Sodium–glucose cotransporter-2 inhibitors and the risk for diabetic ketoacidosis: a multicenter cohort study

Background — Sodium–glucose cotransporter-2 (SGLT-2) inhibitors could increase the risk for diabetic ketoacidosis (DKA).

Objective — To assess whether SGLT-2 inhibitors, compared with dipeptidyl peptidase-4 (DPP-4) inhibitors, are associated with an increased risk for DKA in patients with type 2 diabetes.

Design — Population-based cohort study; prevalent new-user design between 2013 and 2018. (ClinicalTrials.gov: NCT04017221)

Setting — Electronic healthcare databases from 7 Canadian provinces and the United Kingdom.

Patients — 208 757 new users of SGLT-2 inhibitors were matched by using time-conditional propensity scores to 208 757 recipients of DPP-4 inhibitors.

Measurements — Cox proportional hazards models estimated site-specific hazard ratios (HRs) with 95% CIs of DKA comparing receipt of SGLT-2 inhibitors with receipt of DPP-4 inhibitors, which were pooled by using random-effects models. Secondary analyses were stratified by molecule, age, sex, and prior receipt of insulin.

Results — Overall, 521 patients were diagnosed with DKA during 370 454 person-years of follow-up (incidence rate per 1000 person-years, 1.40 [95% CI, 1.29 to 1.53]). Compared with DPP-4 inhibitors, SGLT-2 inhibitors were associated with an increased risk for DKA (incidence rate, 2.03 [CI, 1.83 to 2.25] versus 0.75 [CI, 0.63 to 0.89], respectively; HR, 2.85 [CI, 1.99 to 4.08]). Molecule-specific HRs were 1.86 (CI, 1.11 to 3.10) for dapagliflozin, 2.52 (CI, 1.23 to 5.14) for empagliflozin, and 3.58 (CI, 2.13 to 6.03) for canagliflozin. Age and sex did not modify the association; prior receipt of insulin appeared to decrease the risk.

Limitations — There was unmeasured confounding and no laboratory data were available for the majority of patients, and molecule-specific analyses were conducted at a limited number of sites.

Conclusion — SGLT-2 inhibitors were associated with an almost 3-fold increased risk for DKA, with molecule-specific analyses suggesting a class effect.