Abstract
Background
Whether pre-existing glucose-lowering therapy influences outcomes after sepsis remains uncertain. We compared clinical outcomes among patients with type 2 diabetes (T2D) who developed sepsis according to prior GLP-1 receptor agonists (GLP-1RAs) or SGLT2 inhibitors (SGLT2is) use.
Methods
We conducted a retrospective cohort study using the TriNetX US Collaborative Network. Adults with T2D who developed sepsis had received a GLP-1RA or an SGLT2i within 3 months before sepsis were identified. After 1:1 propensity score matching, 11,969 patients were included in each group. The primary outcome was 90-day all-cause mortality. Secondary outcomes included major adverse cardiovascular event (MACE), major adverse kidney event (MAKE), intensive care unit (ICU) admission, major adverse cardiac and cerebrovascular event (MACCE), septic shock, and acute respiratory failure.
Results
Compared with prior SGLT2i use, prior GLP-1RA use was associated with lower 90-day all-cause mortality (incidence rate, 48.6 vs 60.6 per 100 person-years; HR, 0.81; 95% CI, 0.75–0.88; E-value: 1.6) and lower risks of MACE (HR, 0.77; 0.65–0.91; E-value, 1.7), myocardial infarction (HR, 0.82; 0.67–0.99; E-value, 1.7), ICU admission (HR, 0.89; 0.84–0.94; E-value, 1.4), and MACCE (HR, 0.86; 0.74–0.99; E-value, 1.5). No significant difference was observed for septic shock (HR, 0.87; 95% CI, 0.74–1.02), acute respiratory failure (HR, 0.94; 95% CI, 0.83–1.06), or MAKE (HR, 1.08; 95% CI, 0.92–1.27). The mortality benefit remained consistent through 365 days and across all prespecified subgroups. After Holm correction for multiple comparisons across the seven secondary outcomes, only the associations with MACE (adjusted P=0.012) and ICU admission (adjusted P<0.007) remained statistically significant, while those with MACCE, myocardial infarction, MAKE, septic shock, and acute respiratory failure did not.
Conclusion
Among patients with T2D who developed sepsis, prior GLP-1RA use was associated with lower mortality and select cardiovascular/renal risks than prior SGLT2i use, though roughly half of secondary outcomes showed no significant difference. Exposure was defined at the drug-class rather than individual-agent level, and modest E-values alongside baseline imbalances before matching suggest residual confounding cannot be excluded. These hypothesis-generating findings warrant confirmation in prospective studies.
Keywords: sepsis, type 2 diabetes mellitus, glucagon-like peptide-1 receptor agonist, sodium-glucose cotransporter-2 inhibitor, mortality
Introduction
Sepsis, a life-threatening dysregulated host response to infection, remains one of the largest contributors to global mortality.1 The most recent Global Burden of Disease 2021 analysis estimated 166 million sepsis cases and 21.4 million sepsis-related deaths worldwide in a single year, accounting for nearly one-third of all deaths globally.2 Sepsis incidence and mortality among adults reversed a decades-long decline during 2020–2021, a surge partly attributed to the COVID-19 pandemic. Survivors of sepsis could be also associated with substantial long-term morbidity, including elevated risks of cardiovascular events, chronic kidney disease progression, and death that persist well beyond the acute hospitalization,3–6 underscoring the importance of both short- and long-term outcome measures in sepsis research.
Patients with type 2 diabetes mellitus (T2D) represent a particularly vulnerable subgroup within the sepsis population.7 Chronic hyperglycemia impairs innate and adaptive immune function, promotes endothelial dysfunction, and increases susceptibility to infection, while pre-existing microvascular and macrovascular complications compound the risk of organ failure once sepsis develops. As a result, diabetes is one of the most common comorbidities among patients hospitalized with sepsis and is consistently associated with worse survival and higher rates of septic shock, acute kidney injury, and cardiovascular complications.7–9 Because most patients with T2D who develop sepsis are already receiving glucose-lowering therapy at the time of infection, the choice of background diabetes medication represents a potentially modifiable factor that may influence sepsis outcomes independent of glycemic control itself.
Among available glucose-lowering agents, sodium-glucose cotransporter-2 inhibitors (SGLT2i) have accumulated the most consistent observational evidence of benefit in sepsis.10–12 Several real-world cohort studies have reported that SGLT2i use before or around the time of sepsis onset is associated with lower 90-day mortality, reduced acute kidney injury, and fewer major adverse kidney events.10,11 These findings are thought to reflect the pleiotropic effects of SGLT2i, including reduced oxidative stress, improved mitochondrial efficiency, and anti-inflammatory and natriuretic actions, that may extend organ protection into the setting of acute critical illness.13 Glucagon-like peptide-1 receptor agonists (GLP-1RA), another class of glucose-lowering agents, exhibit similar properties, including anti-inflammatory signaling and endothelial and mitochondrial protection demonstrated across multiple organ systems in preclinical models of critical illness.14 Notably, endogenous GLP-1 levels rise markedly during sepsis, and small physiological studies of exogenous GLP-1 analogues have shown favorable metabolic effects, such as faster lactate clearance, in critically ill patients.15 Despite this biological plausibility, clinical outcome data for GLP-1RA specifically in the setting of sepsis remain scarce.
Although several studies16,17 have reported an association between GLP-1RA and sepsis, whether GLP-1RA confers protection against sepsis similar to, or greater than, that of SGLT2i remains unknown. We therefore used a global federated electronic health record network to conduct a propensity score-matched cohort study comparing 90-day all-cause mortality, along with secondary outcomes including major adverse cardiovascular event (MACE), major adverse kidney event (MAKE), septic shock, acute respiratory failure and intensive care unit (ICU) admission between patients with T2D and sepsis who had recently used a GLP-1RA versus an SGLT2i.
Methods
Study Design and Data Source
We conducted a retrospective, propensity score-matched cohort study using the TriNetX US Collaborative Network (TriNetX, LLC, Cambridge, MA), a federated health research network that aggregates de-identified electronic health record data, including diagnoses, procedures, medications, laboratory values, and vital status, from participating healthcare organizations (HCOs) across the United States.18 The analysis was run on 67 HCOs on July 10, 2026. Because TriNetX provides only aggregated, de-identified counts and statistical outputs and no protected health information is accessible to investigators, the study was conducted under the approval of the Institutional Review Board of Chi Mei Medical Center (IRB approval number 11402-E02).
Study Population
Eligible patients were adults (age ≥18 years) with a diagnosis of sepsis recorded between January 1, 2021 and June 30, 2026, who also had a diagnosis of T2D within 1 year before the sepsis diagnosis. Within this population, two mutually exclusive cohorts were constructed based on antihyperglycemic exposure during the 3 months preceding the sepsis diagnosis. GLP-1RA and SGLT2i exposure were each defined at the drug class level (ATC classification), without distinguishing between individual agents within each class (Supplementary Table 1). The study group (GLP-1RA) comprised patients who had received a GLP-1RA within 3 months before sepsis and had no SGLT2i use during the preceding year. The control group (SGLT2i) comprised patients who had received an SGLT2 inhibitor within the same period and had no GLP-1RA use during the preceding year. The index date was defined as the date of sepsis diagnosis. Patients were excluded if they were younger than 18 years, had not received a GLP-1RA or an SGLT2i within 3 months before sepsis, had prior use of the alternative study drug within 1 year before the index date, or had loss of follow-up.
Propensity Score Matching and Covariate Balance
To reduce confounding by indication, patients in the two cohorts were matched 1:1 using nearest-neighbor greedy propensity score matching with a caliper width of 0.1 of the standard deviation of the logit of the propensity score. The propensity score was estimated from demographic characteristics, comorbid diagnoses, concomitant cardiovascular and renal medications, and laboratory values recorded within one year before the index date, as in previous studies (Supplementary Table 2).19–21 Laboratory covariates were included based on available recorded values within the TriNetX platform; patients without a recorded value for a given laboratory covariate were not matched on that variable, consistent with the platform’s handling of missing data. Covariate balance before and after matching was assessed using standardized mean differences (SMD), with an SMD <0.1 considered indicative of adequate balance. Balance was further examined using a Love plot and variance ratios for continuous covariates.
Outcomes and Follow-Up
The primary outcome was all-cause mortality within 90 days after the index date. Prespecified secondary outcomes, assessed within the same 90-day window, were MACE, the composite of major adverse cardiac and cerebrovascular events (MACCE), myocardial infarction, MAKE, ICU admission, septic shock, and acute respiratory failure.19–21 MACE, MACCE, and myocardial infarction are nested composite outcomes as defined in Supplementary Table 3, with myocardial infarction (acute and subsequent ST-elevation/non-ST-elevation myocardial infarction) forming a component of MACE (which additionally includes cerebral infarction), and MACE in turn forming a component of MACCE (which additionally includes heart failure and cardiac arrest). Outcomes occurring in the time window beginning 1 day after the index date were analyzed; for outcomes for which a prior history could confound ascertainment of an incident event (mortality, MACE, MACCE, myocardial infarction, MAKE, septic shock, and acute respiratory failure), patients with the outcome recorded before the start of the time window were excluded from that specific analysis, consistent with an incident-event framework. Patients without an outcome event were censored at the end of the respective follow-up window or at the last recorded fact in their electronic health record, whichever occurred first.
Statistical Analysis
Continuous variables are presented as means with standard deviations and categorical variables as counts and percentages. Time-to-event outcomes were analyzed using the Kaplan–Meier method and compared using Log rank tests. Cox proportional hazards models were used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs). The proportional-hazards assumption was assessed graphically using complementary log–log plots. E-values were calculated to assess the potential influence of unmeasured confounding. Holm correction was applied separately across the seven secondary outcomes, four subgroup interaction tests, and five landmark time points. All analyses were performed within the TriNetX Advanced Analytics platform. All tests were two-sided, with P<0.05 considered statistically significant. To evaluate the potential impact of unmeasured confounding, we estimated E-values, which quantify the minimum magnitude of association that an unmeasured confounder would need to have with both the exposure and the outcome to account for the observed association.22
Additional Analysis
To assess the consistency of the primary outcome association, we performed prespecified subgroup analyses of 90-day all-cause mortality stratified by sex (female vs male), age (18–64 vs ≥65 years), glycemic control (hemoglobin A1c <9% vs ≥9%), and body mass index (BMI <30 vs ≥30 kg/m2). For each subgroup, HR and 95% CI were estimated separately using the same Cox proportional hazards model specified for the primary analysis. Subgroup-specific cohorts were reconstructed in TriNetX for each analysis. For each subgroup factor, a formal test for interaction was additionally performed using inverse-variance-weighted fixed-effect meta-regression of the log hazard ratio across strata, with the P for interaction derived from a chi-square test of the between-stratum coefficients. In addition, to evaluate the durability of the association with mortality, we performed prespecified landmark analyses of all-cause mortality at 14, 30, 90 (primary time point), 180, and 365 days after the index date, with Holm correction applied across the five time points. A fixed-horizon sensitivity analysis was also performed using the 90-day risk difference and risk ratio.
Results
Patient Selection
On July 10, 2026, we identified 62,715,727 adult individuals who had received health care services within the TriNetX US network between January 1, 2021, and June 30, 2026. Among them, 503,934 had been diagnosed with T2D before the onset of sepsis. After applying the predefined exclusion criteria, patients were categorized according to prior treatment with GLP-1RA or SGLT2i. The initial cohort included 13,101 patients in the GLP-1RA group and 35,122 in the SGLT2i group. Following PSM, 11,969 patients remained in each group for the final analysis (Figure 1).
Figure 1.
Flow Diagram of Cohort Construction.
Abbreviations: HCOs, healthcare organizations; GLP-1RA, glucagon-like peptide-1 receptor agonist; SGLT2i, sodium–glucose cotransporter-2 inhibitor.
Baseline Characteristics
Before matching, patients in the GLP-1RA cohort were younger (mean age 62.6 ± 13.2 vs 67.5 ± 12.3 years; SMD 0.381), more often female (53.2% vs 38.1%; SMD 0.306), and had a substantially lower burden of pre-existing cardiovascular and renal comorbidity than the SGLT2i cohort, including less heart failure (25.3% vs 60.2%; SMD 0.754), ischemic heart disease (28.8% vs 52.8%; SMD 0.503), atrial fibrillation (16.8% vs 37.3%; SMD 0.476), acute kidney failure (27.9% vs 45.8%; SMD 0.378), and chronic kidney disease (35.2% vs 46.7%; SMD 0.235), consistent with the more common use of SGLT2i in patients with established cardiorenal disease. After 1:1 propensity score matching, 11,969 patients remained in each cohort, and all baseline characteristics were well balanced, including age (63.4 ± 12.8 vs 63.7 ± 12.9 years; SMD 0.017), sex (50.4% vs 49.7% female; SMD 0.015), and comorbidity burden (all SMD <0.1) (Table 1 and Figure 1). Mean follow-up after matching was 80.3 ± 23.9 days in the GLP-1RA cohort and 76.7 ± 27.2 days in the SGLT2i cohort, with a median follow-up of 90 days in both cohorts. Matching retained 91.4% of patients in the GLP-1RA cohort. The maximum post-matching SMD was 0.089, and variance ratios for continuous covariates ranged from approximately 0.96 to 1.23 (Supplementary Figure 1).
Table 1.
Baseline Characteristics of Included Subjects
| Variables | Before Matching | After Matching | ||||
|---|---|---|---|---|---|---|
| GLP-1RA Group (n = 13,101) | SGLT2i Group (n = 35,122) | SMD | GLP-1RA Group (n = 11,969) | SGLT2i Group (n = 11,969) | SMD | |
| Age at index, years | ||||||
| Mean (SD) | 62.6 (13.2) | 67.5 (12.3) | 0.381 | 63.4 (12.8) | 63.7 (12.7) | 0.017 |
| Sex, n (%) | ||||||
| Female | 6965 (53.2) | 13,379 (38.1) | 0.306 | 6031 (50.4) | 5944 (49.7) | 0.015 |
| Male | 6132 (46.8) | 21,739 (61.9) | 0.306 | 5934 (49.6) | 6023 (50.3) | 0.015 |
| Demographic, n (%) | ||||||
| White | 8722 (66.6) | 23,635 (67.3) | 0.015 | 8021 (67.0) | 8037 (67.1) | 0.003 |
| Black or African American | 2494 (19.0) | 6443 (18.3) | 0.018 | 2191 (18.3) | 2190 (18.3) | 0.000 |
| Asian | 423 (3.2) | 1571 (4.5) | 0.065 | 410 (3.4) | 425 (3.6) | 0.007 |
| Other Race | 471 (3.6) | 1091 (3.1) | 0.027 | 437 (3.7) | 448 (3.7) | 0.005 |
| Unknown Race | 672 (5.1) | 1618 (4.6) | 0.024 | 627 (5.2) | 626 (5.2) | 0.000 |
| Hispanic or Latino | 1081 (8.3) | 2719 (7.7) | 0.019 | 1005 (8.4) | 1027 (8.6) | 0.007 |
| Not Hispanic or Latino | 10,017 (76.5) | 27,694 (78.9) | 0.057 | 9184 (76.7) | 9221 (77.0) | 0.007 |
| Unknown Ethnicity | 2003 (15.3) | 4709 (13.4) | 0.054 | 1780 (14.9) | 1721 (14.4) | 0.014 |
| Comorbidities, n (%) | ||||||
| Dyslipidemia | 9171 (70.0) | 26,607 (75.8) | 0.130 | 8383 (70.0) | 8399 (70.2) | 0.003 |
| Overweight and obesity | 6722 (51.3) | 15,031 (42.8) | 0.171 | 5801 (48.5) | 5735 (47.9) | 0.011 |
| Chronic kidney disease | 4609 (35.2) | 16,386 (46.7) | 0.235 | 4328 (36.2) | 4385 (36.6) | 0.010 |
| Acute kidney failure | 3651 (27.9) | 16,085 (45.8) | 0.378 | 3529 (29.5) | 3553 (29.7) | 0.004 |
| Ischemic heart diseases | 4117 (31.4) | 20,041 (57.1) | 0.534 | 4036 (33.7) | 4012 (33.5) | 0.004 |
| Chronic ischemic heart disease | 3770 (28.8) | 18,529 (52.8) | 0.503 | 3698 (30.9) | 3690 (30.8) | 0.001 |
| Heart failure | 3319 (25.3) | 21,159 (60.2) | 0.754 | 3313 (27.7) | 3304 (27.6) | 0.002 |
| Obstructive sleep apnea | 3519 (26.9) | 9535 (27.1) | 0.006 | 3148 (26.3) | 3059 (25.6) | 0.017 |
| Atrial fibrillation and flutter | 2199 (16.8) | 13,114 (37.3) | 0.476 | 2166 (18.1) | 2109 (17.6) | 0.012 |
| Cerebrovascular diseases | 1884 (14.4) | 7596 (21.6) | 0.189 | 1792 (15.0) | 1821 (15.2) | 0.007 |
| Chronic obstructive pulmonary disease | 2218 (16.9) | 10,152 (28.9) | 0.288 | 2146 (17.9) | 2121 (17.7) | 0.005 |
| Asthma | 1976 (15.1) | 4305 (12.3) | 0.082 | 1682 (14.1) | 1639 (13.7) | 0.010 |
| Chronic respiratory failure | 636 (4.9) | 3305 (9.4) | 0.178 | 612 (5.1) | 575 (4.8) | 0.014 |
| Pneumonia | 349 (2.7) | 1305 (3.7) | 0.060 | 326 (2.7) | 328 (2.7) | 0.001 |
| Nicotine dependence | 1923 (14.7) | 7321 (20.8) | 0.162 | 1813 (15.1) | 1769 (14.8) | 0.010 |
| Diseases of liver | 2165 (16.5) | 6998 (19.9) | 0.088 | 1988 (16.6) | 1905 (15.9) | 0.019 |
| Fibrosis and cirrhosis of liver | 681 (5.2) | 2364 (6.7) | 0.065 | 643 (5.4) | 629 (5.3) | 0.005 |
| Type 2 diabetes mellitus complications, n (%) | ||||||
| Type 2 diabetes mellitus with ophthalmic complications | 1535 (11.7) | 3122 (8.9) | 0.093 | 1339 (11.2) | 1377 (11.5) | 0.010 |
| Type 2 diabetes mellitus with kidney complications | 4348 (33.2) | 14,770 (42.1) | 0.184 | 4057 (33.9) | 4126 (34.5) | 0.012 |
| Type 2 diabetes mellitus with neurological complications | 4383 (33.5) | 11,023 (31.4) | 0.044 | 3967 (33.1) | 4117 (34.4) | 0.027 |
| Type 2 diabetes mellitus with circulatory complications | 2240 (17.1) | 7911 (22.5) | 0.136 | 2098 (17.5) | 2109 (17.6) | 0.002 |
| Medications, n (%) | ||||||
| Insulins and analogues | 8715 (66.5) | 25,017 (71.2) | 0.102 | 7991 (66.8) | 8089 (67.6) | 0.017 |
| Beta blockers | 6482 (49.5) | 24,105 (68.6) | 0.397 | 6111 (51.1) | 6094 (50.9) | 0.003 |
| Loop diuretics | 4606 (35.2) | 20,702 (58.9) | 0.491 | 4399 (36.8) | 4301 (35.9) | 0.017 |
| Biguanides | 4657 (35.5) | 9733 (27.7) | 0.169 | 4149 (34.7) | 4203 (35.1) | 0.009 |
| Calcium channel blockers | 4309 (32.9) | 13,192 (37.6) | 0.098 | 3984 (33.3) | 4007 (33.5) | 0.004 |
| Angiotensin II inhibitor | 3352 (25.6) | 14,277 (40.6) | 0.324 | 3180 (26.6) | 3143 (26.3) | 0.007 |
| ACE inhibitors | 3453 (26.4) | 8877 (25.3) | 0.025 | 3158 (26.4) | 3206 (26.8) | 0.009 |
| Thiazides | 2618 (20.0) | 7225 (20.6) | 0.015 | 2366 (19.8) | 2342 (19.6) | 0.005 |
| eGFR, mL/min/1.73m2 | ||||||
| Mean (SD) | 66.8 (36.0) | 64.8 (34.3) | 0.055 | 65.6 (35.5) | 68.8 (35.7) | 0.089 |
| ≤ 45, n (%) | 5068 (38.7) | 17,343 (49.4) | 0.217 | 4741 (39.6) | 4775 (39.9) | 0.006 |
| Hemoglobin A1c, % | ||||||
| ≥ 9, n (%) | 3366 (25.7) | 6690 (19.0) | 0.160 | 2944 (24.6) | 3056 (25.5) | 0.022 |
| LDL cholesterol, mg/dL | ||||||
| ≥ 160, n (%) | 348 (2.7) | 735 (2.1) | 0.037 | 301 (2.5) | 283 (2.4) | 0.010 |
| Serum lactate, mmol/L | ||||||
| Mean (SD) | 1.6 (1.0) | 1.6 (1.0) | 0.021 | 1.6 (1.0) | 1.6 (0.9) | 0.069 |
| ≥ 2, n (%) | 2334 (17.8) | 9551 (27.2) | 0.226 | 2220 (18.5) | 2235 (18.7) | 0.003 |
| C-reactive protein, mg/L | ||||||
| Mean (SD) | 69.3 (85.6) | 72.6 (82.6) | 0.039 | 69.4 (85.6) | 76.5 (87.5) | 0.081 |
Abbreviations: BMI, body mass index; CI, confidence interval; CKD, chronic kidney disease; CRP, C-reactive protein; GFR, glomerular filtration rate; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; LDL, low-density lipoprotein; SD, standard deviation; SGLT2i, sodium–glucose cotransporter-2 inhibitor; SMD, standardized mean difference.
Primary Outcome: 90-Day All-Cause Mortality
Within 90 days after the index date, 1,227 patients in the GLP-1RA group died, compared with 1,475 patients in the SGLT2i group. Prior GLP-1RA use was associated with a lower risk of 90-day all-cause mortality than prior SGLT2i use (incidence rate [IR], 48.6 vs 60.6 per 100 person-years; HR, 0.81; 95% CI, 0.75–0.88; Table 2). The corresponding E value was 1.6 with lower confidence limit of 1.4, and the 90-day Kaplan-Meier survival probabilities were 88.9% for the GLP-1RA group and 86.5% for the SGLT2i group (log-rank p<0.001, Figure 2). In the fixed-horizon analysis, mortality was 10.25% versus 12.32%, corresponding to a risk difference of −2.07 percentage points (95% CI, −2.87 to −1.27) and a risk ratio of 0.83 (95% CI, 0.77–0.89; Supplementary Table 4). The E-value was 1.6, with a lower confidence limit of 1.4.
Table 2.
Hazard Ratio of Outcomes Between the GLP-1RA and the SGLT2i Groups
| Outcome | GLP-1RA Group (n = 11,969) | SGLT2i Group (n = 11,969) | HR (95% CI) | P Value | E-Value (95% LCL) | ||
|---|---|---|---|---|---|---|---|
| Events (n) | Incidence Rate Per 100 PYs | Events (n) | Incidence Rate Per 100 PYs | ||||
| Primary Outcome | |||||||
| All-cause mortality | 1227 | 48.6 | 1475 | 60.6 | 0.81 (0.75, 0.88) | <0.001 | 1.6 (1.4) |
| Secondary Outcome | |||||||
| MACE | 243 | 13.1 | 297 | 17.2 | 0.77 (0.65, 0.91) | 0.002 | 1.7 (1.3) |
| MACCE | 336 | 26.9 | 361 | 31.8 | 0.86 (0.74, 0.99) | 0.042 | 1.5 (1.1) |
| Myocardial infarction | 190 | 9.0 | 220 | 11.1 | 0.82 (0.67, 0.99) | 0.042 | 1.7 (1.1) |
| MAKE | 301 | 14.0 | 294 | 13.3 | 1.08 (0.92, 1.27) | 0.350 | 1.4 (1.0) |
| Septic shock | 290 | 14.3 | 322 | 16.8 | 0.87 (0.74, 1.02) | 0.083 | 1.4 (1.0) |
| Acute respiratory failure | 514 | 34.0 | 522 | 37.2 | 0.94 (0.83, 1.06) | 0.293 | 1.3 (1.0) |
| ICU admission | 2281 | 86.7 | 2508 | 99.8 | 0.89 (0.84, 0.94) | <0.001 | 1.4 (1.3) |
Abbreviations: CI, confidence interval; GLP-1RA, glucagon-like peptide-1 receptor agonist; HR, hazard ratio; IR, incidence rate; LCL, lower confidence limit; MACCE, major adverse cardiac and cerebrovascular events; MACE, major adverse cardiovascular events; MAKE, major adverse kidney events; PY(s), person-year(s); SGLT2i, sodium–glucose cotransporter-2 inhibitor.
Figure 2.
Kaplan–Meier curves for all-cause mortality comparing the GLP-1RA and SGLT2i groups during 14-day (A), 30-day (B), 90-day (C), 180-day (D), and 365-day (E) follow-up periods. The shaded areas represent the 95% confidence intervals. Red indicates the SGLT2i group, whereas blue indicates the GLP-1RA group. Log-rank P values are shown in each panel.
Abbreviations: GLP-1RA, glucagon-like peptide-1 receptor agonist; SGLT2i, sodium–glucose cotransporter-2 inhibitor.
Secondary Outcomes at 90 Days
Compared with the SGLT2i cohort, the GLP-1RA cohort had lower risk of, MACE (IR, 13.1 vs 17.2 per 100 person-years; HR, 0.77; 95% CI 0.65–0.91; p=0.002; E-value 1.7), MACCE (IR, 26.9 vs 31.8 per 100 person-years; HR, 0.86; 95% CI 0.74–0.99; p=0.042; E-value 1.5), myocardial infarction (IR, 9.0 vs 11.1 per 100 person-years; HR, 0.82; 95% CI 0.67–0.99; p=0.042; E-value 1.7) and ICU admission (IR, 86.7 vs 99.8 per 100 person-years; HR, 0.89; 95% CI 0.84–0.94; p<0.001; E-value 1.4). No statistically significant between-group difference was observed for septic shock (IR, 14.3 vs 16.8 per 100 person-years; HR, 0.87; 95% CI 0.74–1.02; p=0.083), acute respiratory failure (IR, 34.0 vs 37.2 per 100 person-years; HR, 0.94; 95% CI 0.83–1.06; p=0.293), or MAKE (IR, 14.0 vs 13.3 per 100 person-years; HR, 1.08; 95% CI 0.92–1.27; p=0.350) (Table 2). E-values for all statistically significant secondary outcomes ranged from 1.4 to 1.7. After Holm correction across the seven secondary outcomes, the associations with MACE (adjusted P=0.012) and ICU admission (adjusted P<0.007) remained statistically significant, whereas those with MACCE and myocardial infarction did not. No significant differences were observed for MAKE, septic shock, or acute respiratory failure (Supplementary Table 5).
Landmark Analyses of All-Cause Mortality
The survival advantage associated with prior GLP-1RA use was evident early and persisted throughout 1 year of follow-up. At 14 days, mortality was 4.7% in the GLP-1RA cohort versus 5.7% in the SGLT2i cohort (HR, 0.81; 95% CI 0.72–0.91; p<0.001). At 30 days, mortality was 7.1% versus 8.8% (HR, 0.80; 95% CI 0.73–0.88; p<0.001). At 180 days, mortality was 12.7% versus 15.0% (HR, 0.81; 95% CI 0.75–0.87; p<0.001), and at 365 days, mortality was 15.8% versus 17.7% (HR, 0.84; 95% CI 0.79–0.89; p<0.001). The magnitude and direction of the HR for all-cause mortality therefore remained consistent (HR, range 0.798–0.837) from the earliest landmark at 14 days through 1 year of follow-up (Table 3 and Figure 2). The complementary log–log curves were approximately parallel throughout follow-up, with no evident graphical departure from the proportional-hazards assumption (Supplementary Figure 2).
Table 3.
Landmark Analyses of All-Cause Mortality at Prespecified Time Points
| Landmark Time Point | GLP-1RA Group (n/N) | SGLT2i Group (n/N) | Hazard Ratio (95% CI) | P Value |
|---|---|---|---|---|
| 14 days | 539/11,969 | 665/11,969 | 0.81 (0.72, 0.91) | <0.001 |
| 30 days | 820/11,969 | 1016/11,969 | 0.80 (0.73, 0.88) | <0.001 |
| 90 days | 1227/11,969 | 1475/11,969 | 0.81 (0.75, 0.88) | <0.001 |
| 180 days | 1461/11,969 | 1745/11,969 | 0.81 (0.75, 0.87) | <0.001 |
| 365 days | 1818/11,969 | 2059/11,969 | 0.84 (0.79, 0.89) | <0.001 |
Notes: Nominal P values are shown. After Holm correction across the five landmark time points, all adjusted P values were <0.005.
Abbreviations: GLP-1RA, glucagon-like peptide-1 receptor agonist; SGLT2i, sodium–glucose cotransporter-2 inhibitor.
Subgroup Analyses
The association between prior GLP-1RA use and lower 90-day all-cause mortality was directionally consistent across all prespecified subgroups (Figure 3). The benefit was similar in women (HR, 0.81; 95% CI 0.73–0.90) and men (HR, 0.81; 95% CI 0.74–0.90); in patients aged 18–64 years (HR, 0.81; 95% CI 0.70–0.92) and those aged ≥65 years (HR, 0.80; 95% CI 0.74–0.87); in patients with HbA1c <9% (HR, 0.80; 95% CI 0.73–0.87) and HbA1c ≥9% (HR, 0.84; 95% CI 0.73–0.96); and in patients with BMI <30 kg/m2 (HR, 0.81; 95% CI 0.72–0.91) and BMI ≥30 kg/m2 (HR, 0.80; 95% CI 0.73–0.87). Point estimates across all eight subgroup strata ranged narrowly from 0.80 to 0.84, all p<0.001. Formal tests for interaction showed no evidence of effect modification by sex (P for interaction=1.000), age (P=0.878), glycemic control (P=0.557), or body mass index (P=0.868). After Holm correction across the four effect modifiers, all adjusted P values were 1.000.
Figure 3.
![Table plus forest plot of hazard ratio for 90-day all-cause mortality by subgroup. Forest plot showing the association between glucagon-like peptide-1 receptor agonist (GLP-1RA) use and the study outcome compared with sodium-glucose cotransporter 2 inhibitor (SGLT2i) use, overall and across subgroups. Overall, GLP-1RA use was associated with a lower risk of the outcome (hazard ratio [HR], 0.81; 95% confidence interval [CI], 0.75–0.88; P<0.001). Similar associations were observed across subgroups defined by sex, age, glycated hemoglobin level, and body mass index, with no significant evidence of interaction (P for interaction >0.05).](https://cdn.ncbi.nlm.nih.gov/pmc/blobs/f09f/13633727/b524cf476c11/DDDT-20-642220-g0003.webp)
Forest plot of 90-day all-cause mortality across prespecified subgroups.
Notes: Nominal P values for interaction are shown. After Holm correction across the four effect modifiers, all adjusted P values were 1.000.
Abbbreviations: BMI, body mass index; GLP-1RA, glucagon-like peptide-1 receptor agonist; HbA1c, hemoglobin A1c; SGLT2i, sodium–glucose cotransporter-2 inhibitor.
Discussion
In this large propensity score-matched cohort of patients with T2D who developed sepsis, prior GLP-1RA use was associated with significantly lower 90-day all-cause mortality than prior SGLT2i use. This survival advantage emerged early after sepsis onset, remained consistent through 1 year of follow-up, and was observed across all prespecified subgroups. In addition to lower mortality, prior GLP-1RA use showed nominal associations with lower risks of MACE, myocardial infarction, MACCE, and ICU admission, whereas the risks of septic shock, acute respiratory failure, and MAKE were similar between groups. From a clinical perspective, our findings suggest that the choice of background glucose-lowering therapy may have implications extending beyond glycemic control in patients at risk of sepsis. Current recommendations for selecting GLP-1RA or SGLT2i are largely based on cardiovascular, renal, and metabolic indications. If confirmed in prospective studies, the potential influence of these therapies on sepsis outcomes may represent an additional consideration when individualizing treatment for patients with T2D who are at high risk for severe infection. Nevertheless, our findings should be considered hypothesis-generating and should not be interpreted as evidence supporting initiation of GLP-1RA solely for sepsis prevention.
Our finding was consistent with previous studies about the association between GLP-1RA and sepsis. A large-scale real-world study of 336,863 patients demonstrated that GLP-1RA use was associated with a lower risk of severe sepsis compared with dipeptidyl peptidase-4 inhibitor (DPP-4i) use (HR, 0.61; 95% CI, 0.58–0.62.17 A comprehensive meta-analysis of 136 randomized controlled trials (n=164,322) confirmed that GLP-1RA use was associated with reduced serious infections (RR 0.89, 95% CI 0.86–0.93), with particular reductions in serious respiratory, skin, musculoskeletal, and vascular infections.16 Furthermore, our findings extend the current evidence regarding glucose-lowering therapies in sepsis. Previous observational studies have primarily evaluated SGLT2i against nonuse or other glucose-lowering agents, reporting lower mortality, reduced acute kidney injury, and fewer adverse kidney outcomes among SGLT2i users.10–12 However, whether GLP-1RA provides comparable or even greater protection has remained unclear because direct comparisons between these two cardioprotective drug classes have been lacking. By adopting an active-comparator design,23 our study minimizes confounding related to diabetes severity and treatment indication compared with studies using untreated controls, while addressing a clinically relevant question faced by physicians when selecting glucose-lowering therapy for patients at high cardiometabolic risk. The observed reduction in mortality despite comparison with another therapeutic class known to confer favorable cardiovascular effects suggests that GLP-1RA may possess additional protective mechanisms that become particularly important during sepsis.
Several biological mechanisms may explain these findings. GLP-1RA exert anti-inflammatory effects through inhibition of nuclear factor-κB signaling, suppression of pro-inflammatory cytokine production, and modulation of innate immune cell activation.24–27 Experimental studies have also demonstrated that GLP-1 signaling improves endothelial function, preserves mitochondrial integrity, reduces oxidative stress, and attenuates microvascular dysfunction, all of which are central components of sepsis pathophysiology.24 Furthermore, endogenous GLP-1 concentrations increase during sepsis,14,28 suggesting activation of an adaptive metabolic response to critical illness, while administration of GLP-1 analogues has been associated with improved glucose homeostasis and enhanced lactate clearance in critically ill patients. Although these mechanistic observations cannot establish causality, they provide biological plausibility for the lower mortality observed among patients receiving GLP-1RA before the onset of sepsis. In addition, GLP-1RA may additionally act through incretin-receptor-mediated modulation of systemic and innate immune inflammation and through the endogenous rise in GLP-1 during critical illness,24,29 mechanisms not shared by SGLT2i.30 This broader immunometabolic action, layered on cardiometabolic effects similar to those of SGLT2i, may help explain why GLP-1RA was associated with greater protection than SGLT2i in the present study.
An interesting finding was that the mortality benefit associated with GLP-1RA was accompanied by lower risks of cardiovascular events and ICU admission but not by significant reductions in septic shock, acute respiratory failure, or MAKE. Because MACE, MACCE, and myocardial infarction are nested composite outcomes, and myocardial infarction is a component of MACE, which is in turn a component of MACCE, these should not be interpreted as three independent cardiovascular signals, but as overlapping representations of a reduction that may partly reflect fewer myocardial infarction events. With this caveat, these findings suggest that the survival advantage may be related, at least in part, to reduced myocardial ischemic events and improved cardiovascular status during critical illness, rather than to attenuation of the acute infectious process itself. Given that cardiovascular complications are common among patients with sepsis and contribute substantially to both early and late mortality,3 the established cardioprotective properties of GLP-1RA may partially explain the observed reduction in mortality.31 The persistence of this association through 365 days further supports the possibility that GLP-1RA influences not only the acute phase of sepsis but also the long-term trajectory of recovery after critical illness.
Several strengths enhance the validity of this study. The large, geographically diverse TriNetX network provided substantial statistical power and broad clinical representation. The active-comparator design reduced confounding by treatment indication compared with comparisons involving nonusers, while propensity score matching achieved excellent balance across measured baseline characteristics despite marked differences before matching. The consistency of the mortality association across multiple landmark analyses and prespecified subgroup analyses supports the robustness of the findings. In addition, E-value analyses indicated that an unmeasured confounder would require at least a moderate association with both treatment assignment and mortality to fully explain the observed associations.
This study also has important limitations. First, as a retrospective observational study based on electronic health records, residual confounding cannot be completely excluded despite propensity score matching and E-value analyses. Before propensity score matching, patients receiving SGLT2i had a greater burden of pre-existing cardiorenal comorbidities, including heart failure, ischemic heart disease, and chronic kidney disease, which may reflect differences in underlying clinical profiles and prescribing patterns. Although 1:1 propensity score matching substantially improved the balance of measured covariates between the two treatment groups, unmeasured factors or differences in disease severity, including the severity of cardiorenal disease and other aspects of clinical status not captured in the database, may have persisted and influenced mortality and other clinical outcomes. Therefore, the observed associations should be interpreted cautiously and should not be considered evidence of a causal treatment effect. Second, information regarding sepsis severity, including Sequential Organ Failure Assessment scores, Acute Physiology and Chronic Health Evaluation scores, vasopressor requirements, source of infection, microbiological findings, antimicrobial therapy, and timing of source control, was unavailable. These factors may have influenced both treatment selection and clinical outcomes and therefore represent potential sources of residual confounding. Third, medication adherence, treatment duration, dosage, and whether GLP-1RA or SGLT2i therapy was continued or discontinued after hospitalization could not be determined. Fourth, outcomes and exposures were identified using diagnostic and medication coding, which may be subject to misclassification; in particular, MACE, MACCE, and myocardial infarction are nested composite outcomes that overlap substantially rather than representing independent cardiovascular signals, and “loss of follow-up” refers to patients who left the TriNetX network (eg, through disenrollment or change of care setting) before the end of the follow-up window and could no longer be observed. Fifth, GLP-1RA and SGLT2i exposure were defined at the drug class (ATC) level rather than by individual agent, which may obscure between-drug heterogeneity in efficacy or safety within each class. Sixth, Holm correction was applied separately across the secondary outcomes, subgroup interaction tests, and landmark time points; nevertheless, these analyses should be considered exploratory. Because each subgroup was constructed as an independently rematched sub-cohort rather than a stratification of the overall matched cohort, subgroup-specific HRs are not directly comparable to one another or to the primary estimate. Individual propensity scores were unavailable, and the Love plot therefore demonstrates covariate balance rather than propensity-score overlap. Schoenfeld residuals were also unavailable, so the proportional-hazards assumption was assessed graphically. The fixed-horizon comparison was based on event proportions and did not account for censoring or matched-pair dependence Finally, because the TriNetX network includes participating healthcare organizations in the United States, the generalizability of these findings to other healthcare systems or populations remains uncertain.
In conclusion, among patients with T2D who developed sepsis, prior GLP-1RA use, defined at the drug class (ATC) level rather than by individual agent, was associated with lower short- and long-term mortality than prior SGLT2i use, together with lower risks of cardiovascular events and ICU admission. Given the modest E-values (1.4–1.7) relative to the substantial covariate imbalance observed before matching, a moderate unmeasured confounder could meaningfully affect these estimates, and residual confounding cannot be excluded. These findings should therefore be considered hypothesis-generating rather than causal, and should not be extrapolated to specific individual agents within either drug class. Confirmation in prospective, ideally randomized, studies incorporating measures of sepsis severity, individual-agent-level exposure data, and formal assessment of propensity score overlap will be necessary before these findings can inform clinical practice.
Funding Statement
This research received no external funding.
Data Sharing Statement
Data supporting the findings of this study are available in the article and its Supplementary Information. Source data may be obtained from the corresponding author (Chih-Cheng Lai) upon request. Patient data can also be access on TriNetX on https://trinetx.com/.
Ethics, Consent to Participate, and Consent to Publish Declarations
The study protocol was approved by the Institutional Review Board of Chi Mei Medical Center (IRB No. 11402-E02).
Author Contributions
All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
The authors declare no competing interests.
References
- 1.Prescott HC, Antonelli M, Alhazzani W. et al. Executive Summary: surviving Sepsis Campaign: international Guidelines for Management of Sepsis and Septic Shock 2026. Crit Care Med. 2026;54(4):725–13. doi: 10.1097/CCM.0000000000007089 [DOI] [PubMed] [Google Scholar]
- 2.GBD. 2021 Global Sepsis Collaborators. Global, regional, and national sepsis incidence and mortality, 1990-2021: a systematic analysis. Lancet Glob Health. 2025;13(12):e2013–e2026. doi: 10.1016/s2214-109x(25)00356-0 [DOI] [PubMed] [Google Scholar]
- 3.Kosyakovsky LB, Angriman F, Katz E, et al. Association between sepsis survivorship and long-term cardiovascular outcomes in adults: a systematic review and meta-analysis. Intensive Care Med. 2021;47(9):931–942. doi: 10.1007/s00134-021-06479-y [DOI] [PubMed] [Google Scholar]
- 4.Angriman F, Rosella LC, Lawler PR, Ko DT, Wunsch H, Scales DC. Sepsis hospitalization and risk of subsequent cardiovascular events in adults: a population-based matched cohort study. Intensive Care Med. 2022;48(4):448–457. doi: 10.1007/s00134-022-06634-z [DOI] [PubMed] [Google Scholar]
- 5.Arshad A, Ayaz A, Rehman S, Ukrani RD, Akbar I, Jamil B. Progression of Acute Kidney Injury to Chronic Kidney Disease in Sepsis Survivors: 1-Year Follow-Up Study. J Intensive Care Med. 2021;36(11):1366–1370. doi: 10.1177/0885066620956621 [DOI] [PubMed] [Google Scholar]
- 6.Ou SM, Chu H, Chao PW, et al. Long-Term Mortality and Major Adverse Cardiovascular Events in Sepsis Survivors. A Nationwide Population-based Study. Am J Respir Crit Care Med. 2016;194(2):209–217. doi: 10.1164/rccm.201510-2023OC [DOI] [PubMed] [Google Scholar]
- 7.Costantini E, Carlin M, Porta M, Brizzi MF. Type 2 diabetes mellitus and sepsis: state of the art, certainties and missing evidence. Acta Diabetol. 2021;58(9):1139–1151. doi: 10.1007/s00592-021-01728-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Zheng H, Yu F, Bei C, Zhou Y, Mo Z, Wei B. Diabetes mellitus and mortality in patients admitted to ICU with sepsis: a meta-analysis. Front Med. 2026;13(1743706). doi: 10.3389/fmed.2026.1743706 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Balintescu A, Lind M, Franko MA, et al. Glycemic Control and Risk of Sepsis and Subsequent Mortality in Type 2 Diabetes. Diabetes Care. 2022;45(1):127–133. doi: 10.2337/dc21-1050 [DOI] [PubMed] [Google Scholar]
- 10.Ng PY, Ng AK-Y, Ip A, Wu M-Z, Guo R, Yiu K-H. Risk of ICU Admission and Related Mortality in Patients With Sodium-Glucose Cotransporter 2 Inhibitors and Dipeptidyl Peptidase-4 Inhibitors: a Territory-Wide Retrospective Cohort Study. Crit Care Med. 2023;51(8):1074–1085. doi: 10.1097/ccm.0000000000005869 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Tang J, Jing B, Karunungan K, Badathala A, Wallace A, Legrand M. Association Between Gliflozins Use and Outcomes in Adults with Sepsis: a Multicenter Retrospective Cohort Study Among Veterans. Ann Intensive Care. 2026;16(100021). doi: 10.1016/j.aicoj.2025.100021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.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;29(1):452. doi: 10.1186/s13054-025-05685-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Zampieri FG, Meersch M, See E. Sodium-glucose transporter inhibitors 2 in acutely ill patients. Curr Opin Clin Nutr Metab Care. 2026;29(2):169–176. doi: 10.1097/mco.0000000000001194 [DOI] [PubMed] [Google Scholar]
- 14.Fan W, Zhang Q, Wang C, Sun J, Zhang J, Yin Y. GLP-1 as a regulator of sepsis outcomes: insights into cellular metabolism, inflammation, and therapeutic potential. Int Immunopharmacol. 2025;152(114390). doi: 10.1016/j.intimp.2025.114390 [DOI] [PubMed] [Google Scholar]
- 15.Shah FA, Mahmud H, Gallego-Martin T, Jurczak MJ, O’Donnell CP, McVerry BJ. Therapeutic Effects of Endogenous Incretin Hormones and Exogenous Incretin-Based Medications in Sepsis. J Clin Endocrinol Metab. 2019;104(11):5274–5284. doi: 10.1210/jc.2019-00296 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Han S, Liu Y, Xing B, et al. Association of glucagon-like peptide-1 receptor agonist use with risk of infections: a systematic review and meta-analysis. J Infect. 2025;91(5):106645. doi: 10.1016/j.jinf.2025.106645 [DOI] [PubMed] [Google Scholar]
- 17.Henney AE, Riley DR, Hydes TJ, et al. Comparative estimate of glucose-lowering therapies on risk of incident pneumonia and severe sepsis: an analysis of real-world cohort data. Thorax. 2024;80(1):32–41. doi: 10.1136/thorax-2024-221906 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Palchuk MB, London JW, Perez-Rey D, et al. A global federated real-world data and analytics platform for research. JAMIA Open. 2023;6(2):ooad035. doi: 10.1093/jamiaopen/ooad035 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Huang SC, Wu JY, Kuo CY, Hsu WH, Tsai YW, Lai CC. Association Between Glucagon-Like Peptide-1 Receptor Agonists and Clinical Outcomes in Patients with Chronic Obstructive Pulmonary Disease and Obesity. Drug Des Devel Ther. 2026;20(609054). doi: 10.2147/dddt.S609054 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Wu JY, Hsu WH, Kuo CC, et al. A retrospective analysis of combination therapy with GLP-1 receptor agonists and SGLT2 inhibitors versus SGLT2 inhibitor monotherapy in patients with MASLD. Nat Commun. 2025;16(1):7459. doi: 10.1038/s41467-025-62891-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wu VC, Chen JY, Lin YH, Wang CY, Lai CC. Assessing the cardiovascular events and clinical outcomes of COVID-19 on patients with primary aldosteronism. J Microbiol Immunol Infect. 2023;56(6):1158–1168. doi: 10.1016/j.jmii.2023.09.005 [DOI] [PubMed] [Google Scholar]
- 22.Haneuse S, VanderWeele TJ, Arterburn D. Using the E-Value to Assess the Potential Effect of Unmeasured Confounding in Observational Studies. Jama. 2019;321(6):602–603. doi: 10.1001/jama.2018.21554 [DOI] [PubMed] [Google Scholar]
- 23.Yoshida K, Solomon DH, Kim SC. Active-comparator design and new-user design in observational studies. Nat Rev Rheumatol. 2015;11(7):437–441. doi: 10.1038/nrrheum.2015.30 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Dimic N, Djuric M, Vejapi M, et al. The role of GLP-1 in the pathophysiology and treatment of sepsis: a narrative review. Front Med. 2025;12(1612034). doi: 10.3389/fmed.2025.1612034 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Wang Y, Deng F, Zhong X, et al. Dulaglutide provides protection against sepsis-induced lung injury in mice by inhibiting inflammation and apoptosis. Eur J Pharmacol. 2023;949(175730). doi: 10.1016/j.ejphar.2023.175730 [DOI] [PubMed] [Google Scholar]
- 26.Bray JJH, Foster-Davies H, Salem A, et al. Glucagon-like peptide-1 receptor agonists improve biomarkers of inflammation and oxidative stress: a systematic review and meta-analysis of randomised controlled trials. Diabetes Obes Metab. 2021;23(8):1806–1822. doi: 10.1111/dom.14399 [DOI] [PubMed] [Google Scholar]
- 27.Baer B, Putz ND, Riedmann K, et al. Liraglutide pretreatment attenuates sepsis-induced acute lung injury. Am J Physiol Lung Cell Mol Physiol. 2023;325(3):L368–L384. doi: 10.1152/ajplung.00041.2023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Jogani VG, Mohamed Elfadil O, Edakkanambeth Varayil J, Mundi MS. Exploring the role of glucagon-like peptide-1 receptor agonists in critical illness: mechanisms, benefits, and clinical implications. Curr Opin Crit Care. 2025;31(4):370–378. doi: 10.1097/mcc.0000000000001294 [DOI] [PubMed] [Google Scholar]
- 29.Wong CK, Drucker DJ. Antiinflammatory actions of glucagon-like peptide-1-based therapies beyond metabolic benefits. J Clin Invest. 2025;135(21). doi: 10.1172/jci194751 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Zhang X, Ren L, Li M, et al. Pleiotropic mechanisms and translational trajectory of SGLT2 inhibitors in sepsis-induced cardiomyopathy: navigating the evidence hierarchy. Eur J Pharmacol. 2026;1023(178858). doi: 10.1016/j.ejphar.2026.178858 [DOI] [PubMed] [Google Scholar]
- 31.Staszewsky L, Baviera M, Tettamanti M, et al. Insulin treatment in patients with diabetes mellitus and heart failure in the era of new antidiabetic medications. BMJ Open Diabetes Res Care. 2022;10(2):e002708. doi: 10.1136/bmjdrc-2021-002708 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data supporting the findings of this study are available in the article and its Supplementary Information. Source data may be obtained from the corresponding author (Chih-Cheng Lai) upon request. Patient data can also be access on TriNetX on https://trinetx.com/.