Comparisons of new-onset peripheral arterial disease in Type 2 diabetes mellitus patients exposed to SGLT2I, DPP4I or GLP1a: a population-based cohort study

Study population

This study was approved by the Institutional Review Board of the University of Hong Kong/Hospital Authority Hong Kong West Cluster (HKU/HA HKWC IRB) (UW-20-250) Clinical Research Ethics Committee and complied with the Declaration of Helsinki. This was a retrospective population-based study of prospectively collected electronic health records using the Clinical Data Analysis and Reporting System (CDARS) by the Hospital Authority (HA) of Hong Kong. The records cover both public hospitals and their associated outpatient clinics as well as ambulatory and daycare facilities in Hong Kong. This system has been used extensively by our teams and other research teams in Hong Kong, including diabetes and medication research [9, 10]. The system contains data on disease diagnosis, laboratory results, past comorbidities, clinical characteristics, and medication prescriptions. T2DM patients who were administered with SGLT2I or DPP4I in centres under the HA, between 1^st January 2015, to 31^st December 2020, were included. The GLP1a cohort comprised of patients on GLP1a between 1^st January 2015, to 31^st December 2020 was included for sensitivity analysis to demonstrate the relative effects amongst the second-line oral anti-diabetic agents.

Predictors and variables

Patients’ demographics include gender and age of initial drug use (baseline), clinical and biochemical data were extracted for the present study. Prior comorbidities were extracted by the International Classification of Diseases Ninth Edition (ICD-9) codes (Supplementary Table 1). The diabetes duration was calculated by examining the earliest date amongst the first date of (1) diagnosis using ICD-9; (2) HbA1c ≥6.5%; (3) Fasting glucose ≥7.0 mmol/l or Random glucose 11.1 mmol/l; (4) using anti-diabetic medications. The patients on financial aid were defined as patients on Comprehensive Social Security Assistance (CSSA) scheme, higher disability allowance, normal disability allowance, wavier, other financial aids in Hong Kong. The number of hospitalisations in the year prior to the index days was extracted. The Charlson standard comorbidity index was calculated [11]. The duration and frequency of SGLT2I and DPP4I usage was calculated. The baseline laboratory examinations, including the glucose profiles and renal function tests, were extracted. The estimated glomerular filtration rate (eGFR) was calculated using the abbreviated modification of diet in renal disease (MDRD) formula [12]. Furthermore, the time-weighted lipid and glucose profiles after drug initiation were also calculated by the products of the sums of two consecutive measurements and the time interval, then divided by the total time interval, as suggested previously [13].

The following patient groups were excluded: (1) those with prior PAD (2) patients who died within 30 days after initial drug exposure; (3) without complete demographics; (4) those under 18 years old; (5) developed PAD within 30 days after drug exposure (Figure 1). To account for the incomplete laboratory data, the multiple imputation by chained equations was performed according to a previously published study [14].

• Download figure
• Open in new tab

Figure 1. Procedures of data processing for the study cohort

SGLT2I: Sodium-glucose cotransporter-2 inhibitors; DPP4I: Dipeptidyl peptidase-4 inhibitors. MDRD: modification of diet in renal disease. PAD: Peripheral Arterial Disease

Study outcomes

The primary outcome of this study was new-onset PAD defined using either ICD-9 code, history of amputation due to PAD, and peripheral arterial intervention (Supplementary Table 1), according to the verified published literature [15]. The secondary outcome was all-cause mortality. Mortality data were obtained from the Hong Kong Death Registry, a population-based official government registry with the registered death records of all Hong Kong citizens linked to CDARS. Mortality was recorded using the International Classification of Diseases Tenth Edition (ICD-10) coding. The as-treat approach was adopted which patients were censored at treatment discontinuation or switching of the comparison medications. The endpoint date of interest for eligible patients was the event presentation date. The endpoint for those without primary outcome was the mortality date or the end of the study period (31^st December 2020).

Statistical analysis

Descriptive statistics are used to summarize baseline clinical and biochemical characteristics of patients with SGLT2I and DPP4I use. For baseline clinical characteristics, continuous variables were presented as mean (95% confidence interval/standard deviation), and the categorical variables were presented as total numbers (percentage). Propensity score matching generated by logistic regression with 1:1 ratio for SGLT2I use versus DPP4I use based on demographics, non-SGLT2I/DPP4I medications, number of prior anti-diabetic drugs, prior comorbidities, renal function, duration from T2DM diagnosis initial drug exposure, HbA1c and fasting glucose were performed using the nearest neighbour search strategy with a calliper of 0.1. Propensity score matching was performed using Stata software (Version 16.0). Baseline characteristics between patients with SGLT2I and DPP4I use before and after matching were compared with absolute standardized mean difference (SMD), with SMD<0.10 regarded as well-balanced between the two groups.

The cumulative incidence curves for the primary outcomes and secondary outcomes were constructed. Proportional Cox regression models were used to identify significant risk predictors of adverse study outcomes in the matched cohort, with adjustments for demographics, comorbidities, number of prior hospitalisations, medication profile, renal function, glycaemic tests, and the duration of T2DM. The log-log plot was used to verify the proportionality assumption for the proportional Cox regression models. Subgroup analyses were conducted to confirm the association amongst patients with different clinically important predictors accounting to the diabetic and the metabolic profile, as well as the comorbidities and medications associated with PAD.

Cause-specific and sub-distribution hazard models were conducted to consider possible competing risks. Multiple propensity adjustment approaches were used, including propensity score stratification[16], propensity score matching with inverse probability of treatment weighting (IPTW) [17] and propensity score matching with stable inverse probability weighting [18]. The three arm sensitivity results involving glucagon-like peptide-1 receptor agonist (GLP1a) using stabilized IPTW were conducted to test the association and choice amongst the novel second-line anti-diabetic medications. Sensitivity analyses result with consideration of one-year lag time effects was conducted. Patients with chronic kidney disease (CKD) stage 4/5 (eGFR <30 mL/min/1.73m*2), peritoneal dialysis or haemodialysis who may be contraindicated with SGLT2I were excluded from the sensitivity analysis. The hazard ratio (HR), 95% CI, and P-value were reported. Statistical significance was defined as P-value <0.05. All statistical analyses were performed with RStudio (Version: 1.1.456) and Python (Version: 3.6).

Association between SGLT2I and DPP4I and peripheral arterial diseases

In the matched cohort, 186 SGLT2I users and 256 DPP4I users developed PAD. After a mean follow-up of 314442.4 person-year, the incidence of PAD was lower amongst SGLT2I users (Incidence rate [IR] per 1000 person-year: 1.17; 95% CI: 1.01-1.35) compared to DPP4I users (IR per 1000 person-year: 1.65; 95% CI: 1.45-1.86) (Table 2). SGLT2I users had a 15% lower risk of PAD after adjustment (HR: 0.85; 95% CI: 0.67-0.98, p=0.0464) compared to users of DPP4I regardless of the demographics, comorbidities, medication profile, renal function, glycaemic tests, number of hospitalisations, and the duration of T2DM (Table 2; Supplementary Figure 2; Supplementary Table 3). This was substantiated by the cumulative incidence curves stratified by SGLT2I versus DPP4I (Figure 2).

• Download figure
• Open in new tab

Figure 2. Cumulative incidence curves for new onset Peripheral Arterial Disease (PAD), and all-cause mortality stratified by drug exposure effects of SGLT2I and DPP4I after propensity score matching (1:1).

SGLT2I: Sodium-glucose cotransporter-2 inhibitors; DPP4I: Dipeptidyl peptidase-4 inhibitors; PAD: Peripheral Arterial Disease

Association between SGLT2I and DPP4I and all-cause mortality

Overall, 850 SGLT2I users and 2129 DPP4I users passed away. After a follow-up of 315729.5 person-year, the incidence of all-cause mortality was lower amongst SGLT2I users (IR: 5.33; 95% CI: 4.98-5.70) compared to DPP4I users (IR: 13.63; 95% CI: 13.05-14.22) (Table 2). SGLT2I users had a 41% lower risk of all-cause mortality after adjustment (HR: 0.59; 95% CI: 0.54-0.64, p<0.0001) compared to users of DPP4I regardless of the duration of diabetes mellitus (Supplementary Table 3). This was substantiated by the cumulative incidence curves stratified by SGLT2I versus DPP4I (Figure 2; Supplementary Figure 2).

Subgroup analysis

The results of the subgroup analysis for effects of SGLT2I and DPP4I on the PAD are shown in Figure 3. The result demonstrated that SGLT2I was associated with lower risks of PAD regardless of hyperlipidaemia, stroke/transient ischaemic attack, cancer, uses of ACEI/ARB, insulin, and metformin. SGLT2I was associated with lower risks of peripheral arterial disease only amongst male patients, patient older than 65 years old, with hypertension, without heart failure, without ischaemic heart diseases. Besides, SGLT2I was only associated with lower risks of peripheral arterial disease amongst patients currently on diuretics, lipid lowering drugs, anti-platelets, and anticoagulants, but not sulphonylurea, thiazolidinedione, and GLP1a. Regarding the diabetic control, SGLT2I was associated with lower risks of peripheral arterial disease only amongst the 1^st quartile of mean HbA1c. Interestingly, SGLT2I was associated with higher risks compared to DPP4I amongst those in the 4^th quartile of mean HbA1c. The marginal effects analysis demonstrated that SGLT2I was associated with lower risks of PAD consistently across all eGFR, number of anti-diabetic drugs used, and duration of diabetes, and the effects are more pronounced amongst patients with lower eGFR (Supplementary Figure 3).

• Download figure
• Open in new tab

Figure 3. Forests plot of hazard ratios with 95% CI for SGLT2I v.s. DPP4I on new onset Peripheral Arterial Disease (PAD), and all-cause mortality stratified by drug use in the matched cohort.

ACEI: angiotensin-converting enzyme inhibitors; ARB: angiotensin II receptor blockers; CCB: Calcium channel blocker; SGLT2I: Sodium-glucose cotransporter-2 inhibitors; DPP4I: Dipeptidyl peptidase-4 inhibitors; GLP1a: glucagon-like peptide-1 receptor agonist; TIA: Transient ischaemic attack

Sensitivity analysis

Sensitivity analyses were performed to confirm the predictability of the models. The results of the cause-specific hazard models, sub-distribution hazard models and different propensity score approaches demonstrated that different models did not change the point estimates for both the primary and the secondary outcomes (all p<0.05) (Supplementary Table 4). A 3-arm analysis with the inclusion of GLP1a including patients only on (SGLT2I, DPP4I, and GLP1a) using stabilized IPTW was conducted (Supplementary Table 5). The result between DPP4I and SGLT2I remained consistent with the main result (HR: 0.72; 95% CI: 0.59-0.87). Meanwhile, GLP1a was not associated with higher risks of peripheral arterial disease (p>0.05). Excluding patients with CKD stage 4/5 (eGFR <30), peritoneal dialysis or haemodialysis in the matched cohort likewise demonstrated that SGLT2I was associated with lower risks of peripheral arterial disease outcomes compared to DPP4I (Supplementary Table 6). The sensitivity analysis for 1-year lag time also demonstrated the same trend (Supplementary Table 7). Lastly, a sensitivity analysis was conducted to exclude patients who were on financial aids, and the association between SGLT2I and PAD remained consistent (Supplementary Table 8).

Comparison with previous studies

The prevalence of PAD outcomes in our study for SGLT2I and for DPP4I parallels the values of existing studies, which described an incidence of 2.7 per 1000 person-year [19], lower than the Western counterparts in the ADVANCE trial (12 per 1000 person-year) [20]. In our study, SGLT2I users had relatively reduced risks of PAD outcomes compared to DPP4I users. It should be noted that the observed effect of SGLT2I in our cohort was marginal and may produce a minor modulating effect against new-onset PAD in diabetic patients.

The relationship between SGLT2I on coronary artery disease and cerebral vascular diseases has been well established in large randomised controlled trials recent years [21, 22]. However, the clinical data regarding the relationship between SGLT2I on PAD remain scarce. In our study, we show novel findings that SGLT2I was associated with lower risks of PAD compared to DPP4I.

In the subgroup analysis, the results demonstrated that in terms of PAD, SGLT2I was better amongst patients with low mean HbA1c level, but DPP4I was better amongst patients with high HbA1c levels. The difference extent of glycaemic control contributes to the systemic and renal arterial stiffness, and potentially explaining the differences in the association in PAD [23]. The use of SGLT2I also correlates with a reduction in blood pressure and CRP levels amongst diabetic patients [24]. A study applying the Archimedes model deduced that dapagliflozin usage reduced foot amputations by 13% relative to standard care [25].

Lin et al. demonstrated that canagliflozin correlated with increase in amputation risk and PAD incidence [26], and in the Canagliflozin Cardiovascular Assessment Study (CANVAS) trial the risk of amputation due to vascular disease even increased by twofold [27]. However, meta-analysis of the DAPA-HF and DELIVER trials suggested that dapagliflozin did not increase the risk of amputation compared to placebo regardless of PAD status [28]. These differences could be explained by the differences in the study population and design.

A Taiwanese cohort study found that GLP1a was associated with decreased risks of major adverse limb events and PAD compared to DPP4I [29]. In our sensitivity analysis, the risks of PAD were similar between SGLT2I and GLP1a (Supplementary Table 5). Therefore, we hypothesized that GLP1a might also reduce the risks of PAD amongst T2DM patients, however the results were inconclusive provided the relatively small GLP1a sample size. Our results are aligned with existing analyses on SGTL2I and GLP1a on the risk of PAD amongst hospitalised patients with heart failure (HR: 0.89; 95% CI: 0.57-1.33) [30]. Also, a Danish cohort study showed that GLP1a users had a lower risk of lower limb amputation compared to sulfonylureas [31]. The relationship between SGLT2I and DPP4I with GLP1a all-cause mortality was generally consistent with the existing randomised controlled trials results [32].

However, the results do not necessarily mean that DPP4I increases the risks of new-onset PAD outcomes. It is still plausible to hypothesize that DPP4I may possess protective effects, although further studies are necessary to verify this proposition. DPP4I usage may lower the risk of limb amputation and PAD occurrence amongst diabetic patients [33], but there is limited evidence comparing the role of SGLT2I versus DPP4I on PAD and other crucial patient outcomes [7]. With the available evidence, the results are highly heterogenous and thus difficult to draw any definitive consensus.

On the other hand, a meta-analysis revealed little difference between SGLT2I, DPP4I and GLP1a or other types of oral hypoglycaemic agents on the prognosis of diabetic-foot related limb amputations [34]. One meta-analysis of 12 cohorts suggests that only GLP1a usage, but not SGLT2I, was associated with lower risk of lower limb amputation compared to DPP4I; however, no such correlation was identified with SGLT2I usage [35].

The mechanism(s) behind the relationship between SGLT2I and PAD remain unknown. It was suggested that canagliflozin was associated with enhanced vasodilation of coronary arteries in diabetic mice [36]. The existing literature attributes the possible pleiotropic effects of SGLT2I to the suppression of oxidative stress and autophagy regulation, which may minimize atherosclerosis [37]. SGLT2I usage might also activate NLRP3 inflammasome, associated with inflammatory signalling, thereby reducing the development of atherosclerosis [38]. Alternatively, some found that dapagliflozin could generate anti-inflammatory effects via stimulation of macrophages and cytokines [39].

Clinical implications

The potential protective effects of SGLT2I and DPP4I on various forms of PAD outcomes have attracted increasing attention over recent years. As of 2017, international institutions such as the European Medicines Agency (EMA) issued official warnings against the use of the whole class of SGLT2I treatments due to the potential concern for elevated risks of lower extremity minor and major amputations [30]. Hence, our study results along with existing literature may inspire re-evaluation of current guidelines regulating the use of novel anti-diabetic medications for patients who are at higher risk of developing PAD outcomes. Using data from clinical practice in Hong Kong, our preliminary findings suggest that SGLT2I may exhibit a possible beneficial effect against PAD relative to DPP4I usage. Although it is difficult to establish any definitive claims at this stage, our study may prompt more random clinical trial studies investigate this association. In the long term, this may produce a paradigm shift in diabetic management which have traditionally cantered around glycaemic regulation, to assess the potential secondary benefits of novel anti-diabetic medications for patients at risk of PAD.

Limitations

Several limitations present in this study merits consideration. The observational nature of this study suggests that data on certain variables namely alcohol use, BMI, smoking, family history of PAD and physical activity could not be obtained and analyzed. In addition, the data may also be susceptible to coding errors, under-coding and missing data, resulting in information bias. In compensation of this, laboratory results and comorbidities related to PAD were used to infer for possible risk factors. To corroborate, drug cohorts were matched over a variety of medications and diseases, adjusted with regression, and performed sensitivity analyses to reduce the effect of bias. Moreover, one concern in data on PAD was that the condition was often underdiagnosed as most patients may not present to the hospital until heavily symptomatic. In acknowledgement of this, our study identified clinically significant PAD patients that required medical attentions, amputation, or revascularization. Furthermore, the retrospective nature of the study design suggests that all results derived from our study regarding the relationship between SGLT2I, DPP4I and PAD remain correlational. Hence, further randomized clinical trials are needed to substantiate the potential causal relationship and the consequences of undiagnosed PAD patients.