SGLT2 Inhibitor DKA Prevention: Sick-Day and Surgery Rules

Canagliflozin, dapagliflozin, empagliflozin, and ertugliflozin are sodium-glucose cotransporter-2 (SGLT2) inhibitors, an oral drug class approved for type 2 diabetes and, for some agents, for heart failure or chronic kidney disease independent of diabetes status. They work by blocking glucose reabsorption in the kidney, which lowers blood glucose through glycosuria. That same mechanism is the reason this class carries a distinct ketoacidosis risk that does not behave like typical DKA.

The core answer, and its boundary

Euglycemic diabetic ketoacidosis is a documented, FDA-recognized adverse effect of the SGLT2 inhibitor class, in which significant ketoacidosis (low pH, low bicarbonate, elevated ketones) develops while blood glucose remains below the level clinicians are trained to associate with DKA, sometimes below 250 mg/dL or even close to normal. This is established by FDA post-marketing surveillance and has been described in subsequent clinical reviews of the evidence as it accumulated through 2020. It is not established how often this happens in an unselected outpatient population, and precise incidence figures reported in secondary sources should be treated cautiously until checked against the primary study each number came from. What is not in dispute is the practical implication: a normal-looking glucose meter reading does not rule out DKA in a patient on this drug class who is vomiting, not eating, or acutely ill.

Why the SGLT2 mechanism produces ketosis without high glucose

SGLT2 inhibitors raise the kidney's glucose reabsorption threshold, producing continuous glycosuria. That glucose loss lowers plasma glucose, but it also reduces insulin secretion and raises glucagon, a hormonal shift that favors lipolysis and hepatic ketone production. In effect, the drug can push a patient toward a ketogenic metabolic state at the same time it is masking the hyperglycemic warning sign that would normally prompt a corrective insulin dose. A 2020 clinical review summarizing how the evidence on this phenomenon evolved across trials and post-marketing reports describes this combination of glycosuria-driven insulin suppression and glucagon elevation as the leading proposed mechanism (Musso et al., 2020). The mechanism is well described; the exact contribution of each pathway in an individual patient is not something a glucose meter or a single lab value can settle on its own.

The FDA's May 2015 Drug Safety Communication was issued after a review of post-marketing adverse event reports identified dozens of DKA cases in patients on SGLT2 inhibitors, a number of which occurred at glucose levels lower than clinicians would typically expect for DKA, prompting label changes across the class. This is the primary regulatory basis for the sick-day and perioperative precautions described below. Readers should treat the FDA communication itself, not any secondary paraphrase of it, as the authoritative source for exact case counts and dates.

Who carries the most risk

Risk is not evenly distributed across everyone taking these drugs. Factors that plausibly and, in several cases, clearly increase DKA risk on an SGLT2 inhibitor include:

• Low or absent endogenous insulin. Type 1 diabetes, latent autoimmune diabetes of adulthood, and long-standing type 2 diabetes with significant beta-cell failure remove the physiologic buffer that normally limits ketogenesis. All currently approved SGLT2 inhibitors are contraindicated for glycemic management in type 1 diabetes in the United States; this is a labeled contraindication, not a general caution.
• Very low carbohydrate or ketogenic eating patterns. These diets independently produce mild ketosis. Case reports describe this combining with SGLT2 inhibition to produce frank DKA, though the frequency of this specific combination has not been quantified in a way this draft can cite precisely.
• Heavy alcohol use. Ethanol suppresses gluconeogenesis and independently promotes ketone production, a plausible and pharmacologically coherent interaction with SGLT2-driven ketogenesis, though clinical trial data isolating this specific combination is limited.
• Insulin dose reduction at initiation. Reducing insulin because glucose readings look better after starting an SGLT2 inhibitor removes a safety margin against ketosis. Diabetes guideline bodies generally advise against substantial insulin cuts at SGLT2 inhibitor initiation, particularly in patients with limited beta-cell reserve; the exact wording of any specific guideline should be checked in the current edition of the ADA Standards of Care, published annually in Diabetes Care (ADA Standards of Care, current edition).
• Acute illness, dehydration, or reduced oral intake. Vomiting, diarrhea, fever, and prolonged fasting all raise counter-regulatory hormones and reduce fluid and carbohydrate intake, a combination that can tip a patient into ketosis who would otherwise tolerate the illness without incident.
• Surgery. Discussed separately below, because it is the single highest-risk window for this population.

Sick-day rules: the most actionable prevention tool

Structured sick-day instructions are the intervention with the clearest, most consistent support across diabetes guideline bodies for reducing SGLT2 inhibitor-associated DKA. The general framework instructs patients to hold the SGLT2 inhibitor and contact their care team promptly under any of these conditions:

1. Vomiting, even a single episode
2. Diarrhea lasting several hours or more
3. A prolonged fast, including pre-procedure fasting
4. Fever or an acute febrile illness
5. Any planned surgery or acute injury

A blood ketone (beta-hydroxybutyrate) reading is more reliable than a urine ketone strip for this purpose, because SGLT2-associated ketosis can be predominantly beta-hydroxybutyrate rather than acetoacetate, which urine strips detect. As a general orientation, rising ketones during illness warrant a call to the prescribing clinician, and a clearly elevated reading with fast or labored breathing, confusion, or persistent vomiting is an emergency department situation regardless of the glucose number on the meter. Exact numeric ketone thresholds vary slightly between clinical protocols; a patient's own care team should confirm the specific cutoff used in their sick-day plan.

A decision framework for holding and monitoring an SGLT2 inhibitor

This framework organizes the reader-facing decisions this topic actually requires: whether to hold the drug, what to check, and when the situation has moved past home management. It is a synthesis tool built from the risk factors and thresholds described above, not a substitute for an individualized plan from the prescribing clinician.

The recurring theme in this table is that the trigger for action is illness, intake, and ketone status, not the glucose number. That is the specific point a generic diabetes safety page usually misses: a normal glucose reading on this drug class is reassuring for hyperglycemia, not for ketoacidosis.

Surgery: the highest-risk window, and where the evidence gets less precise

Surgery combines several DKA risk factors at once: prolonged fasting, catecholamine release that suppresses insulin and raises glucagon, and often a temporary change in the patient's usual diabetes regimen. Holding the SGLT2 inhibitor before elective surgery requiring anesthesia or sedation is standard practice recommended by diabetes and anesthesia guideline groups, generally on the order of several days beforehand, though the exact number of days named in different guidelines is not fully consistent and should be confirmed against the current ADA Standards of Care or the relevant anesthesiology society guidance for the specific procedure. Case reports have described postoperative DKA presenting many hours after surgery in patients whose SGLT2 inhibitor was held only briefly beforehand, which is why many centers now favor a longer hold, particularly for major abdominal or thoracic surgery. This draft cannot cite an exact case count or hold-duration recommendation from a specific case series without a verified source, and any such number appearing elsewhere should be checked against the original publication before being treated as authoritative.

A practical point that is easy to overlook: many patients take SGLT2 inhibitors in a combination tablet with metformin. Holding the combination tablet holds both drugs, and the surgical or anesthesia team needs to know that. Patients should not assume the surgical team's medication reconciliation will automatically catch this; disclosing the specific drug by name before any procedure is the patient's most reliable safeguard.

After surgery, restarting is generally deferred until the patient is eating normally, is adequately hydrated, and renal function has returned to baseline, which for most elective procedures means at least several days post-operatively. The exact restart timing should come from the surgical or endocrine team managing the recovery, not from a generic rule.

Insulin dosing when combined with an SGLT2 inhibitor

Adding an SGLT2 inhibitor to an insulin regimen creates two opposing risks that both need active management: cutting insulin too aggressively increases ketosis risk, while leaving insulin unchanged after the SGLT2 inhibitor's glucose-lowering effect takes hold increases hypoglycemia risk. Clinical practice generally favors a modest, cautious insulin adjustment rather than a large upfront cut, with close monitoring in the weeks after starting the SGLT2 inhibitor and further titration based on glucose data. This is a matter of individualized dosing that a patient's own prescriber needs to manage; this article does not provide a dosing formula, because the correct adjustment depends on the individual's insulin regimen, renal function, and hypoglycemia history.

Standard hypoglycemia management still applies for glucose below the individually agreed threshold: fast-acting carbohydrate, recheck in about 15 minutes, and repeat if still low. Severe hypoglycemia, meaning loss of consciousness or inability to self-treat, requires glucagon administered by a trained caregiver and emergency evaluation. Anyone caring for an insulin-using patient on this combination should be trained in glucagon use before it is needed.

Where this overlaps with other diabetes drug risks, briefly

DKA on an SGLT2 inhibitor sits in a differential diagnosis alongside a few other acid-base and hypoglycemia concerns that come up in the same patients, worth naming briefly for context rather than as a full separate topic:

• Metformin-associated lactic acidosis is a distinct and much better-characterized risk, generally considered rare in patients with normal renal function and no acute illness or hypoxia. Widely cited estimates put the background rate in the low single digits to roughly 10 cases per 100,000 patient-years in appropriately selected patients, though exact figures vary by study and should be checked against the primary literature before being quoted precisely. The FDA revised metformin's renal warnings in 2016 to use eGFR rather than serum creatinine thresholds.
• Sulfonylureas, particularly glyburide, carry a meaningfully higher hypoglycemia risk in older adults because of their long-acting metabolites and glucose-independent insulin release; American Geriatrics Society Beers Criteria guidance has flagged glyburide as a drug to avoid in older adults, though the exact current edition and wording should be confirmed directly with the AGS publication rather than a secondary summary.
• Pioglitazone carries cardiovascular benefit data from randomized trial evidence and a separate, smaller bladder cancer signal identified in post-marketing cohort data that led to an FDA label update around 2011. The magnitude of that bladder cancer association, and the exact hazard ratios sometimes quoted for it, need to be verified against the original cohort study rather than repeated from memory, because different cohorts have reported somewhat different effect sizes.

None of these three drugs cause the euglycemic DKA pattern described in this article; they are included only because clinicians evaluating an unexplained acidosis or hypoglycemic event in a diabetic patient need to keep the full medication list in view.

What is established, what is plausible, and what is not established

Established: SGLT2 inhibitors can cause DKA at glucose levels well below the threshold clinicians are trained to associate with ketoacidosis. The FDA issued a class-wide safety communication on this in 2015. The drug class is contraindicated for glycemic management in type 1 diabetes in the United States. Blood ketone testing, not glucose alone, is the appropriate monitoring tool during illness in a patient on this class.

Plausible but not precisely quantified in this draft: The exact incidence of euglycemic DKA in unselected outpatients, the exact optimal number of days to hold an SGLT2 inhibitor before every category of surgery, and the exact magnitude of risk contributed by alcohol or low-carbohydrate diets in combination with these drugs. Guideline bodies offer general ranges and directional advice; treat any single precise number attached to these questions as something to verify against the primary guideline or study before relying on it clinically.

Not established: That any home monitoring strategy eliminates the need for emergency evaluation once ketones are clearly elevated or the patient is symptomatic. Ketone monitoring reduces the chance of missing early ketosis; it does not substitute for emergency care once acidosis is suspected.

When to seek emergency care

Go to an emergency department, rather than attempting home management, if a person on an SGLT2 inhibitor has persistent vomiting, cannot keep fluids down, has rapid or labored breathing, is confused, or has a clearly elevated blood ketone reading, regardless of what the glucose meter shows. Do not wait for glucose to rise before seeking care; that is the specific trap this drug class creates.