Highlight
Key points
– A multinational electronic-health-record cohort (TriNetX) using 1:1 propensity-score matching found baseline SGLT2 inhibitor (SGLT2i) use was associated with a 22% lower hazard of sepsis-induced cardiomyopathy (SICM) within 30 days compared with DPP4 inhibitor (DPP4i) use in patients with type 2 diabetes (T2D) and infection (HR 0.78; 95% CI 0.71–0.86).
– SGLT2i exposure also associated with lower 1-year all-cause mortality (HR 0.58), hospitalization (HR 0.83), and major adverse cardiovascular events (MACEs) (HR 0.86).
– Findings persisted across subgroup and negative-control analyses, but limitations inherent to observational designs (residual confounding, exposure ascertainment, diagnostic coding for SICM) limit causal inference; randomized trials are needed before changing acute sepsis management.
Background: Clinical need and rationale
Sepsis-induced cardiomyopathy (SICM) is an acute, often reversible myocardial dysfunction that develops in a subset of patients with sepsis and contributes to morbidity and mortality in the intensive care unit. SICM typically manifests as reduced ventricular contractility, ventricular dilation, or both, often identified by echocardiography or biomarkers, and has no well-established prophylactic therapy. Patients with type 2 diabetes (T2D) are at higher risk of infection and sepsis and frequently receive glucose-lowering therapies with cardiovascular effects. Sodium-glucose cotransporter-2 inhibitors (SGLT2is) have demonstrated cardiac and renal benefits across trials in chronic heart failure and chronic kidney disease and have pleiotropic effects including natriuresis, improved myocardial energetics, and modulation of inflammation. Whether baseline use of SGLT2is modifies the risk of SICM during an episode of infection or sepsis has direct clinical relevance because SICM is a potentially preventable contributor to sepsis-associated cardiovascular collapse.
Study design
This was a retrospective, propensity-score–matched cohort study using the TriNetX global health research network. The cohort comprised adult patients with T2D who had an infection and who had been prescribed either an SGLT2i or a DPP4i within the three months before the infectious episode. DPP4 inhibitors were chosen as an active comparator because they are commonly used oral glucose-lowering agents without robust evidence for cardioprotective effects similar to SGLT2is, helping to reduce confounding by indication relative to nonuser comparisons.
Key design elements:
– Data source: TriNetX electronic health records from multiple centers (global research network).
– Population: Adults with T2D and an index infection; exposure defined as prescription of SGLT2i or DPP4i within 3 months before infection.
– Matching: 1:1 propensity-score matching (PSM) to balance demographics, comorbidities, and baseline medications between groups.
– Primary outcome: ICD-code–based diagnosis of sepsis-induced cardiomyopathy within 30 days of the index infection.
– Secondary outcomes: 1-year all-cause mortality, all-cause hospitalization, and major adverse cardiovascular events (MACEs).
– Sensitivity analyses: Subgroup analyses and negative-control outcomes to probe robustness to residual confounding.
Key findings
Population and matching
– After PSM, the analytic sample included 73,069 patients in each exposure group (SGLT2i vs DPP4i). Baseline covariates (age, sex, comorbidities, co-medications) were reported as balanced after matching.
Primary outcome: sepsis-induced cardiomyopathy
– SGLT2i exposure was associated with a lower hazard of SICM within 30 days compared with DPP4i (HR 0.78; 95% CI 0.71–0.86; P < .001). This corresponds to a relative risk reduction of about 22% in the matched cohort.
Secondary outcomes: mortality, hospitalization, MACEs
– One-year all-cause mortality was substantially lower in the SGLT2i group (HR 0.58; 95% CI 0.55–0.62; P < .001), a large effect size that is consistent with meaningful clinical benefit but also raises the question of residual or unmeasured confounding.
– Hazard for all-cause hospitalization was lower with SGLT2is (HR 0.83; 95% CI 0.79–0.87; P < .001).
– MACEs at one year were reduced (HR 0.86; 95% CI 0.80–0.93; P < .001).
Robustness checks
– The authors report that subgroup analyses did not materially change the association, and negative-control outcomes (pre-specified outcomes not biologically plausibly affected by SGLT2is) did not show the same pattern of association, lending some support against broad residual confounding.
Safety and adverse events
– The report focuses on effectiveness outcomes; safety endpoints relevant to acute infection and sepsis—particularly risk of euglycemic diabetic ketoacidosis (euDKA) with SGLT2is—were not the primary focus of the analysis. The observational dataset may incompletely capture DKA events or therapy discontinuation during acute illness.
Interpretation and mechanistic plausibility
Biological plausibility
– Several mechanisms could plausibly link chronic SGLT2i exposure to lower risk or severity of SICM: improved myocardial energetics (shift toward ketone utilization), reductions in cardiac preload and afterload through natriuresis and osmotic diuresis, attenuation of inflammation and oxidative stress, inhibition of the NLRP3 inflammasome in experimental models, and direct myocardial effects improving mitochondrial function. These mechanisms have support from preclinical and clinical studies in heart failure but remain incompletely characterized in the setting of sepsis.
Comparator choice
– DPP4 inhibitors were an appropriate active comparator for reducing confounding by indication for glucose-lowering therapy. DPP4is are metabolically neutral with respect to heart failure benefits in randomized trials, though some data have raised safety signals for heart failure with individual agents in the past. Using DPP4is rather than nonuse reduces bias due to differential access to care and prescribing propensity.
Plausibility versus causality
– The magnitude and consistency of associations strengthen plausibility but do not establish causality. The large reductions in 1-year mortality, in particular, may reflect broader differences in unmeasured confounders (socioeconomic factors, frailty, unmeasured disease severity, or differential likelihood of drug discontinuation before severe illness).
Strengths
– Very large sample size with robust 1:1 propensity-score matching on many baseline variables.
– Use of an active comparator (DPP4i) reduces confounding by treatment indication compared with nonuser comparisons.
– Assessment of both short-term SICM and longer-term clinical outcomes (mortality, hospitalization, MACEs).
– Sensitivity analyses, subgroup checks, and negative-control outcomes support the internal consistency of findings.
Limitations
– Observational design: residual confounding and confounding by indication cannot be excluded despite PSM. The large effect sizes for mortality suggest the possibility of unmeasured bias.
– Exposure ascertainment: prescription within 3 months before infection does not guarantee adherence or continued use during the acute sepsis episode; it also leaves open the possibility that drug discontinuation occurred at illness onset.
– Outcome ascertainment: SICM was identified from diagnostic coding and may be subject to misclassification. SICM diagnosis is clinically heterogeneous and often requires echocardiography and adjudication, which is difficult in administrative datasets.
– Safety in acute illness: existing clinical guidance generally recommends holding SGLT2is during serious acute illness because of the risk of euDKA and volume depletion; the study does not change this safety margin without randomized safety data.
– Generalizability: TriNetX captures diverse centers, but findings may not generalize to regions or population subsets not well represented in the dataset.
Clinical implications and practical guidance
– These associative findings are hypothesis-generating rather than practice-changing. They suggest that chronic SGLT2i therapy may be linked to lower risk of SICM and improved long-term outcomes among patients with T2D who develop infection, but causality is unproven.
– Current acute-care recommendations to hold SGLT2is during severe illness, dehydration, or perioperative periods remain prudent given the established risk of euDKA and hemodynamic instability in critically ill patients.
– For ambulatory patients with T2D at increased risk of cardiovascular disease, SGLT2is remain guideline-supported for heart failure and renal protection. These long-term benefits may contribute indirectly to improved resilience in the setting of critical illness, but randomized data in sepsis are lacking.
Research agenda and next steps
– A randomized controlled trial (RCT) would be the definitive test of whether SGLT2is prevent SICM or improve outcomes in patients with T2D at risk for sepsis. Practical and ethical considerations make such a trial challenging: it would likely need to be pragmatic, enroll ambulatory patients with T2D at high risk for infection or with recent infection, randomize to SGLT2i versus placebo/active comparator, and prespecify rules for treatment interruption during acute illness.
– Mechanistic studies: prospective cohorts with systematic echocardiography, biomarkers (troponin, NT-proBNP, inflammatory markers), and metabolic phenotyping during sepsis could clarify how SGLT2is modify myocardial responses to infection.
– Safety-focused studies: explicit evaluation of euDKA and acute kidney injury in the context of infection and sepsis among SGLT2i users is essential before considering starting these drugs in patients at imminent risk of severe infection.
Conclusions
This large propensity-matched cohort study in TriNetX reports that baseline SGLT2i use was associated with a lower risk of sepsis-induced cardiomyopathy and substantially lower 1-year mortality, hospitalizations, and MACEs compared with DPP4i therapy in patients with type 2 diabetes who developed infection. The findings are biologically plausible and consistent with the cardioprotective profile of SGLT2is, but important limitations of observational data—especially the potential for residual confounding and diagnostic misclassification—preclude causal claims. Clinicians should not change practice regarding withholding SGLT2is during severe acute illness based on these data alone. The study provides a strong rationale for prospective randomized and mechanistic studies to evaluate whether SGLT2is can prevent SICM and improve outcomes in sepsis.
Funding and clinicaltrials.gov
The published article (Wu et al., Crit Care. 2025 Oct 27) is cited for the primary analysis; no trial registration applies to this observational study. Funding sources and sponsor declarations should be checked in the original paper (Wu JY et al., Crit Care. 2025). Clinicians and investigators wishing to plan RCTs should register future trials on clinicaltrials.gov prior to enrollment.
Selected references
– Wu JY, Tseng KJ, Kao CL, Hung KC, Yu T, Lin YM. Association between SGLT2 inhibitor use and risk of sepsis-induced cardiomyopathy in patients with type 2 diabetes: a propensity-matched cohort study. Crit Care. 2025 Oct 27;29(1):452. doi: 10.1186/s13054-025-05685-0. PMID: 41146189; PMCID: PMC12557926.
– Zinman B, Wanner C, Lachin JM, et al.; EMPA-REG OUTCOME Investigators. Empagliflozin, cardiovascular outcomes, and mortality in type 2 diabetes. N Engl J Med. 2015;373(22):2117–2128.
– McMurray JJV, Solomon SD, Inzucchi SE, et al.; DAPA-HF Trial Committees and Investigators. Dapagliflozin in patients with heart failure and reduced ejection fraction. N Engl J Med. 2019;381(21):1995–2008.
– Packer M, Anker SD, Butler J, et al.; EMPEROR-Reduced Trial Investigators. Cardiovascular and renal outcomes with empagliflozin in heart failure. N Engl J Med. 2020;383(15):1413–1424.
– Jia G, Hill MA, Sowers JR. Role of insulin resistance and hyperinsulinemia in cardiovascular disease. Circ Res. 2018;122(4):624–641.
– Dapa-CKD and CREDENCE trial publications for renal benefits of SGLT2 inhibitors (see relevant NEJM and Lancet publications).
(Note: For detailed trial citations and the original article methods, readers should consult the primary publications.)
