Ketoacidosis with SGLT2 inhibitors in routine clinical practice of type 2 diabetes: Scandinavian cohort and nested case-control study.
BACKGROUND: SGLT2 inhibitors increase the risk of ketoacidosis, but data from routine clinical practice are scarce. We used nationwide registers with the aim of assessing the incidence, risk factors, and prognosis of ketoacidosis during SGLT2 inhibitor treatment in patients with type 2 diabetes. METHODS: In this cohort and nested case-control study we used data from three Scandinavian countries. In a cohort of SGLT2 inhibitor-treatment episodes among patients with type 2 diabetes aged at least 18 years in Sweden, Denmark, and Norway, we estimated ketoacidosis incidence. Using a nested case-control design (matched on age, sex, region of birth, calendar time, and time since treatment initiation), we assessed risk factors and precipitating or co-occurring events. Changes in diabetes medications were evaluated. FINDINGS: The study period was Jan 1, 2013, to Dec 31, 2021, in Denmark and Sweden, and Jan 1, 2013, to Dec 31, 2022, in Norway. We included 322 597 treatment episodes among 282 282 patients with type 2 diabetes (mean age 63 years, 116 521 [36·1%] of 322 597 women). During a median (IQR) follow-up of 1·3 (0·7-2·8) years, 1452 ketoacidosis events occurred (incidence 2·43 per 1000 person-years). Although highest shortly after initiation, risk persisted throughout follow-up. Strong risk factors included high HbA1c (≥83 vs ≤52 mmol/mol: odds ratio [OR] 15·37 [95% CI 11·50-20·53]), malnutrition (OR 10·54 [7·32-15·18]), previous ketoacidosis (OR 10·40 [7·09-15·24]), low BMI (<20 kg/m2vs 20 to <25 kg/m2: OR 9·98 [5·68-17·53]), and recent hypoglycaemia (OR 5·22 [2·60-10·49]). Infection was the most common precipitating or co-occurring event (454 [31·8%] of 1428 vs 862 [6·1%] of 14 233 for ketoacidosis cases vs controls; OR 7·60 [6·62-8·71]). The strongest associations were observed for alcohol intoxication, acute renal events, acute abdomen, stroke, and major surgery, although associations for some transient exposures, particularly milder conditions, might have been overestimated because of under-registration among controls. Exploratory analyses suggested that the observed associations were largely general to patients with type 2 diabetes rather than specific to SGLT2 inhibitor use. At 1 year after ketoacidosis, 211 (25·1%) of 842 remained on SGLT2 inhibitors and insulin use increased from 269 (31·9%) of 842 to 615 (73·0%) of 842. INTERPRETATION: Ketoacidosis risk with SGLT2 inhibitors varies greatly by patient characteristics and is not confined to early in treatment. Risk should be assessed throughout treatment, and patients should be instructed to pause treatment during acute illness and stress. FUNDING: Region Stockholm, Swedish Society of Medicine, Karolinska Institutet.