FormalPara Key Summary Points

Why carry out this study?

Several guidelines recommend sodium-glucose type 2 cotransporter SGLT2-I as the first pharmaceutical option in treatment-naïve and previously treated patients with type 2 diabetes mellitus (T2DM) owing to the substantial cardiovascular benefits in patients with T2DM and cardiovascular risk factors or established cardiovascular diseases.

However, there are a limited number of specific sub-studies on patients with established extensive coronary artery disease, such as patients with coronary disease affecting the left main coronary artery or multivessel disease.

This real-world study aimed to evaluate the effectiveness of SGLT2-I in patients with T2DM and extensive coronary artery disease from Spain.

What was learned from the study?

This real-world study showed that the patients with T2DM and extensive coronary artery disease treated with SGLT2-I had a reduced risk of all-cause death and statistically significant better renal function in the SGLT2-I group at the final follow-up.

We found a statistically significant increase in unplanned revascularization in the SGLT2-I group in crude and IPTW-adjusted analyses. The follow-up bias could explain this contradictory result due to the lower mortality in the SGLT2-I group. However, further research into this hypothesis-generating secondary outcome is warranted.

Introduction

Type 2 diabetes mellitus (T2DM) is one of the pandemics of the twenty-first century. Sixty million adults in Europe and 45 million in the USA are thought to have T2DM [1], and according to data from the American Diabetes Association, approximately one-quarter of people with diabetes in the USA and nearly half of Asian and Hispanic Americans with diabetes are undiagnosed [2]. The effects of this condition on the cardiovascular (CV) health of the individuals and their offspring create further public health challenges that agencies are attempting to address globally as CV disease, especially coronary artery disease (CAD) and heart failure (HF), is a significant clinical consequence of T2DM and is the leading cause of death in patients with T2DM.

Since 2015, a new group of drugs with novel mechanisms of action have been available—the sodium-glucose cotransporter 2 (SGLT2-I) inhibitors; these increase renal glucose excretion [3]. The first drug in this class that showed a reduction in cardiovascular mortality was empagliflozin in the EMPA-REG OUTCOME study, with a 14% reduction in the primary endpoint (cardiovascular death, nonfatal myocardial infarction [MI], and nonfatal stroke), 38% relative risk reduction in cardiovascular mortality (p < 0.001), and 35% relative risk reduction in admission for HF in patients treated with empagliflozin [3]. Later these findings were confirmed in the CANVAS trial [4] with canagliflozin and the DECLARE TIMI trial [5] with dapagliflozin. Several meta-analyses confirmed that the benefit of reducing cardiovascular events was a class effect [6,7,8,9].

Additional studies, such as the EMPEROR trial [10], DAPA HF trial [11], EMPEROR Preserved trial [12], DELIVER trial [13], EMPA KIDNEY trial [14], and DAPA CKD trial [15], found that the beneficial cardiovascular effects of these drugs were observed in patients with and without diabetes, reduced or preserved ejection fraction, and even at patients with glomerular filtration rates up to 20 ml/min.

This solid evidence positioned these drugs in the latest 2019 European Society of Cardiology Guidelines on Diabetes, Pre-Diabetes and Cardiovascular Diseases as a first pharmaceutical option in treatment-naïve and previously treated patients with T2DM [16]. However, the inclusion criteria in the earliest trials were broad in terms of definition of CV disease, they could include patients at risk but without CV disease, and there are a limited number of specific sub-studies on patients with established, extensive CAD. SGLT2-I treatment in patients after acute MI seems to be associated with prognostic benefit in observational studies [17] and improvement in surrogate markers such as NT-ProBNP, cardiac sympathetic and parasympathetic activity, or in left ventricular mass indexed to body surface area in small randomized trials [18,19,20]. Other observational “real life” study including patients post MI showed that dapagliflozin was a predictive factor for HF rehospitalizations (HR 0.417, 95% CI 0.417–0.838, p = 0.001) in a propensity-matched study including 961 patients [21]. In another study, Salahuddin et al. estimated the potential benefits of SGLT2i treatment using pivotal trial hazard ratios applied to observed deaths in a cohort of SGLT2-I-eligible veterans, finding that a 7.9% (4.5–10.8%) potential mortality reduction might have been expected had the patients been treated with SGLT2-I [22]. With this background, we designed this study including only patients with T2DM and extensive CAD, and aimed to compare all-cause mortality of those patients under the cardiovascular protective effect of SGLT2-I to those without SGLT2-I.

Methods

This study was a single-center, retrospective, observational study including consecutive patients with T2DM and a new diagnosis of extensive CAD. Extensive CAD was defined as angiographic stenosis of at least 50% affecting the left main coronary artery (LMCA) or three main vessels (3MV). Patients with their first diagnosis of extensive CAD were included at discharge, regardless of the revascularization treatment strategy, and had to have a follow-up of at least 12 months available.

Patient Selection and Data Administration

The inclusion criteria were (1) older than 18 years of age at the diagnosis; (2) previous or new T2DM diagnosis as per American Diabetes Association criteria; (3) clinical presentation at index admission as angina, silent ischemia (chronic coronary syndrome or acute coronary syndrome[ACS]); or (4) angiographic stenosis greater than 50% affecting the LMCA or 3MV.

The exclusion criteria were (1) history of previous revascularization (percutaneous or surgical); (2) estimated glomerular filtration rate (eGFR) of less than 45 ml/min (this was the initial threshold for indication in 2015); (3) previous HF hospitalization; (4) death of the patient during the index admission; (5) SGLT2-I treatment less than 6 months after discharge; (6) patients with less than 1 year of follow-up.

The authors obtained anonymous data that included index admission and follow-up visits from the interventional cardiology database of the Hospital Clinico San Carlos (Madrid, Spain) from 2015 (the beginning of the commercialization of SGLT2-I in Spain) until 2020. The database was protected and only accessible to the investigators, and the confidentiality of patient data was protected by assigning a unique study number to each patient. During the follow-up, baseline characteristics, clinical presentation, procedural findings, and information on vital status and major adverse cardiovascular events (death, cardiovascular death, non-fatal MI, non-fatal stroke, unplanned revascularization, and GFR) were recovered from electronic records and telephone calls.

Outcomes

The primary outcome was to compare the adjusted hazard ratio (HR) of the all-cause mortality in patients with T2DM and a new diagnosis of extensive CAD between patients treated with and without SGLT2-I at discharge from the index hospitalization.

The secondary outcomes were to compare the adjusted HR of cardiovascular death, MI, stroke, HF readmission, unplanned revascularization, and renal function in the follow-up between treatment groups (SGLT2-I and non-SGLT2-I).

Estimated glomerular filtration rate was calculated with the Modification of Diet in Renal Disease (MDRD) formula [23]: eGFR = 175 × serum Cr−1.154 (mg/dL) × age−0.203 × 1.212 (if patient’s race is black) × 0.742 (if patient is female).

Statistical Analysis

Categorical variables were reported as percentages and analyzed using the Chi-square test, and quantitative variables were reported as mean ± standard deviation (SD). The main objective was analyzed with Cox regression, and Kaplan–Meier curves were drawn to compare the SGLT2-I and non-SGLT2-I groups. Inverse probability of treatment weighting (IPTW) [24], a propensity score method, was used to adjust baseline variables distributed differently between groups. The propensity score was estimated using a logistic regression model that included all potential confounders that were distributed differently between groups (age [categorized as under and above 70 years], gender, insulin-dependent diabetes mellitus, smoking, neoplasia, body mass index, left ventricular ejection fraction [LVEF], eGFR), or that could have clinical relevance (presentation as ACS). We then performed a balance assessment, comparing the distribution of measured baseline covariates between the treatment groups before and after weighting using standardized differences, which are less sensitive to sample size than conventional p values. As a rule of thumb, a standardized difference of 0.10 or less may be considered a negligible imbalance between groups (Supplementary Table 1). As an added method to assess the balance assessment, we used the overidentification test for covariate balance. IPTW was followed by multivariable Cox regression to control by known potential confounders related to the outcome (complete revascularization and revascularization treatment strategy).

We also calculated the power of the study sample size (420 patients) for the primary outcome. We had a total of 68 observed events, so the probability of an event would be 0.16. Taking an alpha value of 0.05, the coefficient of the model (− 1.156), and the standardized difference of the coefficient (0.482), the estimated power would be 99.54% (or 0.995). Statistical significance was established at P ≤ 0.05 (two-tailed) for the comparisons and measures of association. All statistical analyses were conducted using Stata IC 15.1 (Stata Corp., College Station, TX, USA).

Compliance with Ethics Guidelines

The Clinical Research Ethical Review Board of the Clinico San Carlos Hospital in Madrid, Spain approved this study on May 13, 2021 (Approval number 21/350-O_M_OD), and chaired by Dr. García Arenilla. Informed consent from patients was waived for the use of de-identifed data for this observational analysis, as recommended by the Clinical Research Ethical review board. This study was performed in accordance with the Helsinki Declaration of 1964 and its later amendments. All authors have participated in elaborating this manuscript according to authorship criteria recommended by the International Committee of Medical Journal Editors (ICMJE), had access to full data, and approved the final manuscript for publication.

Results

Study Population and Baseline Characteristics

We recruited 420 consecutive patients (Fig. 1). The mean age of the included patients was 71.2 ± 11.1 years, mostly with an average ejection fraction and a mean glycated hemoglobin of 7.3 ± 1.2% (Table 1). There were some relevant differences in baseline characteristics: the SGLT2-I group included younger patients, more frequently smokers, less frequently with neoplastic diseases, and had a better baseline renal function but without significant differences in the rate of chronic kidney disease (Table 1). Regarding study groups, a total of 104 patients (24.8%) had started treatment after discharge with SGLT2-I, whereas the other 316 (75.2%) patients comprised the non-SGLT2-I group. Within the SGLT2-I group (69 [66.3%] patients received empagliflozin, 34 [32.7%] received dapagliflozin, and only 1 patient received canagliflozin). Patients in the SGLT2-I group were younger, more frequently male, and had slightly higher LVEF (see Table 1 for baseline characteristics.). After IPTW, the two groups were balanced with low standardized differences in the baseline covariates, as shown in Supplementary Table 1. This balance was confirmed by the overidentification test for covariate balance (p = 0.77).

Fig. 1
figure 1

Flowchart of the study. LMCA left main coronary artery, 3MV three main vessels, T1DM type 1 diabetes mellitus, SGLT2-I sodium-glucose cotransporter 2

Table 1 Baseline characteristics

The presentation was ACS in 44.3% (47.1% in the SGLT2-I group and 43.4% in the non-SGLT2-I group; p = 0.503). Regarding the CAD distribution, 86.7% of the patients had 3MV CAD (87.1% in the SGLT2-I group and 85.7% in the non-SGLT2-I group; p = 0.574); 12.1% had LMCA disease + another main vessel (15.4% in the SGLT2-I group and 9.7% in the non-SGLT2-I group; p = 0.481); and 1.2% isolated LMCA disease (2.1% in the SGLT2-I group and 1.4% in the non-SGLT2-I group; p = 0.91). The treatment strategy had no differences among groups (Table 2).

Table 2 Clinical presentation and revascularization treatment strategy

In the overall population, percutaneous coronary intervention + optimal medical treatment (OMT) was the preferred strategy (48.3%), coronary artery bypass graft + OMT in 39.8%, and OMT-only in 11.9%. Notably, only 28.8% of the patients achieved complete revascularization (31.7% in the SGLT2-I group and 27.8% in the non-SGLT2-I group; p = 0.45). Medical therapy at discharge from index hospitalization also had no differences among groups (Table 3), with three out of four patients taking metformin and one out of three patients taking insulin. Only 1.2% of patients were under glucagon-like protein 1 (GLP-1) agonist treatment.

Table 3 Treatment at discharge other than SLGT2-I

Outcomes

The mean follow-up was 3 ± 1.6 years. Table 4 shows cardiovascular treatments and the metabolic parameters and renal function in both groups (SGLT2-I and non-SGLT2-I) at follow-up. The mean time on treatment with SGLT2-I during follow-up was 1.6 ± 0.8 years. With IPTW adjustment, all relevant differences in baseline characteristics (age, body mass index, presentation as ACS, LVEF, eGFR) were well balanced with standardized differences of less than 0.10.

Table 4 Medical treatment and laboratory parameters at follow-up

The primary outcome, all-cause mortality, was 16.4% in the study population (3.8% in the SGLT2-I group vs 20.6% in the non-SGLT2-I group) and crude HR in the SGLT2-I group was 0.18 (95% confidence interval [CI] 0.07–0.48; p = 0.001). This difference remained significant after the IPTW propensity score adjustment (HR IPTW 0.32 [95% CI 0.12–0.81]; p = 0.016). Figure 2 illustrates Kaplan–Meier curves for the primary endpoint, and Table 5 shows hazard ratios for primary and secondary outcomes.

Fig. 2
figure 2

Kaplan–Meier curves for the primary endpoint (all-cause mortality). SGLT2-I sodium-glucose cotransporter 2, HR hazard ratio, IPTW inverse propensity of treatment weighing, CI confidence interval

Table 5 Hazard ratios for primary and secondary outcomes

Concerning the secondary outcomes, overall CV mortality was 9.5%, lower in the SGLT2-I group (2.9%) compared to the non-SGLT2-I group (11.7%), with crude HR of 0.24 (95% CI 0.07–0.74); p = 0.001, but after IPTW adjustment the HR was no longer significant: HR IPTW of 0.52 (0.19–1.44); p = 0.208. There were no differences in new MI (9.6% SGLT2-I group vs 6.0% non-SGLT2-I group), stroke (1.9% SGLT2-I group vs 3.8% non-SGLT2-Igroup), nor HF rehospitalization rates (7.7% SGLT2-I group vs 12.7% non-SGLT2-I group), see Table 5 for crude and IPTW adjusted hazard ratios. The composite endpoints of death or HF readmission and CV death or HF admission were lower in the SGLT2-I group in the crude analysis, but there were no differences after the IPTW analysis (Table 5). We also observed an increase of unplanned revascularizations in the group of patients treated with SGLT2-I with a crude HR of 2.44 (95% CI 1.2–4.93), p = 0.01; and IPTW HR of 3.03 (95% CI 1.43–6.44), p = 0.005. Finally, during follow-up the renal function remained better in the SGLT2-I group, with no significant differences in metabolic parameters (LDL, glycated hemoglobin). Moreover, from those patients with normal renal function at baseline, numerically more patients (53, 17.8%) in the non-SGLT2-I group developed renal failure (eGFR < 60 ml/min), compared to 11 (11.2%) in the SGLT2-I group (p = 0.126).

Discussion

Our study evaluated all-cause mortality and cardiovascular outcomes of consecutive patients with type 2 diabetes mellitus with a recent diagnosis of extensive coronary artery disease (angiographic stenosis affecting the LM or 3MV disease) according to treatment with or without SGLT2-I. The main finding was a statistically significant reduction in the primary endpoint (all-cause mortality) in the group of patients treated with SGLT2-I vs patients treated with standard antidiabetic drugs. This difference was maintained after the IPTW propensity score adjustment. With regard to additional secondary endpoints, we found a statistically significant improvement in renal function in the SGLT2-I group and a statistically significant increase in unplanned revascularization in the SGLT2-I group, both in crude and IPTW-adjusted analyses.

Randomized clinical trials are the cornerstone of evidence-based medicine. However, they often include a selected population, frequently less sick than in daily clinical practice. In this scenario, patients frequently have comorbidities, drug interactions, or even economic limitations that limit the penetration of new beneficial drugs such as SGLT2-I. In this regard, observational evidence might complement randomized trial results. This real-life study showed that only 25% of patients with T2DM received an SGLT2-I as part of their treatment, and the population that received this drug was younger, with better LVEF and fewer comorbidities. We found that even after IPTW adjustment, this group of patients treated with SGLT2-I had a risk reduction of all-cause death. This result is consistent with previous data from randomised trials and meta-analysis [6,7,8,9] but has the interest of focusing specifically on patients with extensive CAD.

Moreover, the reduction in all-cause mortality is driven by a reduction in CV mortality, consistent with previous evidence. For the secondary objectives, we found a non-statistically significant reduction in the composite endpoints of death or HF admission and CV death or HF admissions, although, after IPTW, it was no longer significant, possibly because of the relatively small number of patients in this study. Concerning MI and stroke, there were no statistical differences between both treatment groups and the overall event rate was low. These results are similar to the early pivotal trials of this pharmacological group in patients with diabetes and established CV disease [3,4,5].

Existing research in patients with T2DM and CAD was mainly focused on patients post acute MI or evaluated a CAD cohort but using estimated mortality (theoretical benefit if patients were treated with empagliflozin) [18, 19, 21, 22]; therefore, what our study adds is a population that includes chronic and acute coronary syndromes, extensive CAD, and direct evidence (observed all-cause mortality) in patients treated with empagliflozin, with IPTW adjustment of confounding factors.

The renal benefits [6, 14, 15] of this pharmaceutical group were confirmed in our study, with a trend towards reducing the development of renal failure in the final follow-up independently of the revascularization strategy and the use of contrast during percutaneous coronary intervention or extracorporeal circulation in patients treated with coronary artery bypass graft.

In another observational, propensity-matched study, Mao et al. found a reduction in rehospitalization for HF in those patients treated with dapagliflozin after acute MI [21]. Our study could not find a statistically significant reduction, but a numerical reduction (HF rehospitalization 7.7% SGLT2-I group vs 12.7% non-SGLT2-I group) that aligns with their findings. Differences could be related to the bigger population and the post-MI setting in the Mao et al. study.

Our study also found an increased risk of revascularization in the patients treated with SGLT2-I during the follow-up. No clinical trials have evaluated this issue; however, some observational studies have found a protective effect of these drugs, reducing the revascularizations [25] in patients treated with SGLT2-I or restenosis-related events [26]. This contradictory result could be explained by the follow-up bias due to the lower mortality in the SGLT2 group. However, further research into this hypothesis-generating secondary outcome is warranted.

The mortality benefit found in our study seemed independent of the metabolic control (both groups of patients had similar levels of HbA1c at follow-up, even numerically higher in the SGLT2-I group). This apparently discordant observation amplifies the possibility that the pleiotropic effects of SGLT2-I contribute to cardiovascular and renal benefits beyond blocking the SGLT2 renal receptor [27, 28]. Several off-target effects of SGLT2 inhibitors have been proposed, including the direct effect on the sodium-hydrogen exchanger 1 (NHE1) in the heart, NHE3 in the kidney, extracellular signal-regulated kinases (ERK) 1/2 in the pulmonary endothelium, and NHE9 in inflammatory cells that could influence major adverse cardiovascular events, HF, and kidney outcomes [28,29,30,31]. Left ventricular mass reduction and modulation of cardiac sympathetic and parasympathetic activity could also be responsible for this improvement in prognosis [19, 20].

Finally, only one out of four potential candidates for SGLT2-I were treated with the drug during this study. This underpenetration of the therapy might be explained for a variety of reasons. First, these were not patients with a new diagnosis of diabetes mellitus, so despite the diagnosis of extensive CAD, the therapeutic inertia might leave the patients with their previous antidiabetic treatment, and probably the diagnosis of extensive CAD might divert the focus of the treatment towards revascularization and standard secondary prevention treatment. Second, these drugs started were labelled with some limitations, such as eGFR ≥ 60 ml/min to start the treatment in the absence of HF and ≥ 45 ml/min to continue only if the drug had been started earlier (this label changed recently to allow any patient with ≥ 20 ml/min eGFR in 2021). Third, these drugs are costly, approximately 660 euros per year of treatment in Spain and outside of the HF scenario, physicians might consider them less beneficial. Finally, comorbidities and increased genitourinary infections might limit tolerability. Upcoming cardiology guidelines should reinforce the need to avoid therapeutic inertia and promote the reassessment of optimal antidiabetic treatment in the secondary prevention of CAD, such as recently stated in the European Society of Cardiology CV prevention guidelines [32].

This study has the inherent limitations of an observational retrospective design, and unmeasured confounders could bias results despite the IPTW adjustment. Other cardioprotective drugs such as renin–angiotensin–aldosterone system inhibitors and beta blockers were not recorded in the study database, so we acknowledge a potential bias should these drugs had been prescribed differently to study groups (nonetheless the patients were theoretically treated with optimal medical therapy). As a result of the study design, we chose only endpoints that would be robust in retrospective data gathering, such as death, MI, stroke, or HF readmission recorded. The study database did not record relevant side effects from SGLT2-I, such as ketoacidosis and genitourinary infections, so this study could not assess the balance between cardiovascular benefits and drug side effects. Some SGLT2-I side effects that were recorded (dyslipidemia, renal failure) in the study were not significantly different (dyslipidemia) or even improved (renal function) in the SGLT2-I group. Other limitations are the relatively low number of patients that might limit the power to detect differences in secondary outcomes (although the statistical power for the primary outcome was considered appropriate at 99.54%), and also the possible underreporting of events due to the retrospective nature, or a potential selection bias as in any observational study (although the trial included consecutive patients).

Conclusion

This real-world study found that in patients with type 2 diabetes mellitus and extensive coronary atherosclerotic disease (left main or three-vessel disease) only one out of four patients were treated with SGLT2-I as a part of their secondary prevention treatment. The patients treated with SGLT2-I were associated with a reduction in the risk of all-cause death in the crude analysis and the IPTW adjusted analysis, and with an improvement in renal function at the end of follow-up. Finally, we observed an increase in unplanned revascularizations in the group of patients treated with SGLT2-I in the crude and adjusted analyses.