Empagliflozin Normalizes Fasting Hyperglycemia and Improves Postprandial Glucose Tolerance in Totally Pancreatectomized Patients: A Randomized, Double-Blind, Placebo-Controlled Crossover Study Free

Total pancreatectomy leads to insulin deficiency and so-called pancreatogenic diabetes (previously classified as type 3c diabetes). Because of the altered anatomy after total pancreatectomy, which involves surgical removal of the lower part of the stomach (including the pylorus) and the duodenum, totally pancreatectomized (PX) patients have accelerated gastric emptying contributing to their exaggerated postprandial plasma glucose responses, and because of lack of pancreatic glucagon, these patients are prone to insulin treatment–associated hypoglycemia. Additionally, PX are highly insulin sensitive (they require less exogenous insulin as compared with C-peptide–negative patients with type 1 diabetes [1]), increasing the risk of treatment-associated hypoglycemia further. For these reasons, diabetes secondary to total pancreatectomy is often referred to as “brittle diabetes,” with pronounced fluctuations in plasma glucose concentrations and frequent episodes of hyperglycemia and hypoglycemia. Despite the challenges associated with obtaining glycemic control in these patients, no internationally recognized guidelines for treating diabetes in totally PX patients exist. Because of the shared feature of insulin deficiency between PX patients and patients with type 1 diabetes, most PX patients are managed on multiple-dose insulin injection therapy (basal/bolus regimen) or—in some cases—insulin pump therapy (1,2). In individuals with type 2 diabetes, sodium–glucose cotransporter 2 (SGLT2) inhibitors, in addition to their glucose-lowering effect, have been shown to reduce systolic blood pressure, promote body weight loss, and attenuate the risk of major adverse renal and cardiovascular events and hospitalization for heart failure (3). As the glucose-lowering effect of SGLT2 inhibition is strictly dependent on prevailing plasma glucose concentrations, treatment with SGLT2 inhibitors is associated with low risk of hypoglycemia. Interestingly, SGLT2 inhibition, in some studies, has been shown to increase pancreatic glucagon secretion and hepatic glucose production in patients with type 2 diabetes (4–6). This phenomenon has recently been challenged (7–9), and it is being debated whether it reflects direct stimulation of glucagon secretion from pancreatic α-cells or constitutes an indirect response to lowering of plasma glucose (10,11).

In the current study, we investigated the acute glucoregulatory effects of the SGLT2 inhibitor empagliflozin during a standardized 3-h mixed meal test in PX patients and matched healthy control participants. As PX patients may harbor extrapancreatic glucagon (12–14), we also evaluated the effect of SGLT2 inhibition on glucagon secretion in the two groups of participants.

This study was designed as a randomized, double-blind, placebo-controlled crossover study and was conducted from April 2019 to September 2019 at Center for Clinical Metabolic Research at Gentofte Hospital, University Copenhagen, Hellerup, Denmark. The study also included an experimental day with dipeptidyl-peptidase 4 inhibition (sitagliptin); as prespecified in the protocol, results from this experimental day will be reported separately. The study was approved by the Scientific-Ethical Committee of the Capital of Region of Denmark (Copenhagen, Denmark; reg. no. H-19000992) and registered at ClinicalTrials.gov (identifier NCT04061473). The study was conducted according to the principles of the Helsinki Declaration (seventh revision, 2013). Oral and written consent was obtained from all participants prior to inclusion in the study.

The primary end point was the change in postprandial plasma glucagon response after two 25-mg doses of the SGLT2 inhibitor empagliflozin (administered the night before and 1 h before ingestion of a standardized liquid mixed meal) as compared with placebo in PX patients. Secondary end points included changes in fasting plasma glucose, postprandial plasma glucose excursions, endogenous glucose production (EGP) (derived from glucose tracer methodology), postprandial plasma responses of glucose-dependent insulinotropic polypeptide (GIP), glucagon-like peptide 1 (GLP-1), gastric emptying (assessed by the acetaminophen absorption test), blood pressure and heart rate, appetite and satiety ratings (assessed by visual analog scales), and food intake during an ad libitum meal. To compare the effects of empagliflozin in totally PX patients with its effects in normal physiology, we also evaluated a group of matched healthy control participants.

The study included one group of totally PX patients and one group of matched healthy control participants. The key inclusion criteria for totally PX patients were age >18 years, total pancreatectomy, and hemoglobin >7.0 mmol/L (men)/>6.5 mmol/L (women). Key inclusion criteria for the healthy control participants were age >18 years and hemoglobin in the normal range. Key exclusion criteria for both groups were age >85 years and severe liver, lung, kidney, and/or cardiovascular disease. For the total pancreatectomy group, additional exclusion criteria were pancreatectomy within the last 3 months, ongoing chemotherapy, or chemotherapy within the last 3 months.

In a double-blinded design, each subject underwent two experimental days (one preceded by two doses of empagliflozin and one preceded by two doses of placebo) performed in randomized order. Randomization followed a prespecified random numbers table generated at www.random.org by a third person who was not involved in participant enrolment or data collection. Empagliflozin tablets (Boehringer-Ingelheim, Ingelheim am Rhein, Germany) and similarly sized placebo tablets were encapsulated in indistinguishable capsules by the Central Pharmacy of the Capital Region of Denmark.

After an initial screening visit, participants were examined on two separate experimental days (each involving a 3-h liquid mixed meal test) with at least 72 h between the experimental days. The participants were instructed to refrain from smoking, alcohol consumption, and strenuous physical activities 48 h prior to each experimental day. The PX patients were instructed to take their regular daily basal insulin doses but not to take any short-acting insulin in the morning of experimental days. On the night prior to each of the experimental days, either 25 mg empagliflozin or placebo was administered. Participants met in our department after 10-h fast (overnight) and were placed in a semirecumbent position in a hospital bed, after which a cannula was inserted into each cubital vein, one for infusion of isotope-labeled glucose and one in the contralateral vein for collection of arterialized blood (hand and forearm wrapped in a heating pad [∼40–50°C]). After collection of basal blood specimens, an infusion of stable isotope-labeled glucose ([6,6-D2] glucose; priming dose of 40 µmol × kg⁻¹ × ϒ/5, where ϒ stands for fasting plasma glucose in millimoles per liter, and continuous infusion of 0.76 µmol × kg⁻¹ × min⁻¹) (Cambridge Isotope Laboratories, Tewksbury, MA) was initiated at time −120 min. At time −60 min, either 25 mg empagliflozin or placebo was administered with a sip of water. After 2 h of stable isotope-marked glucose infusion (to obtain tracer steady state), at time 0 min, the participants ingested a 200-mL liquid mixed meal (1,650 kJ [394 kcal]) consisting of glucose (47.2 g + 2.8 g [U13-C6]-glucose), whey protein (15.2 g), rapeseed oil (14.1 g), and 1.5 g acetaminophen (for assessment of gastric emptying [15,16]) over 3 min. Pancreatectomized patients were given Creon 25,000 containing 25,000 United States Pharmacopeia (USP) units of lipase, 18,000 USP units of amylase, and 1,000 USP units of protease (The Abbott Concern, Orifarm A/S, Odense, Denmark) along with the meal. Pulse rate and blood pressure were measured at time −120 min and every 30 min during the experimental day. Resting energy expenditure and respiratory expiratory ratio were measured by indirect calorimetry (Vyntus CPX Canopy; CareFusion, Hoechberg, Germany) for 15 min at time points −90, 30, and 150 min, with the participants in a supine position and awake. At time points 0, 30, 60, 90, 120, 150, and 180 min, appetite, hunger, satiety, fullness, and prospective food consumption were assessed by visual analog scales. Arterialized blood samples were drawn at time points −120, −30, −15, and 0 min before and 15, 30, 45, 60, 90, 120, 150, and 180 min after the liquid meal was ingested. For bedside measurement of plasma glucose, blood was collected in sodium NaF tubes and centrifuged immediately at 8,400g for 30 s at room temperature. For the plasma analysis of tracers, empagliflozin (time −120 min and 90 min), glucagon, GIP, and GLP-1, blood was collected in chilled tubes (on ice) containing EDTA and a specific dipeptidyl-peptidase 4 inhibitor (valine pyrrolidide, a gift from Novo Nordisk, Bagsværd, Denmark; final concentration: 0.01 mmol/L). For serum analysis of acetaminophen, C-peptide, and insulin, blood was collected in tubes containing Li-heparin. All tubes were centrifuged for 15 min at 2,000g and 4°C. Plasma samples for tracer, empagliflozin, glucagon, and GIP analyses were stored at −20°C. Serum samples for insulin, C-peptide, and acetaminophen analyses were stored at −80°C until analysis. At time 180 min, the subjects were served a standardized pasta Bolognese ad libitum meal (50 energy [E]% carbohydrate, 37 E% fat, 13 E% protein) for evaluation of appetite/food intake (17). Pancreatectomized patients were instructed to take their regular pancreatic enzyme supplementation and prandial insulin (according to blood glucose levels) together with the ad libitum meal.

Plasma concentrations of glucose were measured using the glucose oxidase method (Yellow Spring Instruments model 2300 or 2900 Stat Plus analyzer; Yellow Springs, OH). Plasma concentrations of empagliflozin were measured with liquid chromatography–tandem mass spectrometry. (Please see Supplementary Material for details.) Serum acetaminophen concentrations were measured with a photometric method (Atellica CH 930; Siemens Healthineers, Ballerup, Denmark). Serum C-peptide and insulin concentrations were measured with an immunoassay using direct chemiluminescent technology (Atellica IM 1600; Siemens Healthineers, Ballerup, Denmark). Plasma concentrations of glucagon and GIP[1–42] were measured using a specific sandwich ELISA (Mercodia, Uppsala, Sweden) and a specific radioimmunoassay, respectively, as previously described (18,19). Biologically active intact GLP-1 was measured using an in-house sandwich ELISA as previously described (20). Plasma enrichment of [6,6-D2] glucose and [U-13C6] glucose was determined using liquid chromatography–tandem mass spectrometry as previously described (21).

A formal sample size calculation was not performed because of a lack of previous similar studies. Based on previous studies investigating glucometabolic and physiological responses to oral glucose and meal ingestion in totally PX patients (12,13), we assumed that data from 10 PX patients and 10 matched healthy control subjects undergoing the experimental procedures outlined above would provide valuable information on 1) a potential stimulatory effect of SGLT2 inhibition on gut-derived glucagon secretion, and 2) extrapancreatic glucometabolic effects of SGLT2 inhibition in PX patients. Area under the curve (AUC) was calculated using the trapezoid rule and presented as total and baseline-subtracted AUC (bsAUC) values (to take differences in fasting values into account). Glucose rate of appearance (R_a) and glucose rate of disappearance (R_d) were calculated from changes in glucose enrichment using the one-compartment, fixed-volume, non–steady-state model of Steeleand modified for use with stable isotopes and a pool fraction of 70 mL/kg (22). If nothing else is stated, data are presented as mean ± SEM, or, in the case of variables not following normal or log-normal distribution, median with 25th and 75th percentiles. Normally distributed data were compared within and between groups using paired and unpaired sample t tests, respectively. Nonnormally distributed data were analyzed using Mann-Whitney U test for between-group comparisons and Wilcoxon signed-rank test for within-group comparisons. P values ≤0.05 were accepted as statistically significant. Statistical evaluation was performed in SPSS 25.0 (version 25.0, Chicago, IL), and figures were made in GraphPad Prism 9.0 (Boston, MA).

Ten PX patients (8 males; age 65.7 ± 6.4 [mean ± SD] years; BMI 23.8 ± 4.3 kg/m²; hemoglobin A1c [HbA_1c] 7.9 ± 0.8% [63.3 ± 8.8 mmol/mol]; median [25th percentile; 75th percentile] time since operation 2.4 [1.3; 4.2] years) and 10 gender-, age-, and BMI-matched healthy control subjects (8 males; age 65 ± 7.8 years; BMI 24.3 ± 3.5 kg/m²; HbA_1c 5.4 ± 0.4% [35 ± 4.0 mmol/mol]) with no family history of diabetes were included. All participants, except for one PX patient with an estimated glomerular filtration rate of 52 mL/min/1.73 m², had normal kidney function, and no statistically significant differences in estimated glomerular filtration rate between PX patients and control participants were evident. Additional clinical characteristics of the PX patients are displayed in Table 1.

All 20 participants had detectable empagliflozin in plasma on “empagliflozin days.” Concentrations in the pancreatectomy group were similar to those in the control group (time –120 min: 215 [187; 311] vs. 261 [196; 331] pmol/L, P = 0.272; time 90 min: 536 [432; 766] vs. 724 [506; 894] pmol/L, P = 0.281). In both groups, there was a significant increase in the concentration of empagliflozin from −120 min to 90 min (P ≤ 0.007).

Gastric emptying assessed by the acetaminophen absorption test (C_max, T_max, and AUC) was similar on the two experimental days in both groups, but PX patients, in general, exhibited an accelerated gastric emptying of the meal as compared with the healthy control participants (Fig. 1A and B and Table 2).

On placebo days, basal concentrations of plasma glucose were higher in the pancreatectomy group than in the control group (P ≤ 0.03) but not on the empagliflozin days (P ≥ 0.726), because of a more empagliflozin-induced reduction in basal glucose concentrations in the pancreatectomy group (5.0 ± 0.4 vs. 7.9 ± 0.9 mmol/L, P = 0.007) and a smaller effect in the control group (5.1 ± 0.1 vs. 5.5 ± 0.1 mmol/L, P = 0.008). Compared with placebo, empagliflozin reduced postprandial plasma glucose excursions (as assessed by peak concentrations and AUCs) in PX patients, whereas no changes were observed in the control group (Fig. 1C and D and Table 2). One episode of asymptomatic level 1 hypoglycemia (3.2–3.4 mmol/L) was detected in a PX patient before ingestion of the test meal. Baseline levels of total glucose R_a and total glucose R_d, on experimental days 1 and 2, respectively, were similar between the two experimental days in both groups (Supplementary Fig. 2A and B). Compared with placebo, empagliflozin did not affect total glucose R_a or R_d (assessed as bsAUCs) or postprandial EGP in any of the groups (Supplementary Fig. 2).

Serum C-peptide concentrations in the PX participants were below the detection limit of the assay (<16 pmol/L) except in three patients (patients 2, 5, and 8 [Table 1]) showing small increases of C-peptide following meal ingestion with peak values of 23–115 pmol/L (lower than what is needed to define C-peptide negativity [23]). In terms of other end points, these patients did not differ from mean values in the PX patients. In the control group, empagliflozin did not affect fasting serum levels of insulin or C-peptide compared with placebo (Fig. 1E and G and Table 2), but it caused small reductions in the postprandial responses of both insulin and C-peptide (Fig. 1E–H and Table 2).

Empagliflozin did not affect fasting plasma levels of glucagon in any of the groups (Fig. 2A and Table 2). Fasting plasma glucagon concentrations were lower in the pancreatectomy group compared with the control group on placebo days (0.9 [0.3; 1.4] vs. 5.7 [4.5; 7.3] pmol/L, P < 0.001) as well as on empagliflozin days (0.9 [0.1; 2.2] vs. 6.7 [4.3; 8.0] pmol/L, P < 0.001). Both groups exhibited postprandial glucagon responses, which were unaffected by empagliflozin. Still, the pancreatectomy group showed greater postprandial glucagon responses (as assessed by bsAUC taking their lower fasting levels into account) than the control group (Fig. 2A and B and Table 2). However, total AUC for plasma glucagon was higher in control participants because of their higher baseline glucagon levels (Fig. 2A and Table 2).

Compared with placebo, empagliflozin did not affect either fasting plasma GIP or GLP-1 concentrations significantly in any of the groups. Nevertheless, during SGLT2 inhibition, fasting plasma GIP was higher in the pancreatectomy group compared with the control group (P = 0.045) (Fig. 2C and Table 2). The postprandial GIP response during SGLT2 inhibition was lower in the pancreatectomy group compared with the control group (P = 0.043), while there was no difference between groups on the placebo days (Fig. 2C and D and Table 2). There was no difference in postprandial GLP-1 response between days or groups (Fig. 2E and F and Table 2).

Compared with placebo, total diuresis during the 3-h liquid mixed meal test was significantly increased by empagliflozin in both the PX patients (428 ± 83 vs. 193 ± 56 mL, P < 0.001) and the healthy control participants (245 ± 36 vs. 140 ± 29 mL, P = 0.043) (Supplementary Fig. 2I). Compared with placebo, urinary glucose excretion was increased by empagliflozin in both the pancreatectomy group (91.3 ± 12.7 vs. 16.1 ± 7.8 mmol, P < 0.001) and the control group (54.7 ± 0.1 vs. 0.0 ± 0.0 mmol, P < 0.001), with the pancreatectomy group exhibiting greater urinary glucose excretion compared with the control group on both experimental days (P ≤ 0.009) (Supplementary Fig. 2J).

In this randomized, double-blind, placebo-controlled, crossover study, we show that two 25-mg doses of the SGLT2 inhibitor empagliflozin administered, one each, the night before and immediately before a liquid mixed meal test normalize preprandial hyperglycemia and improve postprandial glucose tolerance in totally PX patients, accompanied by increased diuresis and urinary glucose excretion.

To date, the current study is the first randomized placebo-controlled study investigating the glucometabolic effects of SGLT2 inhibition in totally PX patients; thus, no studies are available for comparison. Nevertheless, the effect of SGLT2 inhibition in the treatment of type 1 diabetes has been extensively investigated, and, after being approved as a glucose-lowering drug for this patient group, this indication has recently been withdrawn due to “commercial conflict of interest” (24). However, we still found it relevant to investigate the glucose-lowering effect of SGLT2 inhibitors in totally PX patients, as these patients differ from patients with type 1 diabetes in several aspects. The acute preprandial and postprandial plasma glucose-lowering effects of SGLT2 inhibition in our totally PX patients are on par with previous findings from studies in type 1 diabetes and type 2 diabetes (25,26).

Previous studies on the glucagonotropic effect of SGLT2 inhibition have shown contradicting results (4,5,8,27,28). In the current study, we could not detect any effect of empagliflozin on extrapancreatic glucagon secretion in totally PX patients, and our results in healthy control participants support recent studies showing no glucagonotropic effect of SGLT2 inhibition in healthy individuals (27). In line with these findings, we did not observe any effects of empagliflozin on EGP, supporting findings from recent studies in patients with type 2 diabetes (27,28). Nevertheless, in other studies, SGLT2 inhibition has been shown to stimulate glucagon secretion and EGP in patients with type 2 diabetes (4,5) as well as in healthy control participants (28). The reason for these discrepancies is unclear but may simply relate to different glucose-lowering effects of SGLT2 inhibition in the above-referenced studies, as suggested by Kuhre et al. (10). Also, the present findings may be explained by different tracer methodology. In the current study, resting energy expenditure was not acutely affected by empagliflozin administration, supporting findings from previous studies (5,27).

International guidelines on the management of diabetes secondary to total pancreatectomy are lacking. Presently, insulin treatment remains the only recommended glucose-lowering treatment in the management of totally PX patients, and studies examining the long-term efficacy and safety of adjunctive, and potentially insulin-sparing, glucose-lowering therapy in totally PX patients are warranted, as these patients are notoriously hard to get to glycemic target using insulin-based regimens without triggering an unacceptable risk of hypoglycemic episodes (29). Interestingly, a recent “open label” crossover study in 12 adult outpatients after total pancreatectomy showed that median time spent in euglycemia (3.9–10 mmol/L) was higher during 7 days with a bihormonal artificial pancreas with closed-loop glucose control (78%) compared with a 7-day period with regular diabetes care (57%) ,while time spent in hypoglycemia (<3.9 mmol/L) was shorter (0.0 vs. 1.6%) (30). Larger and longer-term trials are needed to confirm the efficacy and safety of the bihormonal artificial pancreas with closed-loop glucose control in totally PX patients.

The robust glucose-lowering effect of two doses of the SGLT2 inhibitor empagliflozin in totally PX patients reported here raises the question of whether the efficacy and safety of longer-term SGLT2 inhibition in these patients should be investigated. Recently, clinical phase III studies have shown positive effects of the SGLT2 inhibitor dapagliflozin as adjunctive therapy to insulin treatment in overweight type 1 diabetes, with improved glycemic control, decreased insulin requirements, and weight loss without increasing risk of severe hypoglycemia. This initially constituted the foundation of the European Medicine Agency’s approval of dapagliflozin as an adjunctive treatment to insulin therapy in type 1 diabetes (31), later withdrawn by the manufacturer as mentioned (24). In the phase III studies, SGLT2 inhibitor treatment was associated with a small but statistically significant increased risk of diabetic ketoacidosis requiring ketone monitoring (31–33). Nevertheless, in a recent real-life evaluation of 222 patient-years of SGLT2 inhibition in type 1 diabetes, the incidence of ketoacidosis was 0% (34). Interestingly, diabetic ketoacidosis is less frequent in totally PX patients (perhaps because of their lack of pancreatic glucagon secretion known to accelerate ketonemia during insulin withdrawal in type 1 diabetes [35]) as compared with patients with type 1 diabetes (29,36); whether SGLT2 inhibitor treatment will increase the risk of diabetic ketoacidosis in totally PX patients warrants further investigation. Importantly, the weight loss associated with SGLT2 inhibitor treatment is moderate (1–3 kg) (37); it seems to be a consequence of reduced fat mass without effects on muscle mass (38) and to be positively associated with BMI. Therefore, SGLT2 inhibition is considered an important treatment option even in patients with type 2 diabetes who do not have overweight or obesity (as is typical for totally PX patients [29]). Lastly, the recent findings that SGLT2 inhibitor treatment reduces the risk of major adverse cardiovascular events (including hospitalization for heart failure) and renal events in patients with and without type 2 diabetes (39) may be relevant also for totally PX patients.

Recently, Juel et al. (13) reported that a single dose of another noninsulin glucose-lowering agent, the short-acting GLP-1 receptor agonist lixisenatide, caused a robust reduction in postprandial plasma glucose excursions in totally PX individuals. While the plasma glucose-lowering effect of empagliflozin in the current study was associated with increased urinary glucose excretion, the glucose-lowering effect of lixisenatide was associated with decelerated gastric emptying and decreased postprandial glucagon secretion (13).

The randomized, double-blind crossover design is a strength of the current study, but limitations include lack of ketone body measurements, acute treatment, and a relatively small sample size, stressing that nonsignificant findings on secondary end points should be interpreted with caution because of the risk of statistical type 2 errors.

Based on the present results, we believe that SGLT2 inhibition as a plasma glucose-lowering and insulin-sparing treatment modality, not normally associated with an increased risk of hypoglycemia, deserves to be tested in longer-term studies as adjunctive therapy to insulin treatment in selected totally PX patients battling to obtain glycemic control without debilitating hypoglycemia.

Acknowledgments. The authors thank all study participants and acknowledge the laboratory assistance from Lisa H. Jensen and Merete Egeskov from the Steno Diabetes Center Copenhagen; Dorthe B. Nielsen from Center for Clinical Metabolic Research, Gentofte Hospital, University of Copenhagen; and Lene B. Albæk from the Department of Biomedical Sciences, Faculty of Health and Medical Sciences, University of Copenhagen.

Funding. The study was supported by Gentofte Hospital, University of Copenhagen, and partly supported by unrestricted grants from the European Foundation for the Study of Diabetes-Novo Nordisk Research Program for Diabetes Research in Europe, the Novo Nordisk Foundation, and Aase og Ejnar Danielsens Fond.

Duality of Interest. F.K.K. and T.V. have served on scientific advisory panels and given lectures for, served as consultants to, and received research support from Boehringer Ingelheim, the manufacturer of empagliflozin. Boehringer Ingelheim financially supported the analysis of plasma concentrations of empagliflozin. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. M.B. was involved in the planning of the study, obtainment of funding, and conduction of the clinical experiments, and was responsible for the statistical analyses and drafting of the manuscript. S.W.N. conducted clinical experiments. C.P.H. and J.H.S. recruited patients. B.H. and J.J.H. processed and analyzed glucagon and GIP data. T.V. was involved in planning the study. A.L. conceptualized and planned the study, and was involved in the statistical analyses and drafting of the manuscript. F.K.K. conceptualized, planned, supervised, and obtained funding for the study, and wrote the manuscript together with M.B. All authors contributed to interpreting the data, critically reviewed and edited the manuscript, and approved the version to be published. M.B. and F.K.K. are the guarantors of this work and, as such, had full access to all the data in the study and take full responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented at the 80th Scientific Sessions of the American Diabetes Association and the 55th European Association for the Study of Diabetes Annual Meeting in 2020.