Comparative effectiveness of dual antiplatelet therapy after revascularization in diabetes and chronic limb-threatening ischemia: a target trial emulation study

Abstract

Background

Patients with diabetes and chronic limb-threatening ischemia (CLTI) face high risks of adverse limb and cardiovascular events. While post-revascularization dual antiplatelet therapy (DAPT) is generally recommended, the optimal regimen remains uncertain. This study aimed to compare the effectiveness of different DAPT strategies with respect to long-term clinical outcomes.

Methods

We conducted a nationwide 90-day landmark observational analysis informed by a target trial framework based on healthcare claims data from Taiwan. The cohort included adults with type 2 diabetes who underwent lower-extremity revascularization for new-onset CLTI between 2016 and 2019. Patients were classified according to the post-procedural DAPT strategy initiated within a 90-day grace period. Inverse probability of treatment weighting (IPTW) was used to balance baseline covariates. The outcomes assessed were lower-extremity amputation (LEA), major adverse limb events (MALE), major adverse cardiovascular events (MACE), and all-cause mortality.

Results

The final cohort comprised 1,849 patients (aspirin plus clopidogrel, n = 439; aspirin plus cilostazol, n = 753; clopidogrel plus cilostazol, n = 657). Median follow-up years (IQR) were 2.94 (1.97–4.11) for all-cause mortality, 2.79 (1.62–3.96) for LEA, 2.40 (0.98–3.62) for MALE, and 2.55 (1.24–3.70) for MACE. Following IPTW, baseline characteristics were well balanced. Compared with standard DAPT (aspirin plus clopidogrel), aspirin plus cilostazol was associated with a lower risk of LEA (sHR, 0.672; 95% CI, 0.460–0.981) and MACE (sHR, 0.799; 95% CI, 0.641–0.995). Clopidogrel plus cilostazol did not demonstrate significant limb-related advantages but was associated with a lower incidence of MACE (sHR, 0.790; 95% CI, 0.631–0.989). All-cause mortality was comparable across all three regimens. Notably, in stratified analyses, the clinical benefits of cilostazol-based therapies were attenuated and statistically non-significant among patients receiving dialysis.

Conclusions

Among patients with diabetes and CLTI undergoing lower-extremity revascularization, aspirin plus cilostazol was associated with lower risks of LEA and MACE compared with aspirin plus clopidogrel, whereas clopidogrel plus cilostazol was associated with lower MACE but not with significant limb-related benefit. These observational findings suggest that aspirin plus cilostazol may be a favorable DAPT strategy for long-term outcomes in this population; however, further prospective randomized trials are warranted to verify.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12933-026-03290-1.

Introduction

Peripheral artery disease (PAD) is a major contributor to poor wound healing and limb loss in patients with diabetic foot ulcers (DFUs). Contemporary intersocietal guidelines indicate that approximately half of DFU patients have concomitant PAD, and its presence significantly exacerbates the risk of adverse limb outcomes [1]. Diabetes is also a major risk factor for PAD [1] because chronic hyperglycemia promotes a systemic proatherogenic state characterized by endothelial dysfunction, altered vascular permeability, inflammatory cell infiltration, and lipid deposition within the arterial wall, thereby accelerating atherosclerosis and diabetic vasculopathy [2].

In patients with diabetes and PAD, the development of a foot ulcer indicates the onset of chronic limb-threatening ischemia (CLTI) [3]. CLTI represents the most advanced stage of lower-extremity PAD, characterized by persistent hypoperfusion that leads to non-healing ulceration or gangrene. Because untreated CLTI carries a substantial risk of major amputation and mortality, revascularization is a cornerstone of management. Consequently, contemporary guidelines recommend endovascular or surgical revascularization for appropriate candidates to restore limb perfusion, promote wound healing, and prevent limb loss [4, 5]. Nevertheless, high rates of restenosis following revascularization remain a major limitation, with contemporary randomized trials demonstrating a primary patency of only approximately 52% at 12 months after standard percutaneous transluminal angioplasty [6, 7]. This restenosis is a multifactorial process driven by elastic recoil [8], thrombus formation [9], vascular remodeling [10], and subsequent intimal hyperplasia [11].

Furthermore, successful revascularization does not fully normalize a patient's long-term prognosis. PAD is a clinical manifestation of systemic atherosclerotic cardiovascular disease rather than an isolated lower-extremity disorder [12]. Patients with PAD are at a heightened risk not only for major adverse limb events (MALE), including acute limb ischemia and amputation, but also for major adverse cardiovascular events (MACE), including myocardial infarction and stroke, as well as premature death. Consequently, even after technically successful revascularization, patients frequently continue to experience recurrent lower limb ischemia, adverse cardiac events, and excess mortality.

These pathophysiologic and clinical considerations provide the rationale for utilizing dual antiplatelet therapy (DAPT) in selected patients after lower-extremity revascularization. Aspirin and a P2Y12 inhibitor (such as clopidogrel) act through complementary pathways to suppress platelet activation and aggregation, thereby reducing the risk of acute thrombotic occlusion, in-stent thrombosis, and recurrent systemic atherothrombotic events [13]. Clinical evidence indicates that post-revascularization DAPT is associated with lower risks of mortality and MALE [14]. For instance, the MIRROR trial demonstrated that DAPT significantly reduced the target lesion revascularization rates at 6 months (5% vs. 20%) [15], and the CASPAR trial showed that in the subgroup of patients with prosthetic grafts, DAPT was associated with fewer graft occlusions (hazard ratio 0.63) and above-ankle amputations (HR 0.48) [16]. Accordingly, current guidelines generally suggest that short-term DAPT (typically for 1–6 months) may be considered following revascularization to mitigate the risk of adverse limb events (Class IIb recommendation) [4, 5]. While short-term standard DAPT is often considered following revascularization, the evidence regarding its long-term efficacy in preventing MALE and MACE in patients with diabetes and CLTI remains uncertain and limited by high heterogeneity [14].

More intensive antithrombotic strategies that utilize dual-pathway inhibition with low-dose rivaroxaban plus aspirin have demonstrated benefits for composite outcomes (MALE, MACE, and mortality). However, this regimen was evaluated in comparison with aspirin monotherapy rather than standard DAPT [17]. In addition to aspirin and clopidogrel, cilostazol is another important antiplatelet agent. As a phosphodiesterase-3 inhibitor, its mechanism of action involves not only platelet inhibition but also vasodilation [18]. Compared with standard antiplatelet therapy alone, cilostazol has been shown to improve amputation-free survival and reduce the need for repeat revascularization following the procedure [19]. Furthermore, preoperative cilostazol use has been associated with lower perioperative amputation rates and improved long-term survival in patients with CLTI undergoing endovascular revascularization [20]. However, a critical knowledge gap remains: direct, head-to-head comparisons of cilostazol-based DAPT versus standard DAPT (aspirin plus clopidogrel) as a post-revascularization antithrombotic regimen are lacking.

Accordingly, the present study aimed to compare the long-term outcomes associated with three post-procedural DAPT strategies, including aspirin plus clopidogrel, aspirin plus cilostazol, and clopidogrel plus cilostazol, in patients with diabetes and CLTI undergoing lower extremity revascularization, focusing specifically on limb prognosis, cardiovascular events, and overall survival.

Materials and methods

Study design and data source

We conducted a comparative effectiveness study using a target trial emulation framework and nationwide claims data from the Health and Welfare Data Science Center, Ministry of Health and Welfare, Taiwan (HWDC, MOHW), spanning 2011 to 2021 to allow for a minimum 5-year lookback period for incident events. We emulated the target trial using a prespecified 90-day treatment-assignment grace period. Adults with type 2 diabetes who underwent lower-extremity revascularization for new-onset CLTI would be randomly assigned to one of three post-procedural DAPT strategies during this grace period, and outcome follow-up began at day 91 using a landmark design to align treatment assignment and the start of follow-up. Data analysis was performed from January 2024 to March 2026.

Taiwan established its National Health Insurance (NHI) program in 1995, which provides healthcare coverage to nearly 99% of the population. The National Health Insurance Research Database (NHIRD), derived from the NHI system, contains comprehensive longitudinal claims data and is widely used for population-based real-world research [21, 22]. All data access and analyses were conducted in accordance with the privacy and security regulations of the HWDC, MOHW. This study was approved by the Institutional Review Board of Chang Gung Medical Foundation, which waived the requirement for informed consent because all data were de-identified. The study was conducted in accordance with the Declaration of Helsinki as revised in 2024 and is reported in accordance with the STROBE guideline [23] and the TARGET reporting guideline for observational studies emulating a target trial [24].

Specification of the emulated target trial (supplementary appendix 1)

• Eligibility criteria: Adults with type 2 diabetes admitted for a first qualifying inpatient CLTI treatment episode who underwent lower-extremity revascularization during that admission.

• Treatment strategies: Three mutually exclusive post-procedural DAPT regimens, including aspirin plus clopidogrel, aspirin plus cilostazol, or clopidogrel plus cilostazol.

• Treatment assignment: Evaluated during a 90-day grace period following the index date (defined as the admission date for CLTI treatment).

• Start of follow-up (time zero): Day 91 following the index admission.

• Estimand: The intention-to-treat effect of the initially assigned DAPT strategy identified during the 90-day treatment-assignment period among patients who survived and remained free of major LEA and MACE through the landmark date.

• Analysis approach: Inverse probability of treatment weighting (IPTW) was utilized to balance measured baseline covariates across treatment groups, followed by weighted Fine–Gray subdistribution hazard models (for nonfatal outcomes) and Cox regression models (for mortality).

Eligibility criteria

To emulate this target trial, we identified adults aged 18 years or older with type 2 diabetes mellitus who experienced new-onset CLTI between 2016 and 2019 and underwent lower-extremity vascular intervention. Type 2 diabetes mellitus was defined as at least 1 hospitalization with ICD-10 code E11 or at least 3 diabetes-related outpatient visits. We restricted the study population to patients with type 2 diabetes to maintain a relatively homogeneous cohort.

CLTI was defined as peripheral artery disease concurrent with foot disease, including foot ulcer, infection, or gangrene, identified using the ICD-10 codes listed in Supplementary Appendix 2, together with receipt of lower-extremity revascularization. Endovascular treatment was defined by procedure codes for percutaneous transluminal angioplasty, including simple angioplasty (33074B) and complicated angioplasty (33115B). Surgical bypass grafting was identified by procedure code 69004B. The relevant NHI procedure codes are listed in Supplementary Appendix 3. To avoid counting recurrent episodes as incident events, only the first qualifying CLTI episode requiring vascular intervention was considered the index event, and patients were required to have no prior CLTI event during the preceding 5 years.

In routine practice, DAPT initiation after CLTI admission and lower-extremity revascularization may not occur immediately. Therefore, to align eligibility, treatment assignment, and the start of follow-up in the target trial emulation, we applied a 90-day treatment-assignment grace period after the index CLTI admission. Patients who experienced major LEA, MACE, or death during this 90-day period were excluded because they could not be classified consistently into one of the prespecified treatment strategies and would not have remained eligible to initiate follow-up under the emulated protocol. We also excluded patients who received only 1 antiplatelet agent, more than 2 antiplatelet agents, or any anticoagulant during the 90-day treatment-assignment period, because these regimens did not correspond to any of the 3 treatment strategies under comparison.

Treatment strategies and treatment assignment

The 3 treatment strategies in the emulated target trial were defined a priori as follows: (1) aspirin plus clopidogrel, (2) aspirin plus cilostazol, and (3) clopidogrel plus cilostazol. Patients were classified into one of the three mutually exclusive DAPT groups only if prescription records showed concurrent use of the two component antiplatelet agents for at least 28 days within the first 90 days after the index CLTI admission. This 90-day window served as the treatment-assignment grace period and was intended to reflect real-world clinical practice, in which DAPT initiation may not occur on the day of revascularization. Once patients were assigned to one of the three DAPT groups, they remained in that group for outcome analyses regardless of subsequent discontinuation, switching, or continuation after day 90. Therefore, the exposure definition reflects the initial DAPT regimen selected during the early post-procedural period, not sustained treatment exposure throughout follow-up.

The primary analysis estimated the effect of the initially assigned DAPT strategy under an intention-to-treat framework. Patients were not required to remain continuously on DAPT until day 91. Accordingly, once patients were assigned to 1 of the 3 DAPT strategies during the grace period, they remained in that group for subsequent outcome analyses regardless of later treatment changes, discontinuation, or switching after the start of follow-up.

Time zero and follow-up

In the target trial emulation, the eligibility date was the date of the index CLTI admission, whereas time zero for outcome follow-up was defined as day 91 after the index event. This landmark approach ensured that treatment classification was completed before the risk time began and reduced the immortal time bias that would arise if medication exposure were defined during the post-index period while follow-up began earlier. Patients were followed from day 91 after the index CLTI admission until the occurrence of the outcome of interest, death, or December 31, 2021, whichever came first.

Baseline covariates

Baseline covariates included age, sex, comorbidities, and concomitant medications. Comorbidities included hypertension, dyslipidemia, coronary heart disease, cerebrovascular accident, heart failure, renal disease, diabetic retinopathy, and diabetic neuropathy. These conditions were defined by at least 1 inpatient diagnosis or 2 outpatient diagnoses before the index date. Concomitant medications, including antidiabetic agents, antihypertensive agents, and statins, were identified using the same criterion of prescriptions lasting at least 28 days during the treatment-assignment period.

Prespecified kidney disease subgroup

Dialysis status was selected as a clinically relevant potential effect modifier. Patients receiving long-term dialysis represent a distinct high-risk subgroup with advanced uremic vasculopathy, diffuse arterial calcification, endothelial dysfunction, altered platelet function, increased thrombotic and bleeding risks, and potentially different pharmacodynamic responses to antiplatelet agents [25, 26]. These factors may influence the comparative effectiveness of DAPT regimens after lower-extremity revascularization. Therefore, exploratory subgroup analyses were prespecified according to dialysis status. Patients receiving long-term dialysis were classified as having end-stage renal disease, defined as dialysis treatment for at least 3 consecutive months. Dialysis modalities included hemodialysis (NHI procedure codes 58001C, 58019C, 58020C, 58021C, 58022C, 58023C, 58024C, 58025C, and 58029C), peritoneal dialysis (58002C, 58009A, 58009B, 58010A, 58010B, 58011A, 58011AB, 58011B, 58011C, 58012A, 58012B, 58017B, 58017C, 58026C, and 58028C), and other dialysis procedures (58018C, 58027C, and 58030B) [27]. Patients not meeting this definition were classified as non-dialysis patients.

Outcomes

The primary outcome designated for this study was LEA, which was identified using diagnosis or procedure codes for minor amputation (below the ankle) and major amputation (above the ankle), as listed in Supplementary Appendix 2. Secondary outcomes included MALE, MACE, and all-cause mortality. Subgroup analyses stratified by dialysis status were prespecified as exploratory. MALE was defined as a composite of major lower-extremity amputation or repeat lower-extremity revascularization, based on definitions used in prior peripheral artery disease revascularization studies [28, 29]. We defined MACE as a five-component composite cardiovascular endpoint, rather than the conventional three-point MACE, to capture broader clinically relevant cardiovascular events. Similar expanded composite endpoints have been used in large cardiovascular outcome trials, including components such as unstable angina hospitalization, heart failure hospitalization, coronary revascularization, and resuscitated cardiac arrest [30, 31]. All outcomes were ascertained from time zero onward.

Statistical analysis

Baseline characteristics were summarized across the 3 emulated treatment groups as counts with percentages for categorical variables and means with standard deviations for continuous variables. Between-group differences were assessed using the chi-square test for categorical variables and analysis of variance for continuous variables.

To emulate random treatment assignment conditional on measured covariates, we applied inverse probability of treatment weighting (IPTW) based on propensity scores to create a weighted pseudo-population in which baseline characteristics were balanced across treatment groups. Propensity scores were estimated using a multinomial logistic regression model because the treatment variable consisted of three mutually exclusive dual antiplatelet therapy regimens. The model included baseline demographic characteristics, comorbidities, diabetes-related complications, dialysis status, and concomitant medications. Standard inverse probability of treatment weights were then calculated as the inverse of the predicted probability of receiving the actual treatment administered. Covariate balance before and after IPTW was evaluated using standardized mean differences (SMDs). An absolute SMD < 0.10 was generally considered indicative of acceptable balance, which was successfully achieved for the vast majority of variables. Unlike propensity score matching, IPTW does not exclude unmatched individuals; rather, every patient is retained and assigned a specific weight. Therefore, the actual number of patients included in the weighted analysis remained identical to the original cohort size, whereas the weighting procedure modified each individual’s statistical contribution to the outcome models.

The primary estimate was the intention-to-treat effect of the initial DAPT strategy. For all-cause mortality, weighted Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). For nonfatal outcomes, including LEA, MALE, and MACE, death was treated as a competing event, and weighted Fine and Gray subdistribution hazard models were used to estimate subdistribution hazard ratios (sHRs) and 95% CIs. Cumulative incidence functions were generated for LEA, MALE, and MACE, and Kaplan–Meier methods were used for all-cause mortality.

Exploratory subgroup analyses were performed according to dialysis status. Effect modification was assessed by including cross-product interaction terms between treatment strategy and dialysis subgroup in the weighted regression models. As a sensitivity analysis, the primary outcome models were repeated using conventional unweighted multivariable regression. All tests were 2-sided, and a P value < 0.05 was considered statistically significant. All analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC, USA).

Results

Study population and cohort selection

Between January 2016 and December 2019, a total of 95,209 adults (≥ 18 years) were identified in the Taiwan Health and Welfare Data Center as having diabetic foot disease (DFD). After excluding 2,448 individuals with type 1 diabetes and 26,924 with any prior foot disease, 65,832 patients with newly diagnosed DFD remained. Among these, 13,393 patients underwent inpatient treatment for CLTI and were evaluated for inclusion. We further excluded 3,684 patients who experienced major lower-extremity amputation (LEA), MACE, or death within 90 days of the index date. Additional exclusions included patients who received no antiplatelet therapy (n = 3,220), single antiplatelet therapy (n = 2,597), three or more antiplatelet agents (n = 562), or any anticoagulant therapy (n = 559), as well as those who did not undergo a vascular intervention (n = 922). Consequently, the final analytic cohort comprised 1,849 patients treated with DAPT following revascularization. Of these, 439 received standard DAPT (aspirin plus clopidogrel), 753 received aspirin plus cilostazol, and 657 received clopidogrel plus cilostazol (Fig. 1).

Fig. 1.

Flowchart of Study Cohort Selection. Patients with type 2 diabetes and CLTI who underwent revascularization and received post-procedural DAPT were included in the final analysis after the exclusion criteria were applied

Characteristics of patients with diabetes and CLTI across DAPT regimens

Baseline demographic and clinical characteristics, comorbidities, and concomitant medications across the three DAPT regimens are summarized in Table 1. The mean age differed significantly across groups (69.41 ± 9.79 years in aspirin plus clopidogrel, 70.70 ± 12.27 in aspirin plus cilostazol, and 73.17 ± 10.91 in clopidogrel plus cilostazol; P < 0.001), whereas sex distribution was comparable (P = 0.093).

Table 1.

Baseline Clinical Characteristics Across DAPT Regimens

	Aspirin plus clopidogrel (n = 439)	Aspirin plus cilostazol (n = 753)	Clopidogrel plus cilostazol (n = 657)	P value	Post- IPTW SMD† (aspirin plus cilostazol)	Post- IPTW SMD† (clopidogrel plus cilostazol)
Characteristic
Age	69.41 ± 9.79	70.70 ± 12.27	73.17 ± 10.91	< 0.001*	− 0.026	− 0.012
Gender				0.093	− 0.021	− 0.017
Male	271 (61.73%)	458 (60.82%)	368 (56.01%)
Female	168 (38.27%)	295 (39.18%)	289 (43.99%)
Comorbidities
Hypertension	362 (82.46%)	600 (79.68%)	542 (82.5%)	0.316	− 0.004	− 0.027
Dyslipidemia	230 (52.39%)	363 (48.21%)	274 (41.7%)	0.002*	0.022	0.017
Diabetic retinopathy	82 (18.68%)	156 (20.72%)	143 (21.77%)	0.463	− 0.014	− 0.043
Diabetic neuropathy	64 (14.58%)	148 (19.65%)	113 (17.2%)	0.081	− 0.033	− 0.036
Coronary heart disease	264 (60.14%)	285 (37.85%)	314 (47.79%)	< 0.001*	0.024	0.033
Cerebrovascular accident	82 (18.68%)	141 (18.73%)	169 (25.72%)	0.002*	0.031	0.027
Heart failure	59 (13.44%)	80 (10.62%)	77 (11.72%)	0.344	− 0.006	0.006
Dialysis	148 (33.71%)	248 (32.93%)	284 (43.23%)	< 0.001*	0.009	− 0.002
Medications
Anti-Diabetes drugs
Sulfonylurea	133 (30.3%)	266 (35.33%)	213 (32.42%)	0.185	− 0.157	− 0.199
Metformin	185 (42.14%)	307 (40.77%)	201 (30.59%)	< 0.001*	0.011	0.077
Acarbose	28 (6.38%)	75 (9.96%)	55 (8.37%)	0.101	− 0.122	− 0.116
Glinide	69 (15.72%)	97 (12.88%)	100 (15.22%)	0.303	0.113	0.086
Pioglitazone	35 (7.97%)	61 (8.1%)	45 (6.85%)	0.645	0.019	0.042
DPP4 inhibitor	216 (49.2%)	345 (45.82%)	297 (45.21%)	0.393	0.058	0.103
SGLT2 inhibitor	32 (7.29%)	30 (3.98%)	17 (2.59%)	< 0.001*	− 0.006	− 0.012
Insulin	146 (33.26%)	299 (39.71%)	235 (35.77%)	0.067	− 0.132	− 0.035
GLP-1 analog	4 (0.91%)	9 (1.2%)	4 (0.61%)	0.516	− 0.001	0.004
Antihypertensive agents
RAAS blocker	259 (59%)	406 (53.92%)	290 (44.14%)	< 0.001*	0.014	0.012
Calcium channel blocker	236 (53.76%)	413 (54.85%)	352 (53.58%)	0.878	0.019	− 0.014
β-blocker	216 (49.2%)	306 (40.64%)	273 (41.55%)	0.010*	0.035	0.038
Statin	286 (65.15%)	351 (46.61%)	290 (44.14%)	< 0.001*	0.036	0.043

^* P < 0.05

^† IPTW SMD: Inverse probability of treatment weighting standardized mean difference

The number of patients remains unchanged before and after IPTW, as the weighting approach retains the entire cohort for analysis

Before IPTW weighting, baseline comorbidities differed significantly across the three DAPT groups. The prevalence of hypertension was high and comparable among patients receiving aspirin plus clopidogrel, aspirin plus cilostazol, and clopidogrel plus cilostazol (82.46%, 79.68%, and 82.50%, respectively; P = 0.316). Dyslipidemia was more frequent in the aspirin plus clopidogrel group than in the aspirin plus cilostazol and clopidogrel plus cilostazol groups (52.39%, 48.21%, and 41.70%, respectively; P = 0.002). The prevalence of retinopathy and neuropathy did not differ significantly. Coronary heart disease (CHD) was most common in patients receiving aspirin plus clopidogrel (60.14%), followed by those receiving clopidogrel plus cilostazol (47.79%) and aspirin plus cilostazol (37.85%) (P < 0.001). In contrast, cerebrovascular accidents (CVA; P = 0.002) and dialysis (P < 0.001) were most prevalent in the clopidogrel plus cilostazol group (25.72% and 43.23%, respectively). Heart failure was similarly distributed across the treatment arms. After weighting, all comorbidity imbalances were substantially attenuated, with post-weighting standardized mean differences (SMDs) below 0.1, indicating adequate covariate balance (Supplementary Appendix 4).

Medication use also varied across the three DAPT groups before weighting. Among glucose-lowering therapies, metformin use was more common in the aspirin plus clopidogrel and aspirin plus cilostazol groups than in the clopidogrel plus cilostazol group (42.14%, 40.77%, and 30.59%, respectively; P < 0.001). Sodium–glucose cotransporter 2 (SGLT2) inhibitor use was highest in the aspirin plus clopidogrel group and progressively lower in the aspirin plus cilostazol and clopidogrel plus cilostazol groups (7.29%, 3.98%, and 2.59%, respectively; P < 0.001). The use of sulfonylureas, acarbose, glinides, pioglitazone, dipeptidyl peptidase-4 (DPP-4) inhibitors, insulin, and glucagon-like peptide-1 (GLP-1) analogs did not differ significantly among the three groups. Regarding cardiovascular medications, renin–angiotensin–aldosterone system (RAAS) blockers were prescribed more frequently in the aspirin plus clopidogrel group compared with the aspirin plus cilostazol and clopidogrel plus cilostazol groups (59.00%, 53.92%, and 44.14%, respectively; P < 0.001). β-blocker use also differed significantly, with the highest proportion observed in the aspirin plus clopidogrel group (49.20%) versus the aspirin plus cilostazol (40.64%) and clopidogrel plus cilostazol (41.55%) groups (P = 0.010). Calcium channel blocker use was comparable across all treatment arms. Notably, statin use was substantially higher in the aspirin plus clopidogrel group (65.15%) than in the aspirin plus cilostazol (46.61%) and clopidogrel plus cilostazol (44.14%) groups (P < 0.001). After weighting, these major baseline imbalances in medication use were markedly attenuated, indicating successful covariate balance across the treatment arms.

Event analysis between different DAPT regimens

The median follow-up was 2.94 years (IQR 1.97–4.11) for all-cause mortality, 2.79 years (IQR 1.62–3.96) for LEA, 2.40 years (IQR 0.98–3.62) for MALE, and 2.55 years (IQR 1.24–3.70) for MACE. The 5-year event-free rates in the aspirin plus clopidogrel, aspirin plus cilostazol, and clopidogrel plus cilostazol groups were 45.21%, 48.74%, and 36.38%, respectively, for LEA; 34.49%, 39.70%, and 27.04% for MALE; and 35.33%, 40.25%, and 29.51% for MACE (Fig. 2).

Fig. 2.

Five-year event-free rates in patients with diabetes and CLTI undergoing revascularization, stratified by DAPT regimen. The 5-year LEA-free survival rates were 45.21% in the aspirin plus clopidogrel group, 48.74% in the aspirin plus cilostazol group, and 36.38% in the clopidogrel plus cilostazol group. A similar pattern was observed for MALE-free survival, with rates of 34.49%, 39.70%, and 27.04%, respectively. For MACE, the 5-year event-free rates were 35.33% for aspirin plus clopidogrel, 40.25% for aspirin plus cilostazol, and 29.51% for clopidogrel plus cilostazol. The 5-year overall survival rates were 50.93%, 52.48%, and 40.68%, respectively. (LEA: lower-extremity amputation; MALE: major adverse limb event; MACE: major adverse cardiovascular event)

Figure 3 presents the weighted cumulative incidence curves for nonfatal outcomes (LEA, MALE, and MACE) and all-cause mortality across the three DAPT regimens. These curves were adjusted using IPTW to account for baseline demographics (age and sex), comorbidities (hypertension, dyslipidemia, CHD, CVA, and heart failure), diabetes complications, and medication use (RAAS blockers, β-blockers, statins, SGLT2 inhibitors, and GLP-1 analogs). Compared with aspirin plus clopidogrel (the reference group), aspirin plus cilostazol was associated with a significantly lower cumulative incidence of LEA (sHR, 0.672; 95% CI, 0.460–0.981; P = 0.040) and MACE (sHR, 0.799; 95% CI, 0.641–0.995; P = 0.045), alongside a borderline but nonsignificant reduction in MALE (sHR, 0.816; 95% CI, 0.651–1.024; P = 0.079). In contrast, clopidogrel plus cilostazol showed no significant difference regarding LEA (sHR, 0.941; 95% CI, 0.652–1.358; P = 0.745) or MALE (sHR, 0.998; 95% CI, 0.796–1.252; P = 0.988), but it was associated with a lower cumulative incidence of MACE (sHR, 0.790; 95% CI, 0.631–0.989; P = 0.040). Regarding all-cause mortality, the 5-year survival rates were 50.93%, 52.48%, and 40.68% across the three groups, respectively (Fig. 2); however, no statistically significant differences in mortality were observed for either aspirin plus cilostazol (HR, 0.886; 95% CI, 0.739–1.063; P = 0.192) or clopidogrel plus cilostazol (HR, 1.056; 95% CI, 0.879–1.268; P = 0.561) relative to aspirin plus clopidogrel (Fig. 3).

Fig. 3.

Cumulative Incidence and Survival Curves for Clinical Outcomes. Cumulative incidence and survival curves for clinical outcomes. The graphs display the weighted cumulative incidence curves for a lower-extremity amputation (LEA), b major adverse limb events (MALE), and c major adverse cardiovascular events (MACE), alongside the survival curve for (d) all-cause mortality, stratified by DAPT regimen. To provide a clear representation of the actual cohort size over time, the numbers at risk displayed below each plot represent the unweighted (actual) number of patients remaining at risk at each specified time point, whereas the curves reflect the IPTW-weighted event probabilities

(a) LEA: Aspirin plus cilostazol was associated with a significantly lower risk of LEA (sHR, 0.672; 95% CI, 0.460–0.981). (b) MALE: A non-significant reduction in MALE (sHR, 0.816; 95% CI, 0.651–1.024) was observed in the aspirin plus cilostazol group relative to standard DAPT. (c) MACE: A lower cumulative incidence of MACE was noted in both cilostazol-based DAPT regimens (aspirin plus cilostazol: sHR, 0.799; 95% CI, 0.641–0.995; clopidogrel plus cilostazol: sHR, 0.790; 95% CI, 0.631–0.989). (d) All-cause mortality: No significant differences were observed across the treatment arms.

Dialysis-stratified clinical outcomes

Figure 4 presents the dialysis-stratified, IPTW-weighted time-to-event analyses comparing the DAPT regimens. Among patients not receiving dialysis, aspirin plus cilostazol demonstrated a non-significant trend toward a lower risk of LEA (sHR, 0.644; 95% CI, 0.366–1.133; P = 0.127) and MALE (sHR, 0.748; 95% CI, 0.540–1.035; P = 0.080). In contrast, clopidogrel plus cilostazol was associated with a significantly lower risk of MACE (sHR, 0.723; 95% CI, 0.532–0.984; P = 0.039), whereas no significant differences were observed regarding LEA (sHR, 0.843; 95% CI, 0.485–1.465; P = 0.544), MALE (sHR, 0.985; 95% CI, 0.714–1.361; P = 0.929), or all-cause mortality (HR, 1.115; 95% CI, 0.855–1.454; P = 0.421).

Fig. 4.

Dialysis-Stratified Clinical Outcomes Across DAPT Regimens. Among patients not receiving dialysis, aspirin plus cilostazol was associated with a non-significant reduction in the risk of LEA (sHR, 0.644; 95% CI, 0.366–1.133) and MALE (sHR, 0.748; 95% CI, 0.540–1.035) compared with standard DAPT. In contrast, clopidogrel plus cilostazol was associated with a significantly lower risk of MACE (sHR, 0.723; 95% CI, 0.532–0.984), with no significant differences observed for other limb outcomes or overall survival. Among patients receiving dialysis, neither cilostazol-based regimen demonstrated a significant association with improved clinical outcomes relative to aspirin plus clopidogrel

Among patients receiving dialysis, neither aspirin plus cilostazol nor clopidogrel plus cilostazol was significantly associated with reduced risks of LEA (sHR, 0.693; 95% CI, 0.416–1.156; P = 0.160 and sHR, 1.022; 95% CI, 0.625–1.670; P = 0.931, respectively), MALE (sHR, 0.891; 95% CI, 0.649–1.223; P = 0.476 and sHR, 0.998; 95% CI, 0.727–1.370; P = 0.988, respectively), MACE (sHR, 0.792; 95% CI, 0.568–1.103; P = 0.167 and sHR, 0.868; 95% CI, 0.624–1.207; P = 0.399, respectively), or all-cause mortality (HR, 0.904; 95% CI, 0.699–1.169; P = 0.440 and HR, 1.050; 95% CI, 0.809–1.362; P = 0.714, respectively) compared with the reference regimen. Formal interaction testing did not show statistically significant effect modification by dialysis status for LEA, MALE, MACE, or all-cause mortality (Fig. 4). Therefore, the dialysis-stratified findings should be interpreted as exploratory rather than as evidence of confirmed treatment-effect heterogeneity.

Discussion

This cohort study provides important insights into the comparative effectiveness of three DAPT regimens, including standard DAPT (aspirin plus the P2Y12 inhibitor clopidogrel), aspirin plus cilostazol, and clopidogrel plus cilostazol, in patients with diabetes and PAD who underwent revascularization for new-onset CLTI. Overall, we found no significant differences in long-term survival among the three DAPT regimens. Interestingly, compared with standard DAPT, aspirin plus cilostazol was associated with a better limb prognosis. This included a significantly reduced risk of LEA and a non-significant reduction in MALE (a composite of major amputation and repeat revascularization). Furthermore, both aspirin plus cilostazol and clopidogrel plus cilostazol were associated with a lower incidence of MACE than standard DAPT. Under the intention-to-treat framework, however, patients were analyzed according to their initially classified regimen, and subsequent discontinuation, switching, or continuation was not modeled. Therefore, the associations should not be interpreted as the effect of continuous or sustained DAPT use.

The benefits of aspirin plus cilostazol on lower limb prognosis may be partly explained by the dual mechanism of action of cilostazol, which not only inhibits platelets but also promotes vasodilation. This vasodilation is achieved by increasing cyclic adenosine monophosphate (cAMP) levels in vascular smooth muscle cells, thereby enhancing blood flow [18]. Existing literature supports the idea that cilostazol confers clinically meaningful benefits following lower-extremity endovascular revascularization, driven chiefly by reductions in restenosis and the need for repeat revascularization [19, 32, 33]. However, consistent with our study's findings, all-cause mortality does not significantly differ between cilostazol and control groups [19, 32, 33]. In a randomized trial of patients with intermittent claudication undergoing femoropopliteal endovascular therapy, adjunctive cilostazol significantly increased two-year freedom from repeat revascularization (84.6% vs. 62.2%) and reduced restenosis rates (43.6% vs. 70.3%) [34]. Similarly, the STOP-IC trial demonstrated that cilostazol significantly improved 12-month rates of angiographic restenosis (20% vs. 49%) [35]. Importantly, these findings should not be interpreted as resulting from cilostazol monotherapy, because background aspirin therapy was widely used in the aforementioned studies. Therefore, the limb-related benefit observed in our study is more appropriately interpreted as an association specific to a combination strategy with aspirin plus cilostazol rather than as evidence of a uniform class effect of cilostazol-based DAPT.

Unlike aspirin plus cilostazol, clopidogrel plus cilostazol did not demonstrate a limb-related advantage following revascularization in the current study. This finding was unexpected, as current guidelines favor clopidogrel over aspirin as single antiplatelet therapy for symptomatic PAD [4, 5]. However, this recommendation derives largely from the PAD subgroup of the CAPRIE trial, in which clopidogrel reduced a composite of ischemic stroke, myocardial infarction, and vascular death, rather than improving limb-specific outcomes such as amputation or restenosis [36]. Consistent with this observation, subsequent studies have failed to establish a clear limb-related superiority for clopidogrel-based therapies; neither network meta-analyses [37] nor the CASPAR trial [16] demonstrated a significant overall reduction in amputations or improvement in limb outcomes. Another critical factor is the potential for clopidogrel resistance, which is particularly prevalent in Asian populations due to genetic polymorphisms (CYP2C19) that impair the drug's activation, resulting in reduced responsiveness in up to 70% of Asian individuals [38]. Given that our study was conducted in an Asian population, this genetic mechanism may have diminished the effectiveness of clopidogrel-based therapy. Furthermore, residual confounding by indication must be considered. Clopidogrel plus cilostazol may have been preferentially prescribed to patients with aspirin intolerance or a history of prior bleeding, factors that were not completely captured in claims data and may have biased this group toward worse limb outcomes despite statistical weighting. Accordingly, the lack of an apparent benefit with clopidogrel plus cilostazol should be interpreted cautiously and not viewed as definitive evidence of inferiority.

Our study also suggests that cilostazol-containing combination therapies were associated with a reduced risk of MACE. Although cilostazol is an established treatment for intermittent claudication and secondary stroke prevention, its clear benefit in broader atherosclerotic cardiovascular disease has not been firmly established [39]. Nevertheless, accumulating evidence suggests that cilostazol may offer clinically relevant cardiovascular protection. For instance, in a randomized trial involving patients with coronary artery disease or high cardiovascular risk, cilostazol significantly reduced MACE [40]. Furthermore, in patients with type 2 diabetes and subclinical coronary atherosclerosis, the ESCAPE study demonstrated that cilostazol reduced coronary stenosis and non-calcified plaque burden compared with aspirin and also showed a lower long-term incidence of cardiovascular events in its extension study [41, 42]. Several mechanistic theories help explain these clinical observations. Beyond platelet inhibition and vasodilation, other reported potential protective effects include improved endothelial function, the suppression of both vascular smooth muscle proliferation and inflammatory activity, attenuated carotid plaque progression, slower progression of carotid intima-media thickness, and favorable lipid profiles [40, 41, 43, 44]. However, these mechanisms remain inferential in the context of our claims-based observational study. Moreover, because residual confounding and potential survivor-selection bias related to the treatment assignment period cannot be excluded, the observed associations should be interpreted cautiously. Therefore, these findings are best viewed as regimen-specific, hypothesis-generating associations rather than definitive evidence of a broad cilostazol-based DAPT class effect.

An interesting finding is the heterogeneity of effects by dialysis status. In dialysis-stratified analyses, no statistically significant associations between cilostazol-containing regimens and limb or cardiovascular outcomes were observed in the dialysis subgroup. Of note, the interaction tests did not support significant effect modification by dialysis status. Therefore, these subgroup findings should be interpreted cautiously as exploratory and hypothesis-generating. A possible explanation for the attenuated benefit of DAPT in these patients is the extreme severity and complexity of their vasculopathy. In end-stage renal disease, vascular disease is driven not only by thrombosis and atherosclerosis but also by diffuse medial calcification and marked arterial stiffness resulting from uremia-related inflammation [45]. Chronic kidney disease (CKD) promotes a proatherogenic vascular environment characterized by macrophage infiltration and trapping, impaired cholesterol efflux, and increased oxidative stress, all of which accelerate plaque progression [46] and may not be adequately addressed by antiplatelet therapy alone [25]. Furthermore, the pharmacological efficacy of these agents may be inherently diminished in renal disease. Clopidogrel exhibits a weakened antiplatelet effect in patients with CKD, leading to high on-treatment platelet reactivity and potentially worse clinical outcomes [26], while cilostazol has also shown attenuated efficacy in stroke-prevention trials involving patients with advanced CKD [47].

Our findings should be interpreted in the context of several limitations. First, the reliance on a retrospective claims database introduces potential misclassification bias and precludes the assessment of granular clinical parameters, such as precise PAD severity, medication adherence, and drug doses. Second, despite rigorous adjustment using IPTW, residual confounding by indication cannot be entirely excluded given the nonrandomized treatment allocation. Third, a major limitation of this study is the absence of bleeding outcomes. Although we evaluated limb, cardiovascular, and mortality outcomes, we were unable to assess major bleeding, procedure-related bleeding, or bleeding requiring transfusion or hospitalization. Therefore, the present findings should be interpreted as comparative effectiveness estimates for ischemic and limb outcomes only, rather than as evidence of overall safety or net clinical benefit. Another limitation of this 90-day landmark design is that the results should not be interpreted as the total effect of immediate DAPT assignment at admission or revascularization, nor can they be generalized to very early post-admission events. Survivor-selection bias and residual confounding related to early treatment selection remain possible. Nonetheless, this nationwide cohort provides robust real-world evidence regarding the comparative effectiveness of DAPT strategies in a high-risk population.

Conclusion

In conclusion, among patients with diabetes and CLTI undergoing lower-extremity revascularization, different DAPT regimens were associated with varied long-term clinical outcomes. Overall survival was comparable across the evaluated strategies. Aspirin plus cilostazol was associated with lower risks of LEA and MACE compared with standard DAPT (aspirin plus clopidogrel), whereas clopidogrel plus cilostazol was associated with lower MACE but not with a significant limb-related benefit. Exploratory dialysis-stratified analyses did not confirm significant effect modification by dialysis status. These findings do not establish a uniform class effect of cilostazol-based DAPT and should be interpreted cautiously given the observational design, potential survivor-selection bias, and residual confounding. Prospective studies or randomized trials are needed to confirm the optimal post-revascularization DAPT strategy.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

PAD

Peripheral artery disease

DFD

Diabetic foot disease

DFU

Diabetic foot ulcer

CLTI

Chronic limb-threatening ischemia

DAPT

Dual antiplatelet therapy

HWDC

Health and welfare data science center

MOHW

Ministry of health and welfare

NHI

National health insurance

NHIRD

National health insurance research database

IPTW

Inverse probability of treatment weighting

LEA

Lower-extremity amputation

MALE

Major adverse limb event

MACE

Major adverse cardiovascular event

CHD

Coronary heart disease

CVA

Cerebrovascular accident

ESRD

End-stage renal disease

CKD

Chronic kidney disease

sHR

Subdistribution hazard ratio

CI

Confidence interval

SMD

Standardized mean difference

SGLT2

Sodium–glucose cotransporter 2

DPP-4

Dipeptidyl peptidase-4

GLP-1

Glucagon-like peptide-1

RAAS

Renin–angiotensin–aldosterone system

cAMP

Cyclic adenosine monophosphate

Author contributions

C.-W.L. reached data and wrote the manuscript. P.-H.L. researched data and contributed to the study design. C.-H.L. contributed to the study concept. C.-H.H. contributed to interpretation of data and discussion. S.-Y.H. and I-W.C reviewed/edited the manuscript. Y.-C.C. provided statistical and methodological support. Y.-Y.H. contributed to the discussion and reviewed/edited the manuscript. All authors read and approved the final manuscript.

Funding

This study was supported by grants from Linkou Chang-Gung Memorial Hospital (CFRPG3P0011). The funders had no role in the study design, data collection, analysis and interpretation, preparation of the manuscript, and the decision to publish the results.

Data availability

The data that supports the findings of this study are available from the corresponding author upon reasonable request. Restrictions apply to the availability of these data, which were used under license for this study. Data are available from the authors only with the official permission of the Ministry of Health and Welfare, Taiwan.

Declarations

Ethics approval and consent to participate

The studies involving human participants were reviewed and approved by The Institutional Review Board of Chang Gung Memorial Hospital, Taiwan (No. 202301438B0). Written informed consent for participation was not required for this study in accordance with the national legislation and the institutional requirements.

Competing interests

The authors declare no competing interests.

Consent for publication

Not applicable.

Trial registration

Not applicable.