SGLT2 ketoacidosis persists beyond the first months of treatment

Design

Nationwide register cohort with nested case-control analysis

Population

282,282 adults with type 2 diabetes starting SGLT2 inhibitors in Sweden, Denmark and Norway

Primary outcome

Ketoacidosis incidence and risk factors

Effect

2.43 per 1000 person-years; HbA1c ≥83 mmol/mol OR 15.37 (11.50–20.53)

This register study from Sweden, Denmark and Norway followed 322,597 SGLT2 inhibitor treatment episodes in 282,282 adults with type 2 diabetes (2013 to 2022), with a nested case-control analysis of risk factors. It was published in September.

There were 1452 ketoacidosis events, 2.43 per 1000 person-years. Risk peaked after initiation but persisted. The strongest associations were HbA1c ≥83 vs ≤52 mmol/mol (OR 15.37, 95% CI 11.50 to 20.53), malnutrition (OR 10.54), previous ketoacidosis (OR 10.40), BMI under 20 (OR 9.98) and recent hypoglycaemia (OR 5.22). Infection preceded 31.8% of cases against 6.1% of controls (OR 7.60). Alcohol intoxication, acute renal events, acute abdomen, stroke and major surgery showed the strongest transient associations.

The practical shift is from a one-off warning at the start to a standing sick-day plan, and to caution in thin, poorly controlled or insulin-deficient patients. A year after ketoacidosis, only 25% remained on the drug and insulin use had risen from 32% to 73%, which suggests many events unmasked insulin deficiency. In India, where lean type 2 diabetes and delayed insulin are common, the low-BMI signal deserves attention.

• Assess ketoacidosis risk at every review, not only at initiation.
• Be cautious in patients with BMI under 20, malnutrition or very high HbA1c; consider whether insulin is needed first.
• Give written sick-day rules: pause during infection, vomiting, poor intake, heavy alcohol use.
• After a ketoacidosis episode, assess for insulin deficiency before restarting.

Why it matters

Ketoacidosis risk does not fade after the first months, so neither can the counselling.

Don't overread it

These are associations from registers; they identify risk, they do not show that stopping the drug prevents events.

The statistics, in plain English

The odds ratios are large but come from an observational case-control comparison: they show which patients are at risk, not that the factor causes ketoacidosis. The authors note that milder transient conditions may be under-recorded in controls, which would inflate those ratios. Their exploratory analysis suggests the same factors raise ketoacidosis risk in type 2 diabetes generally, not only on SGLT2 inhibitors.

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.