SGLT2 Inhibitor Ketoacidosis Risk Versus Perioperative Mortality and Renal Benefits

Background: SGLT2 inhibitors (SGLT2i) are guideline-directed therapies for T2D, HF, and CKD. Concerns about ketoacidosis (KA) — particularly euglycemic KA (EKA) — have led to routine preoperative discontinuation. Whether KA risk is outweighed by perioperative mortality and renal benefits remains unquantified. Major HF/CKD trials enrolled substantial non-diabetic populations (DAPA-HF 58%, EMPEROR-Reduced ~50%, DELIVER ~50%, EMPA-KIDNEY 54%, DAPA-CKD 33%), and KA risk differs fundamentally by diabetes status, necessitating stratified analysis. Methods: We pooled 22 SGLT2i RCTs (N=102,077; 55,024 SGLT2i vs. 47,053 placebo; ~72,000 diabetic, ~30,000 non-diabetic). The primary outcome was pooled OR for KA in diabetic participants, with prespecified secondary analysis in non-diabetic participants. Subgroup analyses by indication (ASCVD, CKD, HF, glycemic control) were performed. Perioperative mortality and AKI data from two large propensity-matched registry studies were integrated to calculate NNH and NNT, enabling net clinical benefit assessment. Sensitivity analyses modeled underreported EKA (2х / 3х correction factors). Results: Among diabetic participants, 93 vs. 22 KA events occurred (SGLT2i vs. placebo; OR 2.27; 95% CI 1.55-3.33; I 2 =0%), aligning with the Staplin et al. IPD meta-analysis (HR 2.29; 1.42-3.71). Subgroup ORs — ASCVD 2.43 (1.50-3.92), CKD 2.89 (1.38-6.05), HF 2.14 (0.62-7.38) showed no significant heterogeneity (p=0.82), consistent with a class effect. Among ~30,000 non-diabetic participants, only 1 KA event occurred, precluding meaningful risk estimation consistent with KDIGO 2024 guidance stating no excess KA risk without diabetes. Integration with perioperative registry data yielded: KA NNH ~775; 30-day mortality NNT ~72-100 (OR 0.61-0.70); AKI NNT ~36-50 (OR 0.69-0.71). For every 1 additional KA event, approximately 8-11 deaths are prevented and 15-22 AKI events avoided. The VA registry study (JAMA Surgery 2025) found SGLT2i users had increased postoperative EKA (OR 1.11; 1.05-1.17) but reduced mortality (OR 0.70; 0.55-0.88) and AKI (OR 0.69; 0.62-0.78). Up to 58% of SGLT2i-associated KA presents euglycemically, likely underestimated in RCT safety reporting. Sensitivity analyses doubling/tripling KA counts (NNH ~400/~270) preserved a strongly favorable net benefit. Conclusions: SGLT2i confer a rare, manageable KA risk increase in diabetic patients (OR ~2.3, NNH ~775) substantially outweighed by perioperative mortality and renal benefits. In non-diabetic patients, KA risk is negligible, making net benefit even more favorable. These findings support risk-stratified perioperative SGLT2i management rather than blanket discontinuation, with EKA prevention protocols — including point-of-care ketone testing and IV dextrose as recommended by the SPAQI 2026 consensus statement — for diabetic patients undergoing prolonged fasting.