2128 Diabetes Care Volume 43, September 2020

Roselle A. Herring,1 Metabolic Effects of an SGLT2 Fariba Shojaee-Moradie,1 Robert Garesse,1 fl Mary Stevenage,1 Nicola Jackson,2 Inhibitor (Dapagli ozin) During a Barbara A. Fielding,2 Agampodi Mendis,2 Sigurd Johnsen,2 A. Margot Umpleby,2 EMERGING THERAPIES: DRUGS AND REGIMENS Period of Acute Insulin Melanie Davies,3 and Withdrawal and Development David L. Russell-Jones1 of Ketoacidosis in People With Type 1 Diabetes Diabetes Care 2020;43:2128–2136 | https://doi.org/10.2337/dc19-2579

OBJECTIVE To determine the effect of the sodium–glucose cotransporter 2 inhibitor dapagli- flozin on glucose flux, lipolysis, and ketone body concentrations during insulin withdrawal in people with type 1 diabetes.

RESEARCH DESIGN AND METHODS A double-blind, placebo-controlled crossover study with a 4-week washout period was performed in 12 people with type 1 diabetes using insulin pump therapy. Participants received dapagliflozin or placebo in random order for 7 days. Stable

isotopes were infused to measure the glucose Ra,Rd, and lipolysis. At isotopic steady state, insulin was withdrawn, and the study was terminated after 600 min or earlier if blood glucose reached 18 mmol/L, bicarbonate <15 mmol/L, venous pH <7.35, or capillary ketones >5.0 mmol/L. 1 RESULTS Centre for Endocrinology, Diabetes and Research, Royal Surrey County Hospital, Guildford, U.K. 2 At baseline, glucose Ra was significantly higher for the dapagliflozin group than the Department of Nutritional Sciences, University placebo group. Following insulin withdrawal, plasma glucose concentrations at the of Surrey, Guildford, U.K. 3 end point were significantly lower with dapagliflozin than placebo and glucose R Diabetes Research Centre, Department of Health d Sciences, University of Leicester, Leicester, U.K. area under the curve (AUC)0–180 min and b-hydroxybutyrate (BOHB) AUC0–180 min fi fi Corresponding author: Roselle A. Herring, roselle were signi cantly higher. There was a small but signi cantly higher glycerol Ra [email protected] fl fi (measure of lipolysis) AUC0–180 min with dapagli ozin. Nonesteri ed fatty acid Received 2 January 2020 and accepted 30 May concentrations were not different between treatments. When divided by BMI 2020 2 >27 and <27 kg/m ,basalglucoseRa, BOHB, and glycerol Ra AUC0–180 min were Clinical trial reg. no. ISRCTN16404006, www.isrctn significantly higher in the low-BMI group with dapagliflozin treatment versus the .org low-BMI group with placebo. This article contains supplementary material online at https://doi.org/10.2337/figshare.12440042. CONCLUSIONS © 2020 by the American Diabetes Association. During insulin withdrawal, the increase in BOHB with dapagliflozin may be partially Readers may use this article as long as the work is due to increased lipolysis. However, reduced renal excretion, reduced BOHB uptake properly cited, the use is educational and not for profit, and the work is not altered. More infor- by peripheral tissues, or a metabolic switch to increased ketogenesis within the liver mation is available at https://www.diabetesjournals may also play a role. .org/content/license. care.diabetesjournals.org Herring and Associates 2129

Sodium–glucose cotransporter 2 (SGLT2) trials. A recent study estimated real RESEARCH DESIGN AND METHODS inhibitors are selective and reversible world experience rates of DKA with off- A single-center, investigator-led, double- inhibitors of renal SGLT2, the major trans- label use of SGLT2 inhibitors in patients blind, placebo-controlled crossover study porter responsible for renal glucose re- with type 1 diabetes to be higher than with a 4-week washout period was per- absorption. They lower plasma glucose by expected based on the sotagliflozin clin- formed in patients with type 1 diabetes reducing renal glucose reabsorption and ical trials (8). However, in March 2019, using insulin pump therapy. Ethics ap- enhancing urinary glucose excretion. In dapagliflozin was approved in Europe proval was granted from the National type 2 diabetes, this results in a com- for people with type 1 diabetes as an Research Ethics Service committee, South pensatory increase in endogenous glu- adjunct to insulin in people with a BMI Central–Berkshire B. The clinical trial was 2 cose production, lower tissue glucose .27 kg/m (9) and received an evidence- registered with the European Clinical Tri- uptake, and higher glucagon levels based recommendation by the National als Database (EudraCT) under the number (1). Institute for Clinical Excellence in August 2015-002094-38. It has been postulated that the SGLT2 2019 (10). Participants with type 1 diabetes for inhibitorclass mayhave beneficial effects The incidence of DKA is between 0 and .12 months were recruited between in the management of type1 diabetes. 56 cases per 1,000 person-years among February 2018 and October 2018 with Phase 2 and 3 clinical trial studies have people with type1 diabetes (11); the use of the diabetes insulin pump data- shown encouraging results, with lower pathogenesis of DKA is well-known. In base at the Royal Surrey County National HbA1c levels and reduced weight and the absence of adequate direct insulin Health Service Trust. Type 1 diabetes was insulin requirements (2–7). However, sig- action in the liver, the excess glucose established by clinical presentation, treat- nificant safety issues have been high- production is so great that an increase in ment response, and C-peptide level. Ex- lighted in clinical trials, with increased glucose-dependent peripheral glucose clusion criteria included proliferative rates of ketoacidosis, although rates uptake and associated glycosuria fail retinopathy requiring acute treatment differed between trials. In Dapagliflozin to limit progressive hyperglycemia. Li- within the last 3 months, moderate-to- Evaluation in Patients With Inadequately polysis is increased due to lack of insulin’s severe renal impairment (creatinine clear- Controlled Type 1 Diabetes (DEPICT-1), action on hormone-sensitive lipase re- ance ,60 mL/min or estimated glomerular diabetic ketoacidosis (DKA) occurred in sulting in increased release of nonester- filtration rate ,60 mL/min/1.73 m2), 1% and 2% of the dapagliflozin 5 mg and ified fatty acids (NEFA) and glycerol from severe hepatic impairment, New York 10 mg groups, respectively (2). Pooled adipose triglyceride stores (12). In the Heart Association class III–IV cardiac fail- data from the 26-week Empagliflozin as liver, due to the diversion of oxaloacetate ure, uncontrolled cardiac arrhythmia, un- Adjunctive to InSulin thErapy Over 52 toward gluconeogenesis, there is a re- controlled hypertension, mental incapacity, Weeks in Patients With Type 1 Diabetes duced capacity of the Krebs cycle to pregnancy, or breastfeeding. Those with Mellitus (EASE)-2 and EASE-3 studies oxidize acetyl-CoA derived from the child-bearing potential not taking ade- showed DKA occurred in 0.8%, 3.3%, b-oxidation of NEFA. Consequently, there quate contraception precautions and those and 4.3% of the empagliflozin 2.5 mg, is an increased flux of acetyl-CoA toward with suspected allergy to trial products were 10 mg, and 25 mg arms, respectively, and ketogenesis giving rise to increased also excluded. 1.2% in the placebo arm (4). In the Efficacy, acetoacetate, b-hydroxybutyrate (BOHB), Safety, and Tolerability Study of Sotagli- and acetone (13). Hence, in insulin de- Design flozin as Adjunct Therapy in Adult Patients ficiency, there are two parallel metabolic Participants received dapagliflozin (10 With Type 1 Diabetes Mellitus Who Have processes: glucose overproduction and mg daily) or placebo in random order for Inadequate Glycemic Control With Insulin ketone overproduction. 7 days. They were made aware of poten- Therapy (Tandem1), euglycemic keto- The potentially life-threatening state tial changes in glycemic control and were acidosis occurred in 3.4% and 4.2% of of euglycemic DKA is unusual and was asked to record trial medication admin- the sotagliflozin 200 mg and 400 mg first described in 1973 (14). A specula- istration, any concomitant medication groups and 0.4% in the placebo group tive mechanism for its development (to include insulin), hypoglycemia fre- over 24 weeks (6). Finally, in an 18-week with SGLT2 inhibitors is that the insulin- quency (capillary glucose level ,4mmol/ randomized study with canagliflozin, independent removal of glucose from the L), fasting ketone levels, and any adverse ketone-related adverse events occurred body enables glycemic control concur- events. in 5.1% of the 100 mg arm and 9.5% of rent with either an absolute or a relative On day 7, participants attended for a the 300 mg arm (7). deficiency of insulin. In addition, keto- metabolic study. They were asked not to The U.S. Food and Drug Administration acidosis may be driven by an increase in consume food and to drink only water has not approved SGLT2 inhibitors for counterregulatory hormones, i.e., gluca- from 2200 h the day before. They were use as an adjunct therapy in type 1 di- gon, cortisol, or growth hormone. Using also asked not to undertake any stren- abetes and have warned about the risk stable isotope techniques, this study uous exercise or consume alcohol for 24 h of euglycemic DKA. A number of triggers explored these potential mechanisms before the study day. All participants for ketoacidosis were identified. These by studying the effect of the SGLT2 were using continuous subcutaneous in- included reduced calorie intake, lower inhibitor dapagliflozin on glucose flux, sulin therapy and disconnected their in- insulin doses, and illness. It is also impor- lipolysis, and ketone body concen- sulin pumps at 0600 h on the morning of a tant to note that very few participants tration during hyperglycemia in peo- metabolic study. They were transferred experienced a state of illness or intercur- ple with absolute endogenous insulin to a soluble variable insulin infusion to rent stress during the reported clinical deficiency. maintain a whole-blood concentration of 2130 SGLT2 Inhibitor in Type 1 Diabetes Diabetes Care Volume 43, September 2020

2 glucose at 5 mmol/L. [6,6- H2]glucose (Nottingham, U.K.). Serum cortisol Denominator degrees of freedom were 5 and [1,1,2,3,3 H2]glycerol were infused concentration were measured using an adjusted using the Kenward-Roger ap- from 2120 min to the study end point to in-house radioimmunoassay. Serum growth proximations. All other response variables measure glucose Ra and Rd and lipolysis, hormone concentrations were measured measured or derived once per participant respectively (Cambridge isotopes; CK Gas using an immunoradiometric kit purchased in each period were analyzed as per the Products, Ibstock, U.K.). from DRG Instruments, supplied by IDS, primary end point. At 0 min, when isotopic steady state Tyne and Wear, U.K. The isotopic enrich- For assessment of the impact of par- had been achieved, insulin was with- ment of plasma glucose was determined as ticipant BMI level on study conclusions, drawn, participants were given a single the trimethylsilyl-O-methyloxime deriv- the participants were classified as “high dose of study medication, and blood ative (15) using gas chromatography– BMI” (BMI .27 kg/m2, N 5 5) or “low glucose was allowed to increase. mass spectrometry (model 5975C, inert BMI” (BMI ,27 kg/m2, N 5 7). The data Blood samples were taken to measure XL EI/CI MSD; Agilent Technologies, for which a single value per participant plasma glucose and glycerol enrichment Wokingham, U.K.). The isotopic enrich- per period was analyzed were then re- and concentration and concentrations of ment of plasma glycerol was determined as analyzed as described above with the NEFA and plasma BOHB every 20 min the tert-butyl trimethylsilyl glycerol de- modification that the model indepen- until 180 min and then at 30-min inter- rivative (16) using a gas chromatography– dent variables included, additionally, vals. Interval urine collection was also mass spectrometry model 5975 network the BMI classification and the interaction taken, and blood samples for counter- MSD (Agilent Technologies). of the BMI classification with treatment. regulatory hormones were taken to mea- GlucoseRa,Rd,andglycerolproduction The data for which several time point sure plasma glucagon and serum insulin, were calculated using the Steele non– values per participant per period were growth hormone, and cortisol concen- steady state equations modified for sta- analyzed were then reanalyzed as de- trations at 120-min intervals. Urine sam- ble isotopes (17). scribed above with the modification that ples were collected over 2-h intervals for the model independent variables in- measurement of glucose and spot uri- Statistical Analysis cluded, additionally,theBMIclassification, nary ketones. Statistical analysis was performed using the interaction of the BMI classification R, version 3.5.1, and SAS, version 9.4 or with treatment, and the interaction of Rescue and Study Termination above. All hypothesis tests were two the BMI classification, treatment, and At 600 min (10 h) or in the event of blood sided and evaluated at a significance time point. For each separate time point, glucose of 18 mmol/L, bicarbonate ,15 level of 5%. the same additional estimates of effect mmol/L, or venous pH ,7.35 or point-of- The primary end point was glucose and P value for BMI and treatment com- care capillary ketone level of .5.0 mmol/L, concentration as measured at 600 min bined, as immediately above, were reported. the metabolic study was terminated and or at the time of glycemic rescued Area under the curves were calculated participants commenced on rescue intra- whichever occurred first. The study was for study points 0–180 min with and venous insulin infusion and 5% dextrose powered for 12 subjects. In a recent without correction for the baseline. In until blood glucose levels stabilized. study where insulin was withdrawn in the figures, the mean values at each time people with type 1 diabetes, the SD for point only include data where there is a Plasma Measurements plasma glucose concentration was 20% measurement from both arms of the On the study day, plasma glucose con- (18). With 12 subjects, studied with and study at that time point. We have called centration was measured using a glu- without the SGLT2 inhibitor, a differ- these paired measurements. cose oxidase technique on a glucose ence in plasma glucose of 24% can be analyzer (YSI 2300; Yellow Springs Instru- detected with 80% power and two- RESULTS ments, Yellow Springs, OH). Whole-blood sided significance level of 5%. Twelve people (4 male and 8 female) with samples were immediately centrifuged Final glucose concentration was sta- type 1 diabetes, with mean 6 SEM dura- and aliquots of plasma stored at 280°C tistically analyzed as the response vari- tion of diabetes 23.3 6 4.1 years (age 40.7 2 to be analyzed at a later date in a laboratory able in a general linear mixed model 6 3.9 years, BMI 26.8 6 1.4 kg/m ,HbA1c setting. (using the PROC MIXED procedure in 59.9 6 2.3 mmol/mol), completed the Plasma glucose concentrations were SAS software), with treatment, period, study. All participants had C-peptide ,0.2 measured with a Roche Cobas Mira and treatment-by-period interaction as nmol/L, apart from one participant who analyzer using the ABX Pentra glucose fixed effects and the baseline glucose had a C-peptide level of 0.314 nmol/L kit (HORIBA ABX, Northampton, U.K.) concentration as a covariate. The partic- and 0.352 nmol/L in the dapagliflozin and plasma glycerol and BOHB concen- ipant was a random effect in the model. versus placebo arm, respectively, with trations using Randox kits (Glycerol The secondary end points were sta- duration of type 1 diabetes of 17 years. and Ranbut; Randox Laboratories, Co. tistically evaluated as response variables During the 7-day treatment period, Antrim, U.K.). Plasma NEFA concentra- using a general linear mixed model with there was no significant difference in tions were measured using an enzy- treatment, period, time, treatment-by- insulin dose between dapagliflozin ver- matic kit from Wako Chemicals (Neuss, period, and treatment-by-time interactions sus placebo (mean 6 SEM 0.056 6 0.007 Germany). as fixed effects, baseline measurement as units/kg vs. 0.058 6 0.008 units/kg, re- Insulin and glucagon were measured a covariate, participant as a random effect, spectively). Two participants reduced using radioimmunoassays purchased and time as a repeated measure, with their basal setting due to hypoglycemia. from Merck Millipore, Merck Chemicals SP(POW) variance-covariance structure. Other participants did not change their care.diabetesjournals.org Herring and Associates 2131

insulin dose but encountered more epi- SEM glucose concentration was 8.5 6 treatments. AUC0–180 min for glucose sodes of hypoglycemia overnight, which 0.7 mmol/L with dapagliflozin treatment Rd and metabolic clearance rate (MCR) were remedied by taking GlucoGel (for- and 14.3 6 1.1 mmol/L with placebo (P 5 were higher with dapagliflozin compared merly known as HypoStop gel). 0.0005) (Fig. 1). Urinary glucose excre- with placebo (P 5 0.0041, P , 0.0001). The duration of each metabolic study tion between 0 and 120 min (n 5 6) was prior to termination or rescue varied 5.10 6 0.80 mmol/kg/min with dapagli- BOHB, NEFA, and Glycerol Metabolism from 180 to 600 min. All subjects com- flozin and 0.029 6 0.01 mmol/kg/min Baseline pleted 180 min of each metabolic study. with placebo (P 5 0.003). There was no difference in baseline The average time for each metabolic glycerol Ra (a measure of lipolysis) be- study was not different between dapa- Glucose Metabolism tween treatments (Fig. 2 and Table 1). gliflozin (mean 6 SEM 418 6 44 min) and At baseline (0 min), when isotopic en- Baseline BOHB was higher with dapagli- placebo (448 6 38 min). richment and glucose concentration flozin (mean 6 SEM 0.39 6 0.11 vs. 0.16 were in a steady state, glucose Ra was 6 0.05 mmol/L, P 5 0.044), and NEFA Glucose Concentration significantly higher with dapagliflozin concentration was higher (0.61 6 0.09 vs At 0 min, glucose concentration was not compared with placebo (P 5 0.0088) 0.47 6 0.07 mmol/L), although this was different between treatments (Fig. 1). (Fig. 1 and Table 1). At this time point, borderline significant (P 5 0.054). There Following insulin withdrawal, plasma glucose Rd is equal to glucose Ra. During was a significant positive relationship glucose concentration increased in both insulin withdrawal, glucose Ra increased, between NEFA and BOHB with both groups, but at the end of the study peaking at 90 min, and then declined, treatments (Supplementary Fig. 3). Base (600 min or time of rescue) the mean 6 with no difference in AUC between excess concentration and venous bicarbonate

Figure 1—Paired plasma glucose concentration (A), glucose Ra (B), glucose Rd (C), and glycerol Ra (D). All subjects completed 180 min of each metabolic study (n 5 12). By 480 min, n 5 5. ○, placebo; C, dapagliflozin. 2132 SGLT2 Inhibitor in Type 1 Diabetes Diabetes Care Volume 43, September 2020

Table 1—Baseline AUC and incremental AUC from 0 to 180 min for glucose, glycerol, and BOHB metabolism at the end of 7 days’ treatment with dapagliflozin or placebo Dapagliflozin Placebo P Baseline C-peptide concentration (nmol/L) 0.10 6 0.02 0.10 6 0.03 0.470 Glucose concentration at baseline (mmol/L) 5.43 6 0.14 5.74 6 0.17 0.180

Glucose concentration AUC0–180 (mmol/L z min) 1,450 6 70 1,744 6 87 0.015

Glucose concentration incremental AUC0–180 (mmol/L z min) 464 6 77 720 6 81 0.026

Glucose Ra at baseline (mmol/min/kg) 13.0 6 0.77 11.7 6 0.7 0.009

Glucose Ra AUC0–180 (mmol/kg) 3,337 6 271 2,969 6 216 0.157

Glucose Ra incremental AUC0–180 (mmol/kg) 879 6 180 758 6 127 0.596

Glucose Rd at baseline (mmol/min/kg) 13.1 6 0.7 11.8 6 0.6 0.009

Glucose Rd AUC0–180 (mmol/kg) 2,728 6 222 2,014 6 147 0.004

Glucose Rd incremental AUC0–180 (mmol/kg) 593 6 140 275 6 100 0.049 Glucose MCR at baseline (mL/min/kg) 2.43 6 0.14 2.06 6 0.06 0.008

Glucose MCR AUC0–180 (mL/kg) 347 6 26 215 6 9 <0.001

Glucose MCR incremental AUC0–180 (mL/kg) 238.9 6 8.1 278.8 6 8.2 <0.001 Glycerol concentration at baseline (mmol/L) 81.6 6 24.2 75.1 6 26.9 0.392

Glycerol concentration AUC0–180 (mmol/L z min) 15,279 6 1,748 14,684 6 1,388 0.430

Glycerol concentration incremental AUC0–180 (mmol/L z min) 3,412 6 574 3,524 6 549 0.801

Glycerol Ra at baseline (mmol/min/kg) 2.71 6 0.46 2.17 6 0.28 0.134

Glycerol Ra AUC0–180 (mmol/kg) 585 6 77 543 6 56 0.048

Glycerol Ra incremental AUC0–180 (mmol/kg) 66.3 6 15.0 98.2 6 22.5 0.129 BOHB concentration at baseline (mmol/L) 0.39 6 0.11 0.16 6 0.05 0.044

BOHB concentration AUC0–180 (mmol/L z min) 149 6 26 117 6 18 0.035

BOHB concentration incremental AUC0–180 (mmol/L z min) 79.6 6 13.3 88.1 6 15.4 0.588 NEFA concentration at baseline (mmol/L) 0.61 6 0.09 0.47 6 0.07 0.054 Data are means 6 SEM. Boldface type indicates statistical significance where P , 0.05.

were lower with dapagliflozin versus P 5 0.0127). Insulin at arrival was not (18.7 6 5.4 years vs. 29.8 6 5.6 years, placebo (0.61 6 0.0.6 vs. 1.66 6 different between dapagliflozin and pla- respectively) (Supplementary Table 2). 0.48 mmol/L, P 5 0.0095, and 24.1 6 cebo (231 6 40 vs. 264 6 50 pmol/L). Baseline glucose Ra, baseline BOHB, 0.4 vs. 24.9 6 0.30 mmol/L, P 5 0.028, Glucagon-to-insulin ratio at arrival was and glycerol Ra AUC0–180 min were higher respectively). also not significantly different between in the low-BMI group with dapagliflozin dapagliflozin and placebo (0.24 6 0.04 than in the high-BMI group on placebo During Insulin Withdrawal vs. 0.19 6 0.02 ng/pmol). (P 5 0.001, P 5 0.019, and P 5 0.012, BOHB AUC0–180 min was higher with At 0 min, neither insulin (mean 6 SEM respectively). These measurements were dapagliflozin compared with placebo 214 6 45 vs. 235 6 46 pmol/L) nor not different in the high-BMI group with (mean 6 SEM 149 6 26 vs. 117 6 glucagon (35.8 6 2.5 vs. 31.2 6 dapagliflozin versus the high-BMI group 18 mmol/L z min, P 5 0.035). NEFA, 3.8 ng/L) was different between dapa- on placebo (Table 2). NEFA concentra- venous bicarbonate, pH, and capillary gliflozin and placebo, respectively, but tions were not different between dapa- ketones and urinary ketones were not the glucagon-to-insulin ratio was higher gliflozin versus placebo in either group statistically different, but there was a with dapagliflozin (0.27 6 0.06) than (Supplementary Table 2). small but significantly higher AUC0–180 min placebo (0.16 6 0.02 ng/pmol, P 5 For the placebo group, baseline insulin glycerol Ra. 0.04). concentration was significantly higher in BOHB AUC0–180 min and glycerol Ra During insulin withdrawal, insulin, glu- the high-BMI vs. low-BMI group (P 5 AUC0–180 min were negatively correlated cagon, and the glucagon-to-insulin ratio 0.046) (Table 2). with BMI with both dapagliflozin (r 520.58, were not significantly different in the For both the dapagliflozin and placebo P 5 0.048, and r 520.625, P 5 0.030 two treatment arms (Supplementary groups, NEFA concentration at baseline respectively) and placebo (r 520.77, Fig. 2). (both P , 0.05) (Supplementary Table 2) 5 52 5 P 0.003, and r 0.757, P 0.004) There was no statistical difference in and BOHB AUC0–180 min (P , 0.01, P , (Supplementary Fig. 1). growth hormone concentration or cor- 0.05) (Table 2) were significantly lower tisol concentration at any time point. in the high-BMI versus low-BMI group Counterregulatory Hormones (Supplementary Table 2). At arrival (2120 min), glucagon was Subgroup Analysis significantly higher for the dapagliflo- In people with BMI ,27 kg/m2 vs. zin treatment group compared with BMI .27 kg/m2, there was no significant CONCLUSIONS that in the placebo group (mean 6 difference in HbA1c (61 6 1.2 vs. 59 6 This study provides clear evidence that SEM 42.1 6 3.8 vs. 35.2 6 3.9 ng/L, 1.6 mmol/mol) or duration of diabetes when dapagliflozin was used as an adjunct care.diabetesjournals.org Herring and Associates 2133

Figure 2—Paired concentration of capillary ketones (A), BOHB (B), and NEFA (C) in plasma, and venous blood bicarbonate (D). All subjects completed 180 min of each metabolic study (n 5 12). By 480 min, n 5 5. ○, placebo; C, dapagliflozin. therapy in people with type 1 diabetes dapagliflozin versus placebo. However, also due to an increase of tissue glucose on insulin pump therapy, plasma BOHB this highlights that glucose levels in the uptake stimulated by dapagliflozin. was higher in the dapagliflozin group presence of dapagliflozin may not be Although at baseline there was no versus placebo group both in the pres- used as a marker for insulin deficiency difference in glycerol Ra (a measure of ence of insulin treatment and during in patients in the context of patient- lipolysis), NEFA concentration was higher insulin withdrawal. The power of this guided “sick day rules.” with dapagliflozin. The strong relation- study from the clinical perspective was It has been hypothesized that SGLT2 ship between NEFA and BOHB at baseline the crossover design with each individ- inhibition with single-dosing empagliflozin and during insulin withdrawal suggests ual undergoing an identical insulin with- in people with type 2 diabetes results in that NEFA flux to the liver was driving the drawal protocol and the only difference a reduction in insulin secretion and an higher BOHB with dapagliflozin. When being the presence or absence of the augmented glucagon response, which we divided the subjects based on BMI, SGLT2 inhibitor. in turn enhances gluconeogenesis and this effect was only seen in those with a The primary end point was plasma lipolysis (1). The higher baseline glucose low BMI; the BOHB response in the glucose concentration at the end of Ra with dapagliflozin, in the current presence of dapagliflozin at baseline fi study. This was signi cantly lower during study, may be due to augmented gluco- and the rise in glycerol Ra during insulin dapagliflozin treatment than for the pla- neogenesis caused by the increased glu- withdrawal was higher than placebo. This cebo group, while the time to study cagon/insulin. Another possibility is that was not seen in the high-BMI group. termination did not differ between the it may be due to the increase in glucose There was a striking negative relationship two treatments. It is therefore reassuring Rd, although the study cannot determine between lipolysis and BMI and ketone from a clinical perspective that there was whether the increase in glucose Rd is due levels and BMI. While this is only a small no difference in the time to rescue with to increased glucose excretion only or group and we must highlight that the 14SL2Ihbtri ye1Diabetes 1 Type in Inhibitor SGLT2 2134

Table 2—Baseline AUC and incremental AUC from 0 to 180 min for glucose, glycerol, and BOHB metabolism at the end of 7 days’ treatment with dapagliflozin or placebo in BMI subgroups Dapagliflozin Placebo BMI ,27 kg/m2 BMI .27 kg/m2 BMI ,27 kg/m2 BMI .27 kg/m2 BMI ,27 kg/m2 BMI .27 kg/m2 dapagliflozin vs. placebo P dapagliflozin vs. placebo P Age (years) 34.4 6 1.9 49.4 6 1.5§ 34.6 6 1.9 49.4 6 1.5§ BMI (kg/m2) 22.5 6 0.4 31.4 6 0.3** 22.9 6 0.4 31.4 6 0.4** Glucose concentration at baseline (mmol/L) 5.4 6 0.1 5.5 6 0.1 5.8 6 0.1 5.6 6 0.1 0.177 0.694

Glucose concentration AUC0–180 (mmol/L z min) 1,420 6 43 1,498 6 30 1,827 6 53 1,651 6 44 0.019 0.360

Glucose concentration incremental AUC0–180 (mmol/L z min) 415 6 46 535 6 34 778 6 45 648 6 51 0.024 0.414

Glucose Ra at baseline (mmol/min/kg) 14.5 6 0.3 11.0 6 0.3† 12.5 6 0.4 10.5 6 0.2 0.001 0.528

Glucose Ra AUC0–180 (mmol/kg) 3,524 6 161 3,076 6 125 3,250 6 116 2,579 6 98 0.428 0.235

Glucose Ra incremental AUC0–180 (mmol/kg) 790 6 106 1,002 6 94 978 6 53 529 6 106 0.631 0.175 Glucose Rd at baseline (mmol/min/kg) 14.6 6 0.3 11.05 6 0.3† 12.7 6 0.4 10.6 6 0.2 0.001 0.521

Glucose Rd AUC0–180 (mmol/kg) 3,000 6 126 2,347 6 76 2,230 6 76 1,692 6 27 0.009 0.175

Glucose Rd incremental AUC0–180 (mmol/kg) 635 6 87 546 6 56 279 6 42 50 6 16 0.378 0.037 Glucose MCR at baseline (mL/min/kg) 2.7 6 0.1 2.0 6 0.1† 2.22 6 0.1 1.9 6 0.0 0.002 0.285

Glucose MCR AUC0–180 (mL/kg) 389 6 12 286 6 7† 228 6 4 195 6 3 <0.001 0.007

Glucose MCR incremental AUC0–180 (mL/kg) 235.9 6 4.9 242 6 4.0 299.3 6 10.8 292 6 7.0 0.039 0.002 Glycerol concentration at baseline (mmol/L) 112.9 6 14.2 37.9 6 4.2 104.5 6 16.5 33.9 6 2.9 0.426 0.752

Glycerol concentration AUC0–180 (mmol/L z min) 17,428 6 1,007 12,270 6 525 16,477 6 771 12,174 6 491 0.396 0.876

Glycerol concentration incremental AUC0–180 (mmol/L z min) 3,624 6 300 3,115 6 403 3,185 6 239 3,999 6 460 0.455 0.212

Glycerol Ra at baseline (mmol/min/kg) 3.41 6 0.23 1.73 6 0.20§ 2.60 6 0.12 1.57 6 0.17 0.118 0.648 ibtsCare Diabetes Glycerol Ra AUC0–180 (mmol/kg) 6.94 6 41 433 6 31§ 637 6 27 411 6 24 0.012 0.866

Glycerol Ra incremental AUC0–180 (mmol/kg) 59.7 6 5.8 75.8 6 13.8 107.5 6 12.7 85.1 6 13.5 0.119 0.661 BOHB concentration at baseline (mmol/L) 0.6 6 0.05 0.10 6 0.03† 0.23 6 0.06 0.06 6 0.01 0.019 0.703

AUC0–180 BOHB concentration (mmol/L z min) 196 6 11 84 6 12† 156 6 6626 7§ 0.071 0.273

BOHB concentration incremental AUC0–180 2020 September 43, Volume (mmol/L z min) 89 6 7666 8 114 6 8526 6§ 0.219 0.521 Insulin concentration (pmol/L) at baseline 154 6 22 298 6 25 159 6 11 341 6 39§ 0.316 0.924 Data are means 6 SEM. Boldface type indicates statistical significance where P , 0.05. Significance between BMI groups: §P , 0.05, †P , 0.01, **P , 0.001. care.diabetesjournals.org Herring and Associates 2135

study design was not powered to look inhibitors may by nature of their mech- AstraZeneca, and Janssen; has acted as advisory at the difference in BMI, our results anism of action further predispose a board member for Servier and Gilead Sciences suggest a greater risk of ketosis in people person to the risks of euglycemic DKA. and speaker for NAPP Pharmaceuticals, Mitsu- bishi Tanabe Pharma Corporation, and Takeda with type 1 diabetes with a low BMI, SGLT2 inhibitors lower the glucose avail- Pharmaceuticals International; and has received which may be of clinical relevance. Al- ability during insulin deficiency by en- grants in support of research trials from Novo though baseline NEFA concentration was couragingglucoseexcretion andfluidloss Nordisk, Sanofi, Lilly, Boehringer Ingelheim, higher with dapagliflozin, there was no through persistent glycosuria. The base AstraZeneca, and Janssen. No other potential conflicts of interest relevant to this article were difference in NEFA concentration with excess concentration at baseline (mmol/ reported. dapagliflozin in the whole group and the L) was statistically higher for the dapagli- Author Contributions. R.A.H., F.S.-M., and BMI subgroups following insulin with- flozin group, indicating a greater need to M.S. drafted the manuscript. F.S.-M. and R.G. drawal. However, it is possible that an balance acid to ensure a normal body pH carried out the metabolic studies. F.S.-M. carried increase in NEFA production was matched with administration of SGLT2 inhibitors. out sample analysis and interpreted data. N.J. assisted with sample analysis. B.A.F., A.M.U., by an increase in NEFA clearance. The SGLT2 inhibitor was only taken for a M.D., and D.L.R.-J. participated in the design 2 Individuals with a BMI .27 kg/m also short term, and further evidence is re- of the study, interpreted data, and reviewed and required greater insulin doses and had quired to establish any longer-term adap- edited the manuscript. A.M. and S.J. completed the higher plasma insulin concentrations than tions with administration on a regular basis. statistical analysis. R.A.H. is the guarantor of this people with a BMI ,27 kg/m2,leadingto In summary, this study provides evi- work and, as such, had full access to all the data in fl the study and takes responsibility for the integrity lower plasma level of ketones. This can be dence that when dapagli ozin is used as of the data and the accuracy of the data analysis. clinically important because it suggests that an adjunct therapy in people with type 1 Prior Presentation. Parts of this study were the greater insulin requirement in people diabetes on insulin pump therapy, there presented in abstract form at the 55th Annual with a BMI .27 kg/m2 is likely to minimize is a rise in ketones during insulin de- Meeting of the European Association for the – the risk of ketoacidosis caused by SGLT2 ficiency. This has clinical implications. Study of Diabetes, Barcelona, Spain, 16 20 Sep- tember 2019. therapy. The stability of glucose levels indicates While an increase in ketogenesis driven that there will need to be more reliance References by lipolysis is one mechanism for the rise on capillary ketone monitoring with 1. Ferrannini E, Muscelli E, Frascerra S, et al. in BOHB, the rise in lipolysis with dapagli- SGLT2 inhibitors to prevent this predict- Metabolic response to sodium-glucose cotrans- flozin was small and we cannot rule out able and preventable risk (22). Emphasis porter 2 inhibition in type 2 diabetic patients. J thepossibilitythatreducedketoneuptake should be on ensuring adequate basal Clin Invest 2014;124:499–508 by peripheral tissues or reduced urinary insulin levels and avoiding calorie re- 2. Dandona P, Mathieu C, Phillip M, et al.; DEPICT-1 Investigators. Efficacy and safety of da- excretion of ketones may also play a role. striction and dehydration. The patient pagliflozin in patients with inadequately controlled The excretion of ketones occurs by a and clinician will need to be constantly type 1 diabetes (DEPICT-1): 24 week results from a balance between glomerular filtration vigilant, and there must be a low thresh- multicentre, double-blind, phase 3, randomised and tubular reabsorption. Experimental old for stopping this class of drug during controlled trial. Lancet Diabetes Endocrinol 2017; – and clinical data have established that periods of illness or intercurrent stress 5:864 876 3. Pieber TR, Famulla S, Eilbracht J, et al. Empa- SGLT2 inhibition can reduce glomerular (23). A four-step approach to address the gliflozin as adjunct to insulin in patients with filtration rate in people with type 1 and risk of DKA during treatment with SGLT2 type 1 diabetes: a 4-week, randomized, placebo- 2 diabetes, thereby reducing renal ketone inhibitors in people with type 1 diabetes controlled trial (EASE-1). Diabetes Obes Metab excretion (19). In the current study, al- has been suggested as a lifesaving mea- 2015;17:928–935 4. Rosenstock J, Marquard J, Laffel LM, et al. though plasma ketone concentrations sure to help clinicians and patients (24). Empagliflozin as adjunctive to insulin therapy in were higher during dapagliflozin treatment, In addition, an international consensus type 1 diabetes: the EASE trials. Diabetes Care urinary ketone concentrations were not has been published to address the risk 2018;41:2560–2569 different versus the placebo group. This management of DKA in people with 5. Garg SK, Henry RR, Banks P, et al. Effects of fl contrasts with a study in people with type 1 diabetes (25). SGLT2 inhibitors sotagli ozin added to insulin in patients with Type 1 diabetes. N Engl J Med 2017;377:2337–2348 type 2 diabetes that showed increased are now being used in clinical practice 6. Buse JB, Garg SK, Rosenstock J, et al. Sotagli- clearance and fractional urinary excretion both within and outside of license, and flozin in combination with optimized insulin of BOHB after single and 4-week chronic we watch with interest the clinical ben- therapy in adults with type 1 diabetes: the North use of empagliflozin (20). efits and risks of this class of drug as more American inTandem1 study. Diabetes Care 2018; – In the presence of prolonged insulin real-world experience is gained in people 41:1970 1980 7. Peters AL, Henry RR, Thakkar P, Tong C, Alba withdrawal secondary to periods of ill- with type 1 diabetes. M. Diabetic ketoacidosis with canagliflozin, a ness or stress, there may also be upreg- sodium-glucose cotransporter 2 inhibitor, in ulation of the renal reabsorption capacity patients with type 1 diabetes. Diabetes Care of ketones, but this needs further investi- 2016;39:532–538 gation. This also has clinical implications, Funding. This study was funded by Diabetes UK 8. Hampp C, Swain RS, Horgan C, et al. Use of (15/0005178), and the investigational medicinal sodium–glucose cotransporter 2 inhibitors in pa- as historically ketonuria has been used to product was provided by AstraZeneca (study tients with type 1 diabetes and rates of diabetic screen for the presence of ketosis (21). identification ESR-15-10890). ketoacidosis. Diabetes Care 2020;43:90–97 It has been suggested that short-term Duality of Interest. R.A.H. and D.L.R.-J. have 9. Biester T, Kordonouri O, Danne T. Beyond fasting and dehydration predispose peo- received research funding or advisory board or type 2 diabetes: sodium glucose co-transporter- lecturefeehonorariafromAstraZeneca.M.D.has inhibition in type 1 diabetes. Diabetes Obes ple with type 1 diabetes to develop acted as consultant, advisory board member, Metab 2019;21(Suppl. 2):53–61 euglycemic DKA during periods of in- and speaker for Novo Nordisk, Sanofi, Lilly, 10. Dapagliflozin with insulin for treating type 1 sulin withdrawal/deficiency (19). SGLT2 Merck Sharp & Dohme, Boehringer Ingelheim, diabetes. Technology appraisal guidance [TA597] 2136 SGLT2 Inhibitor in Type 1 Diabetes Diabetes Care Volume 43, September 2020

[Internet], 2019. Available from https://www chromatography-mass spectrometry as a mea- 20. Ferrannini E, Baldi S, Frascerra S, et al. Renal .nice.org.uk/guidance/ta597. Accessed 28 Au- sure of glucose transport in vivo. J Mass Spectrom handling of ketones in response to sodium– gust 2019 1996;31:961–966 glucose cotransporter 2 inhibition in patients 11. Fazeli Farsani S, Brodovicz K, Soleymanlou N, 16. Flakoll PJ, Zheng M, Vaughan S, Borel MJ. with type 2 diabetes. Diabetes Care 2017;40: Marquard J, Wissinger E, Maiese BA. Incidence Determination of stable isotopic enrichment and 771–776 and prevalence of diabetic ketoacidosis (DKA) concentration of glycerol in plasma via gas 21. Vallon V. The mechanisms and therapeutic among adults with type 1 diabetes mellitus chromatography-mass spectrometry for the es- potential of SGLT2 inhibitors in diabetes mellitus. (T1D): a systematic literature review. BMJ Open timation of lipolysis in vivo. J Chromatogr B Annu Rev Med 2015;66:255–270 2017;7:e016587 Biomed Sci Appl 2000;744:47–54 22. Taylor SI, Blau JE, Rother KI. SGLT2 inhibitors 12. Jaworski K, Sarkadi-Nagy E, Duncan RE, 17. Steele R, Wall JS, De Bodo RC, Altszuler N. may predispose to ketoacidosis. J Clin Endocrinol Ahmadian M, Sul HS. Regulation of triglyceride Measurement of size and turnover rate of body Metab 2015;100:2849–2852 metabolism. IV. Hormonal regulation of lipolysis glucose pool by the isotope dilution method. Am 23. Herring R, Russell-Jones DL. SGLT2 inhibitors in adipose tissue. Am J Physiol Gastrointest Liver J Physiol 1956;187:15–24 in Type 1 diabetes: is this the future? Diabet Med Physiol 2007;293:G1–G4 18. Herring RA, Shojaee-Moradie F, Umpleby 2018;35:1642–1643 13. Kitabchi AE, Umpierrez GE, Miles JM, Fisher AM, Jones R, Jackson N, Russell-Jones DL. Effect 24. Garg SK, Peters AL, Buse JB, Danne T. Strat- JN; American Diabetes Association. Hyperglyce- of subcutaneous insulin detemir on glucose flux egy for mitigating DKA risk in patients with type 1 mic crises in patients with diabetes mellitus. and lipolysis during hyperglycaemia in people diabetes on adjunctive treatment with SGLT Diabetes Care 2001;24:154–161 with type 1 diabetes. Diabetes Obes Metab 2015; inhibitors: a STICH protocol. Diabetes Technol 14. Munro JF, Campbell IW, McCuish AC, Duncan 17:459–467 Ther 2018;20:571–575 LJ. Euglycaemic diabetic ketoacidosis. BMJ 1973;2: 19. Burge MR, Hardy KJ, Schade DS. Short-term 25. Danne T, Garg S, Peters AL, et al. Interna- 578–580 fasting is a mechanism for the development of tional consensus on risk management of diabetic 15. Shojaee-Moradie F, Jackson NC, Jones RH, euglycemic ketoacidosis during periods of insulin ketoacidosis in patients with type 1 diabetes Mallet AI, Hovorka R, Umpleby AM. Quantitative deficiency. J Clin Endocrinol Metab 1993;76: treated with sodium–glucose cotransporter (SGLT) measurement of 3-O-methyl-D-glucose by gas 1192–1198 inhibitors. Diabetes Care 2019;42:1147–1154