SGLT2 inhibitor ketoacidosis in the ICU

A 2026 case series published in Case Reports in Critical Care describes two patients with type 2 diabetes who developed ^[1] under very different circumstances: one after coronary artery bypass graft (CABG) surgery, and one after semaglutide was added to their SGLT2 inhibitor regimen. In both cases, plasma glucose was below 200 mg/dL, a presentation the literature calls euglycemic diabetic ketoacidosis (EDKA) that disrupts the usual ICU diagnostic reflex.

Why normal glucose creates a diagnostic trap

Classic diabetic ketoacidosis announces itself with hyperglycemia above 250 mg/dL. SGLT2 inhibitors break that rule. By driving urinary glucose excretion, they lower circulating blood sugar even as hepatic ketogenesis accelerates^[1]. The result is a patient with high anion-gap metabolic acidosis, ketonuria, and blood glucose that looks almost normal. In a busy ICU where post-surgical metabolic disturbance is common, that glucose number can suppress the clinical index of suspicion.

The 2022 narrative review by Branco et al., which surveyed 17 published cases of EDKA in perioperative cardiac surgical patients^[2], identified the same pattern: the condition is difficult to recognise precisely because it does not look like textbook DKA. Their core recommendation was early measurement of serum beta-hydroxybutyrate in any cardiac surgery patient on an SGLT2 inhibitor who develops unexplained acidosis, regardless of the glucose level.

The cardiac surgery trigger

Surgery stresses the body in ways that specifically favour ketoacidosis in patients taking SGLT2 inhibitors^[1]. Caloric restriction before and after the procedure, surgical inflammation, catecholamine release, and relative insulin deficiency all tip the balance toward fat catabolism. SGLT2 inhibitors amplify that shift by continuing to suppress insulin secretion and promote glucagon release even in the perioperative period.

A 2025 case report from Cureus^[3] described a 79-year-old man on empagliflozin who was not told to stop the drug before elective CABG. On postoperative day 5, he presented with pH 7.1, bicarbonate of 9 mEq/L, anion gap of 26.4 mEq/L, and near-normal blood glucose. The acidosis triggered acute coronary syndrome and cardiac arrest. Angiography found critical stenosis in the bypass graft, requiring stent placement.

That case underscores why multiple surgical societies now recommend stopping SGLT2 inhibitors at least three to four days before major elective surgery. The recommendation predates this case report and is not specific to cardiac procedures, but cardiac surgery carries the highest metabolic burden of any elective category.

The GLP-1 agonist co-prescription trigger

The second patient in the Garcia Perez et al. series^[1] developed EDKA after semaglutide was added to an existing SGLT2 inhibitor prescription. GLP-1 receptor agonists reduce caloric intake and alter glucagon secretion. When combined with an SGLT2 inhibitor, those effects can compound the substrate shift toward ketogenesis without necessarily producing hyperglycemia as a warning sign.

This drug-drug interaction pattern has been described separately. A 2025 case report in JCEM Case Reports^[4] documented euglycemic ketoacidosis following coadministration of an SGLT2 inhibitor and tirzepatide, a dual GLP-1 and GIP receptor agonist. The authors noted it was the first published case involving tirzepatide specifically, but the underlying mechanism mirrors semaglutide: GLP-1 receptor activation suppresses glucagon, increases insulin secretion, and reduces food intake, all of which interact with the ketogenic shift already promoted by the SGLT2 inhibitor.

Patients starting or escalating a GLP-1 agonist while on an SGLT2 inhibitor warrant a clinical conversation about reduced oral intake in the early titration period. That is when the ketogenic risk is highest^[4]. The combination is not contraindicated, but the ICU and outpatient teams caring for these patients need to treat unexplained metabolic acidosis with a beta-hydroxybutyrate measurement before reassigning it to another cause.

Management: what the literature reports

The management approach in the Garcia Perez et al. cases^[1] and across the broader literature follows the same structure: stop the SGLT2 inhibitor, administer intravenous insulin at a dose sufficient to suppress ketogenesis, and supplement glucose to prevent hypoglycemia while the insulin works. Fluid resuscitation addresses the dehydration that accelerates acidosis.

The unusual wrinkle is that glucose supplementation is required from the start, not as a rescue measure. Because blood glucose may already be low or borderline, the standard DKA insulin drip without glucose support risks hypoglycemia before ketosis resolves. The 2016 expert panel review by Goldenberg et al. flagged this as a consistent feature of SGLT2 inhibitor-associated DKA management that differs from classic hyperglycemic DKA protocols.

The perioperative discontinuation debate

A tension in current guidance deserves mention. The 2026 secondary analysis by Haziri et al. in the British Journal of Anaesthesia^[5] found that patients who stopped SGLT2 inhibitors before noncardiac surgery had substantially higher rates of cardiac complications within 90 days than those who continued: 1.7% vs. 11.5% for cessation three or more days before surgery. Each additional day of discontinuation was associated with a 58% increase in complications.

That finding does not override current guidance, which recommends stopping SGLT2 inhibitors before major surgery to reduce EDKA risk^[2]. It does suggest the risk calculus is more complicated than the ketoacidosis risk alone, and that the optimal pre-surgical discontinuation window is an active research question. The Haziri et al. authors called for randomised controlled trials before revising protocols.

For readers tracking semaglutide or other GLP-1 class agents in combination with SGLT2 inhibitors: the metabolic interaction described here is a reason that prescribers typically review the full drug list before initiating either class in a patient already on the other. For a broader view of the GLP-1 class mechanisms, see the GLP-1 explainer on the semaglutide page.