Comparative Efficacy of GLP‐1 RAs Versus SGLT2 Inhibitors in Patients With Acute Myocardial Infarction Undergoing Percutaneous Coronary Intervention: A Multicenter Propensity Score‐Matched Real‐World Study

ABSTRACT

Background

No head‐to‐head trials have directly compared GLP‐1 receptor agonists (GLP‑1 RAs) and SGLT2 inhibitors (SGLT2is) in patients with acute myocardial infarction (AMI) undergoing percutaneous coronary intervention (PCI).

Aims

To compare short‐, mid‐, and long‐term clinical outcomes of GLP‐1 RAs versus SGLT2 inhibitors in adults with AMI undergoing PCI, using a multicenter propensity score‐matched real‐world cohort from the TriNetX US Collaborative Network.

Methods

This multicenter, retrospective cohort study used de‐identified data from the TriNetX US Collaborative Network. Adults with AMI who underwent PCI and initiated either GLP‑1 RAs (n = 7201) or SGLT2is (n = 4252) within 14 days were included. One‐to‐one greedy nearest‐neighbor propensity score matching (caliper 0.1) yielded balanced cohorts across > 50 covariates, resulting in 1752 patients per group. Kaplan−Meier and Cox models assessed mortality, heart failure, hospitalization, recurrent myocardial infarction (RMI), stroke, atrial fibrillation, major adverse cardiovascular events (MACE), acute kidney injury (AKI), and cardiac arrest.

Results

Compared with SGLT2is, GLP‐1 RAs were associated with lower 1‐year risks of acute heart failure (HR 0.415 [0.343–0.501]), all‐cause hospitalization (HR 0.559 [0.495–0.631]), RMI (HR 0.799 [0.710–0.899]), stroke (HR 0.800 [0.667–0.959]), atrial fibrillation (HR 0.804 [0.693–0.932]), MACE (HR 0.788 [0.706–0.881]), and AKI (HR 0.534 [0.444–0.643]). At 30 days, absolute risk reduction was lower with GLP‐1 RAs for acute heart failure (7.19% [5.66%–8.72%]) and all‐cause hospitalization (16.61% [14.24%–18.98%]). All‐cause mortality did not differ at 30 or 90 days but was lower at 1 year (HR 0.700 [0.507−0.967]). Cardiac arrest did not differ significantly.

Conclusions

In this real‐world, propensity score‐matched study of patients with AMI undergoing PCI, GLP‐1 RA initiation linked to lower risk of short‐ and long‐term adverse outcomes than SGLT2i. These are observational findings needing confirmation in randomized trials.

Abbreviations

ACEI

angiotensin‐converting enzyme inhibitor

AKI

acute kidney injury

AMI

acute myocardial infarction

ARB

angiotensin receptor blocker

ARR

absolute risk reduction

BMI

body mass index

CI

confidence interval

CPT

current procedural terminology

CV

cardiovascular

eGFR

estimated glomerular filtration rate

EHR

electronic health record

GDPR

General Data Protection Regulation

GLP1‐RA/GLP‑1 RAs

glucagon‐like peptide‐1 receptor agonist(s)

GLP‐1

glucagon‐like peptide‐1

HCO

healthcare organization

HCPCS

Healthcare Common Procedure Coding System

HIPAA

Health Insurance Portability and Accountability Act

HR

hazard ratio

ICD‐10‐CM

International Classification of Diseases, 10th Revision, Clinical Modification

KM

Kaplan−Meier

LVEF

left ventricular ejection fraction

MACE

major adverse cardiovascular events

PCI

percutaneous coronary intervention

SGLT2

sodium‐glucose cotransporter‐2

SGLT2is

SGLT2 inhibitor(s)

SMD

standardized mean difference

STROBE

Strengthening the Reporting of Observational Studies in Epidemiology

1. Introduction

Despite significant advances in reperfusion strategies and secondary prevention therapies, acute myocardial infarction (AMI) remains a major cause of morbidity and death worldwide. Patients who survive AMI and undergo percutaneous coronary intervention (PCI) still face high risks of recurrent ischemic events, heart failure, arrhythmias, hospitalization, and early death. This emphasizes the need for additional cardioprotective treatments beyond those currently guideline‐recommended [1, 2]. Additionally, the burden of cardiometabolic disease has increased sharply globally, with drugs originally meant for blood sugar control now also recognized for their broad cardiovascular and renal benefits in individuals with and without diabetes [3, 4].

Glucagon‑like peptide‑1 receptor agonists (GLP‑1 RAs) and sodium‐glucose cotransporter‑2 inhibitors (SGLT‐2is) are the two major cardiovascular‑ and kidney‑targeted drug classes that have revolutionized management aiming at cardiometabolic risk reduction [5, 6, 7]. Large cardiovascular outcome trials have demonstrated that several GLP‑1 RAs reduce major adverse cardiovascular events (MACE) and provide modest benefits for kidney outcomes in patients with type 2 diabetes mellitus and established atherosclerotic cardiovascular disease or high cardiovascular risk [8, 9, 10]. SGLT2is, in turn, consistently lowers the risk of heart failure hospitalization (HHF) and progression of chronic kidney disease, with favorable effects on cardiovascular mortality across broad populations that now extend to individuals without diabetes [10, 11]. Based on this evidence, current diabetes and cardiology guidelines endorse both classes as essential initial treatments for patients with type 2 diabetes and atherosclerotic cardiovascular disease [12, 13, 14].

Despite these advancements, significant uncertainty remains about whether to prefer GLP‑1 RAs or SGLT2is for patients with AMI undergoing PCI. Most key trials focused on stable outpatients rather than those in the early post‐infarction phase, often excluding patients with recent MI or hemodynamic instability [8, 9, 10, 11, 15]. A recent meta‐analysis of RCTs found that initiating SGLT2i soon after MI was associated with significantly lower rates of future HHF and improved left‐ventricular ejection fraction (LVEF), with no notable impact on cardiovascular deaths or all‐cause mortality (ACM) compared to placebo [16]. Another meta‐analysis indicates that in AMI patients undergoing PCI, very early GLP‑1 RAs slightly improve LVEF and decrease infarct size, although there is no clear evidence of reduced short‐term clinical events [17]. Observational registries and cohort studies of both SGLT2i and GLP‑1 RAs in post‑AMI patients (often after PCI) generally show lower HF events and composite CV outcomes. Mortality and MACE benefits are especially consistent with GLP‐1 RAs, while the most reproducible signal with SGLT2i is a reduction in HF hospitalizations, with mortality benefits less uniform across studies [18, 19, 20]. However, there is limited head‐to‐head comparative effectiveness data for these two drug classes, and no randomized trial has directly compared GLP‑1 RAs with SGLT2is in post‐AMI PCI survivors. As a result, clinicians currently need to personalize therapy based on extrapolated evidence, comorbidities, and practical factors, rather than relying on solid comparative data specific to this high‐risk group.

Real‐world evidence from large electronic health record (EHR) networks provides an opportunity to fill this gap by enabling active‐comparator, new‐user analyses that approximate randomized comparisons in everyday clinical settings. We thus conducted a multicenter, propensity score‐matched cohort study using a large US EHR network to compare short‐, mid‐, and long‐term clinical outcomes linked to GLP‐1 RA versus SGLT2i initiation in adults with AMI undergoing PCI, regardless of diabetes status. We hypothesized that GLP‐1 RA therapy would be associated with different risks of ACM, heart failure, recurrent ischemic events, arrhythmias, and kidney injury compared with SGLT2i therapy in this vulnerable post‐PCI population.

2. Materials and Methods

2.1. Study Design and Data Source

We conducted a multicenter, retrospective, observational, active‐comparator, new‐user cohort study using the TriNetX US Collaborative Network (TriNetX, Cambridge, MA, USA), a federated health research platform aggregating de‐identified EHRs from 67 healthcare organizations across the United States as of January 2026 [21]. The network provides longitudinal, real‐time access to comprehensive patient‐level data, including diagnoses (coded via International Classification of Diseases, Tenth Revision, Clinical Modification [ICD‐10‐CM] and Unified Medical Language System [UMLS]), procedures (coded via ICD‐10‐PCS, Current Procedural Terminology [CPT], and Healthcare Common Procedure Coding System [HCPCS]), medications (coded via RxNorm), laboratory values, vital signs, and demographic information [22]. Patient privacy is maintained through cryptographic hashing, and the platform fully complies with the Health Insurance Portability and Accountability Act (HIPAA) and General Data Protection Regulation (GDPR). All analyses were performed on aggregated, de‐identified data within the TriNetX Analytics suite, eliminating the need for institutional review board approval. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines [23]. The analysis was executed on January 4, 2026.

2.2. Study Population

Adult patients (aged ≥ 18 years) were included if they had a recorded diagnosis of AMI (ICD‐10‐CM code I21) and underwent PCI, regardless of diabetes status, identified by a comprehensive set of ICD‐10‐PCS procedure codes for coronary artery dilation (full code listed in Appendix S1). The index event was defined as the first recorded PCI procedure meeting these criteria, with the index date serving as the day of the procedure. Patients were assigned to two mutually exclusive cohorts based on medication exposure initiated within 14 days following the index PCI:

• Cohort 1 (GLP‑1 RAs PCI): Patients with a recorded prescription or administration of a GLP‐1 RAs (semaglutide [RxNorm 1991302] or tirzepatide [RxNorm 2601723]).

• Cohort 2 (SGLT2is PCI): Patients with a recorded prescription or administration of an SGLT2 inhibitor (dapagliflozin [RxNorm 1488564] or empagliflozin [RxNorm 1545653]).

To ensure new‐user design and minimize exposure misclassification, patients were excluded if they had any prior recorded use of the comparator class (i.e., GLP‑1 RAs excluded from the SGLT2is cohort, and SGLT2is excluded from the GLP‑1 RAs cohort). No minimum look‐back period for continuous observation was required, but index events were restricted to those occurring within the past 20 years to ensure contemporary relevance.

2.3. Outcome Measures

Outcomes were assessed across three distinct time windows, each beginning 1 day after the index PCI date: short‐term (30 days), mid‐term (90 days), and long‐term (1 year). All outcomes were evaluated using Kaplan−Meier (KM) survival analysis, with patients included regardless of whether the outcome occurred prior to the respective time window, unless otherwise specified. For each outcome, event counts, event‐free survival probabilities at the end of the time window, hazard ratios (HRs) with 95% confidence intervals (CIs), and log‐rank p values were calculated to compare the GLP‑1 RAs PCI and SGLT2is PCI cohorts (Central Illustration iillustration 1, 1).

Central iillustration 1.

First real‐world head‐to‐head: GLP‐1 receptor agonists (GLP‐1 RAs) versus SGLT2 inhibitors (SGLT2is) in patients with acute myocardial infarction (AMI) undergoing percutaneous coronary intervention (PCI). Early and sustained benefits of GLP‐1 RAs across multiple cardiovascular, renal, and hospitalization endpoints in this high‐risk post‐AMI PCI cohort, with emerging mortality advantage by 1 year. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 1.

Patient selection flow diagram illustrating the step‐by‐step identification, inclusion, exclusion, and propensity score matching process for the GLP‑1 RAs PCI cohort (n = 7201 before matching) and SGLT2is PCI comparator cohort (n = 4252 before matching) from the TriNetX US Collaborative Network, resulting in 1:1 matched cohorts of 1752 patients each. [Color figure can be viewed at wileyonlinelibrary.com]

ACM was ascertained primarily through TriNetX‐integrated linkage to national death registries, supplemented by diagnosis codes for ill‐defined or unknown cause of mortality (ICD‐10‐CM R99). Acute heart failure events were identified using ICD‐10‐CM codes indicative of acute or acute‐on‐chronic presentations, including I50.21 (acute systolic congestive heart failure), I50.23 (acute on chronic systolic congestive heart failure), I50.31 (acute diastolic congestive heart failure), I50.33 (acute on chronic diastolic congestive heart failure), I50.41 (acute combined systolic and diastolic congestive heart failure), I50.43 (acute on chronic combined systolic and diastolic congestive heart failure), I50.811 (acute right heart failure), and I50.813 (acute on chronic right heart failure). All‐cause hospitalization was captured using a comprehensive set of procedure and service codes encompassing inpatient admissions, observation stays, and discharge management, including relevant CPT codes (e.g., 99234–99236 for hospital inpatient or observation care with admission and discharge on the same date, 99238–99239 for discharge day management), HCPCS code G0378 (hospital observation service per hour), and SNOMED code 50699000 (hospital admission, short‐term).

Recurrent myocardial infarction (RMI) was defined by ICD‐10‐CM codes I21 (AMI) and I22 (subsequent ST elevation and non‐ST elevation myocardial infarction). Stroke events were identified using the broad range of ICD‐10‐CM codes I60–I69, encompassing all cerebrovascular diseases. Atrial fibrillation (Afib) was ascertained via ICD‐10‐CM code I48 (Afib and flutter). The composite MACE endpoint included myocardial infarction (MI) (I21, I22), nontraumatic intracranial hemorrhage (I61, I62), cerebral infarction (I63), and cardiac arrest (I46.2, I46.9). Acute kidney injury (AKI) was defined by ICD‐10‐CM code N17 (acute kidney failure). Cardiac arrest was identified using ICD‐10‐CM code I46 (cardiac arrest, including I46.2 for cardiac arrest due to an underlying cardiac condition and I46.9 for cardiac arrest with unspecified cause).

2.4. Covariates and Propensity Score Matching

Baseline covariates were assessed in the 365‐day period preceding and including the index PCI date. These included demographics (age, sex, race/ethnicity), comorbidities (Afib/flutter [I48], cardiomyopathy [I42], hypertensive diseases [I10–I1A], ischemic heart diseases [I20–I25], chronic kidney disease [N18], diabetes mellitus [E08–E13], disorders of lipoprotein metabolism [E78], etc.), medication classes (β‐blockers, loop diuretics, thiazides, ACE inhibitors, angiotensin II inhibitors, antiarrhythmics, anticoagulants, antiplatelet agents, insulin, oral hypoglycemics, etc.), and laboratory parameters (sodium, potassium, creatinine, glucose, hemoglobin, albumin, prothrombin time, lipids, troponin I, natriuretic peptides, hemoglobin A1c, C‐reactive protein, left ventricular ejection fraction [LVEF], heart rate, body weight, body mass index [BMI], blood pressure).

Propensity scores were estimated using multivariable logistic regression incorporating all listed covariates. One‐to‐one greedy nearest‐neighbor matching without replacement was performed with a caliper width of 0.1 on the logit scale of the propensity score. Post‐matching balance was evaluated using standardized mean differences (SMD); an SMD < 0.10 was targeted and largely achieved for most variables (p > 0.05 for nearly all comparisons). Propensity score density plots confirmed overlapping distributions post‐matching, supporting comparability between cohorts.

2.5. Statistical Analysis

Time‐to‐event analyses were performed using KM estimates with log‐rank tests for between‐cohort comparisons. Patients were censored at the date of the last recorded fact in their EHR. HRs with 95% CIs were derived from Cox proportional hazards models, with the GLP‑1 RAs PCI cohort as the reference group. Proportional hazards assumptions were evaluated graphically and via Schoenfeld residuals (with proportionality test p values reported for each outcome and time window). Absolute risk differences were calculated in addition to HRs. No formal competing risks analysis was performed for mortality, given low competing event rates and primary focus on ACM. Survival probability at the end of each time window and the number of events were summarized. All analyses were conducted within the TriNetX Analytics suite using built‐in algorithms. Two‐tailed p < 0.05 was considered statistically significant. No imputation was performed for missing data, as the federated platform uses observed records only.

3. Results

3.1. Cohort Selection and Propensity Score Matching

This retrospective multicenter study identified 7201 patients initiating GLP‑1 RAs PCI and 4252 initiating SGLT2is PCI following AMI with PCI from the TriNetX US Collaborative Network, regardless of diabetes status. Pre‐matching, the cohorts exhibited substantial imbalances across multiple domains (Figure iillustration 1, 1). Patients in the GLP‑1 RAs PCI cohort were younger (mean age 64.2 ± 11.0 vs. 69.1 ± 11.3 years; SMD 0.443), less likely to be male (58.9% vs. 72.9%; SMD 0.299), and showed differences in race/ethnicity distribution. Comorbidities were also markedly imbalanced: Afib/flutter was less prevalent in the GLP‑1 RAs group (13.0% vs. 31.4%; SMD 0.454), as was cardiomyopathy (5.4% vs. 24.9%; SMD 0.564), chronic kidney disease (19.5% vs. 36.8%; SMD 0.392), and pulmonary heart disease. Medication use differed significantly, with lower utilization of β‐blockers (68.8% vs. 86.9%; SMD 0.446), loop diuretics (22.5% vs. 60.1%; SMD 0.824), anticoagulants, and oral hypoglycemic agents (28.8% vs. 80.6%; SMD 1.218) in the GLP‑1 RAs cohort. Laboratory parameters showed additional imbalances, including higher mean BMI (35.5 ± 6.8 vs. 29.5 ± 6.4 kg/m²; SMD 0.922), higher body weight, and differences in natriuretic peptides, hemoglobin, and other markers (many SMDs > 0.2; all p < 0.001 for key variables).

One‐to‐one greedy nearest‐neighbor propensity score matching with a caliper of 0.1 was performed on demographics (age, sex, race/ethnicity), comorbidities (Afib, cardiomyopathy, hypertension, ischemic heart disease, chronic kidney disease, diabetes, etc.), medication classes (β‐blockers, loop diuretics, ACE inhibitors, anticoagulants, antiplatelet agents, etc.), BMI, and available laboratory values (sodium, potassium, creatinine, glucose, hemoglobin A1c, lipids, etc.). After matching, balanced cohorts of 1752 patients each were obtained. Post‐matching, excellent covariate balance was achieved for the vast majority of variables (most SMDs < 0.10; p > 0.05 for nearly all comparisons). Mean age was nearly identical (67.9 ± 10.4 vs. 67.7 ± 11.3 years; SMD 0.013), sex distribution was similar (male 67.4% vs. 69.1%; SMD 0.037), and race/ethnicity proportions aligned closely. Key comorbidities showed minimal residual differences: Afib/flutter (21.7% vs. 22.4%; SMD 0.018), chronic kidney disease (29.6% vs. 30.2%; SMD 0.012), diabetes mellitus (72.5% vs. 71.4%; SMD 0.024), and ischemic heart disease (92.2% vs. 92.9%; SMD 0.024). Medication utilization was well‐balanced across major classes, including β‐blockers (77.1% vs. 78.5%; SMD 0.033), anticoagulants, antiplatelet agents, and oral hypoglycemic agents (57.3% vs. 56.0%; SMD 0.026). Laboratory values also improved in balance, though minor residual differences remained for some parameters such as creatinine (SMD 0.236), sodium (SMD 0.136), and potassium (SMD 0.208). BMI remained moderately imbalanced (34.9 ± 7.0 vs. 30.3 ± 6.5 kg/m²; SMD 0.685), reduced substantially from pre‐matching levels but still notable. The propensity score density plots demonstrated clear overlap post‐matching (cohort 1 in purple, cohort 2 in green as shown in Figure 2), confirming adequate comparability between the GLP‑1 RAs PCI and SGLT2is PCI groups for unbiased outcome comparison.

Figure 2.

Propensity score distribution before and after 1:1 matching. Left panel (before matching): Overlapping but markedly shifted propensity score distributions, with the GLP‑1 RAs PCI cohort skewed toward lower scores (purple) and the SGLT2is PCI cohort toward higher scores (green), reflecting baseline imbalances and confounding by indication. Right panel (after matching): Excellent overlap with nearly identical distributions in the propensity score‐matched cohorts (n = 1752 per group), demonstrating successful balance across measured covariates using a 0.1‐standard deviation caliper. [Color figure can be viewed at wileyonlinelibrary.com]

After propensity score matching, the final analytic cohort consisted of 1752 patients in the GLP‑1 RAs PCI group and 1752 patients in the SGLT2is PCI group, with baseline characteristics well balanced across most measured covariates (Table 1). Post‐matching balance was good for most covariates (SMD < 0.10 for the majority; Love plot in Figure 3). However, moderate residual imbalances persisted for BMI (SMD 0.685) and selected laboratory parameters (e.g., creatinine SMD 0.236, sodium SMD 0.136, potassium SMD 0.208). These imbalances were further addressed through multivariate adjustment (Table 5). Forest plot and KM survival analyses were performed for each outcome at 30 days, 90 days, and 1‐year follow‐up, revealing consistent patterns favoring GLP‑1 RAs PCI across multiple endpoints (Figures 4, 5, 6, 7, 8, 9). Event counts, KM event‐free survival probabilities, HRs with 95% CIs, and log‐rank p values are summarized in Tables 2, 3, and 4 for short‐term (30 days), mid‐term (90 days), and long‐term (1 year) outcomes, respectively, highlighting the temporal evolution of treatment effects.

Table 1.

Baseline characteristics and comparison of demographics, comorbidities, medication use, and laboratory/vital sign parameters in patients initiating GLP1‐RAs PCI versus SGLT2is PCI before and after 1:1 propensity score matching (n = 7201 and n = 4252 before matching; n = 1752 per group after matching).

	Before propensity score‐matching						After propensity score‐matched patients
Characteristic name	GLP1‐RAs PCI cohort (n  = 7201)		SGLT2is cohort (n  = 4252)		p value	Standardized mean difference (SMD)	GLP1‐RAs PCI cohort (n  = 1752)		SGLT2is cohort (n  = 1752)		p value	Standardized mean difference (SMD)
Current age, mean (SD)	64.1676	10.9699	69.1014	11.3011	< 0.0001	0.4430	67.8858	10.4332	67.7449	11.2847	0.7010	0.0130
Age at index, mean (SD)	62.1683	10.8902	67.0640	11.3482	< 0.0001	0.4402	65.7300	10.4000	65.5668	11.3029	0.6565	0.0150
Female, n (%)	2952	40.99%	1147	26.98%	< 0.0001	0.2993	568	32.42%	539	30.77%	0.2920	0.0356
Male, n (%)	4241	58.90%	3100	72.91%	< 0.0001	0.2989	1181	67.41%	1211	69.12%	0.2762	0.0368
Unknown gender, n (%)	10	0.14%	10	0.24%	0.2330	0.0223	10	0.57%	10	0.57%	1.0000	< 0.0001
Ethnicity, n (%)
Not Hispanic or Latino	5646	78.41%	3336	78.46%	0.9485	0.0013	1362	77.74%	1372	78.31%	0.6833	0.0138
Hispanic or Latino	427	5.93%	330	7.76%	0.0001	0.0726	135	7.71%	137	7.82%	0.8995	0.0043
Unknown ethnicity	1128	15.66%	586	13.78%	0.0064	0.0532	255	14.56%	243	13.87%	0.5615	0.0196
Race, n (%)
White	5777	80.23%	3073	72.27%	< 0.0001	0.1877	1303	74.37%	1293	73.80%	0.6998	0.0130
Black or African American	724	10.05%	603	14.18%	< 0.0001	0.1267	207	11.82%	220	12.56%	0.5020	0.0227
Unknown race	200	2.78%	149	3.50%	0.0288	0.0417	61	3.48%	60	3.43%	0.9263	0.0031
Asian	200	2.78%	181	4.26%	< 0.0001	0.0804	87	4.97%	77	4.40%	0.4238	0.0270
Native Hawaiian or Other Pacific Islander	58	0.81%	59	1.39%	0.0028	0.0559	20	1.14%	19	1.08%	0.8721	0.0054
American Indian or Alaska Native	43	0.60%	32	0.75%	0.3190	0.0190	10	0.57%	15	0.86%	0.3156	0.0339
Diagnosis, n (%)
Hypertensive diseases	5795	80.48%	3702	87.07%	< 0.0001	0.1794	1491	85.10%	1483	84.65%	0.7060	0.0127
Disorders of lipoprotein metabolism and other lipidemias	5737	79.67%	3588	84.38%	< 0.0001	0.1230	1448	82.65%	1440	82.19%	0.7226	0.0120
Diabetes mellitus	4114	57.13%	2627	61.78%	< 0.0001	0.0949	1270	72.49%	1251	71.40%	0.4749	0.0241
Ischemic heart diseases	6389	88.72%	4085	96.07%	< 0.0001	0.2800	1616	92.24%	1627	92.87%	0.4791	0.0239
Chronic kidney disease (CKD)	1402	19.47%	1564	36.78%	< 0.0001	0.3924	519	29.62%	529	30.19%	0.7122	0.0125
Atrial fibrillation and flutter	935	12.98%	1334	31.37%	< 0.0001	0.4539	380	21.69%	393	22.43%	0.5964	0.0179
Cerebrovascular diseases	779	10.82%	803	18.89%	< 0.0001	0.2283	268	15.30%	269	15.35%	0.9626	0.0016
Other peripheral vascular diseases	611	8.49%	689	16.20%	< 0.0001	0.2363	242	13.81%	242	13.81%	1.0000	< 0.0001
Cardiomyopathy	390	5.42%	1057	24.86%	< 0.0001	0.5636	208	11.87%	205	11.70%	0.8751	0.0053
Chronic rheumatic heart diseases	396	5.50%	806	18.96%	< 0.0001	0.4197	169	9.65%	182	10.39%	0.4645	0.0247
Pulmonary heart disease and diseases of pulmonary circulation	381	5.29%	773	18.18%	< 0.0001	0.4087	162	9.25%	165	9.42%	0.8617	0.0059
Medications, n (%)
Antilipemic agents	5552	77.10%	3814	89.70%	< 0.0001	0.3436	1460	83.33%	1466	83.68%	0.7848	0.0092
Beta blockers/related	4952	68.77%	3693	86.85%	< 0.0001	0.4459	1351	77.11%	1375	78.48%	0.3293	0.0330
Platelet aggregation inhibitors	4504	62.55%	3571	83.98%	< 0.0001	0.4992	1287	73.46%	1310	74.77%	0.3750	0.0300
Anticoagulants	3249	45.12%	3332	78.36%	< 0.0001	0.7279	1113	63.53%	1130	64.50%	0.5496	0.0202
Oral hypoglycemic agents, oral	2077	28.84%	3428	80.62%	< 0.0001	1.2179	1004	57.31%	981	55.99%	0.4330	0.0265
Antianginals	3283	45.59%	2770	65.15%	< 0.0001	0.4012	963	54.97%	990	56.51%	0.3585	0.0310
Antiarrhythmics	3260	45.27%	2886	67.87%	< 0.0001	0.4683	958	54.68%	998	56.96%	0.1736	0.0460
Insulin	2533	35.18%	2148	50.52%	< 0.0001	0.3138	870	49.66%	901	51.43%	0.2949	0.0354
Angiotensin II inhibitor	2297	31.90%	2545	59.85%	< 0.0001	0.5845	741	42.30%	768	43.84%	0.3570	0.0311
Calcium channel blockers	2642	36.69%	1911	44.94%	< 0.0001	0.1685	709	40.47%	732	41.78%	0.4297	0.0267
Loop diuretics	1623	22.54%	2554	60.07%	< 0.0001	0.8244	685	39.10%	697	39.78%	0.6783	0.0140
Metformin	1665	23.12%	869	20.44%	0.0008	0.0651	651	37.16%	645	36.82%	0.8337	0.0071
ACE inhibitors	1862	25.86%	1182	27.80%	0.0231	0.0438	512	29.22%	509	29.05%	0.9112	0.0038
Potassium sparing/combinations diuretics	601	8.35%	1804	42.43%	< 0.0001	0.8510	340	19.41%	339	19.35%	0.9659	0.0014
Thiazides/related diuretics	1041	14.46%	614	14.44%	0.9812	0.0005	255	14.56%	242	13.81%	0.5290	0.0213
Alpha blockers/related	711	9.87%	630	14.82%	< 0.0001	0.1507	230	13.13%	232	13.24%	0.9205	0.0034
Glipizide	357	4.96%	251	5.90%	0.0292	0.0417	171	9.76%	166	9.48%	0.7745	0.0097
Laboratory values, mean (SD)
Heart rate, bpm	74.8570	13.9732	75.9398	15.5123	0.0007	0.0733	74.7663	14.4837	76.3361	15.0409	0.0057	0.1063
Body weight, lbs (pounds)	228.1298	50.9413	192.0783	47.5091	< 0.0001	0.7319	226.0663	52.1756	195.1400	48.6603	< 0.0001	0.6130
BMI, kg/m2	35.5461	6.8141	29.4520	6.3989	< 0.0001	0.9220	34.9377	7.0311	30.2939	6.5270	0.0001	0.6846
Systolic blood pressure, mmHg	129.0681	17.9548	121.5944	20.5010	< 0.0001	0.3878	128.6371	18.9140	125.3011	20.6201	< 0.0001	0.1686
Diastolic blood pressure, mmHg	74.7746	11.6620	69.9070	12.5206	< 0.0001	0.4023	73.2684	11.6813	71.5059	12.6355	0.0001	0.1448
Glucose, mg/dL	143.8543	65.2806	137.2861	62.3390	< 0.0001	0.1029	150.9206	67.5345	151.3280	72.0988	0.8712	0.0058
Creatinine, mg/dL	1.3216	3.9639	1.1977	0.5820	0.0523	0.0437	1.3748	1.4313	1.1206	0.5173	< 0.0001	0.2363
Potassium, mmol/L	4.2745	0.4535	4.1851	0.4942	< 0.0001	0.1886	4.2866	0.4803	4.1854	0.4939	< 0.0001	0.2079
Sodium, mmol/L	138.7069	2.9616	137.9914	3.3589	< 0.0001	0.2260	138.3922	3.1347	137.9582	3.2401	0.0002	0.1361
Albumin, g/dL	4.0332	0.4694	3.7729	0.5712	< 0.0001	0.4980	3.9651	0.5111	3.8787	0.5635	< 0.0001	0.1607
Prothrombin time (PT) in plasma or blood	13.2218	3.8694	14.4604	4.7817	< 0.0001	0.2848	13.6334	4.1511	13.9683	4.5465	0.1461	0.0769
Natriuretic peptide B, pg/mL	448.8215	2240.4097	1197.1365	3016.4385	< 0.0001	0.2816	607.8028	3081.7275	807.4425	2009.9113	0.2948	0.0767
Natriuretic peptide. B prohormone N‐Terminal, pg/mL	2636.2220	8713.7250	4308.1740	6823.2050	< 0.0001	0.2136	3100.4258	9006.3910	2619.0652	4026.5588	0.3610	0.0690
Left ventricular ejection fraction (LVEF) (%)	57.2110	9.7844	40.8439	14.7342	< 0.0001	1.3087	56.0877	11.0670	45.6941	14.1013	< 0.0001	0.8200
Troponin I (cardiac), ng/mL	4.9980	14.8403	7.9949	27.1925	0.0109	0.1368	5.7105	16.0907	9.7541	33.8489	0.0975	0.1526
C‐reactive protein, mg/L	28.3027	51.7068	40.6249	58.6237	0.0001	0.2229	34.6549	51.1516	35.2723	53.8222	0.9063	0.0118
Hemoglobin, g/dL	13.3822	2.0191	12.7280	2.3468	< 0.0001	0.2988	13.0993	2.0897	13.0813	2.2740	0.8262	0.0082
Hemoglobin A1c %	7.3064	1.9082	7.2543	1.9632	0.2405	0.0269	7.6915	2.0044	7.7455	2.0336	0.5012	0.0268
HDL cholesterol, mg/dL	39.7182	14.3465	37.4673	17.2155	< 0.0001	0.1420	38.4229	13.9419	38.3097	15.6099	0.8502	0.0076
Cholesterol, mg/dL	148.6684	47.0307	147.9482	49.6201	0.5148	0.0149	144.2391	45.2085	153.7042	50.6124	< 0.0001	0.1972
Triglyceride, mg/dL	172.5905	136.5725	142.4982	102.4645	< 0.0001	0.2493	164.7369	114.2271	165.6065	121.2207	0.8552	0.0074
LDL, cholesterol, mg/dL	75.4200	37.7607	79.6254	42.0571	< 0.0001	0.1052	73.4645	36.8457	81.7184	42.0636	< 0.0001	0.2087

Figure 3.

Love plot: Covariate balance before and after 1:1 propensity score matching between the GLP‐1 receptor agonist (GLP1‐RA) and SGLT2 inhibitor cohorts in patients with acute myocardial infarction undergoing percutaneous coronary intervention. Absolute standardized mean differences (SMDs) are shown for key covariates. Red circles represent pre‐matching imbalance, and teal diamonds represent post‐matching balance. The dashed vertical line at SMD = 0.1 indicates the threshold for good balance. Post‐matching, excellent balance (SMD). [Color figure can be viewed at wileyonlinelibrary.com]

Table 5.

Multivariate Cox proportional hazards model for all‐cause mortality in the propensity score‐matched cohort. The model adjusted for key demographic and clinical covariates, including sex, chronic kidney disease, type 2 diabetes mellitus, hypertension, dyslipidemias, body mass index (BMI) categories, and left ventricular ejection fraction (LVEF) categories.

Covariate	Adjusted hazard ratio (aHR)	95% confidence interval	Coefficient	Standard error	z	p value
GLP1‐RAs cohort	0.391	(0.320, 0.478)	−0.939	0.103	9.130	0.0001
Male	1.060	(0.884, 1.270)	0.058	0.092	0.626	0.531
Chronic kidney disease (CKD)	2.831	(2.367, 3.386)	1.041	0.091	11.384	0.0001
Left ventricular ejection fraction (LVEF) (%)
< 40%	1.340	(1.023, 1.755)	0.293	0.138	2.123	0.034
40%−50%	0.804	(0.536, 1.205)	−0.218	0.206	−1.057	0.291
50%−60%	1.168	(0.819, 1.665)	0.155	0.181	0.858	0.391
60%−70%	1.015	(0.668, 1.541)	0.015	0.213	0.069	0.945
> 70%	0.958	(0.420, 2.184)	−0.043	0.421	−0.102	0.919
Type 2 diabetes mellitus	1.463	(1.195, 1.793)	0.381	0.103	3.679	0.0001
BMI
< 20 kg/m2	2.266	(1.695, 3.030)	0.818	0.148	5.517	0.0001
20−25 kg/m2	1.460	(1.162, 1.833)	0.378	0.116	3.256	0.001
2−30 kg/m2	0.953	(0.785, 1.158)	−0.048	0.099	−0.481	0.630
30−35 kg/m2	1.040	(0.861, 1.255)	0.039	0.096	0.405	0.686
> 35 kg/m2	1.018	(0.823, 1.260)	0.018	0.109	0.169	0.866
Hypertension	1.448	(0.982, 2.136)	0.370	0.198	1.866	0.062
Dyslipidemias	0.831	(0.624, 1.107)	−0.185	0.146	−1.265	0.206

Figure 4.

Forest plot displaying HRs (GLP‑1 RAs PCI vs. SGLT2is PCI) with 95% CIs in propensity score‐matched cohorts (n = 1752 per group) at 30‐day follow‐up. Outcomes are sorted by increasing HR. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 5.

Kaplan–Meier curves comparing event‐free survival for GLP‑1 RAs PCI versus SGLT2is PCI in propensity score‐matched cohorts (n = 1752 per group) at 30‐day follow‐up. Higher curves indicate superior event‐free survival with GLP‑1 RAs PCI for most endpoints (log‐rank p values provided per panel). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 6.

Forest plot displaying HRs (GLP‑1 RAs PCI vs. SGLT2is PCI) with 95% CIs in propensity score‐matched cohorts (n = 1752 per group) at 90‐day follow‐up. Outcomes are sorted by increasing HR. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 7.

Kaplan–Meier curves comparing event‐free survival for GLP‑1 RAs PCI versus SGLT2is PCI in propensity score‐matched cohorts (n = 1752 per group) at 90‐day follow‐up. Higher curves indicate superior event‐free survival with GLP‑1 RAs PCI for most endpoints (log‐rank p values provided per panel). [Color figure can be viewed at wileyonlinelibrary.com]

Figure 8.

Forest plot displaying HRs (GLP‑1 RAs PCI vs. SGLT2is PCI) with 95% CIs in propensity score‐matched cohorts (n = 1752 per group) at 1‐year follow‐up. Outcomes are sorted by increasing HR. [Color figure can be viewed at wileyonlinelibrary.com]

Figure 9.

Kaplan–Meier curves comparing event‐free survival for GLP‑1 RAs PCI versus SGLT2is PCI in propensity score‐matched cohorts (n = 1752 per group) at 1‐year follow‐up. Higher curves indicate superior event‐free survival with GLP‑1 RAs PCI for most endpoints (log‐rank p values provided per panel). [Color figure can be viewed at wileyonlinelibrary.com]

Table 2.

Propensity score‐matched comparison of short‐term outcomes: GLP1‐RAs PCI versus SGLT2is PCI cohort at 30‐day follow‐up.

	Outcome	GLP1‐RAs PCI cohort events (n = 1752)	SGLT2is PCI cohort events (n = 1752)	KM survival probability at end of follow‐up GLP1‐RAs PCI cohort	KM survival probability at end of follow‐up SGLT2is PCI cohort	Hazard ratio (GLP1‐RAs vs. SGLT2is)	95% CI	Log‐rank p value	Risk difference (GLP1RAs ‐SGLT2is)	95% CI for risk difference	Proportionality test p value	Proportionality Test χ 2 (df = 1)
1	All‐cause mortality (ACM)	15	12	99.09%	99.27%	1.233	0.577–2.635	0.587	+0.0017	(−0.0041, 0.0075)	0.133	2.253
2	Acute heart failure	39	165	97.68%	90.17%	0.224	0.158–0.317	< 0.001	−0.0719	(−0.0872, −0.0566)	0.006	7.682
3	All‐cause hospitalization	136	427	92.05%	75.32%	0.290	0.239–0.351	< 0.001	−0.1661	(−0.1898, −0.1424)	0.001	30.759
4	Recurrent myocardial infarction (RMI)	205	308	87.90%	81.67%	0.628	0.526–0.749	< 0.001	−0.0588	(−0.0821, −0.0355)	0.003	8.875
5	Stroke	61	100	96.38%	94.01%	0.592	0.431–0.814	0.001	−0.0223	(−0.0361, −0.0084)	0.264	1.249
6	Atrial fibrillation (Afib)	128	214	92.41%	87.21%	0.568	0.456–0.707	< 0.001	−0.0491	(−0.0687, −0.0295)	0.002	9.816
7	MACE	233	348	86.25%	79.30%	0.626	0.530–0.739	< 0.001	−0.0656	(−0.0902, −0.0411)	0.001	10.243
8	Acute kidney injury (AKI)	40	111	97.63%	93.33%	0.347	0.242–0.498	< 0.001	−0.0405	(−0.0539, −0.0271)	0.589	0.293
9	Cardiac arrest	12	18	99.29%	98.93%	0.654	0.315–1.358	0.251	−0.0034	(−0.0095, 0.0027)	0.285	1.145

Table 3.

Propensity score‐matched comparison of mid‐term outcomes: GLP1‐RAs PCI versus SGLT2is PCI cohort at 90 days follow‐up.

	Outcome	GLP1‐RAs PCI cohort events (n = 1752)	SGLT2is PCI cohort events (n = 1752)	KM survival probability at end of follow‐up GLP1‐RAs PCI cohort	KM survival probability at end of follow‐up SGLT2is PCI cohort	Hazard ratio (GLP1‐RAs vs. SGLT2is)	95% CI	Log‐rank p value	Proportionality test p value	Proportionality test χ 2 (df = 1)
1	All‐cause mortality (ACM)	26	30	98.37%	98.09%	0.854	0.505–1.444	0.556	0.842	0.040
2	Acute heart failure	73	233	95.51%	85.74%	0.290	0.223–0.378	< 0.001	0.003	8.870
3	All‐cause hospitalization	223	494	86.53%	71.04%	0.400	0.342–0.469	< 0.001	0.000	61.224
4	Recurrent myocardial infarction (RMI)	317	420	80.84%	74.47%	0.701	0.606–0.812	< 0.001	0.002	9.848
5	Stroke	113	145	93.07%	91.08%	0.751	0.587–0.960	0.022	0.009	6.828
6	Atrial fibrillation (Afib)	222	293	86.43%	82.06%	0.712	0.598–0.848	< 0.001	0.000	12.853
7	MACE	361	479	78.21%	70.87%	0.693	0.604–0.794	< 0.001	0.002	9.539
8	Acute kidney injury (AKI)	75	177	95.39%	89.04%	0.401	0.306–0.525	< 0.001	0.078	3.110
9	Cardiac arrest	19	25	98.84%	98.46%	0.744	0.410–1.351	0.330	0.224	1.476

Table 4.

Propensity score‐matched comparison of long‐term outcomes: GLP1‐RAs PCI versus SGLT2is PCI cohort at 1‐year follow‐up.

	Outcome	GLP1‐RAs PCI cohort events (n = 1752)	SGLT2is PCI cohort events (n = 1752)	KM survival probability at end of follow‐up GLP1‐RAs PCI cohort	KM survival probability at end of follow‐up SGLT2is PCI cohort	Hazard ratio (GLP1‐RAs vs. SGLT2is)	95% CI	Log‐rank test p value	Proportionality test p value	Proportionality Test χ 2 (df = 1)
1	All‐cause mortality (ACM)	63	89	95.39%	93.30%	0.700	0.507–0.967	0.029	0.499	0.457
2	Acute heart failure	158	345	88.86%	76.76%	0.415	0.343–0.501	< 0.001	0.001	14.537
3	All‐cause hospitalization	430	669	70.23%	57.56%	0.559	0.495–0.631	< 0.001	0.001	56.294
4	Recurrent myocardial infarction (RMI)	515	595	65.71%	60.80%	0.799	0.710–0.899	< 0.001	0.002	9.148
5	Stroke	213	257	85.40%	82.06%	0.800	0.667–0.959	0.016	0.897	0.017
6	Atrial fibrillation (Afib)	325	381	78.63%	75.08%	0.804	0.693–0.932	0.004	0.005	8.036
7	MACE	583	677	61.43%	55.61%	0.788	0.706–0.881	< 0.001	0.005	8.025
8	Acute kidney injury (AKI)	177	309	87.52%	78.54%	0.534	0.444–0.643	< 0.001	0.087	2.935
9	cardiac arrest	31	47	97.86%	96.72%	0.645	0.410–1.015	0.056	0.809	0.058

3.2. All‐Cause Mortality

GLP‑1 RAs were associated with a numerically higher ACM risk at 30 days compared with SGLT2is (15 vs. 12 events; KM event‐free survival 99.09% vs. 99.27%; HR 1.233, 95% CI 0.577–2.635; log‐rank p = 0.587). The absolute number of events remained very low in both groups during this early post‐procedural period, and the wide CI reflected limited statistical power for detecting differences at such low event rates. By 90 days, the direction of the association reversed, showing a nonsignificant trend favoring GLP‑1 RAs (26 vs. 30 events; KM survival 98.37% vs. 98.09%; HR 0.854, 95% CI 0.505–1.444; log‐rank p = 0.556). At 1 year, GLP‑1 RAs demonstrated a statistically significant reduction in ACM (63 vs. 89 events; KM survival 95.39% vs. 93.30%; HR 0.700, 95% CI 0.507–0.967; log‐rank p = 0.029), corresponding to a 30% relative risk reduction. The survival curves remained closely aligned through the first 90 days but began to separate progressively thereafter, with the gap widening steadily between 3 and 12 months. This temporal pattern suggests that any protective effect of GLP‑1 RAs on mortality in this post‐AMI population required time to manifest fully. Schoenfeld's proportional hazards assumptions were not violated at any time point (all proportionality test p > 0.05).

3.3. Acute Heart Failure

GLP‑1 RAs showed a large and highly significant reduction in acute heart failure events compared with SGLT2is, evident from the earliest time point. At 30 days, the difference was striking (39 vs. 165 events; KM survival 97.68% vs. 90.17%; HR 0.224, 95% CI 0.158–0.317; log‐rank p < 0.001), representing a greater than 75% relative risk reduction. This substantial early benefit persisted at 90 days (73 vs. 233 events; KM survival 95.51% vs. 85.74%; HR 0.290, 95% CI 0.223–0.378; log‐rank p < 0.001) and remained highly significant at 1 year (158 vs. 345 events; KM survival 88.86% vs. 76.76%; HR 0.415, 95% CI 0.343–0.501; log‐rank p < 0.001). The KM curves diverged sharply within the first 30 days and continued to separate gradually over the subsequent months, resulting in a progressively larger absolute difference in event‐free survival by the end of the first year. The Schoenfeld proportional hazards assumption was violated at 30 days and 90 days (p = 0.006 and p = 0.003, respectively) but held at 1 year.

3.4. All‐Cause Hospitalization

GLP‑1 RAs were associated with a strong and early reduction in all‐cause hospitalization risk. At 30 days, hospitalization events were substantially fewer in the GLP‑1 RAs group (136 events vs. 427 events; KM survival 92.05% vs. 75.32%; HR 0.290, 95% CI 0.239–0.351; log‐rank p < 0.001). This represented one of the most pronounced early differences observed across all endpoints. The benefit was maintained at 90 days (223 vs. 494 events; KM survival 86.53% vs. 71.04%; HR 0.400, 95% CI 0.342–0.469; log‐rank p < 0.001) and remained clearly evident at 1 year (430 vs. 669 events; KM survival 70.23% vs. 57.56%; HR 0.559, 95% CI 0.495–0.631; log‐rank p < 0.001). The survival curves showed the steepest separation during the first month, after which the rate of divergence slowed, but the absolute gap continued to widen through the remainder of the year. Proportionality test p values indicated potential violation of the proportional hazards assumption.

3.5. Recurrent MI

GLP‑1 RAs were associated with a significantly lower incidence of RMI compared with SGLT2is. At 30 days, fewer events occurred in the GLP‑1 RAs cohort (205 vs. 308 events; KM survival 87.90% vs. 81.67%; HR 0.628, 95% CI 0.526–0.749; log‐rank p < 0.001). This reduction persisted at 90 days (317 vs. 420 events; KM survival 80.84% vs. 74.47%; HR 0.701, 95% CI 0.606–0.812; log‐rank p < 0.001) and at 1 year (515 vs. 595 events; KM survival 65.71% vs. 60.80%; HR 0.799, 95% CI 0.710–0.899; log‐rank p < 0.001). The event‐free survival curves demonstrated consistent separation across the entire follow‐up period, with no evidence of convergence or attenuation of the relative benefit over time.

3.6. Stroke

GLP‑1 RAs were associated with a reduced risk of stroke beginning at 30 days (61 vs. 100 events; KM survival 96.38% vs. 94.01%; HR 0.592, 95% CI 0.431–0.814; log‐rank p = 0.001). The lower risk continued at 90 days (113 vs. 145 events; KM survival 93.07% vs. 91.08%; HR 0.751, 95% CI 0.587–0.960; log‐rank p = 0.022) and remained statistically significant at 1 year (213 vs. 257 events; KM survival 85.40% vs. 82.06%; HR 0.800, 95% CI 0.667–0.959; log‐rank p = 0.016). The KM curves separated early and maintained a consistent gap throughout the observation period, with no indication of diminishing relative benefit over the year. The proportional hazards assumption was violated at 90 days (p = 0.009) but held at 30 days and 1 year.

3.7. Afib

GLP‑1 RAs were associated with a lower incidence of Afib at 30 days (128 vs. 214 events; KM survival 92.41% vs. 87.21%; HR 0.568, 95% CI 0.456–0.707; log‐rank p < 0.001). This reduction persisted at 90 days (222 vs. 293 events; KM survival 86.43% vs. 82.06%; HR 0.712, 95% CI 0.598–0.848; log‐rank p < 0.001) and at 1 year (325 events vs. 381 events; KM survival 78.63% vs. 75.08%; HR 0.804, 95% CI 0.693–0.932; log‐rank p = 0.004). The greatest relative difference was observed in the early post‐procedural period, with a gradual narrowing of the relative risk reduction over subsequent months, although the absolute benefit continued to accumulate.

3.8. MACE

GLP‑1 RAs were associated with a significantly lower risk of the composite MACE endpoint at 30 days (233 vs. 348 events; KM survival 86.25% vs. 79.30%; HR 0.626, 95% CI 0.530–0.739; log‐rank p < 0.001). The reduction remained significant at 90 days (361 events vs. 479 events; KM survival 78.21% vs. 70.87%; HR 0.693, 95% CI 0.604–0.794; log‐rank p < 0.001) and at 1 year (583 vs. 677 events; KM survival 61.43% vs. 55.61%; HR 0.788, 95% CI 0.706–0.881; log‐rank p < 0.001). The event‐free survival curves diverged early and continued to separate steadily, reflecting the cumulative contribution of multiple individual endpoints to the overall MACE advantage.

3.9. AKI

GLP‑1 RAs were associated with a lower risk of AKI at 30 days (40 vs. 111 events; KM survival 97.63% vs. 93.33%; HR 0.347, 95% CI 0.242–0.498; log‐rank p < 0.001). The reduction persisted at 90 days (75 vs. 177 events; KM survival 95.39% vs. 89.04%; HR 0.401, 95% CI 0.306–0.525; log‐rank p < 0.001) and remained significant at 1 year (177 vs. 309 events; KM survival 87.52% vs. 78.54%; HR 0.534, 95% CI 0.444–0.643; log‐rank p < 0.001). The most pronounced relative difference occurred within the first 30 days, after which the curves continued to separate but at a slower rate. The proportional hazards assumption held at 30 days and 1 year but was borderline at 90 days (p = 0.078).

3.10. Cardiac Arrest

GLP‑1 RAs were associated with a numerically lower risk of cardiac arrest across all time points, although statistical significance was not reached. At 30 days, events were 12 versus 18 (KM survival 99.29% vs. 98.93%; HR 0.654, 95% CI 0.315–1.358; log‐rank p = 0.251). Similar patterns were observed at 90 days (19 vs. 25 events; KM survival 98.84% vs. 98.46%; HR 0.744, 95% CI 0.410–1.351; log‐rank p = 0.330) and at 1 year (31 vs. 47 events; KM survival 97.86% vs. 96.72%; HR 0.645, 95% CI 0.410–1.015; log‐rank p = 0.056). The point estimates consistently favored GLP‑1 RAs, with the trend becoming more apparent by the end of the first year. Proportional hazards assumptions were not violated at any time point.

3.11. Multivariate Cox Proportional Hazards Model for ACM

In the multivariate Cox proportional hazards model for ACM in the propensity score‐matched cohort, the model adjusted for key demographic and clinical covariates, including sex, chronic kidney disease, type 2 diabetes mellitus, hypertension, dyslipidemias, BMI categories, and LVEF categories (Table 5).

Initiation of GLP‑1 RAs was independently associated with a substantially lower risk of ACM compared with SGLT2is (aHR 0.391, 95% CI 0.320–0.478, p < 0.0001). This corresponded to a 60.9% relative risk reduction after full adjustment. Among the adjusted covariates, several factors were significantly associated with increased mortality risk. Chronic kidney disease was a strong independent predictor (aHR 2.831, 95% CI 2.367–3.386, p < 0.0001). Type 2 diabetes mellitus was also associated with higher mortality (aHR 1.463, 95% CI 1.195–1.793, p < 0.0001). Lower BMI categories showed a particularly pronounced association with worse survival: BMI < 20 kg/m² (aHR 2.266, 95% CI 1.695–3.030, p < 0.0001) and BMI 20–25 kg/m² (aHR 1.460, 95% CI 1.162–1.833, p = 0.001). Severely reduced LVEF ( < 40%) was associated with increased mortality risk (aHR 1.340, 95% CI 1.023–1.755, p = 0.034). Male sex, hypertension, and dyslipidemias were not independently associated with ACM in the adjusted model (p > 0.05 for all). LVEF categories between 40% and > 70% and BMI categories ≥ 25 kg/m² did not show statistically significant associations with mortality after adjustment. These findings indicate that, independent of the balanced baseline characteristics achieved through propensity score matching and after further adjustment for key prognostic factors, treatment with GLP‑1 RAs remained strongly associated with improved survival compared with SGLT2is in patients with AMI undergoing PCI.

4. Discussion

This large, multicenter, propensity score‐matched real‐world analysis showed that initiation of GLP‐1 RAs within 14 days following PCI for AMI was associated with reductions in several adverse outcomes compared with SGLT2is. GLP‐1 RAs produced early marked reductions in acute heart failure (78% relative risk reduction at 30 days; HR 0.224) and all‐cause hospitalization (71% at 30 days; HR 0.290), with these advantages persisting and widening through 1 year (HR 0.415 and 0.559, respectively). Significant and sustained reductions were also observed for RMI (HR 0.799 at 1 year), stroke (HR 0.800), Afib (HR 0.804), the composite MACE endpoint (HR 0.788), and AKI (HR 0.534). ACM showed no early difference but 30% relative reduction by 1 year (HR 0.700), while cardiac arrest events trended lower without reaching statistical significance. These findings indicate that, in the vulnerable early post‐PCI phase after AMI, GLP‐1 RAs confer clinically meaningful protection that emerges rapidly and extends across ischemic, arrhythmic, heart failure, renal, and survival domains. However, these observational associations should be interpreted cautiously, as residual imbalances remained (particularly in BMI [SMD ~0.685] and selected laboratory values), even though these were explicitly addressed through additional multivariate Cox modeling, which confirmed an independent association favoring GLP‐1 RAs for mortality.

These findings extend evidence from cardiovascular outcomes trials in T2D, where several GLP‑1 RAs have reduced MACE and, in some cases, ACM (e.g., liraglutide in LEADER, dulaglutide in REWIND) [8, 9, 10, 24]. In contrast, SGLT2is have demonstrated particularly strong effects on reducing HHF and improving renal outcomes across a wide range of cardiovascular conditions risk [25, 26]. Unlike prior trials that mainly enrolled stable outpatients, our study centers on the early and intermediate post‐infarction period, a phase marked by increased risk of pump failure, recurrent ischemia, arrhythmias, and kidney injury. The sharp early divergence in the curves for acute heart failure and hospitalization suggests that GLP‐1 RAs may offer clinically meaningful protective effects during this high‐risk period. Possible mechanisms include beneficial effects on weight, blood pressure, endothelial function, inflammation, and myocardial substrate use, which, together, may enhance hemodynamic stability following PCI [27].

The consistent benefit across RMI, stroke, Afib, and AKI is also noteworthy. GLP1RAs have been linked to improved atherothrombotic and inflammatory profiles that could reduce plaque instability and cerebrovascular risk, while modest natriuretic and renal effects may support standard heart failure and nephroprotective therapies [10, 28]. In contrast, although SGLT2is are highly effective in populations with chronic heart failure and chronic kidney disease, their relative benefit may be reduced when started immediately after AMI, especially in patients without advanced heart failure or with temporary peri‐procedural renal dysfunction [29]. The consistent trend toward lower cardiac arrest risk with GLP‑1 RAs, although not statistically significant, aligns with the broader pattern of benefit.

This aligns with dedicated post‐MI trials and meta‐analyses. The DAPA‐MI trial showed that dapagliflozin initiated early after MI in patients without diabetes or chronic heart failure provided significant cardiometabolic benefits (e.g., reduced new‐onset type 2 diabetes and weight loss) but no reduction in the composite of cardiovascular death or HFF (HR, 0.95; 95% CI, 0.64−1.40) [30]. This divergence aligns with dedicated post‐MI trials and meta‐analyses. A 2025 meta‐analysis of early SGLT2i initiation following acute MI reported significantly lower future HHF (OR 0.75; 95% CI 0.62–0.90) and higher LVEF (MD 1.65%; 95% CI 0.34–2.96) compared with placebo, but no beneficial impact on cardiovascular death, ACM, stroke, or all‐cause hospitalization [16]. Consistent findings emerged from the SWEDEHEART registry, where SGLT2i prescription within 3 days post‐discharge (or 120 days pre‐discharge) in 11,271 patients with type 2 diabetes mellitus after MI was associated with lower rates of the composite outcome of death or first HHF at 1 year (adjusted HR 0.70; 95% CI 0.59–0.82) [18].

In contrast, real‐world evidence increasingly supports GLP‐1 RAs in the post‐MI setting. In a nationwide Czech registry analysis (2015–2024) of patients with incident nonfatal MI or ischemic stroke and confirmed type 2 diabetes, GLP‐1 RA initiation within 12 months post‐event (propensity score‐matched to non‐users) was associated with lower MACE risk in MI survivors and stroke survivors [31]. Similarly, in a large US real‐world cohort of early GLP‐1 RA use (semaglutide or tirzepatide) after AMI, GLP‐1 RA initiation was not associated with reduced recurrent coronary events (3.7% vs. 3.8%; HR 0.913; 95% CI 0.750–1.111), but conferred significant reductions in ACM (HR 0.484; 95% CI 0.413–0.567), HF hospitalizations (HR 0.578; 95% CI 0.531–0.630), AKI (HR 0.688; 95% CI 0.610–0.732), cardiac arrest (HR 0.549; 95% CI 0.418–0.722), and all‐cause hospitalizations (HR 0.673; 95% CI 0.636–0.713) [20]. These patterns highlight differential class‐specific effects in the acute post‐MI phase: SGLT2is primarily attenuates HHF risk without clear mortality or broader ischemic benefits, whereas GLP‐1 RAs demonstrate more consistent advantages across mortality, HF events, renal outcomes, and hospitalizations.

Mechanistic differences likely explain the observed benefit of GLP‐1 RAs in the acute post‐reperfusion phase. GLP‐1 RAs exert direct cardioprotective effects beyond glucose lowering, including anti‐inflammatory actions, improved endothelial function, reduced oxidative stress, and favorable shifts in myocardial substrate utilization [27]. Small randomized trials of exenatide and liraglutide administered peri‐PCI in STEMI have shown reductions in infarct size, improved myocardial salvage index, and better LVEF recovery; effects not consistently seen with SGLT2is [17]. A 2025 scoping review of GLP‐1 RAs in post‐STEMI care, encompassing 10 studies of exenatide and liraglutide in adults with STEMI, confirmed safety and good tolerability. Exenatide was associated with reduced infarct size, improved myocardial salvage, and better cardiac function, although benefits were inconsistent in two broader STEMI trials. Liraglutide, evaluated across three trials, consistently improved myocardial salvage, infarct size, LVEF, stroke volume, and no‐reflow phenomenon, accompanied by favorable biomarker shifts, albeit without significant reductions in major cardiovascular events [32]. In contrast, SGLT2is primarily mediate benefits through natriuresis, volume offloading, and metabolic reprogramming; mechanisms most effective in established chronic heart failure or chronic kidney disease. These effects may be attenuated in the immediate post‐MI period due to hemodynamic instability, transient renal impairment, or concurrent diuretic use [29].

Collectively, these data suggest that, among patients with AMI undergoing PCI, GLP‐1 RAs may offer a broader and more immediate cardioprotective profile than SGLT2is when a single agent is chosen. In patients without a compelling indication for SGLT2is (e.g., advanced heart failure with reduced ejection fraction or albuminuric chronic kidney disease), GLP‑1 RAs may be an attractive option in the early post‑PCI period, though this requires confirmation in randomized trials. Combination therapy, which recent cohort studies have associated with additive reductions in MACE and renal events [33], may ultimately prove optimal but requires dedicated randomized evaluation in the acute post‐MI setting.

Randomized head‐to‐head trials comparing GLP‐1 RAs versus SGLT2is (and their combination) specifically in patients with recent AMI and PCI are urgently needed to confirm these observational findings and guide personalized therapy. Until such trials are completed, the present real‐world evidence provides actionable comparative effectiveness data to inform shared decision‐making in this high‐risk population.

4.1. Strengths of the Study

This study has several important strengths that support the reliability and relevance of the findings. After propensity score matching, the analytic cohorts comprised 1752 patients in each group, providing a reasonably powered comparison in a real‐world post‐PCI setting. The active‐comparator, new‐user design minimized indication bias and exposure misclassification by excluding patients with prior use of the comparator class, ensuring clean initiation cohorts free from carryover effects. The inclusion of three distinct time windows—short‐term (30 days), mid‐term (90 days), and long‐term (1 year)—allowed for a detailed evaluation of the temporal pattern of treatment effects, revealing early and sustained advantages of GLP‑1 RAs across multiple endpoints. Rigorous one‐to‐one greedy nearest‐neighbor matching with a caliper of 0.1 achieved a good to excellent balance across more than 50 baseline covariates, with most SMD < 0.10 and p > 0.05 for nearly all variables. This substantial reduction in measured confounding enhances the validity of comparative effectiveness inferences in this observational context.

4.2. Limitations of the Study

Although the study possesses several methodological strengths, including a large multicenter cohort, active‐comparator new‐user design, and extensive propensity score matching on over 50 covariates, its retrospective observational nature using EHR data inherently limits the establishment of causal relationships. Despite comprehensive propensity score matching with a caliper of 0.1, residual confounding from unmeasured or incompletely captured factors cannot be excluded. Variables not routinely recorded in EHRs, such as smoking status, physical activity level, dietary patterns, socioeconomic status, medication adherence, specific dose titration schedules, post‐discharge cardiac rehabilitation participation, or granular details on procedural complexity during PCI, may have influenced outcomes. Post‐matching imbalances in certain key variables persisted, most notably BMI (SMD ≈ 0.685), creatinine, sodium, and potassium levels. These imbalances raise concerns for residual confounding and potential channeling bias, whereby GLP‐1 RAs (particularly agents associated with greater weight loss) may have been preferentially prescribed to patients with higher BMI or specific metabolic profiles, possibly contributing to regression‐to‐the‐mean effects or exaggerated early outcome differences.

Reliance on EHR introduces the possibility of ascertainment and misclassification biases. All outcomes, including acute heart failure, RMI, stroke, Afib, MACE, AKI, and cardiac arrest, were defined using ICD‐10‐CM codes. While such administrative definitions are standard in large real‐world studies and any misclassification is likely non‐differential between cohorts (biasing results toward the null), the absence of validated coding algorithms or supporting data (e.g., echocardiography for heart failure severity, imaging for stroke etiology, or biomarker trends) limits diagnostic precision. Granular data on heart failure severity (e.g., NYHA class), LVEF trajectory over time, or cause‐specific mortality (cardiovascular vs. non‐cardiovascular) were unavailable, restricting mechanistic insights into the observed benefits.

The timing of drug initiation was restricted to within 14 days following the index PCI to focus on the early post‐procedural period; however, exact start dates (in‐hospital vs. immediate post‐discharge) and subsequent treatment persistence, discontinuation, dose adjustments, or switching to the comparator class were not directly measurable. The analysis followed an intention‐to‐treat framework based on initial prescription records, which may not fully reflect real‐world exposure. Standard Cox proportional hazards models were employed, with Schoenfeld residuals used to evaluate assumptions (reported per outcome and time window); competing risk approaches (e.g., Fine−Gray models) were considered but not applied in the primary analysis given the focus on ACM and low rates of competing events.

Patient follow‐up depended on continued engagement within the contributing healthcare organizations of the TriNetX US Collaborative Network, with censoring at the last recorded activity; this may introduce informative censoring bias favoring individuals with more consistent healthcare utilization. Mortality ascertainment relied primarily on linkage to national death registries, supplemented by diagnostic codes, which may underdetect events occurring outside networked systems. Finally, the study population was derived exclusively from US healthcare organizations within the TriNetX network, which may limit generalizability to other countries or healthcare systems with differing demographics, prescribing practices, access to GLP‐1 RAs or SGLT2is, reimbursement policies, or comorbidity profiles. These limitations, combined with the observational design, mean the findings should be interpreted as associations rather than definitive evidence of superiority or causality. Randomized head‐to‐head trials are essential to confirm the observed patterns.

4.3. Conclusion

In this large propensity score‐matched real‐world cohort of adults with AMI undergoing PCI, initiation of GLP‐1 RAs within 14 days post‐PCI was associated with reductions in acute heart failure, all‐cause hospitalization, RMI, stroke, Afib, MACE, and AKI compared with SGLT2is. These observational associations suggest potential broader cardioprotective benefits during the vulnerable post‐PCI period but require confirmation in randomized, head‐to‐head trials.

Funding

The authors have nothing to report.

Ethics Statement

This study utilized fully de‐identified, aggregated electronic health records from the TriNetX US Collaborative Network. All analyses were performed on anonymized data in compliance with the Health Insurance Portability and Accountability Act (HIPAA) and General Data Protection Regulation (GDPR), obviating the need for institutional review board approval or individual patient consent.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File

Data Availability Statement

All data generated and analyzed during this study were derived from the TriNetX US Collaborative Network under a limited‐use agreement that prohibits the redistribution of individual‐level or raw data. Aggregate‐level results, including cohort characteristics, outcome counts, event‐free survival probabilities, and statistical summaries, are available within the published article and its supplementary materials.