Prognostic impact of metformin in diabetic patients undergoing a percutaneous coronary intervention (PCI): protective effect is modified by procedural complexity

Abstract

Background

Metformin, a widely prescribed glucose-lowering agent, has demonstrated cardiovascular benefits in patients with diabetes who do not have established atherosclerotic cardiovascular disease. However, evidence regarding its role specifically in patients with diabetes undergoing percutaneous coronary intervention (PCI) remains limited. This study therefore aimed to evaluate the prognostic association of metformin use in this high‑risk population and to explore its potential interaction with procedural complexity.

Methods

From January 2017 to December 2018, 11,585 diabetic patients undergoing PCI at Fuwai hospital were consecutively enrolled in our study. Patients were categorized into four groups according to metformin use and PCI complexity. The primary endpoint was major adverse cardiovascular and cerebrovascular events (MACCEs), including cardiovascular death, non-fatal myocardial infarction (MI), non-fatal stroke, and unplanned revascularization.

Results

After a follow-up of 3 years, a total of 1292 MACCEs were recorded. Overall, metformin use was observed to be associated with a lower incidence of 3-year MACCEs (adjusted HR 0.80, 95%CI 0.70–0.92) after multivariable adjustment. Significantly lower incidence of MACCEs was observed in patients undergoing non-complex PCI (adjusted HR 0.65, 95% CI: 0.53–0.79), while such protective effect of metformin didn’t exist in complex PCI patients (adjusted HR 1.00, 95%CI: 0.83–1.21). Significant interaction between metformin and PCI complexity was found with regard to the 3-year MACCE rate (P_interaction = 0.003; adjusted P_interaction = 0.002).

Conclusions

In this observational study, there was significant difference in the efficacy of metformin in diabetic patients undergoing complex or non-complex PCI. Metformin use was associated with improved prognosis in patients with diabetes undergoing non-complex PCI.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s12933-026-03102-6.

Introduction

Diabetes mellitus (DM) is a prevalent comorbidity in patients with coronary artery disease (CAD) and constitutes more than one‑third of all individuals undergoing percutaneous coronary intervention (PCI) [1]. Even in the era of contemporary drug‑eluting stents (DES), DM remains associated with a heightened risk of adverse cardiovascular events post‑PCI [2, 3]. This elevated ischemic risk has been attributed to multiple pathophysiological mechanisms, including endothelial dysfunction, impaired fibrinolysis, platelet hyperactivity, uneven neointimal strut coverage after DES implantation, smaller coronary vessel dimensions, and greater residual plaque burden [4–6]. Consequently, the optimal use of glucose‑lowering agents remains a central consideration in the management of diabetic patients undergoing PCI, prompting a series of cardiovascular outcome trials designed to evaluate the safety and efficacy of these therapies.

Metformin is a cornerstone glucose‑lowering agent and is recommended as first‑line therapy for patients with type 2 diabetes (T2DM) who require initiation or intensification of glycemic control [7]. Its widespread use is supported by a favorable safety profile, accessibility, and cost‑effectiveness, alongside evidence of reduced risks of cardiovascular (CV) events, microvascular complications, and mortality [7]. Prior studies have further suggested that metformin may confer secondary cardiovascular protection [4, 8]. For instance, in patients with diabetes and acute myocardial infarction (MI), metformin use has been associated with lower mortality and reduced risks of non‑fatal MI and stroke [9, 10]. Similarly, a randomized clinical trial demonstrated that, compared with glipizide, 3‑year metformin treatment significantly lowered major cardiovascular events over a median 5‑year follow‑up in patients with T2DM and CAD [11]. Nevertheless, robust evidence regarding the cardiovascular benefits of metformin in patients with established atherosclerotic cardiovascular disease (ASCVD) remains limited.

Although percutaneous coronary intervention (PCI) is a standard revascularization strategy in CAD, data on metformin use specifically in diabetic patients undergoing PCI are sparse. Several reports indicate that metformin may attenuate the risk of in‑stent restenosis or the no‑reflow phenomenon after PCI in this population [12, 13], yet broader cardiovascular outcomes have not been comprehensively evaluated. Furthermore, given that complex and non‑complex PCI procedures may differentially influence treatment responses [14, 15], it is pertinent to examine whether the cardiovascular effects of metformin vary according to procedural complexity. Therefore, this study aimed to assess the effectiveness and safety of metformin in diabetic patients undergoing PCI and to explore its potential interaction with PCI complexity.

Methods

Study design and population

We conducted a prospective cohort study at a single tertiary center. Between January 2017 and December 2018, a total of 11,585 consecutive diabetic patients undergoing PCI were enrolled from Fuwai Hospital, Chinese Academy of Medical Sciences. Diabetes was defined according to standard criteria [16], including a documented history of diabetes, active use of glucose‑lowering therapy, fasting blood glucose (FBG) ≥ 7.0 mmol/L, glycosylated hemoglobin A1c (HbA1c) ≥ 6.5%, or a 2‑hour plasma glucose ≥ 11.1 mmol/L during an oral glucose tolerance test. Key exclusion criteria comprised missing essential laboratory data (FBG or HbA1c), age < 18 or ≥ 80 years, severe hepatic or renal impairment, decompensated heart failure, systemic inflammatory disease, malignancy, acute infection, or loss to follow‑up. After applying these criteria, 11,585 eligible patients were included in the final analysis (Fig. 1).

Fig. 1.

Flow chart of the enrolled patients

The study was performed in accordance with the Declaration of Helsinki and received approval from the Institutional Review Board of Fuwai Hospital, National Center for Cardiovascular Diseases. Written informed consent was obtained from all participants prior to enrollment.

PCI procedure and medication

All PCI procedures and concomitant medical therapies were performed in accordance with contemporary guidelines and at the discretion of the treating cardiologists, as previously described [17, 18]. Specific PCI strategies, including device selection and the use of adjunctive imaging (e.g., intravascular ultrasound and optical coherence tomography), were determined by experienced interventional cardiologists. Prior to PCI, all patients received a loading dose of aspirin (300 mg) and a P2Y12 inhibitor (clopidogrel 300–600 mg or ticagrelor 180 mg). Procedural anticoagulation was achieved with unfractionated heparin or bivalirudin. Following the intervention, indefinite aspirin therapy (100 mg daily) was prescribed, and clopidogrel (75 mg daily) was typically continued for 12 months. All data were prospectively entered into a dedicated database by independent research personnel.

Data collection and biochemical analysis

Baseline demographic and clinical data were collected prospectively for all participants by independent research staff. Demographic variables included age, sex, body mass index (BMI), comorbidities, smoking status, history of myocardial infarction, and prior revascularization (PCI or coronary artery bypass grafting [CABG]). Clinical information encompassed the admission diagnosis, physical examination findings, medical imaging results, laboratory tests, and discharge medications. Angiographic and procedural data were obtained independently from catheterization laboratory records by three trained interventional cardiologists.

Venous blood samples were collected after a minimum 12‑hour fast at the time of admission and analyzed in the clinical chemistry laboratory of Fuwai Hospital. Total cholesterol (TC), high‑density lipoprotein cholesterol (HDL‑C), triglycerides, fasting blood glucose (FBG), and serum creatinine were measured using an automated biochemical analyzer (Hitachi 7150, Tokyo, Japan). Low‑density lipoprotein cholesterol (LDL‑C) was calculated by the Friedewald method [19]. Glycated hemoglobin (HbA1c) was assayed with a Tosoh Automated Glycohemoglobin Analyser (HLC‑723G8, Tokyo, Japan). High‑sensitivity C‑reactive protein (hs‑CRP) was determined using standard biochemical techniques in the hospital’s core laboratory. Baseline estimated glomerular filtration rate (eGFR) was derived from the Chinese‑modified Modification of Diet in Renal Disease Eq.  [19].

Follow-up, study endpoints and definitions

Patients were followed up at 6-month intervals for a duration of 3 years. Follow-up information was collected via medical records, clinical visits, and/or telephone interviews conducted by well-trained investigators who were blinded to the patients’ clinical data and study design.

The primary endpoint was major adverse cardiovascular and cerebrovascular event (MACCE), defined as a composite of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, and unplanned revascularization. The secondary endpoint was a composite of cardiovascular events, cardiovascular death, and the individual component of the primary endpoint. Death was considered cardiac unless other unequivocal non-cardiovascular causes were confirmed. Non-fatal MI was defined according to the third universal definition of MI, excluding periprocedural MI [20]. Non-fatal stroke was defined as new neurological deficits, either ischemic or hemorrhagic, confirmed by neurologists based on imaging findings. Unplanned revascularization was defined as any unplanned repeated PCI or CABG of any coronary vessel. Cardiovascular events included cardiovascular death, non-fatal myocardial infarction and non-fatal stroke. Complex PCI was defined by the presence of at least one of the following features: 3 or more stents implanted, 3 or more lesions treated, bifurcation with 2 stents implanted, total stent length more than 60 mm, or chronic total occlusion [21].

Statistical analysis

Continuous variables are expressed as mean ± standard deviation (SD) or median [interquartile range (IQR)], and were compared using Student’s t‑test or the Mann–Whitney U test, as appropriate. Categorical variables are reported as frequency and percentage, and were compared using the chi‑square test or Fisher’s exact test, as appropriate. Cumulative incidence of clinical endpoints was illustrated by Kaplan–Meier curves and compared with the log‑rank test. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated using univariable and multivariable Cox proportional‑hazards regression models. The multivariable Cox model was adjusted for the following potential confounders: age, sex, BMI, current smoking, systolic blood pressure, estimated glomerular filtration rate (eGFR), left ventricular ejection fraction (LVEF), previous MI, previous PCI, previous CABG, previous stroke, acute coronary syndrome (ACS) presentation, HbA1c, total cholesterol (TC), triglycerides, LDL‑C, high‑sensitivity C‑reactive protein (hsCRP), statin use, aspirin use, and SYNTAX score. An interaction term between metformin use (yes vs. no) and PCI complexity (complex vs. non‑complex) was included in the Cox model to formally test for effect modification on 3‑year outcomes. All analyses were performed on a first‑event basis. A two‑sided P value < 0.05 was considered statistically significant. Covariate selection was informed by clinical relevance, prior evidence [3, 4], and statistical considerations. To mitigate potential multicollinearity, variance inflation factors (VIFs) were calculated for all adjusted variables. A series of sensitivity analyses, including parsimonious modelling and Least Absolute Shrinkage and Selection Operator (LASSO) regression, were conducted to verify the robustness of the primary models. All statistical analyses were performed using R software version 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Baseline characteristics

Among the 11,585 diabetic patients undergoing PCI included in this study, 5307 (45.8%) patients underwent complex PCI. Baseline characteristics of participants are shown in Table 1. Patients taking metformin were younger, with higher BMI, less likely to be treated with insulin, and had a longer diabetes duration, higher levels of FBG and HbA1c compared with non-metformin users in both complex or non-complex PCI group. In addition, metformin users had lower levels of TC and LDL-C, higher eGFR, and lower SYNTAX score II. Among patients undergoing complex PCI, those taking metformin were more likely to have hypertension, less likely to have stroke history, had lower SYNTAX score and a higher proportion of IVUS use. As for patients undergoing non-complex PCI, those taking metformin were more likely to have family history of CAD, and had lower HDL-C level.

Table 1.

Baseline Characteristics of Diabetic Patients Undergoing PCI Stratified by Metformin Use and PCI Complexity

	Non-complex PCI			Complex PCI
	Non-metformin N = 4716	Metformin N = 1562	P value	Non-metformin N = 4079	Metformin N = 1228	P value
Age, years	61.44 ± 9.81	59.24 ± 9.23	< 0.001	60.78 ± 9.89	59.87 ± 8.79	0.004
Male sex	3423 (72.6)	1168 (74.8)	0.097	3085 (75.6)	943 (76.8)	0.426
BMI	26.23 ± 3.21	26.64 ± 3.38	< 0.001	26.25 ± 3.22	26.60 ± 3.02	0.001
ACS	3113 (66.0)	1037 (66.4)	0.807	2439 (59.8)	702 (57.2)	0.107
Hypertension	3263 (69.2)	1076 (68.9)	0.846	2824 (69.2)	890 (72.5)	0.032
Diabetes with insulin treated	968 (20.5)	189 (12.1)	< 0.001	829 (20.3)	148 (12.1)	< 0.001
DM duration	7.24 ± 6.74	8.25 ± 6.28	< 0.001	7.32 ± 6.97	8.60 ± 6.63	< 0.001
Previous MI	1288 (27.3)	405 (25.9)	0.301	1160 (28.4)	322 (26.2)	0.138
Previous PCI	1565 (33.2)	486 (31.1)	0.138	1060 (26.0)	316 (25.7)	0.888
Previous CABG	87 (1.8)	30 (1.9)	0.933	178 (4.4)	41 (3.3)	0.133
Previous stroke	706 (15.0)	212 (13.6)	0.189	683 (16.7)	156 (12.7)	0.001
PAD	341 (7.2)	96 (6.1)	0.161	347 (8.5)	99 (8.1)	0.664
Family history of CAD	508 (10.8)	203 (13.0)	0.018	474 (11.6)	144 (11.7)	0.96
Current smoker	1378 (29.2)	497 (31.8)	0.056	1230 (30.2)	366 (29.8)	0.842
LVEF, %	61.32 ± 7.12	61.91 ± 5.97	0.003	60.69 ± 7.59	61.95 ± 5.60	< 0.001
Glu, mmol/L	8.06 ± 3.00	8.75 ± 3.07	< 0.001	8.05 ± 2.86	8.69 ± 3.16	< 0.001
HbA1C, %	7.39 ± 1.22	7.75 ± 1.30	< 0.001	7.47 ± 1.24	7.77 ± 1.33	< 0.001
TC, mmol/L	3.97 ± 0.98	3.89 ± 0.98	0.004	4.00 ± 0.99	3.93 ± 1.00	0.029
TG, mmol/L	1.76 ± 1.15	1.82 ± 1.16	0.105	1.77 ± 1.14	1.80 ± 1.07	0.345
LDL-C, mmol/L	2.35 ± 0.80	2.27 ± 0.82	< 0.001	2.40 ± 0.85	2.33 ± 0.85	0.023
HDL-C, mmol/L	1.08 ± 0.27	1.06 ± 0.26	0.019	1.06 ± 0.26	1.05 ± 0.26	0.373
hsCRP, mg/L	2.55 ± 2.88	2.41 ± 2.70	0.165	2.61 ± 2.94	2.50 ± 2.82	0.229
eGFR	84.70 ± 19.20	87.66 ± 17.62	< 0.001	84.86 ± 19.27	87.23 ± 18.50	< 0.001
SYNTAX score	11.72 ± 4.77	11.54 ± 4.57	0.202	13.88 ± 5.80	13.49 ± 5.23	0.034
SYNTAX score II	29.97 ± 10.04	27.69 ± 9.68	< 0.001	30.38 ± 10.14	28.46 ± 9.65	< 0.001
IVUS use	257 (5.4)	105 (6.7)	0.071	627 (15.4)	222 (18.1)	0.026
Medication at discharge
Aspirin	4668 (99.0)	1551 (99.3)	0.336	4031 (98.8)	1219 (99.3)	0.244
P2Y12 inhibitor	4648 (98.6)	1546 (99.0)	0.264	4005 (98.2)	1215 (98.9)	0.089
Dual-antiplatelet therapy	4612 (97.8)	1536 (98.3)	0.231	3978 (97.5)	1207 (98.3)	0.144
CCB	1512 (32.1)	526 (33.7)	0.25	1346 (33.0)	412 (33.6)	0.745
β-blocker	3425 (72.6)	1135 (72.7)	1.0	2975 (72.9)	875 (71.3)	0.263
Statins	4641 (98.4)	1530 (98.0)	0.271	4013 (98.4)	1210 (98.5)	0.807
ACEI/ARB	2736 (58.0)	953 (61.0)	0.04	2432 (59.6)	744 (60.6)	0.568
GLP-1 RA	20 (0.4)	11 (0.7)	0.254	15 (0.4)	19 (1.5)	< 0.001
SGLT2i	7 (0.2)	3 (0.2)	1.000	3 (0.1)	2 (0.2)	0.742
GLP-1 RA or SGLT2i	27 (0.6)	14 (0.9)	0.241	18 (0.5)	21 (1.7)	< 0.001

Values are presented as mean ± standard deviation, median (interquartile range), or n (%). P values were derived from Student’s t-test, Mann–Whitney U test, or chi-square test as appropriate. BMI, body mass index; ACS, acute coronary syndrome; MI, myocardial infarction; PCI, percutaneous coronary intervention; CABG, coronary artery bypass grafting; PAD, peripheral artery disease; CAD, coronary artery disease; LVEF, left ventricular ejection fraction; HbA1c, glycated hemoglobin; TC, total cholesterol; TG, triglycerides; LDL-C, low-density lipoprotein cholesterol; HDL-C, high-density lipoprotein cholesterol; hsCRP, high-sensitivity C-reactive protein; eGFR, estimated glomerular filtration rate; IVUS, intravascular ultrasound; CCB, calcium channel blocker; ACEI/ARB, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker. GLP-1 RA, glucagon-like peptide-1 receptor agonist; SGLT2i, sodium-glucose cotransporter-2 inhibitor. *

Clinical outcomes by metformin use

After a follow-up of 3 years, a total of 1292 MACCEs were recorded, including 317 deaths, 204 MI, 68 strokes, and 894 unplanned revascularizations. Metformin users had lower incidence of 3-year MACCEs compared with non-metformin users in the overall cohort (9.5% vs. 11.7%, HR 0.80, 95%CI 0.70–0.92, p < 0.001), which remained significant (adjusted HR 0.80, 95%CI 0.70–0.92, p = 0.002, Table 2) after adjustment for potential confounding factors including age, sex, BMI, current smoking, systolic blood pressure, eGFR, LVEF, previous MI, previous PCI, previous CABG, previous stroke, acute coronary syndrome presentation, HbA1c, TC, triglycerides, LDL-C, hsCRP, statin use, aspirin use, and SYNTAX score. Patients taking metformin had lower incidence of CV events (3.0% vs. 4.6%), all-cause death (1.7% vs. 3.1%), CV death (1.2% vs. 2.1%) and MI (1.2% vs. 1.9%) compared with non-metformin users (all p value < 0.05, Supplementary Table S1). The multivariable Cox regression analyses showed that compared with non-metformin users, the adjusted HRs for CV events, MI and unplanned revascularization in all patients taking metformin were 0.73 (95% CI: 0.57–0.92), 0.59 (95% CI: 0.40–0.86) and 0.81 (95% CI: 0.69–0.96, Table 2), respectively.

Table 2.

Cox proportional hazard model for metformin and clinical outcomes in the overall cohort

	Unadjusted HR	P value	Adjusted HR	P value
MACCE	0.80 (0.70–0.92)	0.001	0.80 (0.70–0.92)	0.002
CV events	0.65 (0.52–0.83)	< 0.001	0.73 (0.57–0.92)	0.009
All-cause death	0.58 (0.43–0.79)	< 0.001	0.78 (0.57–1.06)	0.117
CV death	0.59 (0.41–0.85)	0.004	0.80 (0.55–1.17)	0.250
MI	0.60 (0.42–0.88)	0.008	0.59 (0.40–0.86)	0.006
Stroke	0.60 (0.32–1.15)	0.127	0.63 (0.33–1.22)	0.170
Unplanned revascularization	0.85 (0.73–1.00)	0.052	0.81 (0.69–0.96)	0.013

Multivariable models were adjusted for age, sex, BMI, current smoking, systolic blood pressure, eGFR, LVEF, previous MI, previous PCI, previous CABG, previous stroke, ACS presentation, HbA1c, TC, TG, LDL-C, hsCRP, statin use, aspirin use, and SYNTAX score

The cumulative incidence of MACCEs over the 3-year follow-up period is illustrated in Kaplan–Meier curves (Fig. 2), demonstrating significantly lower event rates among metformin users (log-rank P < 0.001).

Fig. 2.

Kaplan–Meier Curves for MACCE According to Metformin Use in Different Cohorts. (A) Overall cohort (n = 11,585). (B) Subgroup undergoing non-complex PCI (n = 6,278). (C) Subgroup undergoing complex PCI (n = 5,307). Complex PCI was defined by the presence of at least one of the following angiographic/procedural features: implantation of 3 or more stents, treatment of 3 or more lesions, bifurcation intervention with 2 stents implanted, total stent length > 60 mm, or treatment of a chronic total occlusion. Non-complex PCI included all other procedures. The primary endpoint was 3-year MACCE, a composite of cardiovascular death, non-fatal myocardial infarction, non-fatal stroke, and unplanned revascularization. Cumulative event incidence was estimated using the Kaplan–Meier method. Between-group comparisons within each cohort were performed using the log-rank test, and the corresponding P-values are displayed on each panel. Abbreviations: MACCE, major adverse cardiovascular and cerebrovascular events; PCI, percutaneous coronary intervention.”

Clinical outcomes by metformin use and PCI complexity

Subsequently, all participants were divided according to the complexity of PCI. In patients with non-complex PCI, the unadjusted 3-year rate of MACCEs of metformin users was lower (7.4% vs. 11.2%, HR 0.65, 95% CI: 0.53–0.80) than those not taking metformin, while there was no significant difference of MACCE incidence between metformin and non-metformin users (12.1% vs. 12.3%, HR 0.98, 95%CI: 0.81–1.18, Supplementary Table S2) among patients undergoing complex PCI. The lower incidence of MACCEs of metformin users compared with those not taking metformin in the non-complex PCI cohort remained significant (adjusted HR 0.65, 95% CI: 0.53–0.79, p < 0.001, Table 3) in the multivariable Cox regression analysis. Stratified Kaplan–Meier analyses revealed divergent effects of metformin according to PCI complexity (Fig. 2). In non-complex PCI patients, metformin use was associated with significantly lower MACCE incidence throughout follow-up (log-rank P < 0.001), whereas no such benefit was observed in complex PCI patients (log-rank P = 0.868). Significant heterogeneity existed in terms of MACCE rate in the metformin and non-metformin group among patients undergoing complex and non-complex PCI (P_interaction = 0.003, Supplementary Table S2; adjusted P_interaction = 0.002, Table 3).

Table 3.

Multivariable-Adjusted Cox Proportional Hazards Analysis for Metformin Use and Clinical Outcomes Stratified by PCI Complexity

	Noncomplex PCI		Complex PCI		Pinteraction
	Adjusted HR	P value	Adjusted HR	P value
MACCE	0.65 (0.53–0.79)	< 0.001	1.00 (0.83–1.21)	0.999	0.002
CV events	0.65 (0.46–0.91)	0.011	0.82 (0.59–1.16)	0.261	0.341
All-cause death	0.81 (0.53–1.24)	0.339	0.76 (0.48–1.21)	0.250	0.779
CV death	0.69 (0.40–1.20)	0.190	0.96 (0.57–1.61)	0.869	0.487
MI	0.55 (0.32–0.93)	0.025	0.63 (0.36–1.08)	0.091	0.698
Stroke	0.53 (0.22–1.27)	0.152	0.77 (0.29–2.06)	0.606	0.632
Unplanned revascularization	0.66 (0.51–0.84)	< 0.001	1.01 (0.81–1.26)	0.911	0.011

Models were adjusted for age, sex, BMI, current smoking, systolic blood pressure, eGFR, LVEF, previous MI, previous PCI, previous CABG, previous stroke, ACS presentation, HbA1c, TC, TG, LDL-C, hsCRP, statin use, aspirin use, and SYNTAX score. P for interaction was calculated from multivariable Cox proportional hazards models

Among non-complex PCI patients, patients taking metformin had lower incidence of CV events (2.7% vs. 4.5%), all-cause death (1.7% vs. 2.9%), CV death (1.0% vs. 2.0%), MI (1.1% vs. 1.9%) and unplanned revascularization (5.2% vs. 7.5%) compared with non-metformin users (all p value < 0.05, Supplementary Table S1). After adjusting confounding factors, metformin use was associated with lower rates of CV events (adjusted HR 0.65, 95% CI: 0.46–0.91), MI (adjusted HR 0.55, 95% CI: 0.32–0.93) and unplanned revascularization (adjusted HR 0.66, 95% CI 0.51–0.84, Table 3) among patients undergoing non-complex PCI.

As for complex PCI patients, the rate of all-cause death was lower in patients taking metformin compared with those not (1.7% vs. 3.2%, p = 0.008, Supplementary Table S1), while there was no significant difference in rates of CV events, CV death, MI, stroke or unplanned revascularization between metformin and non-metformin users. In the multivariable adjusted Cox regression model, the significant difference of all-cause death between metformin and non-metformin users didn’t exist (adjusted HR 0.76, 95% CI: 0.48–1.21). There was a statistically significant interaction between metformin use and PCI complexity for the rate of unplanned revascularization (P_interaction = 0.013, Supplementary Table S2; adjusted P_interaction = 0.011, Table 3).

Subgroup analysis

For all subgroups of patients undergoing non-complex PCI, no significant interaction was observed, and the association between metformin use and the incidence of MACCEs was consistent with the overall complex PCI cohort (all p value > 0.05, Table 4). No significant heterogeneity of metformin use for MACCEs was detected by age, sex, CAD presentation, hypertension or insulin use in the complex PCI cohort (all p value > 0.05, Supplementary Table S3). Metformin was associated with a numeric reduction of MACCEs (adjusted HR 0.72, 95% CI: 0.51–1.04) in patients undergoing non-complex PCI with BMI ≥ 28. However, there was no significant heterogeneity as for the protective effect of metformin on MACCEs between patients with BMI < 28 and BMI ≥ 28 in the non-complex PCI cohort.

Table 4.

Subgroup Analyses in patients undergoing non-complex PCI for the MACCE According to metformin prescriptions or not

	Metformin	Non-metformin	Adjusted HR (95%CI)	P value
Age, years				0.259
< 65	83/1133 (7.3)	299/3031 (9.9)	0.70 (0.55–0.90)
≥ 65	33/429 (7.7)	228/1685 (13.5)	0.57 (0.39–0.82)
Sex				0.969
Male	85/1168 (7.3)	378/3423 (11.0)	0.65 (0.51–0.83)
Female	31/394 (7.9)	149/1293 (11.5)	0.63 (0.43–0.94)
BMI				0.434
< 28	76/1100 (6.9)	384/3462 (11.1)	0.62 (0.48–0.79)
≥ 28	40/462 (8.7)	143/1254 (11.4)	0.72 (0.51–1.04)
CAD presentation				0.738
CCS	39/525 (7.4)	168/1603 (10.5)	0.67 (0.47–0.96)
ACS	77/1037 (7.4)	359/3113 (11.5)	0.63 (0.49–0.81)
Hypertension				0.459
Absent	29/486 (6.0)	153/1453 (10.5)	0.56 (0.38–0.84)
Present	87/1076 (8.1)	374/3263 (11.5)	0.68 (0.54–0.87)
Insulin treated				0.197
No	102/1373 (7.4)	387/3748 (10.3)	0.69 (0.56–0.87)
Yes	14/189 (7.4)	140/968 (14.5)	0.49 (0.28–0.86)

Models were adjusted for age, sex, BMI, current smoking, systolic blood pressure, eGFR, LVEF, previous MI, previous PCI, previous CABG, previous stroke, ACS presentation, HbA1c, TC, TG, LDL-C, hsCRP, statin use, aspirin use, and SYNTAX score. P for interaction was calculated from multivariable Cox proportional hazards models

Consistent results were observed across most prespecified subgroups, with no significant interactions detected for age, sex, clinical presentation, hypertension status, or insulin use in either complex or non-complex PCI cohorts.

Sensitivity and robustness analyses

To assess the stability of our findings, we performed several sensitivity analyses. First, The VIFs for the majority of covariates were below 5, indicating acceptable collinearity. Elevated VIFs were observed for total cholesterol (VIF = 14.21) and low-density lipoprotein cholesterol (VIF = 12.07), reflecting their known high biological correlation (r = 0.912). The mean VIF across all covariates was 2.38. (Supplementary Table S4). Second, two pre-specified parsimonious Cox models—one adjusting only for a core set of clinically imperative variables (age, sex, ACS presentation, eGFR, SYNTAX score, and statin use) and another excluding the highly correlated lipid parameters (TC and LDL-C)—yielded hazard ratios for metformin use consistent with the fully adjusted models (Supplementary Table S5). In particular, the protective association of metformin with MACCE in non-complex PCI patients remained significant across specifications (adjusted HR range 0.67–0.70), and the interaction with PCI complexity was consistently preserved (P for interaction ≤ 0.021). Finally, covariate selection via LASSO regression identified a robust subset of predictors; the resulting model (12 variables) demonstrated predictive performance (C-index = 0.570) virtually identical to that of the original adjustment set (20 variables, C-index = 0.571), and 60% of the original covariates were automatically selected by LASSO, indicating good agreement between clinical and data-driven selection (Supplementary Table S6). Collectively, these analyses support the robustness of our multivariable adjustment strategy and the reliability of the reported associations.

Discussion

In this large-scale, real-world prospective cohort study, we investigated the association between metformin use and clinical outcomes in diabetic patients undergoing either complex or non-complex PCI. The principal findings can be summarized as follows: (1) In the overall cohort, metformin use was associated with a lower incidence of MACCEs, CV events, MI, and unplanned revascularization. (2) A significant heterogeneity in the effect of metformin was observed according to PCI complexity. Specifically, metformin use was linked to a reduction in MACCEs, CV events, MI, and unplanned revascularization among patients undergoing non-complex PCI, whereas no such protective association was evident in those undergoing complex PCI. (3) No statistically significant interactions were detected between metformin use and age, sex, BMI, CAD presentation, hypertension, or insulin use with respect to MACCEs in either the complex or non-complex PCI subgroups. Notably, however, the magnitude of metformin’s protective effect on MACCEs appeared attenuated in non-complex PCI patients with BMI ≥ 28, suggesting potential variability in its efficacy among obese individuals undergoing less complex procedures. It is noteworthy that in the complex PCI cohort, some unadjusted associations lost statistical significance after multivariable adjustment. This indicates that the observed effects, particularly the absence of benefit in complex PCI, should be interpreted with caution and are hypothesis-generating. They highlight a potential differential effect of metformin based on procedural complexity that merits further investigation in prospective studies.

The major finding of this study is that the heterogeneity in the effect of metformin according to PCI complexity was most evident for the composite endpoint of MACCE and for unplanned revascularization, whereas no significant interaction was observed for other individual endpoints such as all-cause death or myocardial infarction. This pattern resonates with previous literature suggesting that the cardiovascular benefits of insulin-sensitizing therapies like metformin are often more apparent for composite endpoints and repeat revascularization, while consistent effects on individual hard outcomes remain less established [14]. Consequently, the differential effect of metformin may be particularly relevant to processes influencing restenosis or plaque progression—mechanisms that directly affect repeat revascularization risk—rather than to atherothrombotic events per se. The disparity of influence of metformin on prognosis may be explained by different characteristics of complex or non-complex PCI. Patients who undergo complex PCI are at higher ischemic risk [22–24], which has been found proportional to the degree of procedural complexity [25]. A greater number of stents implanted and stenting of more complex lesions in complex PCI may directly influence the propensity to platelet activation and coronary thrombosis [26], which may act as mediators of the stent-related thrombotic risk. Additionally, patients undergoing complex PCI are more likely to have incomplete myocardial revascularization, which is associated with higher risk of recurrent cardiac ischemic events [27, 28]. A former analysis of pooled patient-level data from 6 randomized controlled trials suggested that prolonged dual antiplatelet therapy significantly reduced major adverse cardiovascular events (MACEs) in the complex PCI group versus the noncomplex PCI group, and the magnitude of the benefit increased with procedural complexity [27]. Therefore, it may be deduced that high thrombotic risk is a main factor resulting in adverse prognosis in complex PCI, which probably hinders the protective effect of metformin in patients undergoing complex PCI. The absence of a clear benefit in complex PCI patients could be explained by the higher thrombotic burden and more frequent incomplete revascularization characteristic of such procedures, which may outweigh any potential advantages of metformin on endothelial function or plaque stability. However, this explanation remains hypothetical, and the observed interaction should be interpreted as hypothesis-generating rather than definitive. Further studies are needed to elucidate the pathophysiological interplay between procedural complexity and the cardiometabolic actions of metformin. While in non-complex PCI patients with relatively low thrombotic risk, the protective effect of metformin is more prominent, since hyperglycemia seems to be an important risk factor influencing prognosis.

Our study revealed that metformin was associated with lower incidence of MACCEs, CV events, MI and unplanned revascularization in overall patients undergoing PCI. These results corroborate the findings of a series of previous work investigating whether metformin influences the clinical outcomes of PCI patients. An early retrospective study found that metformin decreased adverse clinical events, especially death and MI, in diabetic patients undergoing PCI compared with insulin and/or sulfonylureas [29]. A former study found that chronic pretreatment with metformin was associated with the reduction of the no-reflow phenomenon in patients with diabetes after primary angioplasty for acute MI, which might be independent of the glucose lowering effect [12]. Chen et al. revealed that metformin was beneficial in reducing stent restenosis by a dose-dependent manner after PCI in T2DM patients [13]. Additionally, significantly reduced postprocedural myocardial injury and improved 1-year clinical outcomes were observed with a 7-day metformin pretreatment regimen in metabolic patients undergoing PCI [30]. Furthermore, the observed association between metformin use and improved outcomes in non-complex PCI patients, while robust to multivariable adjustment, may still be influenced by unmeasured residual confounding. Additionally, the lack of serial glycemic control data during the follow-up period may preclude a definitive distinction between the gluco-metabolic and potential pleiotropic effects of metformin on the reported outcomes.

There have been several retrospective studies evaluating the survival benefits of metformin for patients with diabetes and coronary artery diseases, and current evidence remains controversial. Jong et al. found metformin was associated with lower all-cause mortality from an analysis of 1157 patients with diabetes and acute coronary syndrome [31]. Similarly, Abualsuod et al. demonstrated lower 30-day all-cause mortality and tendency for a lower 12-month all-cause mortality in patients with diabetes using metformin following MI [32]. In accordance with the present results that metformin was associated with lower incidence of CV events and MI, Mellbin et al. found metformin use had a protective effect on non-fatal MI and stroke in patients with DM and MI [33]. And the reduced risk of unplanned revascularization associated with metformin treatment observed in our study was in line with the previous study [34]. Although not all studies have observed the cardiovascular benefits of metformin in patients with diabetes and coronary artery disease [35, 36], evidence from clinical and preclinical studies provides possible explanations for the cardiovascular protective effect of metformin. Previous studies have indicated that metformin use may reduce myocardial infarct size [37], improve plaque stability [38, 39], and attenuate pericoronary adipose tissue and inflammation [40, 41] in diabetic patients, which possibly contribute to better prognosis of metformin use. Numerous preclinical studies have also verified the cardiac protection of metformin in myocardial ischemia/reperfusion injury, with the mechanism of attenuation of mitochondrial dysfunction, decrease of myocardial oxidative damage, reduction of apoptosis signaling and endoplasmic reticulum stress [42–44]. Therefore, our study should be hypothesis-generating and the effect of metformin on the cardiovascular system independent of glucose-lowering effect should be reconsidered.

Metformin has been recommended as the first-line medication for T2DM patients without cardiovascular diseases [7], with cardiovascular benefits confirmed by former studies. An early multicenter, randomized clinical trial, the United Kingdom Prospective Diabetes Study (UKPDS), suggested metformin significantly reduced diabetes-related death, MI, coronary death, and stroke in 342 overweight/obese people with newly diagnosed T2DM without previous cardiovascular diseases [45]. And lower incidence of MI and longer survival were observed in those randomized to metformin in the UKPDS during 8–10 years of follow-up [46]. As for diabetic patients with ASCVD or at high CV risk, current guidelines recommend sodium–glucose co-transporter-2 (SGLT2) inhibitors or glucagon-like peptide-1 receptor agonists (GLP-1 RAs), which have been verified with CV risk reduction independent of glucose management considerations, as first-line glucose-lowering medications [7, 47]. Metformin has been recommended if additional glucose control is warranted on the premise of SGLT2 inhibitors or GLP-1 RAs in diabetic patients with ASCVD [47], due to its uncertain CV effects and lack of dedicated randomized cardiovascular outcome trials. Despite the emergence and increasing use of novel glucose-lowering agents with demonstrated cardioprotective properties such as empagliflozin and semaglutide, metformin is still preferred by a significant number of patients due to economic burden or reluctance to subcutaneous injection of semaglutide. Therefore, whether metformin has a direct cardiovascular protective effect on diabetic patients with ASCVD or those undergoing PCI is still worth exploring.

Limitations

Our study should be interpreted cautiously due to several potential limitations. Firstly, this was a single‑center study conducted in a Chinese population; thus, the generalizability of the results may be constrained by the homogeneous demographic and local clinical practice patterns, and caution is warranted when extrapolating the findings to other ethnic or healthcare settings. The observed interaction between metformin and PCI complexity requires validation in broader, multiethnic cohorts. Secondly, although our study has adjusted a considerable number of potential confounding factors, there are still unknown or unmeasured confounding factors which can’t be fully adjusted due to the observational nature of study design. Thirdly, the medication compliance of metformin during the follow-up period was unknown, and former studies reported relatively low adherence rate of metformin [31, 48, 49], which is possibly related to its gastrointestinal intolerance. Finally, although we captured and reported the use of contemporary cardioprotective glucose-lowering agents (SGLT2 inhibitors and GLP-1 receptor agonists), their prescription rates during the study period (2017–2018) were uniformly low (< 2% in all subgroups, Table 1). While we adjusted for the use of these agents in a sensitivity analysis (Supplementary Table S7), and the principal findings remained unchanged, the limited number of exposed patients precludes a meaningful assessment of their potential interaction with metformin or their independent effect on outcomes. This low utilization rate, however, reflects the real-world clinical practice in China at the time, which preceded the widespread guideline-recommended adoption of these therapies for patients with established atherosclerotic cardiovascular disease. Consequently, the potential interplay between metformin and these newer agents within the contemporary guideline-directed treatment landscape warrants further dedicated investigation.

Conclusions

Metformin use was associated with a lower incidence of 3-year MACCEs in diabetic patients undergoing non-complex PCI, while such protective effect didn’t exist in those undergoing complex PCI. The heterogeneity of metformin effect stratified by PCI complexity gives us more insights into metformin use in the specified population.

Author contributions

Z.Z. and Z.W. contributed equally to this work and share first authorship. Z.Z., Z.W., and J.H. were responsible for data curation, formal analysis, and validation. Y.S. assisted with investigation and methodology. W.S. and K.D. contributed equally as senior authors and were responsible for conceptualization, project administration, resources, supervision, and funding acquisition. All authors participated in writing – original draft preparation, review, and editing, and approved the final manuscript.

Funding

This study was supported by the Noncommunicable Chronic Diseases–National Science and Technology Major Project (2025ZD0548200), the Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (CIFMS) (2025‑I2M‑XHCL‑022, 2023‑I2M‑C&T‑B‑055), Beijing Clinical Key Specialized Projects (PM202401220002), Beijing Natural Science Foundation (24L60308), CAMS Fund for Clinical and Translational Medicine Research Project (2024‑I2M‑C&T‑B‑042), Noncommunicable Chronic Diseases–National Science and Technology Major Project (2024ZD0539300), and the National High Level Hospital Clinical Research Funding (2022‑GSP‑QN‑06, 2023‑GSP‑GG‑02, 2023‑GSP‑QN‑17, 2023‑GSP‑QN‑34, 2023‑GSP‑RC‑05).

Data availability

The datasets used during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Ethics approval and consent to participate

The study process was in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Fuwai hospital. All subjects provided informed written consent for long-term follow-up before intervention.

Trial registration

Not applicable.