Publication

• Title: Dapagliflozin and Acute Kidney Injury Following Cardiac Surgery: A Randomized Clinical Trial
• Acronym: MERCURI-2 (Promoting Effective Renoprotection in Cardiac Surgery Patients by Inhibition of SGLT-2)
• Year: 2026
• Journal published in: JAMA
• Citation: Oosterom-Eijmael MJP, Hulst AH, Monteiro de Oliveira NP, Niesten ED, Wietsma NE, Gerritse BM, et al; MERCURI-2 Study Group. Dapagliflozin and acute kidney injury following cardiac surgery: a randomized clinical trial. JAMA. Published online July 30, 2026.

Context & Rationale

• Background

  • Cardiac surgery-associated acute kidney injury (AKI) is common, with reported incidences ranging from 2% to 50%, and is associated with subsequent chronic kidney disease and increased mortality.
  • Its pathobiology is multifactorial, involving renal hypoperfusion and hypoxia, ischaemia-reperfusion injury, oxidative stress, inflammation, haemolysis and neurohormonal activation.
  • No pharmacological strategy had convincingly prevented cardiac surgery-associated AKI.
  • Sodium-glucose cotransporter 2 (SGLT2) inhibitors slow chronic kidney disease progression and reduce serious AKI in large chronic cardiorenal trials; a 2026 meta-analysis reported a 26% relative reduction in serious AKI (95% CI 18% to 34%).¹
  • Two small perioperative randomised studies had also suggested benefit: AKI occurred in 20% versus 66.7% in an open-label pilot study and in 22.5% versus 39.1% in the POST-CABGDM trial.²³
• Research Question/Hypothesis

  Would four perioperative doses of dapagliflozin, initiated on the day before elective cardiac surgery, reduce the incidence of full Kidney Disease: Improving Global Outcomes (KDIGO)-defined AKI during the first seven postoperative days compared with placebo?
• Why This Matters

  A short, oral and readily scalable treatment that genuinely prevented kidney injury after cardiac surgery would address a major unmet clinical need and could alter both cardiac anaesthesia and perioperative medicine.

Design & Methods

• Research Question: Among adults undergoing elective cardiac surgery, does dapagliflozin 10 mg once daily for four perioperative doses reduce AKI within seven days compared with placebo? The primary endpoint was any KDIGO AKI: serum creatinine increase of at least 26.5 μmol/L within 48 hours, serum creatinine at least 1.5 times baseline within seven days, or urine output below 0.5 mL/kg/h for at least six hours, with higher stages assigned for longer or more severe oliguria or anuria.
• Study Type: Investigator-initiated, multicentre, parallel-group, randomised, double-blind, placebo-controlled phase IV superiority trial conducted at two academic and five non-academic cardiac surgical hospitals in the Netherlands. Enrolment occurred from June 8, 2023, to January 27, 2025, and final follow-up was completed on May 16, 2025.⁴
• Population:
  • Adults aged 18 years or older scheduled for elective coronary artery bypass grafting, valve surgery, aortic surgery or combined cardiac procedures.
  • Key exclusions were type 1 diabetes; type 2 diabetes with body mass index below 25 kg/m² treated with multiple daily insulin injections; estimated glomerular filtration rate (eGFR) below 20 mL/min/1.73 m²; previous diabetic ketoacidosis; systolic blood pressure below 100 mm Hg; current SGLT2 inhibitor treatment; suspected SGLT2 inhibitor allergy; and pregnancy, breastfeeding, intended pregnancy or childbearing potential without adequate contraception.
  • In total, 784 participants were randomised, 392 to each group.
• Intervention:
  • Dapagliflozin 10 mg orally once daily for four doses.
  • The first dose was administered in hospital between 15:00 and 20:00 on the day before surgery, the second on the morning of surgery, the third on postoperative day 1 and the fourth on postoperative day 2.
  • Administration was observed by hospital or study staff and documented electronically.
  • If postoperative AKI was diagnosed before the final dose, further study medication was to be withheld.⁴
• Comparison:
  • Matching placebo capsules, administered at the same four time points.
  • All other perioperative care was delivered by the treating teams; the trial did not impose a study-specific AKI-prevention bundle.
• Blinding: Participants, treating clinicians, investigators and outcome assessors were masked. Dapagliflozin tablets were encapsulated and matched with microcrystalline-cellulose placebo; allocation was accessible only to trial pharmacists who had no other study role.
• Statistics: Computer-generated 1:1 randomisation used variable blocks of 4, 6 and 8, stratified by sex, type 2 diabetes and study site, through Castor Electronic Data Capture. A total of 784 participants was required to detect a 7.9% absolute reduction in AKI, from 22.0% with placebo to 14.1% with dapagliflozin (RR 0.64), with 80% power (β=0.20) and a two-sided α of 0.05. The primary analysis included all randomised participants; an exploratory per-protocol analysis included those taking all four doses. No missing-data imputation or multiplicity adjustment was used. The prespecified adjusted analysis included sex; adjustment for all stratification factors was post hoc.
• Follow-Up Period: Serum creatinine was measured daily while participants remained hospitalised for up to seven days, and urine output was recorded hourly while a urinary catheter remained in place. The primary outcome was assessed through postoperative day 7. Major cardiac and kidney events, days at home, disability and quality of life were assessed at 30 days; no longer-term kidney follow-up was reported.

Key Results

This trial was not stopped early. It completed its planned enrolment of 784 participants.

• The primary composite fell by approximately 24 percentage points and remained robust after adjustment for sex (OR 0.36; 95% CI 0.27 to 0.48; P<0.001) and post hoc adjustment for sex, type 2 diabetes and site (adjusted event rates 28.3% (95% CI 23.9 to 32.7) versus 52.3% (95% CI 47.4 to 57.2)); there was no evidence of interaction by sex (P=0.89). The signal was nevertheless concentrated in the urine-output component: creatinine-defined AKI was 14% versus 15%.
• Dapagliflozin reduced post hoc stage 1 and stage 2 KDIGO AKI, yet did not reduce stage 3 AKI, MAKE30, MACE30, length of stay, days at home, disability or quality of life.
• The single ketoacidosis event is clinically relevant because the intervention deliberately exposed participants to an SGLT2 inhibitor immediately before and after surgery.

Internal Validity

• Randomisation and Allocation: Central computer-generated randomisation, variable block sizes, prespecified stratification and pharmacist-only access to allocation make selection bias unlikely.
• Dropout or Exclusions: Follow-up testing was completed for 778/784 participants (99%), and all 784 randomised participants were retained in the primary analysis. Six did not undergo surgery; the manuscript does not clearly explain how a postoperative AKI endpoint was assigned to those patients. Patient-reported outcomes were incomplete for 61 EQ-5D-3L and 72 WHO-DAS assessments, limiting precision for those secondary outcomes but not the primary outcome.
• Performance and Detection Bias: Identical encapsulated tablets, observed administration and masking of participants, clinicians, investigators and assessors substantially reduce performance and detection bias. Creatinine and measured urine output were objective outcomes.
• Protocol Adherence: All four doses were received by 344/392 participants (88%) assigned dapagliflozin and 339/392 (87%) assigned placebo; four versus one received no dose. The per-protocol analysis reproduced the primary result: 27% versus 51%; RR 0.52; 95% CI 0.43 to 0.64.
• Per-Protocol Caveat: The protocol required withdrawal of further study medication once AKI developed. Completion of all four doses could therefore depend on a post-randomisation outcome, so the per-protocol result is supportive but not a clean causal sensitivity analysis.⁴
• Baseline Characteristics: The groups were well balanced. Median age was 68 years in both groups; median eGFR was 79 versus 80 mL/min/1.73 m²; EuroSCORE II was 1.4% versus 1.6%; cardiopulmonary bypass was used in 90% of both groups; and median bypass duration was 110 versus 108 minutes.
• Underlying Risk: Participants were stable elective surgical patients with relatively preserved renal and cardiac function: only 12% had type 2 diabetes, 6% had heart failure and 76%–81% had left ventricular ejection fraction above 50%. Surgical exposure was nevertheless substantial, with 40%–42% undergoing at least two procedures and 90% receiving cardiopulmonary bypass. This population generated many urine-output AKI events but few severe renal events.
• Heterogeneity: Seven hospitals and a mix of coronary, valvular, aortic and combined procedures improve clinical heterogeneity. There was no evidence of treatment-effect heterogeneity by sex (interaction P=0.89), but few prespecified subgroup analyses were reported.
• Timing: The regimen achieved exposure before the operative insult, with doses on the evening before and morning of surgery, followed by two postoperative doses. This timing was appropriate for prophylaxis but also maximised the possibility that acute tubular sodium and glucose effects would influence perioperative urine output.
• Dose: Dapagliflozin 10 mg is the standard therapeutic dose. Four doses tested a discrete perioperative strategy, not chronic treatment or continuation in established SGLT2 users.
• Separation of the Variable of Interest: Medication exposure was clearly separated: 98%, 97%, 93% and 93% of dapagliflozin participants received the preoperative, operative-day, postoperative-day-1 and postoperative-day-2 doses, compared with 100%, 98%, 94% and 90% receiving placebo. No pharmacokinetic confirmation was reported.
• Key Delivery Aspects: Administration was supervised, but other AKI-prevention practices were not protocolised. This improves pragmatism while allowing centre- and clinician-level variation in fluid, vasopressor and haemodynamic management.
• Adjunctive Therapy Use: Full-cohort numerical data on intravenous fluid, diuretics, noradrenaline and other haemodynamic co-interventions were not reported. The editorial described a 105-patient Amsterdam subset with numerically lower fluid volumes and noradrenaline doses in the dapagliflozin group, but detailed estimates were not presented in the primary report.⁵
• Crossover: No crossover or open-label SGLT2 inhibitor use was reported.
• Outcome Assessment: Full KDIGO surveillance was unusually meticulous, with daily creatinine and hourly catheter urine output. The primary outcome was therefore reliably detected, but one component was also biologically susceptible to the intervention. Urine-output surveillance ended when the urinary catheter was removed and creatinine surveillance at hospital discharge, so later events could have been missed, although this limitation should have affected both groups similarly.
• Statistical Rigor: The planned sample size was achieved and the intention-to-treat primary analysis was straightforward. However, the control-event rate was 52%, far above the assumed 22%; secondary outcomes were exploratory without multiplicity control; missing data were not imputed; and adjustment for all three stratification factors was post hoc, although it did not materially alter the result.

Conclusion on Internal Validity: Internal validity is strong for the narrow conclusion that assignment to dapagliflozin reduced the measured seven-day full KDIGO-defined endpoint. It is only moderate for the stronger claim that dapagliflozin prevented intrinsic kidney injury, because the benefit was urine-output driven, creatinine-defined AKI was unchanged, and severe or patient-centred renal outcomes did not improve.

External Validity

• Population Representativeness: Participants were 97% White and 76% male, with median eGFR 80 mL/min/1.73 m², only 12% type 2 diabetes and 6% heart failure. They were healthier and less renally vulnerable than many contemporary cardiac-surgery populations. The protocol envisaged a 50:50 male:female balance, but women comprised only 24% of the final cohort; the report did not explain this change.⁴
• Recruitment and Selection: Of 2551 patients assessed, 1369 were eligible but not enrolled, including 500 who declined and 396 excluded for logistical reasons. This does not threaten randomised comparisons but may select a more research-engaged and logistically straightforward elective population.
• Important Exclusions: Emergency or urgent surgery, current SGLT2 inhibitor use, eGFR below 20 mL/min/1.73 m², systolic blood pressure below 100 mm Hg, type 1 diabetes and previous ketoacidosis were excluded. Current SGLT2 therapy alone accounted for 269 exclusions.
• Surgical Setting: Ninety per cent underwent cardiopulmonary bypass, and the cohort included coronary, valve, aortic and combined procedures. Applicability is therefore reasonable across elective on-pump cardiac surgery, but not to non-cardiac surgery or off-pump-dominant programmes.
• Healthcare-System Applicability: The intervention is oral, widely available and operationally simple, so delivery is feasible in both high- and lower-resource systems. Safe implementation would nevertheless require perioperative glucose, ketone, volume and haemodynamic protocols.
• Drug and Treatment Strategy: The result applies specifically to newly initiated dapagliflozin 10 mg for four doses. It does not establish a class effect, support continuation of chronic SGLT2 therapy through surgery, or define treatment in established AKI.

Conclusion on External Validity: Generalisability is moderate within stable, SGLT2-naïve adults undergoing elective cardiac surgery, but limited for emergency surgery, advanced chronic kidney disease, haemodynamic instability, chronic SGLT2 users and ethnically diverse populations.

Strengths & Limitations

• Strengths:
  • Large, adequately powered, multicentre, placebo-controlled and rigorously blinded randomised design.
  • Central concealed allocation and excellent, balanced treatment adherence.
  • Near-complete ascertainment of the primary outcome with daily creatinine and meticulous hourly urine-output measurement.
  • Clinically pragmatic delivery across academic and non-academic centres.
  • Public funding with no reported funder role in design, analysis, interpretation or publication; relevant external industry relationships, including with SGLT2 manufacturers, were transparently disclosed.
• Limitations:
  • The primary endpoint was a short-term physiological composite rather than a patient-centred kidney outcome.
  • The large treatment effect was almost entirely attributable to urine output, while creatinine-defined AKI was unchanged and the acute eGFR fall was greater with dapagliflozin.
  • No kidney-injury biomarkers or histological data were analysed, preventing direct distinction between altered renal physiology and reduced tissue injury.
  • The trial was underpowered for stage 3 AKI, kidney replacement therapy, MAKE30 and uncommon harms.
  • No improvement occurred in hospital or ICU stay, days at home, disability or quality of life.
  • The homogeneous, predominantly White male, elective and SGLT2-naïve population restricts generalisability; the protocol-planned 50:50 sex balance was not achieved.
  • Stage 1, stage 2 and component analyses were post hoc; secondary outcomes were not multiplicity-adjusted.
  • Six randomised participants did not undergo surgery, and handling of their postoperative primary outcome was insufficiently described.
  • The trial did not record how many participants lacked a serum creatinine measurement on the day before surgery, so the timing and comparability of the baseline creatinine could not be fully assessed.
  • Several prespecified mechanistic and economic outcomes remained unreported, including glucose, haemodynamics, cardiac function, urinary oxygenation and costs.
  • The Methods narrative reversed the conventional MACE30 and MAKE30 definitions, whereas the outcome table footnotes used the clinically coherent definitions, creating a reporting inconsistency.

Interpretation & Why It Matters

• What MERCURI-2 Establishes

  Four perioperative doses of dapagliflozin substantially reduced the incidence of the prespecified full KDIGO-defined AKI composite after elective cardiac surgery.
• What It Does Not Establish

  The trial does not demonstrate less creatinine-defined kidney injury, dialysis, stage 3 AKI, MAKE30, mortality, disability, reduced length of stay or improved quality of life.
• Clinical Meaning

  The result is compelling evidence that the regimen altered postoperative renal physiology and KDIGO classification, but the discordance between urine output, creatinine and patient-centred outcomes means that “reduced AKI” should not automatically be equated with established renal protection.
• Practice Implication

  Routine perioperative initiation of dapagliflozin should await independent confirmation using creatinine-based, biomarker and major kidney outcomes, together with a clearly defined ketoacidosis-monitoring strategy.
• Conceptual Advance

  MERCURI-2 challenges the assumption that perioperative SGLT2 exposure is uniformly undesirable, but it does not answer whether patients already taking an SGLT2 inhibitor should continue therapy through surgery because chronic users were excluded.

Controversies & Other Evidence

• Endpoint–intervention entanglement: The relative treatment effect exceeded 40%, while the placebo AKI rate was 52% rather than the 22% assumed for power. Meticulous urine measurement explains part of the high event rate, but SGLT2 inhibition directly alters proximal tubular glucose and sodium handling and may increase early urine volume. Although the investigators cited evidence that sustained diuresis is modest and usually short-lived, this regimen delivered two doses immediately before surgery. This creates a risk that transient pharmacological changes in urine output altered the diagnostic component used to declare benefit even if structural kidney injury was unchanged.
• Creatinine and eGFR discordance: Creatinine-defined AKI was virtually identical at 14% versus 15%, while the reported maximum eGFR change was −7.4 versus −1.75 mL/min/1.73 m². The supplemental creatinine trajectory also showed no biochemical advantage and numerically higher early postoperative creatinine with dapagliflozin. An acute haemodynamic eGFR dip is a recognised SGLT2 effect and need not represent tissue damage, but the discordance remains the central reason to resist interpreting the primary composite as definitive tissue-level renoprotection.
• Stage 2 signal: The reduction in post hoc stage 2 AKI, 4.1% versus 13%; RR 0.33; 95% CI 0.19 to 0.56, argues that the result was not limited to transient six-hour oliguria. However, creatinine-only stage 2 AKI occurred in 3 versus 4 participants, whereas urine-output stage 2 AKI occurred in 17 versus 48; the separation was therefore predominantly urine-output based, and stage 2 was not a prespecified standalone efficacy endpoint.
• Reporting and analytic caveats: The prespecified adjusted model included sex but not all three randomisation strata; complete stratification adjustment was post hoc. Secondary outcomes were not multiplicity-adjusted, several prespecified mechanistic outcomes remained unpublished, and the Methods narrative reversed the conventional MACE30 and MAKE30 definitions despite correct definitions in the outcome-table footnotes. These issues do not explain the primary effect but reduce confidence in broader secondary claims.
• Editorial interpretation: The accompanying editorial urged “healthy scepticism” about such a large effect but judged the likelihood of no benefit or harm to be very low. It cited a 105-patient Amsterdam subset with numerically lower intravenous-fluid and noradrenaline exposure in the dapagliflozin group, and lower urinary TIMP-2 × IGFBP7 and KIM-1 on postoperative day 1 in the earlier pilot study, as evidence against inappropriate forced diuresis. These supportive observations were not fully quantified in the main report, and the editorial still restricted inference to stable elective cardiac-surgery patients pending confirmation.⁵
• Conflict with perioperative safety guidance: Contemporary multidisciplinary guidance advises withholding SGLT2 inhibitors before elective surgery because of perioperative euglycaemic ketoacidosis risk.⁶ MERCURI-2 instead initiated dapagliflozin on the day before surgery and continued it afterwards; one participant developed ketoacidosis. Systematic perioperative ketone surveillance was not reported, and the sample was far too small to exclude a clinically important increase in this uncommon complication.
• Clinical relevance of oliguria: Urine-output-defined AKI should not be dismissed as a meaningless endpoint: in cardiac-surgery cohorts, oliguria-defined AKI has been associated with subsequent major adverse kidney events.⁷ The specific concern in MERCURI-2 is not that oliguria lacks prognostic importance, but that the intervention may directly modify the criterion used to diagnose it.
• Prior randomised evidence: The direction of effect is consistent with the 55-patient open-label pilot trial, in which AKI occurred in 5/25 versus 20/30 participants, and with POST-CABGDM, in which AKI occurred in 16/71 versus 29/74 participants.²³ Both were smaller and less broadly representative than MERCURI-2.
• Broader SGLT2 evidence: Large chronic cardiorenal trials show a consistent reduction in serious AKI with SGLT2 inhibitors, supporting genuine renoprotective biology.¹ Those trials do not resolve whether a four-dose perioperative regimen prevents intrinsic surgical kidney injury.
• Subsequent evidence: MERCURI-2 was published on July 30, 2026; at the time of this appraisal, no independent confirmatory trial or guideline incorporating its findings was yet available.

Summary

• MERCURI-2 randomised 784 stable, SGLT2-naïve adults undergoing elective cardiac surgery to four doses of dapagliflozin 10 mg or matching placebo.
• Seven-day full KDIGO-defined AKI occurred in 28% versus 52%; RR 0.54; 95% CI 0.45 to 0.65; P<0.001.
• The treatment effect was driven by urine-output-defined AKI, 21% versus 48%; creatinine-defined AKI was 14% versus 15%; RR 0.93; 95% CI 0.66 to 1.32.
• There was no demonstrated benefit in stage 3 AKI, MAKE30, MACE30, length of stay, days at home, disability or quality of life; one dapagliflozin participant developed ketoacidosis.
• The trial is methodologically strong and practice-challenging, but independent confirmation with harder renal outcomes and robust safety assessment is required before routine perioperative initiation.

Overall Takeaway

MERCURI-2 is a provocative and potentially practice-changing proof-of-concept trial showing a large reduction in seven-day full KDIGO-defined AKI with four perioperative doses of dapagliflozin. Because the effect was almost wholly urine-output driven, with no creatinine-defined benefit and no improvement in severe or patient-centred outcomes, it should not yet be regarded as definitive evidence for routine perioperative SGLT2 initiation.

Bibliography

• 1.Neuen BL, Fletcher RA, Anker SD, et al; SGLT2 Inhibitor Meta-Analysis Cardio-Renal Trialists’ Consortium (SMART-C). SGLT2 inhibitors and kidney outcomes by glomerular filtration rate and albuminuria: a meta-analysis. JAMA. 2026;335(3):233-244.
• 2.Snel LIP, Oosterom-Eijmael MJP, Rampanelli E, et al. The effects of sodium-glucose transporter 2 inhibition on cardiac surgery-associated acute kidney injury: an open-label randomized pilot study. J Clin Anesth. 2025;103:111811.
• 3.Pitta FG, Lima EG, Tavares CAM, et al. Empagliflozin in patients with type 2 diabetes undergoing on-pump CABG: the POST-CABGDM randomized clinical trial. Diabetes Care. 2025;48(6):988-995.
• 4.Oosterom-Eijmael M, Monteiro de Oliveira NP, Niesten ED, et al; MERCURI-2 Study Group. Study protocol of the multicentre, randomised, triple-blind, placebo-controlled MERCURI-2 trial: promoting effective renoprotection in cardiac surgery patients by inhibition of sodium glucose cotransporter (SGLT)-2. BMJ Open. 2025;15(5):e095504.
• 5.Winkelmayer WC, Chertow GM. Dapagliflozin to reduce the risk of perioperative acute kidney injury in elective cardiac surgery. JAMA. Published online July 30, 2026.
• 6.El-Boghdadly K, Dhesi J, Fabb P, et al. Elective peri-operative management of adults taking glucagon-like peptide-1 receptor agonists, glucose-dependent insulinotropic peptide agonists and sodium-glucose cotransporter-2 inhibitors: a multidisciplinary consensus statement: a consensus statement from the Association of Anaesthetists, Association of British Clinical Diabetologists, British Obesity and Metabolic Surgery Society, Centre for Perioperative Care, Joint British Diabetes Societies for Inpatient Care, Royal College of Anaesthetists, Society for Obesity and Bariatric Anaesthesia and UK Clinical Pharmacy Association. Anaesthesia. 2025;80(4):412-424.
• 7.Priyanka P, Zarbock A, Izawa J, Gleason TG, Renfurm RW, Kellum JA. The impact of acute kidney injury by serum creatinine or urine output criteria on major adverse kidney events in cardiac surgery patients. J Thorac Cardiovasc Surg. 2021;162(1):143-151.e7.