Highlights
– A retrospective cohort analysis of 2,694 T2DM patients from the MIMIC‑IV database presented at AHA 2025 reported substantially lower short‑ and long‑term mortality among those with metformin exposure around cardiac surgery (30‑day HR 0.19; 95% CI 0.07–0.54).
– Absolute 30‑day mortality difference was 4.1 percentage points (0.4% vs 4.5%); protective associations persisted at 90 and 360 days and across sensitivity and propensity‑matched analyses.
– Findings are hypothesis‑generating: biologic plausibility exists, but residual confounding, exposure misclassification, and perioperative management variables limit causal inference; randomized trials are needed.
Background and clinical context
Type 2 diabetes mellitus (T2DM) increases perioperative risk after cardiac surgery through multiple mechanisms: greater burden of coronary and microvascular disease, impaired myocardial tolerance to ischemia, altered wound healing, and higher rates of postoperative complications. Identifying modifiable treatments that reduce perioperative mortality in this high‑risk group is a major clinical priority.
Metformin is the most widely prescribed first‑line glucose‑lowering agent worldwide. Beyond glucose control, observational and mechanistic studies suggest cardiovascular benefits that may extend to reduced infarct size, improved endothelial function, and attenuated inflammatory responses—mechanisms relevant to perioperative myocardial injury. However, perioperative management of metformin is controversial because of concerns about metformin‑associated lactic acidosis in settings of acute kidney injury or hemodynamic instability. Contemporary practice varies widely.
Study design
The study reported at AHA 2025 is a retrospective cohort analysis using the Medical Information Mart for Intensive Care IV (MIMIC‑IV) electronic health record (EHR) database. Inclusion criteria were adult ICU patients with T2DM who underwent cardiac surgery. Patients were classified into a metformin exposure group and a non‑exposure group based on documented medication records in proximity to the index surgical admission.
Primary endpoint: 30‑day all‑cause mortality.
Secondary endpoints: 90‑ and 360‑day all‑cause mortality.
Analytic approach: Multivariable Cox proportional hazards models adjusting for demographic, comorbidity, and perioperative covariates; subgroup analyses; sensitivity checks including propensity score methods (matching/weighting) to address confounding by indication.
Data source: MIMIC‑IV — a large, publicly available critical care EHR dataset enabling granular ICU research (PhysioNet/ MIMIC‑IV resource).
Key findings and interpretation
Population: 2,694 T2DM patients undergoing cardiac surgery; 1,127 in the metformin exposure group and 1,567 in the non‑exposure group.
Unadjusted outcomes: 30‑day mortality was 0.4% in the metformin group versus 4.5% in the non‑metformin group (absolute risk reduction 4.1 percentage points, p < 0.001).
Adjusted associations (primary multivariable Cox model):
– 30‑day mortality: HR 0.19 (95% CI 0.07–0.54), p = 0.002 — interpreted as an 81% relative risk reduction.
– 90‑day mortality: HR 0.24 (95% CI 0.12–0.47), p < 0.001 — a 76% relative reduction.
– 360‑day mortality: HR 0.35 (95% CI 0.23–0.52), p < 0.001 — a 65% relative reduction.
Robustness checks: Results were consistent across subgroup analyses, multiple sensitivity tests, and propensity score matched analyses, suggesting that the association is not driven solely by a single analytic choice.
Clinical magnitude: The absolute risk reduction for 30‑day mortality (4.1%) is clinically meaningful in a high‑risk surgical cohort. Even if attenuated after accounting for residual confounding, these effect sizes merit attention and further study.
Biological plausibility and mechanisms
Plausible mechanisms could explain a protective perioperative effect of metformin:
– Metabolic and cardioprotective signaling: Metformin activates AMP‑activated protein kinase (AMPK), which can improve myocardial energy handling and reduce ischemia‑reperfusion injury in preclinical models.
– Anti‑inflammatory effects: Metformin modulates inflammatory cytokines and may limit systemic inflammatory responses triggered by cardiopulmonary bypass.
– Endothelial and microvascular effects: Improvements in endothelial function and microcirculatory flow could reduce perioperative myocardial or end‑organ injury.
– Indirect benefits via glycemic control: Better baseline glycemic control in metformin users may contribute to lower complication rates.
Mechanistic reviews support these pathways but do not prove that perioperative exposure to metformin drives improved outcomes; they offer biologically plausible hypotheses to test in interventional studies (see References).
Strengths of the analysis
– Large, contemporary critical‑care database with granular perioperative and medication records.
– Clear, clinically important primary endpoint (30‑day all‑cause mortality) with extended follow‑up to 360 days.
– Multiple analytic approaches (multivariable adjustment, propensity score methods, sensitivity and subgroup analyses) that consistently pointed to benefit.
– Relevance to a high‑risk, frequently encountered clinical population.
Limitations and threats to causal inference
– Observational design: Even with careful adjustment, residual confounding by indication, unmeasured disease severity, frailty, or baseline functional status may bias results. For example, metformin users may have been healthier, had better preoperative care, or differed systematically in unmeasured ways from non‑users.
– Exposure definition ambiguity: ‘Metformin exposure’ in EHR data can reflect prior outpatient prescriptions, in‑hospital administration, or documentation artifacts. The timing, dosing, and perioperative withholding practices (preop stop vs continuation) are critical but may be incompletely captured.
– Immortal time bias: If exposure windows include time during which death could not occur for metformin‑classified patients, hazard ratios may be biased in favor of the exposure group unless carefully handled.
– Confounding by glycemic control and co‑therapies: Hemoglobin A1c, intra‑ and postoperative glucose management, insulin use, and other cardioprotective medications may not be fully adjusted for or accurately recorded.
– Selection bias and generalizability: MIMIC‑IV reflects ICU patients within participating centers and may not generalize to all cardiac surgical populations or healthcare systems.
– Safety outcomes: The study focused on mortality; data on metabolic complications (acute kidney injury, lactic acidosis), glycemic events, or readmissions were not reported in the summary and remain important for clinical decision‑making.
Clinical implications: what should clinicians do now?
– Do not change practice solely on this observational finding. The magnitude of association is striking but not proof of causality.
– Continue individualized perioperative medication management that balances the (low) absolute risk of metformin‑associated lactic acidosis with the potential benefits of continued glycemic control. Current perioperative guidance often recommends withholding metformin in patients at high risk for renal hypoperfusion, acute kidney injury, or contrast exposure; institutional protocols vary.
– Consider that metformin exposure may be a marker for better baseline disease control or access to care; clinicians should review the broader clinical context rather than interpret the finding as a mandate to universally continue metformin perioperatively.
– Support enrollment of suitable patients in randomized trials designed to evaluate perioperative metformin continuation versus withholding with predefined safety and efficacy endpoints.
Research agenda and next steps
– Randomized controlled trials: Definitive assessment of causality requires RCTs comparing continuation versus withholding of metformin in the perioperative period for cardiac surgery, with endpoints including mortality, myocardial injury, acute kidney injury, lactic acidosis, and functional recovery.
– Mechanistic studies: Translational studies should test metformin’s effects on myocardial ischemia‑reperfusion injury, systemic inflammation, and endothelial function in the perioperative context.
– Detailed observational work: Well‑designed emulated target trials using time‑aligned exposure definitions, instrumental variable approaches, and richer adjustment for preoperative status and intraoperative variables could help triangulate causal inference.
– Safety assessment: Large datasets and prospective registries should systematically capture lactic acidosis, renal outcomes, and hypoglycemic events related to perioperative metformin use.
Conclusion
The AHA 2025 MIMIC‑IV cohort analysis reports a strong and consistent association between metformin exposure and lower 30‑, 90‑, and 360‑day mortality among T2DM patients undergoing cardiac surgery. The signal is biologically plausible and robust across multiple analytic approaches, but the observational nature and potential biases preclude causal claims. These results are important and hypothesis‑generating: they support prioritizing randomized trials to determine whether perioperative metformin provides true cardioprotective benefit and to clarify safety parameters.
Funding and trial registration
The presented study used the publicly available MIMIC‑IV critical care dataset. The AHA abstract did not report specific external funding in the summary provided here. No clinicaltrials.gov registration is associated with the retrospective analysis; future interventional studies should be prospectively registered.
References
1. AHA 2025 abstract: Metformin exposure and mortality after cardiac surgery in T2DM patients (MIMIC‑IV retrospective cohort). Available at: https://aha.apprisor.org/epsAbstractAHA.cfm?id=82
2. Johnson AEW, Pollard TJ, Shen L, et al. MIMIC‑IV (version 1.0). PhysioNet; 2021. https://physionet.org/content/mimiciv/1.0/
3. UK Prospective Diabetes Study (UKPDS) Group. Effect of intensive blood‑glucose control with metformin on complications in overweight patients with type 2 diabetes (UKPDS 34). Lancet. 1998;352(9131):854‑865.
4. Rena G, Hardie DG, Pearson ER. The mechanisms of action of metformin. Diabetologia. 2017;60(9):1577‑1585.
5. American Diabetes Association. Standards of Care in Diabetes—2024. Diabetes Care. 2024;47(Suppl 1):S1‑S200. (Perioperative management recommendations summarized in the Standards.)
6. Lalau JD. Lactic acidosis induced by biguanides (metformin): a pharmacological and clinical perspective. Diabetes Metab. 2010;36(2):101‑107.
(Clinicians and researchers should consult the full AHA abstract and subsequent peer‑reviewed publication for full methodological details.)