Background

Sodium-glucose co-transporter-2 inhibitors (SGLT2i) have shown benefits for different endpoints, such as renal outcomes, heart failure, and major cardiovascular (CV) events among patients with type-2 diabetes mellitus (T2DM) treated with antihyperglycemic agents [1,2,3,4]. Although the two available trials with dapagliflozin and empagliflozin did not report a significant increase in amputations, the Canagliflozin Cardiovascular Assessment Study (CANVAS) program indicated a higher rate of amputations in the canagliflozin group compared with the placebo group (6.3 vs. 3.4 patients per 1000 patient-years) [1,2,3]. Nevertheless, the CANVAS results raised concerns regarding the suitability of SGLT2i for patients with T2DM with a high risk of amputation, such as those with concomitant peripheral artery disease (PAD). A few observational studies have investigated the association of SGLT2i with the risk of lower limb amputation; however, these studies have reported inconsistent and conflicting findings. For example, Yuan et al. reported no increased risk of amputations (hazard ratio [HR] 0.98; 95% confidence interval [CI] 0.68–1.41); Adimadhyam et al. reported increased risk (HR 1.38, 95% CI 0.83–2.31); and Udell et al. reported an increased risk (HR 1.99, 95% CI 1.12–3.51) for SGLT2i treatment compared with nonSGLT2i agents [5,6,7].

T2DM is a major risk factor for CV disease and PAD, and the prevalence of PAD in patients with T2DM has been estimated to reach 20% [8,9,10]. Patients with T2DM and concomitant PAD have an increased risk of CV events and amputation compared with those without PAD [11]. Subgroup analyses of the landmark studies on empagliflozin revealed consistent CV benefits in patients with T2DM and concomitant PAD without an increased risk of amputation [12]. However, real-world data on the effectiveness, safety, and limb outcomes for such a specific population treated with SGLT2i are scarce. Dipeptidyl peptidase-4 inhibitors (DPP4i) improve glycemic control by increasing the serum levels of glucagon-like peptide 1 (GLP-1) and exhibit a neural effect in CV composite outcomes, that are clinically widely prescribed as second-line agents in the management of hyperglycemia for patients with T2DM [13, 14]. Our study investigated the outcomes of patients with T2DM and concomitant PAD treated with SGLT2i compared with those treated with DPP4i in a large, real-world setting.

Methods

Study population

This retrospective nationwide cohort study analyzed data from the Taiwan National Health Insurance (NHI) Research Database (NHIRD), which contains detailed health-care information for more than 23 million enrollees with a > 99% coverage rate of residents of Taiwan [15]. This study was approved by the Institutional Review Board of Chang Gung Medical Foundation, Taiwan (104-8079B and 201801427B0). Informed consent was waived because the original identification number of each patient in the NHIRD had been encrypted and de-identified to protect their privacy.

Study cohort

The study identified a total of 3,623,527 patients with T2DM diagnosed using International Classification of Diseases (ninth revision) Clinical Modification (ICD-9-CM) codes (250) between January 1, 1998 and December 31, 2015, or ICD-10-CM codes (E10.0, E10.1, E10.9, E11.0, E11.1, and E11.9) between January 1, 2016 and December 31, 2017. To identify patients with T2DM who had diagnoses indicating PAD, patients with PAD were required to fulfill with at least one of the following the diagnoses or treatments, which have been registered using medical records, ICD-9-CM or ICD-10-CM diagnostic codes, or ICD-9/10-CM procedural codes (Additional file 1: Table S1). Among the 452,149 patients with T2DM and concomitant PAD, 12,355 patients received first prescriptions of SGLT2i (empagliflozin and dapagliflozin; approval date in Taiwan: May 1, 2016) between May 1, 2016 and December 31, 2017. Canagliflozin has not been included in the present study because it is approved after March 1, 2018 in Taiwan. Of the other 439,794 patients not receiving SGLT2i treatments, 93,972 patients received first prescriptions for DPP4i (saxagliptin, sitagliptin, linagliptin, or alogliptin) during the same period. Patients with T2DM are not allowed to use SGLT2i and DPP4i simultaneously according to Taiwan’s NHI regulations. For each study group, the index date was defined as the first date of prescription for SGLT2i or DPP4i after May 1, 2016. The follow-up period was from the index date until the independent occurrence of any study outcome, discontinuation of the index drug, or end date of the study period (December 31, 2017), whichever occurred first. The flowchart of study enrollment is summarized in Fig. 1.

Fig. 1
figure 1

Enrollment of patients with concomitant type-2 diabetes mellitus (T2DM) and peripheral artery disease (PAD). From May 1, 2016 to December 31, 2017, a total of 11,431 patients with T2DM and comorbid PAD treated with sodium-glucose co-transporter-2 inhibitors (SGLT2i) and 11,431 1:1 propensity score matched patients treated with dipeptidyl peptidase-4 inhibitors (DPP4i) were enrolled in the present study. Abbreviations: DPP4i dipeptidyl peptidase-4 inhibitor, PAD peripheral artery disease; SGLT2i sodium-glucose co-transporter-2 inhibitor, T2DM type 2 diabetes mellitus

Full size image

Covariates and study outcomes

Baseline covariates were obtained from all claim records with diagnoses, procedures, or medication codes prior to the index date. A history of all prescription medications was confined to medications used at least once within 3 months before the index date. We reported the following outcomes in the present study: (i) ischemic stroke (IS), (ii) acute myocardial infarction (AMI), (iii) congestive heart failure (CHF), (iv) lower limb ischemia requiring revascularization, (v) lower limb amputation, (vi) all-cause mortality, and (vii) cardiovascular mortality. All study outcomes should be the primary discharge diagnosis to avoid misclassification. The diagnostic codes of the NHIRD were shifted from the ICD-9-CM to ICD-10-CM after January 1, 2016. The ICD-9-CM and ICD-10-CM codes used to identify study outcomes along with the baseline covariates are summarized in Additional file 1: Tables S1 and S2.

Statistical analysis

The propensity score matching (PSM) method, which simulates the design of a randomized clinical trial for observational cohort data by forming matched sets of treated and untreated subjects who share a similar value of the propensity score [16], was used to compare the study outcomes between the SGLT2i and DPP4i group. We calculated propensity score, the predicted probability of treatment conditional on all the covariates in Table 1, by using the generalized boosted model (GBM). The GBM involves an iterative process with multiple regression trees to capture complex and nonlinear relationships between treatment assignment and the pretreatment covariates without over-fitting the data and leading the best balance across study groups [17]. The PSM ratio between the SGLT2i users and DPP4i users was 1:1 without replacement and nearest neighbor matching within a caliper width (8-to-1 digit matching) [18]. The balance of potential confounders at the baseline (index date) between study groups was assessed using the absolute standardized mean difference (ASMD) rather than statistical testing because balance is a property of the sample and not of the underlying population. An ASMD value of ≤ 0.1 would indicate a nonsignificant difference in potential confounders between the two study groups [19]. Incidence rates were estimated using the total number of study outcomes during the follow-up period divided by person-years at risk. The risk of study outcomes occurring over the follow-up duration for SGLT2i versus DPP4i (reference) was obtained using survival analysis (Kaplan–Meier method and log-rank test for univariate analysis and Cox proportional hazards model for multivariate analysis), and they were presented as HRs with 95% CIs. Statistical significance was defined as a p value of < 0.05. All statistical analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA).

Table 1 Clinical characteristics of patients with concomitant type-2 diabetes mellitus (T2DM) and peripheral artery disease (PAD) treated with sodium-glucose co-transporter-2 inhibitors (SGLT2i) and dipeptidyl peptidase-4 inhibitors (DPP4i) before and after propensity score matching (PSM)
Full size table

Results

Baseline characteristics of SGLT2i and DDP4i groups

Among the 452,149 patients with T2DM and concomitant PAD, a total of 12,355 and 93,972 were treated with SGLT2i and DDP4i, respectively, from May 1, 2016 to December 31, 2017 (Fig. 1). The mean follow-up periods were 0.96 ± 0.57 and 0.66 ± 0.45 years for SGLT2i and DDP4i, respectively. In the SGLT2i group, 6,915 (56.0%) and 5,440 (44.0%) patients were treated with dapagliflozin and empagliflozin, respectively. In the DDP4i group, 29,782 (31.7%), 24,833 (26.4%), 28,534 (30.4%), 10,636 (11.3%), and 187 (0.2%) patients were treated with sitagliptin, vildagliptin, linagliptin, saxagliptin, and alogliptin, respectively. Before PSM, the SGLT2i group was younger and had a lower prevalence of chronic kidney disease (CKD), hypertension, CHF, hypertension, and stroke history compared with the DDP4i group. The SGLT2i group had a higher rate of dyslipidemia, higher rate of prescriptions for statins, metformin, sulfonylurea, acarbose, and glitazones and a lower rate of prescriptions for glinides. Both study groups were well balanced in all characteristics after PSM (all ASMD < 0.1) (Table 1).

Main analysis of SGLT2i versus DDP4i

The SGLT2i group had comparable cumulative risks of IS and AMI compared with the DPP4i group after PSM. The SGLT2i group was associated with a lower cumulative risk of CHF (log-rank p = 0.0059), all-cause (log-rank p < 0.0001) and cardiovascular mortality (log-rank p = 0.0085) compared with the DPP4i group after PSM. Regarding major adverse limb events, the use of SGLT2i was associated with a lower cumulative risk of lower limb revascularization (log-rank p = 0.0359) and amputation (log-rank p < 0.0001) compared with the use of DPP4i (Fig. 2).

Fig. 2
figure 2

Cumulative incidence curves of outcomes for patients with concomitant type-2 diabetes mellitus (T2DM) and peripheral artery disease (PAD) treated with SGLT2i versus DPP4i after propensity score matching (PSM). Cumulative incidence curves of effectiveness outcomes including ischemic stroke (IS) (a), acute myocardial infarction (AMI) (b), congestive heart failure (CHF) (c), lower extremity revascularization (d) or amputation (e), all-cause mortality (f), and cardiovascular mortality (g) for patients with T2DM and concomiant PAD taking SGLT2i versus DPP4i after PSM are presented. SGLT2i were associated with lower cumulative risks of CHF, all-cause and cardiovascular mortality, and lower extremity revascularization or amputation compared with DPP4i among patients with T2DM and concomitant PAD. Abbreviations: AMI acute myocardial infarction, CHF congestive heart failure, IS ischemic stroke, PSM propensity score matching. Other abbreviations are the same as those in Fig. 1

Full size image

The incidence rates (per 100 person-years) of IS (1.26 vs. 1.54, p = 0.1213) and AMI (0.66 vs. 0.77, p = 0.3702) were comparable between the SGLT2i and DDP4i groups. The SGLT2i group had a significantly lower incidence rate of CHF (0.96 vs. 1.43; HR: 0.66; 95% CI 0.49–0.89; p = 0.0062), lower limb ischemia requiring revascularization (0.97 vs. 1.32; HR: 0.73; 95% CI 0.54–0.98; p = 0.0367), lower limb amputation (0.54 vs. 1.23; HR: 0.43; 95% CI: 0.30–0.62; p < 0.0001), all-cause mortality (3.19 vs. 5.44, HR: 0.58; 95% CI 0.49–0.67; p < 0.001), and cardiovascular mortality (0.91 vs. 1.33, HR: 0.67; 95% CI 0.49–0.90; p = 0.0089) compared with the DDP4i group (Table 2 and Fig. 3). The use of SGLT2i was not associated with an increased risk of bone fracture or urinary tract infection compared with DDP4i use after PSM (Table 2 and Additional file 1: Table S1).

Table 2 Number of events, event rates, and hazard ratio (HR) among patients with type-2 diabetes mellitus and concomitant peripheral artery disease (PAD) using sodium-glucose co-transporter-2 inhibitors (SGLT2i) versus dipeptidyl peptidase-4 inhibitors (DPP4i) after propensity score matching
Full size table
Fig. 3
figure 3

Forest plot of the hazard ratios of clinical outcomes for SGLT2i versus DPP4i among patients with type-2 diabetes mellitus (T2DM) comorbid with peripheral artery disease (PAD) after propensity score matching (PSM). SGLT2i were associated with a comparable risk of thromboembolic events and with lower risks of CHF, lower limb revascularization or amputation, and all-cause or cardiovascular mortality compared with DPP4i among patients with T2DM and concomitant PAD after PSM. Abbreviation: CI confidence interval, HR hazard ratio. Other abbreviations are the same as those in Figs. 1, 2

Full size image

Subgroup analysis of high-risk patients

The subgroup analysis indicated that SGLT2i reduced the risk of AMI in patients with concomitant CKD but not in those without CKD (p interaction = 0.02; Fig. 4). In general, the subgroup analysis revealed consistent results for CHF, major adverse limb outcomes, and mortality for SGLT2i versus DPP4i among patients aged ≥ 75 years, the presence of CKD and established CV disease, consistent with the main analysis (Figs. 4, 56 ).

Fig. 4
figure 4

Subgroup analysis of the hazard ratios for the risks of ischemic stroke (IS) (a) acute myocardial infarction (AMI) (b), and congestive heart failure (CHF) (c) for SGLT2i versus DPP4i among T2DM patients with concomitant peripheral artery disease after propensity score matching. In general, the subgroup analysis revealed consistent results for the risks of IS (a) AMI (b), and CHF (c) for SGLT2i versus DPP4i among patients aged \(\ge \) 75 years, the presence of chronic kidney disease (CKD) and established CV disease, consistent with the main analysis. The subgroup analysis indicated that SGLT2i reduced the risk of IS and AMI in patients with concomitant CKD but not in patients without CKD (p interactions = 0.02). Abbreviations: CKD chronic kidney disease, CV cardiovascular disease. Other abbreviations as in Figs. 1, 2, 3.

Full size image
Fig. 5
figure 5

Subgroup analysis of hazard ratios for the risk of major adverse lower limb events including lower limb revascularization procedure (a) and amputation (b) for SGLT2i versus DPP4i among T2DM patients with concomitant with peripheral artery disease after propensity score matching. The subgroup analysis revealed consistent results for lower limb revascularization (a) or amputation (b) for SGLT2i versus DPP4i among patients aged ≥ 75 years, the presence of CKD and established CV disease, consistent with the main analysis (all p interactions > 0.05). The abbreviations are the same as those in Figs. 1, 2, 3, 4

Full size image
Fig. 6
figure 6

Subgroup analysis of hazard ratios for the risk of all-cause mortality (a) and cardiovascular mortality (b) for SGLT2i versus DPP4i among T2DM patients with peripheral artery disease after propensity score matching. The subgroup analysis revealed consistent results for all-cause mortality (a) and cardiovascular mortality (b) for SGLT2i versus DPP4i among patients aged ≥ 75 years, the presence of chronic kidney disease and established CV disease, consistent with the main analysis (all p interactions > 0.05). The abbreviations are the same as those in Figs. 1, 2, 3, 4

Full size image

Discussion

To the best of our knowledge, the present study is the first and largest population-based cohort study to investigate the outcomes of patients with T2DM and concomitant PAD treated with SGLT2i compared with those treated with DPP4i. Our findings indicate that SGLT2i was associated with comparable risks of IS and AMI, and significantly lower risks of CHF, lower limb ischemia requiring revascularization or amputation, and all-cause or cardiovascular mortality when compared with DPP4i. This study suggests that SGLT2i is an effective and safe alternative to DPP4i for patients with T2DM and concomitant PAD.

DPP4i improves glycemic control by increasing the serum levels of glucagon-like peptide 1 (GLP-1) through the inhibition of GLP-1 degradation, which indirectly stimulates insulin secretion and enhances beta-cell function. Previous large-scale clinical trials, including EXAMINE, SAVOR-TIMI53, and TECOS, have indicated that the use of DPP4i has a neutral effect in CV composite outcomes for patients with T2DM, except for a higher risk of CHF for those treated with saxagliptin [20,21,22]. Those clinical studies did not explore the risk of lower limb outcomes for patients with T2DM treated with DPP4i. A previous meta-analysis of the three clinical trials confirmed the benefit of SGLT2i on CHF (HR 0.69; 95% CI 0.61–0.79), all-cause death (HR 0.85; 95% CI 0.78–0.93), and reduced risk of major adverse CV events (composite of myocardial infarction, stroke, and cardiovascular death; HR 0.89, CI 0.83–0.96]) [23]. One large retrospective cohort study also indicated that SGLT2i were associated with lower risks of CHF and death compared with DPP4i in patients with diabetes [24]. Other cohort studies investigating SGLT2i versus other nonSGLT2i antidiabetic agents have consistently reported that SGLT2i reduces the risk of CHF [25,26,27]. A recent clinical trial also showed the beneficial role of SGLT2i on reducing the risk of HF hospitalization [28, 29].

Patients with T2DM have a higher prevalence of PAD compared with those without T2DM, and patients with T2DM and concomitant PAD have a higher risk of mortality and amputation [11, 30]. However, evidence supporting the benefits of SGLT2i in patients with diabetes and concomitant PAD is limited. The CANVAS program reported a higher rate of amputations in the canagliflozin group compared with the placebo group (0.63 vs. 0.34 per 100 person-years, p < 0.001), but not in the pivot studies of empagliflozin (0.65 vs. 0.65 per 100 person-years, p = 1.000) and dapagliflozin (1.4 vs. 1.3 per 100 person-years, p = 0.53) [1,2,3]. A clear mechanism explaining why canagliflozin contributes to amputation is lacking; this adverse event may be related to volume depletion, which might accordingly cause circulatory failure in the distal peripheral vasculature [1, 31]. Although a meta-analysis showing patients treated with SGLT2i without a significant association with increased risk of amputation, a large-scale cohort study revealed that SGLT2is were associated with an increased risk of amputation compared with other antihyperglycemic agents for type 2 diabetes [32, 33]. Subgroup analyses from the pivot study of empagliflozin for patients with T2DM and concomitant PAD also revealed benefits of reduced risks of death and CHF without an increased risk of amputation [12]. Consistent with the data, our results indicated that SGLT2i can reduce the risks of CHF and mortality in such a high-risk population. Notably, SGLT2i (dapagliflozin and empagliflozin) were associated with a lower risk of adverse limb events (lower limb ischemia requiring revascularization and lower limb amputation) compared with DDP4i in our study (0.97 vs. 1.32 per 100 person-years, p = 0.0367 and 0.54 vs. 1.23 per 100 person-years, p < 0.0001). In the assessment of patients with T2DM and concomitant PAD with a relatively high risk of amputation, the absolute risk of amputation in patients treated with SGLT-2i was similar or lower than those seen in the pivot studies, and there is no increase in the probability of amputation [1,2,3]. In animal or human studies, SGLT2i have been reported to have many benefits for vasculature, such as improved endothelial function, vasodilatation, and attenuated oxidative stress, suggesting that SGLT2i may be able to halt the progression of atherosclerosis and improve vascular outcomes [34,35,36]. In addition, SGLT2i had been reported to improve cardiometabolic risk factors than DDP4i [37]. PAD is a manifestation of systemic atherosclerosis, and because SGLT2i could reduce the risk of adverse atherosclerotic events, it may also be beneficial in reducing the risk of adverse limb events for patients with PAD [23, 38, 39]. However, studies investigating SGLT2i in patients with T2DM and concomitant PAD are scarce. A subgroup analysis revealed a trend of a lower risk of lower limb amputation (HR: 0.84, 95% CI 0.54–1.32) in the empagliflozin group among patients with T2DM and concomitant PAD [12]. Because patients with T2DM have a high prevalence of PAD [8,9,10], further randomized or prospective studies should investigate the effect of SGLT2i on lower limb outcomes in such a high-risk population.

Limitations

To avoid time-lag bias from the prescriptions of study drugs, which may lead to false positive or negative associations depending on the treatments for patients with early or advanced disease, we selected the same second-line hypoglycemic agents of DDP4i as the comparator in our study [7, 40]. To avoid immortal time bias, our study only included new descriptions of study drugs of SGLT2i or DDP4i without baseline use [40, 41]. Nevertheless, the present study had several limitations. First, although PSM with several variables allowed the matching of baseline comorbidities among the study groups, residual confounding by unmeasured variables and prescribing behavior could not be excluded in this retrospective cohort study. Second, the NHIRD does not contain several crucial types of laboratory data such as body weight, glycohemoglobin (HbA1c), and serum creatinine, all of which are associated with the risk of CV events and death among patients with T2DM [42]. In addition, even with adjustment for CKD, the diagnosis of CKD by coding could not reflect the severity of renal disease, which may interfere with SGLT2i or DDP4i selection for each patient. Third, although we utilized some criteria for the selection of the PAD population, our PAD study patients included only part of the PAD population. Thoroughly screening patients with PAD is difficult because PAD populations are typically underrecognized or undertreated in clinical practice, and the incidence of asymptomatic PAD is higher than that of symptomatic PAD [43, 44]. Fourth, miscoding and misclassification of underlying comorbidities and outcomes registered by each physician were another limitation. Therefore, we only considered primary discharge diagnoses to improve the outcome accuracy. However, minor cardiovascular or limb events without admission may have been missed in the present study. Fifth, we did not analyze canagliflozin because of its approval date after March 1, 2018 in Taiwan. Finally, we only investigated Asian patients, and whether our results can be extrapolated to other races remains unclear.

Conclusions

Our data indicated that SGLT2i, compared with DDP4i, were associated with lower risks of CHF, lower limb ischemia requiring revascularization or amputation, and all-cause death for patients with T2DM and concomitant PAD. Further prospective studies are necessary to evaluate the effects of SGLT2i on lower limb outcomes among such patients in the future.