SGLT2 inhibitor ketoacidosis: the risk persists and concentrates in identifiable patients

Design

Nationwide register cohort with nested case-control analysis, three Scandinavian countries

Population

282,282 adults with type 2 diabetes; 322,597 SGLT2 inhibitor treatment episodes, 2013–2022

Primary outcome

Ketoacidosis incidence, risk factors and precipitating events

Effect

2.43 events per 1000 person-years; HbA1c ≥83 vs ≤52 mmol/mol OR 15.37 (95% CI 11.50–20.53); infection OR 7.60 (6.62–8.71)

Nationwide registers from Sweden, Denmark and Norway captured 322,597 SGLT2 inhibitor treatment episodes in 282,282 adults with type 2 diabetes between 2013 and 2022. Over a median 1.3 years there were 1452 ketoacidosis events, an incidence of 2.43 per 1000 person-years. A nested case-control analysis looked for risk factors and precipitants.

Risk was highest shortly after starting but did not disappear: it persisted throughout follow-up. The strongest associations were with HbA1c of 83 mmol/mol or more, malnutrition, previous ketoacidosis, BMI under 20 and recent hypoglycaemia. Infection was present in about a third of cases against 6% of controls. Alcohol intoxication, acute renal events, acute abdomen, stroke and major surgery carried the strongest associations of all.

The practical point is that sick-day counselling is not a one-off at initiation. The patient profile that should prompt caution — thin, poorly controlled, possibly insulin-deficient — is common in Indian clinics, where lean type 2 diabetes with high HbA1c is frequent. Reassess that profile at each review, and make the pause-during-acute-illness instruction explicit.

• Before starting, and at each review, check HbA1c, BMI, nutritional state and any history of ketoacidosis
• Consider whether a lean patient with very high HbA1c is actually insulin-deficient before choosing an SGLT2 inhibitor
• Tell patients to pause the drug during infection, vomiting, poor intake, heavy drinking or before surgery
• Check ketones, not just glucose, in an unwell patient on an SGLT2 inhibitor, since glucose may be near normal
• After an episode, three-quarters of patients in this cohort did not restart the drug and most went on insulin

Why it matters

Ketoacidosis on SGLT2 inhibitors is not only an early-treatment problem, so counselling at the first prescription is not enough.

Don't overread it

This was observational — it identifies who is at risk, not that stopping the drug in those patients prevents harm.

The statistics, in plain English

An incidence of 2.43 per 1000 person-years means roughly one event for every 400 patients treated for a year — uncommon for any individual. The odds ratios for risk factors are very large (around 10 to 15), but they come from a case-control design, which overstates some associations when minor illnesses are under-recorded in controls, as the authors note. Exploratory analyses suggested much of the risk profile applies to type 2 diabetes generally, not only to SGLT2 inhibitor users.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.