Glucagon-like peptide-1 receptor agonists and major limb events in adults with type 2 diabetes and peripheral artery disease: a systematic review and meta-analysis of RCTs and cohort studies

Abstract

Background

The impact of glucagon-like peptide-1 receptor agonists (GLP-1RAs) on peripheral artery disease (PAD) remains uncertain. We assessed the association between GLP-1RA use and major limb events in individuals with type 2 diabetes (T2D) and PAD.

Methods

We performed a literature search from inception to 31 March 2026 for randomized controlled trials and cohort studies comparing GLP-1RA use versus placebo, active comparators or non-use on lower limb outcomes in individuals with T2D and PAD. The primary outcome was a composite of major limb events, as defined by the investigators of the original studies included in the meta-analysis. Secondary outcomes included lower extremity amputation (LEA), revascularization, gangrene, major adverse cardiovascular events (MACE) and all-cause mortality. Subgroup analyses, leave-one-out sensitivity analyses, and meta-regression analysis were also performed.

Results

Twelve studies (2 RCTs and 10 matched cohort studies) with thirteen arms involving 418,282 participants were included. Treatment with GLP-1RAs was associated with a significantly lower risk of major limb events by 27% (thirteen arms, RR 0.73, 95% CI 0.65–0.82). Reduced risks were also observed for LEA (RR 0.76, 95% CI 0.66–0.87), revascularization (RR 0.81, 95% CI 0.77–0.86), gangrene (RR 0.80, 95% CI 0.77–0.85), MACE (RR 0.76, 95% CI 0.63–0.90) and all-cause mortality (RR 0.67, 95% CI 0.61–0.73). Subgroup and sensitivity analyses did not substantially change point estimates.

Conclusion

Among adults with T2D and PAD, treatment with GLP-1RAs was associated with a lower risk of major limb events. There is a need of dedicated studies with standardized limb-specific endpoints to confirm the protective role of GLP-1RAs on limb events in people with T2D and PAD.

Graphical abstract

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12933-026-03283-0.

Introduction

Peripheral artery disease (PAD) affects about 236 million individuals worldwide [1]. PAD is a progressive condition characterized by stenosis and occlusions in the peripheral arterial bed and is associated with a higher risk of myocardial infarction, stroke and cardiovascular mortality [2]. The prevalence of PAD among individuals with type 2 diabetes (T2D) ranges from 15.2% to 40% which is much higher compared to the general population (5%–5.6%) [3]. Compared with individuals without diabetes, people with diabetes are five times more likely to have an amputation and two times more likely to die, with death occurring at a younger age [4]. Moreover, individuals with T2D have 94% higher odds for having atherosclerotic disease in the tibial vessels compared with those without diabetes [5].

Considering the large prevalence of diabetes worldwide and its future projection to the outstanding number of 1.3 billion by 2050 [6], it is relevant to improve the management of PAD in people with diabetes. Recent guidelines from the American Diabetes Association [7], the American College of Cardiology/American Heart Association [8], and the European Society of Cardiology [9] recommend the use of glucagon-like peptide-1 receptor agonists (GLP-1RAs) and sodium-glucose cotransporter 2 inhibitors (SGLT-2 inhibitors) in individuals with T2D and PAD to lower their elevated risk of major adverse cardiovascular events (MACE). While overwhelming evidence support the cardiovascular benefits of these drugs in T2D, the specific impact of GLP-1RAs in individuals with T2D and PAD remains uncertain [10–13]; one main reason for the uncertainty may lie in mixing diabetic patients with or without PAD at baseline.

To address this, we performed an updated systematic review and meta-analysis of randomized controlled trials (RCTs) and cohort studies to evaluate the effects of GLP-1RAs on lower extremity outcomes in individuals with T2D and PAD.

Methods

This systematic review with meta-analysis was performed according to PRISMA (preferred Reporting Items for Systematic Reviews and Meta-Analysis) guidelines [14]. The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO, CRD420261284846). Ethical approval was not obtained for this study because it consisted of a study‐level systematic review and meta‐analysis.

Search strategy and study selection

A comprehensive search of MEDLINE (via PubMed), Embase, and Cochrane Central Register of Controlled Trials (CENTRAL) was performed from inception to 31 March 2026 using a combination of search terms related to T2D, GLP-1RAs and major limb events, with no restrictions on language or publication status. The complete search strategy is provided in the Supplementary Appendix (Additional file 1: Tables S1-S3). Reference lists of prior reviews and former meta-analyses were also manually searched to identify other relevant studies.

Two investigators (P.C. and D.G.) independently screened records at the title, abstract and full text level. Discrepancies were resolved by a third investigator (L.S.). Eligible studies were RCTs and cohort studies in individuals with T2D and PAD that evaluated the effects of GLP-1RAs in comparison with placebo, other antidiabetic therapies or GLP-1RA non-use on major limb events, MACE and all-cause mortality. The prespecified selection criteria included: (1) RCTs comparing GLP-1RAs with placebo; (2) cohort studies comparing GLP-1RA use versus other antidiabetic drugs or non-use, and assessing matched groups obtained with methods to reduce bias due to confounding variables that could affect the treatment effect estimate (propensity score matching, inverse probability weighting); and (3) follow-up duration of at least 9 months.

Data extraction and risk of bias assessment

Data were extracted by P.C. and D.G. using a standardized extraction form, with conflicts over study inclusion resolved by consensus. The retrieved data included study characteristics, characteristics of patients, interventions, and outcome measures that included the hazard ratios (HR) or risk ratios (RR) and 95% confidence intervals (CI). The Cochrane risk-of-bias tool version 2 (RoB 2) was used for quality assessment of the RCTs [15] which includes five domains: randomization, interventions, missing outcome data, measurement of the outcome and selection of the reported result. The overall risk of bias for each study was judged as low risk, some concerns, or high risk. For cohort studies, we used the Risk of Bias in Non-randomised Studies—of Interventions (ROBINS-I) tool which covers seven domains: confounding, selection of participants, classification of intervention, deviation from intended intervention, missing data, measurement of the outcome, and selection of reported results [16]. The overall risk of bias for each study was judged as low, moderate, serious, critical, or no information. Two investigators (N.D.M. and M.I.M.) independently assessed the risk of bias for all included studies. Disagreements were resolved by a third investigator (M.L.).

Outcomes of interest

The primary outcome was a composite of major limb events as defined by the investigators in the original RCTs or cohort studies included in the meta-analysis (Additional file 1: Table S4). Secondary outcomes were selected only in those studies (RCTs or cohorts) that reported on major limb events, and included lower extremity amputation (LEA), revascularization, gangrene, MACE and all-cause mortality.

Statistical methods

A random-effects model was used to calculate outcome measures; effect estimates were expressed as risk ratios (RR) and the corresponding 95% CI. As the observational cohorts reported time-to-event data expressed as Hazard Ratios (HR) or sub-distribution Hazard Ratios (sHR), and one study (30) reported outcomes as RR, values were converted to RR to establish a uniform summary metric and to avoid overestimation of the treatment effect in the presence of high event rates (> 10%). Therefore, HR and sHR were converted to RR using the formula described by Symons and Moore [17] which adjusts for the cumulative incidence in the control group (P0), as follows: RR = [1 – eHR x ln(1-P0)]/P0. P0 was calculated based on the reported number of events and sample size or derived from incidence rates (events per person-years) over the mean follow-up period. Heterogeneity was evaluated using the Cochrane Q-test, with a P value < 0.10 indicating significant heterogeneity. The I² statistic was used to quantify the extent of inconsistency across studies, classified as low (< 25%), moderate (25%-75%) or high (> 75%) [18]. To assess heterogeneity, subgroup analyses were performed for the primary outcome based on the type of study (RCT or cohort), comparators and length of treatment (meta-regression). Publication bias was assessed for study outcomes with more than ten studies using a funnel plot and Egger’s test, with a P value < 0.10 considered indicative of significant asymmetry [19]. When potential bias was detected, the trim and-fill method was applied to estimate and adjust the pooled effect size for hypothetical missing studies [20]. A leave-one-out sensitivity analysis was conducted to evaluate the influence of individual studies on the overall pooled effect size by iteratively excluding one study at a time. Finally, to address the potential loss of the time-to-event component resulting from the RR conversion, a sensitivity analysis was performed by directly pooling HRs and sHRs for the primary outcome, restricted to the subset of studies reporting these measures. All statistical tests were two-sided, and P values < 0.05 were regarded as significant. Analyses were performed using IBM SPSS Statistics software (version 30.0).

Certainty of evidence

The certainty of evidence was assessed with GRADE (Grading of Recommendations Assessment, Development and Evaluation) [21]. Two authors independently applied the GRADE criteria for each outcome (N.D.M. and D.G.), with any disagreements resolved by a third reviewer (L.S.).

Results

After removing duplicates and records screened on the basis of title and abstract, 114 study reports were assessed for eligibility. Twelve studies were included, two RCTs [22, 23] and ten cohort studies [24–33], with thirteen arms (Additional file 1: Fig. S1). The study of Yehualashet et al. [34] was ineligible because the two cohorts were not matched.

An overview of the study and participant characteristics are summarized in Table 1. The overall participants’ number was 418,282 (3592 participants in RCTs and 414,690 participants in matched cohort studies), follow-up ranged from 9.6 to 60 months and participants’ mean age was 63.9 ± 11.7. Two arms compared GLP-1RAs versus placebo [22, 23], four arms [24–26, 29] compared GLP-1RAs vs SGLT-2 inhibitors, two arms [28, 32] compared GLP-1RAs vs therapies without GLP-1RAs, two arms [29, 33] compared GLP-1RAs vs DPP-4 inhibitors, two arms [27, 30] compared semaglutide use versus non-use, one arm [31] compared tirzepatide use versus non-use. The primary outcome of the included studies and the outcomes of this meta-analysis are also given in Table 1.

Table 1.

Characteristics of studies included in the meta-analysis

First author, year (ref.)	Study Design	Follow-up, months	Population	Age, years (Int/Ctrl)	No. of patients (Int/Ctrl)	Intervention	Comparator	Study Primary Outcome	Outcome included in Meta-Analysis
Badjatiya [22]	RCT, double blind (post-hoc analysis)	38.4	T2D with PAD	61.9	1400/1400	Exenatide	Placebo	The first occurrence of death from cardiovascular causes, all-cause mortality	Major limb events, LEA, gangrene, revascularization, MACE, all-cause mortality
Hsiao [24]	Retrospective cohort (IPTW)	9.6/18a	T2D with LEAD	59.0 ± 12.6/58.8 ± 12.1	109/361	GLP-1 RAs	SGLT-2 inhibitors	EVT for LEAD, newly developed LEAD-related foot ulcers, or non-traumatic LLA	Major limb events
Rodionov [25]	Retrospective cohort (IPW)	18.2/13.4a	T2D with PAD	67.9 ± 9.0/68.0 ± 9.1	2074/6032	GLP-1 RAs	SGLT-2 inhibitors	Minor and major LEA, HHF	LEA
Lin [26]	Retrospective cohort (PSM)	23.5/23.2b	T2D with LEAD	56.4 ± 13.2/56.9 ± 13.0	1644/6759	GLP-1 RAs	SGLT-2 inhibitors	Newly diagnosed CLI, PTA or peripheral bypass surgery for PAD, and non-traumatic amputation	Major limb events
Bonaca [23]	RCT, double blind	12	T2D with PAD		396/396	Semaglutide	Placebo	Ratio to baseline of the maximum walking distance	Major limb events, LEA, all-cause mortality, revascularization
Caruso [27]	Retrospective cohort (PSM)	30.1c	T2D with PAD or foot ulcers	60.8 ± 10.6/61.5 ± 10.8	167/167	Semaglutide	Other GLTs (no GLP1/DPP4)	PTA and/or CLI	Major limb events, LEA
Go [28]	Retrospective cohort (PSM)	12	T2D with PAD	62.4 ± 12.7e	41,627/41,627	GLP-1 RAs	No GLP1-RAs	Major amputation or acute limb ischemia requiring intervention, MACE and 1-year mortality	Major limb events, MACE
Hong [29]	Retrospective cohort (PSM)	36	T2D and PAD	66.5 ± 10.1/66.6 ± 10.6	77,393/77,393	GLP-1 RAs	SGLT-2 inhibitors	Major LEA (above ankle)	Major LEA, any LEA, all-cause mortality
		36	T2D and PAD	67.4 ± 10.3/67.5 ± 10.6	39,907/39,907	GLP-1 Ras	DPP4 inhibitors	Major LEA (above ankle)	Major LEA, any LEA, all-cause mortality
Lewis [30]	Retrospective cohort (PSM)	12	T2D and DFU	56.7 ± 11.4e	6329/6329	Semaglutide	No Semaglutide	Wound healing complications and amputation	Amputation
Wu [31]	Retrospective cohort (PSM)	12	T2D and PAD	63.6 ± 11.3/63.3 ± 12.6	4023/4023	Tirzepatide	No Tirzepatide	Amputation and repair	Major limb events
Hsiao [33]	Retrospective cohort (IPTW)	49.2	T2D with MALE	68.9 ± 11.6/70.8 ± 11.9	10,731/17,072	GLP-1 Ras	DPP4 inhibitors	Lower limb revascularization or nontraumatic major or minor amputation	Major limb events, MACE, all-cause mortality
Yahyavi [32]	Retrospective cohort (PSM)	24.8d	T2D with CLTI	65.4 ± 10.5/65.4 ± 11.5	15,743/15,743	GLP-1 RAs	No GLP-1 RA	Major lower limb amputation (below or above-the-knee amputation)	Major amputation, all cause-mortality

CLI critical limb ischemia, CLTI critical limb-threatening ischemia, DFU diabetic foot ulcers, EVT endovascular therapy, GLTs glucose-lowering therapies, HHF hospitalization heart failure, IPW inverse probability weighting, IPTW inverse probability of treatment weighting, LEA lower extremity amputation, LEAD lower extremity arterial disease, LLA lower limb amputation, MACE major adverse cardiovascular events, MALE major adverse limb events, OR observational retrospective, PAD Peripheral Artery Disease, PAT percutaneous transluminal angioplasty, PSM propensity score matching, RCT randomized controlled trial, T2D type 2 diabetes

^aMedian follow-up intervention/control groups

^bMean follow-up intervention/control groups

^cMedian follow-up

^dMean follow-up

^e overall population

Major limb events

All thirteen arms of the 12 studies [22–33] reported on major limb events. Compared with placebo, other antidiabetic drugs or no treatment, treatment with GLP-1RAs was associated with a lower risk of major limb events (RR 0.73, 95% CI 0.65 to 0.82) (Fig. 1, Table 2), high heterogeneity (I², 81%), low certainty of evidence (Table 2), and risk of bias and inconsistency. Additional file 1: Figs. S2 and S3 summarize the risk of bias across included studies, highlighting studies with serious or high risk and the reasons for these assessments. There is some publication bias at the Egger’s test (P = 0.014); however, the effect size didn’t change after applying the trim-and-fill methodology that found no missing study (Additional file 1: Fig. S4). The RR value in the two RCTs was numerically higher (RR 0.85, 95% CI 0.78 to 1.08) than that of cohort studies (RR 0.72, 95% C to 0.64–0.81), but without difference between the two populations (P_interaction = 0.19) (Fig. 1). The effect size of the two RCTs was not significant owing to the small number of trials and the wide confidence intervals, which suggest imprecision.

Fig. 1.

Effect of GLP-1RAs on major limb events in RCTs and matched cohort studies of individuals with type 2 diabetes and PAD. a (GLP-1RAs versus SGLT-2i) and b (GLP-1RAs versus DPP-4i) refer to the two arms of the same study (Ref. 29)

Table 2.

Summary of results of the meta-analysis, with Risk Ratios, heterogeneity and quality of evidence using the GRADE approach

Outcome measure	No. of participants (arms)	RR (95% CI)	Heterogeneity I2 (%)	GRADE
Major limb events	418,752 (13)	0.73 (0.65, 0.82)	81	⊕⊕OOa,b Low
LEA	318,579 (9)	0.76 (0.66, 0.87)	83	⊕⊕OOa,b Low
Revascularization	238,526 (5)	0.81 (0.77, 0.86)	0.0	⊕⊕⊕Oa Moderate
Gangrene	237,734 (4)	0.80 (0.77, 0.85)	0.0	⊕⊕⊕Oa Moderate
MACE	121,903 (4)	0.76 (0.63, 0.90)	97	⊕⊕OOa,b Low
All-cause mortality	305,527 (7)	0.67 (0.61, 0.73)	92	⊕⊕OOa,b Low

^aEvidence certainty was downgraded one level due to risk of bias

^bEvidence certainty was downgraded one level due to inconsistency (heterogeneity in point estimates with I² up to 75%)

Lower extremity amputation

Nine arms of eight studies [22, 23, 25, 27, 29, 30, 32, 33] reported on LEA (n = 318,579). Compared with participants not receiving GLP-1 RAs, GLP-1RA use was associated with a lower risk of LEA (RR 0.76, 95% CI 0.66 to 0.87) (Fig. 2, Table 2), high heterogeneity (I², 82.9%), and low certainty of evidence (Table 2). There was no difference in the effect estimate between RCTs and cohort studies (P_interaction = 0.08, Fig. 2).

Fig. 2.

Effect of GLP-1RAs on low extremity amputation (LEA) in RCTs and matched cohort studies of individuals with type 2 diabetes and PAD. a (GLP-1RAs versus SGLT-2i) and b (GLP-1RAs versus DPP-4i) refer to the two arms of the same study (Ref. 29)

Revascularization

Five arms of four studies [22, 23, 27, 29] reported on revascularization (n = 238,526). Treatment with GLP-1RA was associated with a lower risk of revascularization (RR 0.81, 95% CI 0.77 to 0.86) (Additional file 1: Fig. S5, Table 2), no heterogeneity (I², 0%), and moderate certainty of evidence (Table 2). There was no difference in the effect estimate between RCTs and cohort studies (P_interaction = 0.94, Additional file 1: Fig. S5).

Gangrene

Four arms of three studies [22, 27, 29] reported on gangrene (n = 237,734). Compared with GLP-1RA non-use, GLP-1RA use was associated with a lower risk of gangrene (RR 0.80, 95% CI, 0.77 to 0.85) (Additional file 1: Fig. S6, Table 2), no heterogeneity (I², 0%), and moderate certainty of evidence (Table 2). There was no difference in the effect estimate between RCTs and cohort studies (P_interaction = 0.90, Additional file 1: Fig. S6).

MACE and all-cause mortality

Four studies [22, 28, 31, 33] reported on MACE (n = 121,903). Treatment with GLP-1RA was associated with a lower risk of MACE (RR 0.76, 95% CI 0.63 to 0.90) (Fig. 3, Table 2), high heterogeneity (I², 97%), and low certainty of evidence (Table 2). There was no difference in the effect estimate between RCTs and cohort studies (P_interaction = 0.25, Fig. 3).

Fig. 3.

Effect of GLP-1RAs on MACE and all-cause mortality in RCTs and matched cohort studies of individuals with type 2 diabetes and PAD. a (GLP-1RAs versus SGLT-2i) and b (GLP-1RAs versus DPP-4i) refer to the two arms of the same study (Ref. 29)

Seven arms of six studies [22, 23, 29, 31–33] reported on all-cause mortality (n = 305,527). Treatment with GLP-1RA was associated with a lower risk of all-cause-mortality (RR 0.67, 95% CI 0.61 to 0.73) (Fig. 3, Table 2), high heterogeneity (I², 92%), and low certainty of evidence (Table 2). There was no difference in the effect estimate between RCTs and cohort studies (P_interaction = 0.26, Fig. 3).

Sensitivity and subgroup analyses

The leave-one-out analyses found that no single study was associated with the high heterogeneity found in major limb events (primary outcome), LEA, MACE and all-cause mortality (Additional file 1: Tables S5-S8). The subgroup analysis relative to major limb events showed no significant difference (P = 0.134) among the comparators of GLP-1RAs (DPP-4 inhibitors, SGLT-2 inhibitors, non-users, placebo), although the heterogeneity was null for SGLT-2 inhibitors and placebo (Additional file 1: Table S9; Fig. S7). Meta-regression analysis showed that longer follow-up was associated with an attenuation of the protective effect of GLP-1RAs on major limb events (ß = 0.012, P = 0.006). Follow-up duration explained about one-half of the between-study variance (R² = 54.5%); however, substantial residual heterogeneity remained (I² = 62.1%). The meta-regression bubble plots are presented in Additional file 1: Fig. S8. The sensitivity analysis directly pooling HR and sHRs, excluding only one study reporting RR (30), showed consistent results (HR = 0.76, 95% CI 0.70–0.83, I² = 64%), confirming the benefit of GLP-1RAs on major limb events (Additional file 1: Fig. S9).

Discussion

The present systematic review and meta-analysis included 12 studies, with 13 arms and 418,282 individuals with T2D and PAD. Compared with placebo, other antidiabetic drugs (SGLT-2 inhibitors, DPP-4 inhibitors) or no treatment, treatment with GLP1-RAs was associated with a significantly lower risk of major limb events (RR of 0.73). The analysis of individual limb events showed a lower risk of amputation, revascularization, and gangrene; moreover, GLP-1RA therapy was also associated with a lower risk of MACE and all-cause mortality (24% and 33%, respectively). Although heterogeneity was high for the primary outcome, it was null for revascularization and gangrene, while meta-regression suggested a significant attenuation with time of the beneficial effects of GLP-1RAs on the primary outcome.

Statistical heterogeneity was observed mainly across comparator types and outcome definitions. In particular, heterogeneity was null when GLP-1RAs were compared with SGLT-2 inhibitors or placebo (Additional file 1: Table S9) and highest when compared with non-users, although this last comparison should be interpreted cautiously because it is more vulnerable to confounding. Moreover, when the analysis was limited to only one specific outcome (revascularization or gangrene) heterogeneity was null, suggesting that the different outcome definitions across the different RCTs or cohort studies were implicated. Lastly, the different length of the studies seems also have played a role since the meta-regression analysis (Additional file 1: Fig. S8) found that about 54.5% of the variance was associated with follow-up duration. The vascular benefit seems to appear early in the disease course, suggesting a protective effect beyond glycemic control.

In the last two years (2025–2026), at least seven systematic review and meta-analyses [10–13, 35–37] have been published evaluating the role of GLP-1RAs on PAD in people with T2D. Although most meta-analyses agree that cohort studies tend to report more limb benefits compared to RCTs, this seems more due to a flaw in the RCTs than an advantage of the cohort ones. In particular, the overwhelming majority of RCTs were not specifically designed to assess limb-related endpoints, nor were they exclusively done in patients with diabetes and PAD because limb-related outcomes were studied in a cohort where only a tiny fraction had PAD. Moreover, lower-extremity events, such as amputations or critical limb ischemia, were often not predefined endpoints, were reported as safety events or in post-hoc analyses. The incremental novelty of the present meta-analysis is based on two specific points: the inclusion of studies investigating the effect of GLP-1RAs on lower extremity outcomes in people with T2D and PAD at entry, and the large number of observational studies included, making it the largest meta-analysis carried out so far.

Pharmacotherapy for diabetes has changed dramatically with introduction of GLP-1RAs and SGLT-2 inhibitors that are now recommended as first-line treatment for individuals with T2D and either established or high-risk for cardiovascular disease, including in patients with PAD [7–9]. However, GLP-1RAs have demonstrated specific advantages over SGLT-2 inhibitors in the management of PAD in individuals with T2D [24, 26, 29]. Management of PAD based on recent guidelines [8, 38] should include two major facets: lowering of the risk of cardiovascular events and optimization of cardiometabolic risk factors (i.e. hyperglycemia, hypertension, hyperlipidemia); and management of symptoms. Until recently, medical therapies for PAD have not had significant impact on both of these areas. An ideal therapy for individuals with T2D and PAD would control cardiometabolic risk factors, improve symptoms and decrease the risk of adverse cardiovascular outcomes. Not surprisingly, GLP-1RAs may help control cardiometabolic risk factors, such as hypertension, hyperglycemia and obesity, improve symptoms of PAD by increasing walking distance [23, 39], and decrease the risk of MACE and all-cause mortality [40]. Even in individuals without diabetes, GLP-1RAs can reduce major limb events, MACE and all-cause mortality [28, 41]. However, GLP-1RAs are not all the same. In people with T2D and PAD, only liraglutide and semaglutide improve walking distance [23, 39], whereas semaglutide and tirzepatide reduce amputation [27, 30, 31]. Interestingly enough, Hong et al. [29] report that tirzepatide and semaglutide, compared with other GLP-1RAs, were associated with significantly lower risks of amputation, lower extremity revascularization and mortality, including a 24% reduction in major amputation risk with semaglutide and a 23% reduction in any amputation with tirzepatide. Lastly, tirzepatide demonstrated comparable efficacy to individual GLP-1RAs on cardiovascular outcomes [42].

Although all the mechanisms underlying the benefits of GLP-1RAs on peripheral vasculature have not been completely elucidated, research suggests important effects on inflammation and vascular function. Potential mechanisms by which GLP-1RAs may improve vessel function in individuals with PAD include decreased inflammation, reduced immune activation, improved endothelial function, and effects on vascular regenerative progenitor cells [43]. The extended follow-up of the STARDUST trial [44] demonstrated that in people with T2D and PAD, liraglutide ameliorated circulating markers of angiogenesis (vascular endothelial growth factor, circulating endothelial progenitor cells) and inflammation (C-reactive protein, interleukin-6) over an 18-month follow-up. Moreover, the importance of glycemic control to prevention of adverse lower extremity outcomes in patients with T2D cannot be overlooked [45, 46].

Limitations

This meta-analysis has limitations. Cohort studies are limited by their retrospective design and reliance on health records, which might introduce residual confounding, misclassification of diagnoses and under-ascertainment of outcomes. However, we selected cohort studies that used methods to reduce bias due to confounding variables that could affect the treatment effect estimate (propensity score matching, inverse probability weighting). Statistical heterogeneity was observed across types of studies and some subgroups, reflecting differences in populations, comparator groups, study design, and study sample heterogeneity, although sensitivity analyses did not alter the overall direction of the results. Due to insufficient data, we were unable to assess the comparative effectiveness of individual GLP-1 RAs. Furthermore, the follow-up durations in many studies may be insufficient to capture the full trajectory of limb complications, although the meta-regression seems to suggest that the putative beneficial effect of GLP-1RAs on lower limb circulation may occur early during treatment, as also suggested by the improved walking capacity in RCTs of short duration [23, 39]. As promising as our results are, the overall estimate may be substantially influenced by observational evidence and therefore causal interpretation remains limited.

Conclusion

In this meta-analysis, GLP-1RA use was associated with a significantly lower risk of major limb events in individuals with T2D and PAD with a low certainty of evidence. However, current evidence remains largely observational, and causality cannot yet be firmly established. Dedicated outcome trials with specific limb outcomes are needed to obtain direct evidence for the limb benefits of GLP-1RAs in people with T2D and PAD.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

P.C., D.G., M.I.M., and K.E. were responsible for the conceptualization and design of the study. P.C., N.D.M., D.G., M.I.M., and K.E. analyzed data and drafted the first manuscript. P.C., N.D.M., M.L., L.S., M.I.M., G.B., K.E., and D.G. critically reviewed and edited the manuscript. D.G., K.E., G.B., and M.I.M. supervised the study. All authors contributed to the article, and read and approved the final manuscript. D.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Funding

This study was supported in part by the “Associazione Salute con Stile”. The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.

Data availability

No new, previously unpublished patient data were generated or analyzed in support of this research.

Declarations

Ethics approval and consent to participate

Not applicable. This study is a meta-analysis and did not involve direct recruitment of human participants.

Generative AI and AI-assisted technologies in the writing process

The authors didn’t use ChatGPT for assistance in the preparation of the manuscript.

Competing interests

The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: [Dario Giugliano reported receiving consultancy fee from Eli Lilly, Sanofi, Novartis, Astrazeneca, and Novo Nordisk; Maria Ida Maiorino reported receiving personal fees from Novo Nordisk, Eli Lilly, and Sanofi; Katherine Esposito reported receiving consultancy fees from Eli Lilly and giving lectures for Eli Lilly, Sanofi, Novo Nordisk, Roche, Bayer, and Lifescan. All other authors declare that they have no competing interests.