Case of sodium–glucose cotransporter-2 inhibitor-associated euglycaemic diabetic ketoacidosis

Abstract

Following non-elective orthopaedic surgery, a 61-year-old man with poorly controlled type 2 diabetes mellitus on empagliflozin developed high anion gap metabolic acidosis in the high-dependency unit. Metabolic acidosis persisted despite intravenous sodium bicarbonate, contributing to tachycardia and a run of non-sustained ventricular tachycardia. He was euglycaemic throughout hospital admission. Investigations revealed elevated urine and capillary ketones, and a diagnosis of sodium–glucose cotransporter-2 inhibitor-associated euglycaemic diabetic ketoacidosis was made. He was treated with an intravenous sliding scale insulin infusion and concurrent dextrose 5% with potassium chloride. Within 24 hours of treatment, his arterial pH, anion gap and serum bicarbonate levels normalised. After a further 12 hours, the intravenous insulin infusion was converted to a basal/bolus regimen of subcutaneous insulin, and he was transferred to the general ward. He was discharged well on subcutaneous insulin 6 days postoperatively.

Background

With increasing sodium–glucose cotransporter-2 inhibitor (SGLT2i) use, greater awareness is needed among healthcare professionals to ensure timely recognition and appropriate diagnosis and treatment of SGLT2i-associated diabetic ketoacidosis (DKA). Healthcare professionals also need to ensure SGLT2is are adequately managed perioperatively to minimise the risk of SGLT2i-associated DKA.

Case presentation

A 61-year-old Chinese man was transferred from another hospital following a road traffic accident 2 days prior with fractures of the L1 vertebral body, left ulna shaft and left radial head. He had a 30-year history of type 2 diabetes mellitus (T2DM) (glycated haemoglobin, HbA1c, 10.0%; table 1), hyperlipidaemia and stable ischaemic heart disease post-coronary artery bypass graft 12 years ago.

Table 1.

Patient’s blood results

	Reference Range	Day 0 12:00	POD1 01:00	POD1 08:30	POD1 10:40	POD1 14:00	POD1 17:00	POD1 22:00	POD2 02:00	POD2 08:00	POD2 14:00	POD2 20:00	POD3 08:00
Urea	2.8–7.7 mmol/L	7.1	9.0	8.4									5.2
Creatinine	53–115 μmol/L	67	80	89									69
Albumin	37–51 g/L	40	41	40									38
Haematocrit	38%–52%	47.2	50.4	44.2			40
Lactic acid	0.5–2.2 mmol/L			0.9
pH (arterial)	7.35–7.45		7.12	7.17	7.32	7.35	7.34	7.35	7.34		7.39	7.43
PCO2 (arterial)	35–45 mm Hg		25.8	22.4	25.5	25.1	24.5	25.5	32		38.2	33
PO2 (arterial)	75–100 mm Hg		147.3	150	94	96.5	106.6	100.1	97		78.4	82.7
HCO3 (arterial)	22–26 mmol/L		8.3	8.3	12.8	13.4	13	13.7	17.4		22.4	21.3
Base excess (arterial)	−2 to +2 mmol/L		−19.4	−20	−1.4	−10.4	−10.9	−10.3	−8		−2.2	−2.2
SaO2 (arterial)	>95%		98.2		96.8	97.2	97.7	97.4			95.5	96.5
Sodium (arterial)	135–150 mmol/L		137	137	143.8	145.4	141.4	138.9			133.3	133.7
Potassium (arterial)	3.5–5 mmol/L		5.28	5.0	4.09	4.09	4.11	3.5			3.96	4.17
Chloride (arterial)	98–106 mmol/L		110	110	110	113	113	115			105	104
Anion gap	3–11 mmol/L		18.7	18.7	21	19	15.4	10.2			5.9	8.4
Haemoglobin (ABG)	13.5–16.5 g/dL	15.3	14.3	13.7	13.6	13.3	12.1			13.0	13.3
HbA1c	<7%	10.0
Glucose, random	4–7.8 mmol/L			8.3		5.9	8.7
Urine ketones	0						2+
Capillary β-hydroxybutyrate	≤0.6 mmol/L						6			2.1	1.7		0.5

Bold values for those outwith reference range.

ABG, arterial blood gas; HbA1c, glycated haemoglobin; PCO₂, partial pressure of carbon dioxide; PO₂, partial pressure of oxygen; POD, postoperative day; SaO₂, oxygen saturation (arterial).

His preadmission medications were as follows:

1. Metformin XR 1000 mg at night.

2. Empagliflozin 10 mg at night (started 3 months ago).

3. Aspirin 100 mg at night (stopped 1 week prior to admission due to gastric upset).

4. Bisoprolol 2.5 mg at night (stopped 1 week prior to admission as supplies ran out).

5. Atorvastatin 20 mg at night.

6. Fish oil capsules 1 capsule at night.

7. Calcium supplement 1 tablet at night as needed.

On the morning of surgery (also day of transfer), day 0, he was well hydrated, both clinically and on blood tests (table 1). He received metformin, empagliflozin and aspirin that morning, before fasting in preparation for surgery. No presurgical endocrinologist input was sought. Variable dose subcutaneous insulin was initiated three times a day, and his blood glucose levels remained below 10 mmol/L throughout his hospital stay (table 1).

That evening, he underwent percutaneous pedicle screw fixation of the L1 fracture from T11 through L3 and open reduction and internal fixation of the left ulna shaft. Intravenous dextrose saline (D5%-NaCl 0.9%) was started preoperatively with four-hourly blood glucose monitoring, followed by Hartmann’s solution postoperatively. Clear oral feeds were started at 02:00 the next day (postoperative day 1, POD1), and full feeds were resumed on the evening of POD1.

At 01:00 on POD1, arterial blood gases (ABGs) showed high anion gap metabolic acidosis (HAGMA) (pH 7.12, anion gap 18.7 mmol/L, base excess −19.4 mmol/L; table 1). He was given 200 mL of sodium bicarbonate 8.4% over 1 hour, and his intravenous fluid regimen was changed to NaCl 0.9% 2 L/day and Hartmann’s solution 1 L/day. At 06:00, he was noted to be tachycardic, and at 08:30, he was still profoundly acidotic (table 1). His blood glucose was slightly elevated (8.3 mmol/L) and blood lactate was normal (table 1). About 500 mL sodium bicarbonate 8.4% was infused over 2 hours. A repeat ABG showed persistent HAGMA (table 1), and a further 200 mL of sodium bicarbonate 8.4% was infused over 2 hours. At 14:00, he had a run of non-sustained ventricular tachycardia, and ABG revealed a persistently high anion gap (pH 7.35, anion gap: 19 mmol/L; table 1). His blood glucose was 5.9 mmol/L, and his intravenous fluid regimen was changed to dextrose 5%. Empagliflozin was stopped, and the patient was referred to an endocrinologist. At 17:00, investigations revealed elevated urine ketones (2+), elevated capillary ketones (6 mmol/L), mildly elevated blood glucose (8.7 mmol/L) and a persistent HAGMA (table 1). A diagnosis of euglycaemic diabetic ketoacidosis (EuDKA) was made.

Differential diagnosis

Alcoholic ketoacidosis was excluded on the patient’s negative alcohol history.

Starvation ketosis usually results in a relatively mild degree of metabolic acidosis, hence is unlikely to be the cause.

Lactic acidosis was excluded as serum lactate was normal, and renal causes of ketoacidosis were excluded as kidney function was normal.

Treatment

Empagliflozin was stopped, and, according to international guidelines, intravenous insulin sliding scale infusion was started alongside concurrent Dextrose 5% infusion with potassium chloride.

Outcome and follow-up

Within 24 hours of intravenous insulin treatment, the patient’s arterial pH, anion gap and serum bicarbonate levels normalised (table 1). The intravenous insulin sliding scale infusion was converted to a basal/bolus subcutaneous insulin regimen, and the patient was subsequently discharged on POD6. During his outpatient endocrinology follow-up, antiglutamic acid decarboxylase antibody was negative, making a diagnosis of type 1 diabetes or latent autoimmune diabetes of adulthood unlikely, and the patient was referred back to his primary physician.

Discussion

DKA occurs most commonly in type 1 diabetes mellitus, and is characterised by the triad of hyperglycaemia, HAGMA (arterial pH <7.3, serum bicarbonate <18 mmol/L, anion gap >10 mmol/L) and ketosis (positive urine or serum ketones, or serum β-hydroxybutyrate >3.0 mmol/L), often accompanied by varying degrees of circulatory volume depletion.^1–5 DKA occurs when absolute or relative insulin deficiency and concomitant elevation of counter-regulatory hormones (glucagon, catecholamine, cortisol) result in excessive lipolysis with increased hepatic beta-oxidation of fatty acids to ketone bodies, leading to ketonaemia and metabolic acidosis.^{1 2 6} Symptoms include nausea, vomiting, abdominal pain, excessive thirst, breathing difficulties, confusion, and unusual fatigue or sleepiness.^{2 6}

A warning that SGLT2is may increase the risk of ketoacidosis was issued in May 2015 by the U.S. Food and Drug Administration,⁷ followed in 2016 by the European Medicines Agency, Health Canada, and Health Services Authority Singapore.^8–10 Relevant risk factors for SGLT2i-associated DKA in this case included a low reserve of insulin-secreting cells (long-standing poorly controlled T2DM), increased insulin requirements due to trauma and surgery, and a low-carbohydrate perioperative diet.1 6 8 9 11 As SGLT2is promote glycosuria and decrease urinary excretion of ketone bodies, SGLT2i-associated DKA can present with normal or lower-than-expected blood glucose concentrations (<14 mmol/L) without ketonuria, resulting in a delay in diagnosis and treatment.^{1 6}

The underlying mechanism for SGLT2i-associated ketoacidosis is not fully established,^{9 12–14} and several mechanisms may contribute:

1. Increased ketogenesis:

   1. Due to increased renal clearance of glucose, SGLT2is promote a shift from carbohydrate to lipid metabolism and a reduction in insulin secretion, resulting in increased lipolysis and ketogenesis.^{12 15 16}

   2. Inhibition of SGLT2 expressed on pancreatic α-cells increases glucagon secretion, resulting in hyperglucagonaemia.^{17 18} The resulting increase in glucagon:insulin ratio promotes ketone production via lipolysis and hepatic beta-oxidation of fatty acids.11–15 18 19

   3. SGLT2is promote a negative fluid and sodium balance. Hypovolaemia drives elevations in glucagon, cortisol and epinephrine, further increasing insulin resistance, lipolysis and ketogenesis.¹²

   4. In animal studies, glucagon promotes hepatic secretion of kiss-peptin-1, which suppresses glucose-stimulated insulin secretion, and could further promote ketogenesis by decreasing endogenous glucose-stimulated insulin secretion.²⁰

2. Decreased renal clearance and utilisation of ketone bodies:

   1. Inhibition of SGLT2-mediated Na⁺ reabsorption increases Na⁺ concentration in the renal tubular fluid, thereby increasing the electrochemical gradient driving carrier-mediated reabsorption of negatively charged ketone bodies.^{13 19}

   2. By increasing urinary glucose excretion, SGLT2is may mimic starvation conditions, resulting in increased renal reabsorption of ketones and simultaneous reduction in renal utilisation of ketoacids.^{1 21 22}

   3. Reduced sodium reabsorption in the proximal renal tubule results in a relative excess of ATP, shifting the kidneys away from ATP-generating metabolic pathways, such as ammoniogenesis and ketone body oxidation, and worsening metabolic acidosis.²³

In addition to SGLT2i cessation, SGLT2i-associated EuDKA is managed as for DKA, and centres around fluid replacement (large volumes of NaCl 0.9% may be required as SGLT2is compound hypovolaemia due to DKA²⁴), adequate insulin delivery (to suppress lipolysis and ketone production) and normalising pH, glucose, and ketone and electrolyte concentrations.2–4 24 25

Although rare in T2DM, the incidence of SGLT2i-associated EuDKA may be higher in the perioperative period.^{6 24 26} Emerging guidance recommend stopping SGLT2is at least 1–3 days preoperatively, only restarting postoperatively once the patient is eating and drinking normally (table 2).^{1 25 27} Patients should be monitored closely, as Lau et al reported three cases of EuDKA occurring on POD1 after elective coronary artery bypass grafting despite the discontinuation of empagliflozin 24–48 hours pre-operatively.6 24 25 27 28

Table 2.

Published recommendations for perioperative management of patients on SGLT2is

Year/date	6 June 2016	14 February 2018	January 2020
Association/Society	AACE-ACE1	ADS27	ADS, NZSSD, ANZCA, Diabetes Australia, ADEA25
When to consider DKA/ EuDKA	Symptoms suggestive of DKA (eg, abdominal pain, nausea, vomiting, fatigue, dyspnoea), regardless of blood glucose level—confirm diagnosis with appropriate workup	Symptoms of DKA, regardless of plasma glucose level Finger prick ketone (or blood β-hydroxybutyrate) levels >0.6 mmol/L in the perioperative period or >1.5 mmol/L at any other time Low pH on VBG or ABG and low bicarbonate with a high anion gap, indicating metabolic acidosis	One or more of: Symptoms of DKA, regardless of plasma glucose level Finger prick capillary blood ketone (or blood β-hydroxybutyrate) levels >1.0 mmol/L Low (negative) BE less than −5 mmol/L indicating metabolic acidosis on arterial or venous blood gases
Preoperative cessation of SGLT2i	Consider stopping SGLT2i at least 24 hours before planned surgery or invasive procedures	At least 3 days preoperatively (2 days prior to surgery and day of surgery)*	Procedures requiring one or more days in hospital, and/or requiring ‘bowel preparation’ including colonoscopy: at least 3 days preprocedure (2 days prior to surgery and day of surgery)* Day-stay procedures (including gastroscopy): SGLT2i can be stopped just for the day of procedure, but fasting before and after the procedure should be minimised
When to postpone non-urgent surgery	–	If SGLT2is have not been ceased preoperatively and either blood ketones >0.6 mmol/L, or HbA1c >9.0%, as these are indicators of insulin insufficiency and a higher risk of DKA	If patient is unwell In clinically well patients who have NOT ceased SGLT2is, blood ketones >1 mmol/L, and BE less than −5 mmol/L or if blood gas analysis not available
Preoperative or perioperative monitoring in patient on an SGLT2i prior to surgery	–	Routinely check both blood glucose and blood ketone levels if patient is unwell, fasting or has limited oral intake. If blood ketone level is >0.6 mmol/L in an unwell preoperative or perioperative patient, perform an urgent VBG to measure the pH	Measure preoperative blood glucose and blood ketone levels— proceed with procedure if patient is clinically well and blood ketone levels <1.0 mmol/L Consider blood glucose and blood ketone testing hourly during procedure and every two hours following procedure until eating and drinking normally
Restarting SGLT2i postoperatively	–	Only when patient is eating and drinking and close to discharge (usually 3–5 days postsurgery) Day surgery/procedures: only if on full oral intake†	Only when patient is eating and drinking normally or close to discharge Day surgery/procedures: only if on full oral intake†
Postoperative advice for patients on SGLT2i prior to or following surgery	–	Check blood glucose and blood ketone levels if patient is unwell in the week following surgery.	Provide patients with written advice to seek medical advice if unwell in the week following the procedure

*This may require an increase in other glucose-lowering drugs during this time.

†Consider delaying SGLT2i recommencement for a further 24 hours, but also consider potential for hyperglycaemia.

AACE, American Association of Clinical Endocrinologists; ABG, arterial blood gas; ACE, American College of Endocrinology; ADEA, Australian Diabetes Educators Association; ADS, Australian Diabetes Society; ANZCA, Australian and New Zealand College of Anaesthetists; BE, base excess; DKA, diabetic ketoacidosis; EuDKA, euglycaemic diabetic ketoacidosis; HbA1c, glycated haemoglobin; NZSSD, New Zealand Society for the Study of Diabetes; SGLT2i, sodium–glucose cotransporter-2 inhibitor; VBG, venous blood gas.

Learning points.

• Although rare, SGLT2is may cause severe DKA with lower-than-anticipated glucose levels.

• Direct measurement of blood ketones (β-hydroxybutyrate) and arterial pH is recommended to diagnose SGLT2i-associated DKA.

• SGLT2is should be stopped at least 1–3 days preoperatively and only started postoperatively when the patient is eating and drinking normally or close to discharge.

Ethics statements

Patient consent for publication

Obtained.