Disproportionate reporting of euglycaemic and diabetic ketoacidosis with SGLT-2 inhibitors across expanding cardiorenal indications: a cross-database study of FAERS and JADER.

SGLT-2 inhibitors are now used beyond type 2 diabetes (T2DM) in heart failure (HF) and chronic kidney disease (CKD), where glucose is monitored less routinely. Whether their established ketoacidosis signal-particularly euglycaemic diabetic ketoacidosis (euDKA)-is maintained as the indications expand, and whether it reproduces across independent reporting systems, is untested. We analysed the US FAERS (2020Q1-2026Q1) and Japanese JADER using one pipeline. Signals required consensus across four methods (ROR, PRR, IC, EBGM/EB05). Analyses were run overall, within report-level indication strata [T2DM, HF, CKD, off-label type 1 diabetes (T1DM)] and against an active comparator (DPP-4 inhibitors), with two pre-specified negative controls, time-to-onset modelling and sensitivity analyses. The DKA signal met four-method consensus within every indication stratum, including HF and CKD, and was highest in off-label T1DM; it was therefore maintained, not diluted, as indications expanded. Overall RORs were 67.4 (95% CI 65.7-69.2) in FAERS and 112.3 (104.8-120.3) in JADER. SGLT-2 inhibitors accounted for 85.8% (FAERS) and 91.7% (JADER) of all euDKA reports, though euDKA was only 4.1% and 7.8% of SGLT-2 inhibitor reports. Both negative controls were null in both databases. Median time-to-onset was 60 days (Weibull β=0.48, 95% CI 0.46-0.50). Across two independent reporting systems, the SGLT-2 inhibitor ketoacidosis signal was maintained across the expanding cardiorenal indications, with euDKA concentrated within this class and early onset. Ketone-based assessment is warranted when ketoacidosis is suspected, particularly soon after initiation. As a disproportionality analysis, these are hypothesis-generating signals that cannot establish incidence, relative risk, or causality.