolism also is affected by incretins (43). Results of previous studies with use of sham feeding (44) or atropine infusion in humans (20,45) or monkeys (17) have shown that vagal activation increases incretin secretion and PNS blockade delays it. Consistent with previous publications, cholinergic blockade in our study delayed prandial GIP secretion in SG and CN, but it had no effect on the rates of GIP release in GB subjects, suggesting that GIP secretion is altered in parallel with the corresponding R_aO in each group. However, elimination of the PNS signal markedly reduced the plasma GLP-1 response in GB subjects and delayed the GLP-1 response in SG and CN. This was not associated with any changes in R_aO in GB subjects, suggesting a direct PNS contribution to enhanced GLP-1 secretion by L cells after GB. Previously, preclinical studies in rodents have shown that GLP-1 secretion by intestinal L cells is mediated partly through PNS signal (46); our findings are the first to confirm the role of PNS in GLP-1 secretion in humans. It is unlikely that changes in GLP-1 or GIP concentrations in these experiments were involved with ISR reducing effects of PNS blockade, since β-cell effect of incretins in nonsurgical subjects are glucose dependent (47,48). Nonetheless, further studies are needed to examine the glucose dependency of GLP-1 action on β-cell secretion in subjects with rerouted gut due to bariatric surgery.

There are several limitations to this study that warrant consideration. Our interpretation of β-cell effects of PNS activity is based on the assumption that clearance of C-peptide is similar among surgical and nonsurgical subjects and is not affected by hyperinsulinemia. However, contribution of liver or splanchnic tissues to removal of C-peptide is negligible; therefore, it is unlikely that bariatric surgery has a significant effect on C-peptide clearance (49). Antecedent hypoglycemia can blunt subsequent counterregulatory responses to future hypoglycemia (50). But counterregulatory adaptation to antecedent hypoglycemia should be similar in studies with and without atropine infusion. Therefore, the reported PNS effects on outcomes are not likely to be affected by antecedent hypoglycemia. Further, post-meal glucose concentrations in one CN, one SG, and four GB subjects increased above the target level but remained under basal glycemic concentrations despite discontinuation of glucose infusion. While our study is not designed to address the hypoglycemic threshold that can elicit a counterregulatory response in the fed state, we have found that the glycemic rise after meal ingestion is likely due to a reduced action of insulin in these subjects (Supplementary Fig. 1). Therefore, prandial EGP or glucagon responses were unaffected by variation in glycemic concentrations among the subjects. Also, the number of subjects in each group was relatively small, diminishing the power of the present study to examine differences in the outcomes of interest among the groups as well as the generalizability of these findings to a broad range of subjects after bariatric surgery. However, the groups were similar in BMI, sex, and age, and the surgical groups were similar regarding time and weight loss since surgery. Furthermore, there was a significant difference in prandial R_aO, EGP, and glucose response among the groups, primarily based on the type of GI surgery. Finally, we had no information regarding alteration of vagal nerve anatomy caused by the bariatric surgery. However, the plasma PP concentration, an index of pancreatic innervation (34,51), was consistently decreased in response to hypoglycemia and meal ingestion in all bariatric subjects, suggesting that vagal innervation to pancreas is intact.

In summary, in subjects without history of GI surgery, acute elimination of cholinergic signal dramatically diminishes the physiologic responses (R_aO, EGP, and glucagon) that are vital for recovery from hypoglycemia after meal ingestion. In contrast, cholinergic effect on R_aO or glucagon response to hypoglycemia is minimal after bariatric surgery, whereas eliminated PNS increases EGP in SG subjects. Taken together with our previous results showing a diminished EGP response to hypoglycemia after GB and SG, these findings demonstrate a distinct cholinergic contribution to the attenuated prandial counterregulatory responsiveness to hypoglycemia in patients adapted to GB or SG.

Acknowledgments. The authors thank Andrea Hansis-Diarte, Nancy Yegge, and John Adams from the Department of Medicine of University of Texas Health Science Center at San Antonio for technical support and the nursing staff as well as nutritionist from Bartter Clinical Research Unit, Audie L. Murphy Memorial Veterans’ Hospital, South Texas Veterans Health Care System, for expert technical assistance. The authors owe a great debt to the research participants.

Funding. This work was supported by National Institutes of Health grant DK105379 (to M.S.) and in part by National Center for Advancing Translational Sciences, National Institutes of Health, grant UL1 TR002645. A.G. acknowledges the financial support from the European Union’s Horizon 2020 Research and Innovation Programme for the project Stratification of Obesity Phenotypes to Optimize Future Obesity Therapy (SOPHIA). SOPHIA has received funding from the Innovative Medicines Initiative 2 Joint Undertaking (IMI2 JU) under grant agreement no. 875534. IMI2 JU received support from the European Union’s Horizon 2020 Research and Innovation Programme, European Federation of Pharmaceutical Industries and Associations (EFPIA), T1D Exchange, JDRF, and Obesity Action Coalition.

The communication reflects the authors’ view. The Innovative Medicines Initiative, European Union, EFPIA, and any associated partners are not responsible for any use that may be made of the information contained herein.

Duality of Interest. No potential conflicts of interest relevant to this article were reported.

Author Contributions. M.S. designed and supervised the study, obtained data, analyzed and interpreted data, and wrote the manuscript. D.T., R.P., H.H., and S.P. contributed to conducting studies and obtaining data. R.D. and A.G. contributed to interpretation of data and review and editing of the manuscript. A.G. analyzed data. M.S. 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 80th Scientific Sessions of the American Diabetes Association, 12–16 June 2020.