Effects of Sodium-Glucose Co-Transporter-2 (Sglt2) Inhibitors on Major Cardiovascular Events in Type 2 Diabetic Patients a Meta-Analysis of Randomized Controlled Trials

Background: Sodium-glucose co-transporter-2 inhibitors (SGLT2i) reduce cardiovascular (CV) events in diabetic patients, with a consistent effect on heart failure (HF) related outcomes. However, the effects on ischemic CV events appear less certain, in particular in patients with history of HF. The aim of this meta-analysis is to investigate CV benets of SGLT2i and to assess the effects in patients with and without established atherosclerotic cardiovascular disease (ASCVD), with and without HF, and with estimated glomerular ltration rate (eGFR) < or ³60 ml/min. Methods: We searched PubMed, Embase, Cochrane, ISI Web of Science, SCOPUS, and clinicaltrial.gov databases. We performed a systematic review and meta-analysis of randomised, placebo-controlled, cardiovascular outcome trials (CVOT) of SGLT2i in diabetic patients, assessing the effects of SGLT2i on 3-point major adverse cardiac events (MACE) (CV death, non fatal myocardial infarction (MI), non fatal stroke) and composite of HF hospitalization or CV death. Of 205 articles, 7 CVOTs were included in the meta-analysis. Results: Compared to placebo, SGLT2i signicantly reduced by 10% the risk of 3-point MACE (HR 0.90; p=0.025) and the risk of CV death or HF hospitalization by 24% (Hazard Ratio (HR) 0.76; p<0.001). SGLT2i signicantly reduced HF hospitalization by 30% (HR 0.70; p<0.001), with consistent effects in all subgroups analyzed, CV death by 17% (HR 0.83; p=0.035) and all-cause mortality by 18% (HR 0.82; p=0.024). No signicant effects were observed on MI and stroke. Conclusions: SGLT2i signicantly reduce CV outcome in diabetic patients. SGLT2i remarkably and consistently reduce HF hospitalization, in patients with and without HF at baseline and independently on the presence of ASCVD.


Background
Type 2 diabetes mellitus (DM) is a major risk factor for atherosclerotic cardiovascular disease (ASCVD) and heart failure (HF) (1). This latter condition may ensue as either preserved or reduced ejection fraction cardiomyopathy, as consequence of ischemic cardiac events and/or associated comorbidities, or representing a peculiar type of diabetic cardiomyopathy, that has been linked to metabolic, neurohormonal and in ammatory effects of DM and insulin resistance on cardiac function (2) (3). Sodium-glucose co-transporter-2 inhibitors (SGLT2i) have shown to reduce major cardiovascular (CV) events and slow renal function impairment in patients with type 2 DM (4). Among bene cial CV effects of SGLT2i, the most clinically relevant appears the remarkable reduction of HF hospitalizations, in patients without history of HF at baseline, that has been consistently shown in all CV outcome trials (CVOTs) (5) (6) (7) (8) (9). More recently, the DAPA-HF trial (10) purposely enrolled patients with HF and reduced ejection fraction, with and without type 2 DM at baseline, to assess the effects of dapagli ozin, compared to placebo, on a primary composite endpoint of CV death and worsening HF. The results of this trial for the rst time demonstrated that the bene ts of SGLT2i also extend to patients with systolic HF, with or without DM and independently on etiology. Similarly, the EMPEROR-Reduced (11), enrolled patients with HF and reduced ejection fraction, with and without type 2 DM at baseline, documented lower risk of CV death or hospitalization for HF in patients treated with empaglifozin than those in the placebo group, regardless the presence or the absence of DM.
Currently, based on four CVOTs (5) (6) (7) (8) and one meta-analysis (12), ESC/EASD guidelines (13) and ADA guidelines (14) recommend SGLT2i for primary prevention of CV events, including HF, in DM patients at high and very high CV risk. Yet, although the effects of SGLT2i on HF were consistent in clinical trials and registry data (15) (16) for all agents of the class, the effects on CV ischemic events appear less robust. In fact, a signi cant reduction of 3points major CV events (MACE) (CV death, myocardial infarction (MI), stroke) was observed in three trials (EMPA-REG OUTCOME (5), CANVAS Program (6) and CREDENCE (7)) but not in two other trials (DECLARE-TIMI 58 (8), VERTIS CV (9). In addition, differently from HF endpoint, effects on MACE were only evident in patients with ASCVD at enrollment.
A previous meta-analysis (12), including 34322 patients enrolled in three CVOTs on SGLT2i, reported a signi cant 11% reduction of MACE with a statistically signi cant effect on MI (11%) and CV death (16%), with high heterogeneity among trials for this latter endpoint. Notably, no effects on MACE were observed in patients without ASCVD at baseline. In contrast, a substantial 29% and 36% reduction of HF hospitalizations was observed, respectively, in patients with and without ASCVD, without heterogeneity among trials. More recently, a meta-analysis (17) on 46969 diabetic patients enrolled in ve trials documented a signi cant reduction of major adverse CV events risk in diabetic patients treated with SGLT2-inhibitors. Indeed, this meta-analysis demonstrated that the largest bene t across the class was for an associated reduction in risk for HF hospitalization and kidney outcomes. However, this meta-analysis (17), did not included data of DAPA-HF trial (10) and EMPEROR-Reduced trial (11), that provided yhe most robust data in patients with systolic HF.
Thus, the aim of the present meta-analysis was to investigate CV outcomes of SGLT2i in 7 CVOTs including 49108 diabetic patients and to assess the effects in subgroups of patients with and without established ASCVD, as well in those with and without HF at baseline, and in patients with estimated glomerular ltration rate (eGFR) < or ≥60 ml/min.

Search strategy and selection criteria
The meta-analysis was conducted according to PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analysis) reporting guidelines (18). We searched PubMed, Embase, Cochrane, ISI Web of Science, SCOPUS, and clinicaltrial.gov databases to identify all eligible trials with a primary outcome including 3-point MACE and composite of HF hospitalization or CV death, comparing e cacy and safety of SGLT2i to placebo in patients with type 2 DM at high CV risk. The terms used for the research, as suggested by an expert medical librarian (A.P.), were: 'SGLT2 inhibitors', 'sodium-glucose-cotransporter-2 inhibitors, 'gli ozin', 'empagli ozin', 'canagli ozin', 'dapagli ozin', 'ertugli ozin', 'placebo', 'cardiovascular disease', 'atherosclerotic cardiovascular disease', 'MACE', 'major adverse cardiovascular event', 'death', 'diabetes mellitus', 'heart failure', 'randomized controlled trials'. The reference lists of included studies were searched for additional studies. Searches were done up from 01 May 2020 until 30 August 2020.
Study inclusion criteria were Phase 3 randomized and controlled allocation to SGLT2i versus placebo enrolling more than 500 patients, articles published from January 2012 to June 2020, assessing at least one of following major CV outcomes: 3-point MACE, HF hospitalization, CV death, non-fatal myocardial infarction and stroke, all-cause mortality, in patients with type 2 DM. No language, background medication therapy, background CV risk or disease restrictions were applied. We excluded observational non-randomized studies registries, ongoing trials without results, duplicate series, meta-analysis, abstracts and oral communications. Articles were screened for ful llment of inclusion criteria by four independent reviewers (P.D.N., S.D.A., I.E., L.B.).
Reviewers compared selected trials and discrepancies were resolved by four authors (P.G., S.P., A.P., G.D.). Corresponding author was asked to provide fulltext articles if not available. Neither ethics approval nor patient consent was required for this analysis. This meta-analysis was previously registered in PROSPERO (CRD42020189257).

Data analysis
The primary e cacy outcomes were: 1) 3-point MACE (composite of CV mortality, non-fatal MI or non-fatal stroke); 2) CV death or HF hospitalization. The secondary e cacy outcomes were:1) CV death 2) non-fatal MI 3) non-fatal stroke 4) HF hospitalization 5) all-cause death. The CANVAS Program (6) consisted of two trials, CANVAS and CANVAS-R, but are presented combined. In this trial the number of events, strati ed for drug and placebo, is presented as patients per 1000 patient-years, differently from other studies. So, for this trial we don't reported the number of events. In addition, from DAPA-HF (10) and EMPEROR-Reduced (11), enrolling HF patients with and without type 2 DM, we extracted only data of type 2 DM patients (19) for available endpoints.
In the rst step the meta-analysis was performed in the overall population, subsequently patients were strati ed by presence of established ASCVD or multiple CV risk factors, previous history of HF, and by renal function (eGFR < 60 or >/= 60 ml/min), in order to explore the effects of SGLT2i in speci c subgroups. In case of renal function, when required, effect estimates for subgroups within the same study were merged by use of a xed-effects model.
When the data were not presented in main study, we extracted data from supplementary analyses and sub-studies of main trials (20), (21), (22), (23), (24), (25). The data of VERTIS CV (9) trial were extracted from presentation of study results, available online, because the article has not yet published.
Safety endpoints of interest consisted in lower limb amputations, bone fractures, diabetic ketoacidosis, genital infections, pancreatitis, severe hypoglycemia, urinary tract infections, and volume depletion.
Statistical heterogeneity between studies was assessed using the Cochrane Q statistic and I 2 statistic. Standard thresholds were considered for judging the percentage of total variability across studies not due to sampling error (I 2 ): 25% or lower for low heterogeneity, 26-50% for moderate heterogeneity and greater than 50% for high heterogeneity. The random effects model was a-priory selected to obtain pooled estimates of treatment effect (Hazard Ratios -HRs) with the 95% Con dence Intervals (95% CIs) for all e cacy outcomes. The Paule-Mandel method for estimating the between-study variance τ 2 was used. Due to the small number of studies, the Hartung and Knapp (HK) adjustment was employed.
Subgroup analyses for e cacy outcomes to assess effect modi cation by presence of established ASCVD or multiple CV risk factors, previous history of HF, and by renal function (eGFR < 60 or >/= 60 ml/min) were based on random effects models, applying the Paule-Mandel method and HK adjustment. A random effect model, with HK adjustment, was further applied to obtain pooled estimates of Risk Ratio (RR) with the corresponding 95% C.I. for the main safety outcomes. Due to the small number of studies no attempt to assess publication bias was made. Statistical analysis was performed using the R statistical programming environment, Version 3.5.2 (http://www.r-project.org). Package meta, Version 4.11 (26) was used for all the meta-analysis elaborations.
Effect of SGLT2 inhibitors on e cacy outcomes in selected subgroups.
Effect of SGLT2i on CV outcomes in patients with and without ASCVD.

Discussion
The ndings of the current meta-analysis, including 49108 diabetic patients enrolled in seven CVOT, show that SGLT2i signi cantly reduce the 3-point MACE endpoint (10%), as well the composite endpoint of CV death or HF hospitalization (23%) in patients with type 2 DM.
Among the single components of 3-point MACE, CV death was signi cantly reduced (17%), con rming previous analysis (12). However, no signi cant effects were observed for non-fatal MI at variance with a previously reported signi cant reduction (12). Similar to previous analysis, our ndings demonstrate a neutral effect on stroke, and a signi cant reduction of all-cause mortality (18%). Notably, the favourable effect on MACE was mostly evident in patients with ASCVD with no signi cant effects in those without.
The second primary endpoint of this meta-analysis was the composite of CV death or HF hospitalization, that was substantially reduced (23%  (27), to clinically manifest HF in diabetic patients, thus representing a novel opportunity for primary prevention of HF in this high risk group of patients. In addition, the 30% reduction of HF hospitalization in patients with HF at baseline, together with the 25% reduction of the composite of CV death and HF hospitalization, extend the bene t of SGLT2i to treatment of patients with HF and reduced ejection fraction on top of optimized HF therapy. These effects on HF prevention and therapy are of potentially relevant impact to reduce the escalating clinical and economic burden of HF worldwide.
Insights on the pathogenetic mechanisms of clinical bene t are beyond the scope of the meta-analysis. In fact, several hypotheses have been made to explain CV effects of SGLT2i (28) (29) but no de nitive explanations have been reached. Conceivably, the concurrent favourable effect on HF and renal disease likely explain the signi cant bene t on CV and all-cause mortality observed in the current and previous meta-analysis (30). In contrast, no signi cant effects are apparent on ischemic non-fatal events, including stroke and MI, that were not signi cantly reduced in any CVOT.
Use of SGLT2i appear to be generally safe, although a signi cant increase of genital infections and of ketoacidosis has been observed.

Comparison with previous meta-analysis
Two recent meta-analysis (17,31) reported the effects of SGLT2i in diabetic and non diabetic patients, with and without HF. McGuire et al. (17) reported the effects of SGLT2i in 46969 patients included in 5 studies, excluding two trials (10, 11) from analysis. On the other hand, Zannad et al. (31) in a metaanalysis, including only these two trials for a total of 3995 diabetic plus HF patients, but not including relevant subgroups of similar patients from other 5 studies that were included in our analysis. Thus, the current study reports the largest number of patients and, in particular, more than twice patients with diabetes and HF.

Limitations
We acknowledge that use of trial rather than patient data is a limitation of meta-analysis. In addition, subgrouping of patients is based on investigator reporting of ASCVD and HF, that may be responsible for misclassi cation of individual patients. Within the subgroup of patients with HF, we could not make a separate analysis in patients with reduced or not reduced systolic function and therefore the effects of SGLT2i in patients with preserved or mildly reduced ejection fraction remain not adequately investigated. Finally, due to the small number of studies included in this meta-analysis, no attempt was made to formally evaluate potential publication bias.

Conclusions
In summary, our ndings further strengthen ESC/EASD (13) and ADA guidelines (14), that recommend SGLT2i in high and very high risk diabetic patients, supporting the bene ts of this class of drugs at early stages of diabetic disease. The bene cial effects on HF and mortality should prompt adherence to guidelines and implementation of SGLT2i therapy in clinical practice to reduce the burden of CV disease in type 2 diabetic patients.