The cardiovascular outcomes, heart failure and kidney disease trials tell that the time to use Sodium Glucose Cotransporter 2 inhibitors is now

Abstract Sodium glucose contrasporter 2 inhibitors (SGLT2i) were initially introduced as a novel class of modestly effective antiglycemics. Over the last 5 years, multiple members of this class have been examined for their cardiovascular safety, effects on heart failure with reduced ejection fraction (HFrEF) and chronic kidney disease (CKD) in diverse populations with or without diabetes type 2. The plethora of studies and outcomes examined make it difficult for the practitioner to track the entirety of the evidence. SGLT2i improve cardiorenal outcomes and have a beneficial risk benefit ratio across populations with cardiovascular disease, HFrEF and kidney disease. In this quantitative review, we synthesize the data from the large outcomes trials about the benefits and risks of SGLT2i. SGLT2i reduce all cause, cardiovascular mortality, heart failure hospitalizations, need for dialysis and acute kidney injury as a class effect across a broad range of populations with diabetes Type 2 at risk for cardiovascular disease, patients with HFrEF or CKD with or without diabetes. While certain adverse events for example, diabetic ketoacidosis and genital mycotic infections are reproducibly increased by SGLT2i, the absolute increase in the risk of these complications is smaller than the absolute risk reductions conferred by SGLT2i. Other complications such as amputations, fractures and urinary tract infections are increased to a lesser degree, or not at all (e.g., hypoglycemia). Overall, SGLT2is appear to have a favorable safety profile and thus should be used by cardiologists, nephrologists, endocrinologists, primary care physicians when managing the cardiorenal risk of their patients.

kidney disease (CKD) have shown a dramatic improvement in cardiovascular, kidney and survival outcomes, across diverse populations of patients with and without diabetes type 2 (T2D).
The CVOTs of SGLT2s were conducted to comply with the Food and Drug Administration (FDA)'s 2008 mandate of that any new antidiabetic agents should be shown to be free of excess cardiovascular risk. 4 The notion that antiglycemics may be associated with cardiovascular risk has been around since tolbutamide, a first generation sulfonylurea, was linked to higher cardiovascular mortality in the 1970s 5 . Nevertheless, the cardiovascular risk of specific antiglycemics remained underappreciated until the Phases 2 and 3 trials of the proliferator-activated receptor agonists muraglitazar 6 and rosiglitazone 7,8 showed increases in congestive heart failure, death and cardiovascular events. In combination with the 22% increased risk of all-cause mortality in ACCORD trial, 9 this data made it clear that the cardiovascular safety of antiglycemics should be proven in controlled trials.
The CVOTs reported to date include empagliflozin's EMPA-REG Outcome, 10 canagliflozin's integrated CANVAS Program 11 (comprised of two trials: CANVAS and CANVAS-R), dapagliflozin's DECLARE-TIMI-58 12 and ertugliflozin's VERTIS-CV. 13 The EMPA-REG OUT-COME trial, 10,14 demonstrated superiority of empagliflozin for the three-point major cardiovascular event (MACE-3: a composite of cardiovascular death, nonfatal myocardial infarction [MI] or nonfatal stroke), with significantly lower rates of cardiovascular death, hospitalization for heart failure (HHF), all cause death and kidney outcomes.
Similar to empagliflozin, canagliflozin 11 was demonstrated to be superior to placebo for the MACE-3. On the other hand, dapagliflozin 12 and ertugliflozin 13 only achieved non-inferiority in their respective cardiovascular safety trials, DECLARE-TIMI-58 and VERTIS-CV.
DAPA-HF 15 and EMPEROR-reduced 16 recruited patients with or without T2D but with reduced ejection fraction (EF < 35%). Both studies showed highly significant reductions in the primary outcome of HHF or cardiovascular death irrespective of baseline kidney function, proteinuria, gender, race, diabetic status. CREDENCE 17 and DAPA-CKD 18 examined the effects of canagliflozin and dapagliflozin in patients with persistent kidney damage, that is, persistent macroalbuminuria and impaired eGFR (between 30 and 90 ml/ min/1.73 m 2 ) on a background of maximally tolerated therapy with RASi. While CREDENCE recruited patients with type 2 diabetes, DAPA-CKD also included patients without T2D. These CKD trials not only demonstrated that the drugs reduced a composite that involved worsening kidney function, the need for dialysis, and reduced cardiovascular outcomes.
The rapidly expanding landscape of the SGLT2i trials can be rather daunting to follow, given the variable study designs, populations, and the sequential testing strategy employed by the study authors. There remains understandable confusion about the relative merits of one SGLT2i versus another since individual study results tend to be presented as "significant" or "nonsignificant" without consideration of the totality of evidence across trials. In the present meta-analysis we will synthesize the primary outcomes and the high-profile adverse events, which to the authors' opinion seem to act as a barrier against the wider adoption of the SGLT2i.

| Eligibility criteria
We included all randomized, placebo-controlled studies involving SGLT2i that included cardiovascular endpoints, such as cardiovascular (CV) death, MACE-3 and its components, HHF, or composite renal end-points of the four commercially available selective SGLT2i in the United States (canagliflozin, dapagliflozin, empagliflozin, ertugliflozin).
Other prospective and nonrandomized studies were all excluded.

| Search methods
We built an incremental version of the "renal outcomes" metaanalysis, 19

| Selection of studies
The two authors independently verified the abstract of the identified studies for meeting the inclusion and the exclusion criteria. For studies reported in multiple publications, the authors merged the baseline and follow up data from all publications. The Cochrane Risk of Bias tool 20 was used to assess the potential for bias for each individual study analyzed.

| Data abstraction and statistical analyses
Information about trial characteristics, primary and secondary outcomes efficacy measures (Hazard Ratios [HR] and their confidence intervals) were extracted from the source publications and converted to log-HR and standard errors by standard formulas. For the analysis of adverse events, the number of events in the SGLT2i and placebo arms in each trial, were analyzed via a logistic regression and the log-odds (log-OR) ratio and standard error were extracted from the logistic model output.
Log-HRs and log-ORs were synthesized via fixed (FE) and random effects (RE) meta-analysis, to provide an overall Treatment Effect. The fixed effects model assumes that the treatment effect will be the same across all trials, while the random effects model assumes that the treatment effect in each of the trials, are similar but not identical. Statistical heterogeneity among the trials synthesized was quantified by the I 2 statistic and the between study variance(τ 2 ), which was calculated by REstricted Maximum Likelihood (REML). The p-value of the Q test was used to test for the statistical significance of the observed heterogeneity. To explore replicability of these findings we calculated the prediction interval 21 : this is the 95% confidence interval for the treatment effects in a future trial.

| RESULTS
The initial quantitative synthesis 19 had screened 2085 papers to include four main trial publications, eight secondary analysis papers, and one correspondence letter for three cardiovascular safety trials EMPA-REG Outcome, CANVAS Program, DECLARE-TIMI-58 and one F I G U R E 2 Effects of SGLT2i on the composite outcome of cardiovascular death or heart failure hospitalization (a) hospitalization for heart failure (b), the three point major adverse cardiovascular events (c) and nonfatal myocardial infarction (d). Random effects model synthesizes the effect across all studies, and the prediction interval gives the 95% range for the result of a future SGLT2i trial. Event rates (per 1000 patient years) are shown for both the SGLT2i and the placebo arms. CKD, chronic kidney disease; CVOT, cardiovascular outcome trial; HFrEF, heart failure with reduced ejection fraction; HHF, hospitalization for heart failure; HR, hazard ratio F I G U R E 3 SGLT2i and adverse events: hypoglycemia (a) diabetic ketoacidosis (b), amputation (c) and fracture (d). Fixed and Random effects model synthesize the odds ratio across all studies, and the prediction interval gives the 95% range for the result of a future SGLT2i trial. Side effects for canagliflozin were available either from the CANVAS trial or the integrated dataset of the CANVAS/CANVAS-R trials (CANVAS Program) as event rates; they were converted to events by multiplying the event rate and the sample size in each arm of the study. CKD, chronic kidney disease; CVOT, cardiovascular outcome trial; DKA, diabetic ketoacidosis; HFrEF, heart failure with reduced ejection fraction; HHF, hospitalization for heart failure; OR, odds ratio F I G U R E 4 SGLT2i and renal/infectious adverse events: volume depletion (a) acute kidney injury (b), genital mycotic infections (c) and urinary tract infection (d). Fixed and random effects model synthesize the odds ratio across all studies, and the prediction interval gives the 95% range for the result of a future SGLT2i trial. Side effects for canagliflozin were available either from the CANVAS trial or the integrated dataset of the CANVAS/CANVAS-R trials (CANVAS Program). AKI, acute kidney injury; CKD, chronic kidney disease; CVOT, cardiovascular outcome trial; DKA, diabetic ketoacidosis; HFrEF, heart failure with reduced ejection fraction; HHF, hospitalization for heart failure;OR, odds ratio; UTI, urinary tract infection kidney outcomes trial, CREDENCE. Out of these scientific publica-  Table 1). Our final dataset consisted of eight randomized controlled trials of 59 747 patients: 4 were CVOT safety trials, two targeted individuals with HFrEF, and two were kidney outcomes trials in patients with type 2 diabetes (Table 1).
On average these studies followed participants for a median of 1.5-4.2 years; most participants (>65%) were men and of middle age (average age 63.1-66.8). All participants had type 2 diabetes except for three trials that recruited individuals with HFrEF and the kidney outcome trial DAPA-CKD. Even though these trials recruited throughout the globe most participants were white (53.2% -87.8%). Inhibitors of the Renin Angiotensin System (RASi) were used by the majority of the study participants (80% in the trials not requiring such an inhibitor at baseline and > 98% in the kidney outcomes trials thar required participants to be on maximally tolerated RASi).
SGLT2i reduce cardiovascular (Figure 1(A)) and all-cause mortality ( Figure 1 Figure 3). None of these outcomes showed any evidence for heterogeneity, so the effects of SGLT2i were examined as a class, rather than by drug. Whereas SGLT2i reduced MACE-3 by 10% (p value of FE and RE were both .0002) and nonfatal MI by a similar amount (FE and RE p value were both .024), they had no effect on stroke. SGLT2i reduced worsening kidney function or ESKD by 39% (p < .0001). The beneficial effect on MACE-3/MI and kidney outcomes were observed irrespective of the study drug, and trial type. Examination of the prediction intervals, shows that the effects on MACE-3 and kidney function are very likely to be recapitulated in future SGLT2i trials, while some uncertainty exists about the replication of the effect on MI.
Whereas the rate of amputations appears to be increased in participants receiving SGLT2i, the magnitude of the effect appears to be small (OR 1.22), the statistical significance of the association dependent on the statistical model used (Figure 3 i.e. these appear consistent with class rather than drug specific effects. Similar to anti-hypertensive in-class differences, the challenge in gaining stronger evidence of in-class differences is the extremely large sample size that would be required for the expected small differences.
While the benefits of SGLT2i are large, clinically meaningful, and robust, they come with certain side effects that appear to be rather unpalatable e.g. mycotic infections, or serious, disfiguring, or even potential lethal e.g. amputations, fractures or DKA. Providers may thus hesitate to prescribe SGLT2i, and patients resistant to take them.
However, any apprehension about SGLT2is disappears once the frequency of these side effects is contrasted to the benefits. This is best understood via specific examples. In EMPEROR-Reduced, 24.7% versus 19.3% individuals experienced the primary outcome of HHF or cardiovascular death in the placebo and SGLT2i arms. On the other hand, the impressive OR of 3.9 for mycotic infections translates to an absolute increase in the probability of this event from 0.6% to 1.7%.
A patient or a provider may value one death/HHF event averted, to be almost five times (since [24.7-19.3 limitation of our meta-analysis is that it omitted smaller studies from the phase 3 programs that led to the introduction of SGLT2i as antiglycemics and from those SGLT2i not currently licensed in North America, for which outcomes studies are still lacking. We addressed this limitation by reporting prediction intervals, that is, plausible ranges of hazard ratios that may be observed in future trials. This statistical computation, which is not commonly reported in meta-analyses, provides a snapshot of outcomes for which some uncertainty remains (e.g., all cause mortality) and others (e.g., HHF or kidney outcomes) for which minimal residual uncertainty exists about future attempts to replicate these findings.
Given the favorable risk benefit ratio, how do we get SGLT2i to be used more widely? We hope that by synthesizing the data, our meta-analysis will inform the practice of the "four" key specialties (cardiologists, nephrologists, endocrinologists and primary care).
When using these drugs, a few commonsense rules should be employed to reduce risk of adverse events. Providers should consider temporarily discontinuing SGLT2i in clinical situations that predispose to ketoacidosis, for example, prolonged fasting, post-surgery or an acute illness, a strategy known as the "sick day rule." Peripheral arterial disease (PAD) is a major risk factor for atherosclerotic cardiovascular disease, and nearly 20%-25% of participants in these trials had PAD. Given the results of CANVAS, there has been concern about using SGLT2i in individuals with PAD. However, withholding SGLT2i from patients with PAD does not seem wise given the known benefits and the relative risk of amputations in other trials (including canagliflozin's CREDENCE). Such considerations led the FDA to remove the black box warning of amputations in August 2020. Nevertheless, it is reasonable to avoid SGLT2i in patients with active PAD (e.g. critical ischemia, non-healing ulcers as was done in CREDENCE) and discuss the potential for this complication in all other patients initiated on SGLT2i, while following them clinically for signs of incident or worsening PAD. Understanding the effects of these drugs on kidney function would also seem important; nearly all SGLT2i still carry warnings about AKI because of a major misunderstanding about the effects of these drugs on the eGFR during their launch as antiglycemics. All SGLT2i will cause an immediate and reversible 4-6 ml/min/1.73m 2 drop in eGFR after initiation 17,18 and thus kidney function should be checked within 3-4 weeks to establish the patient's new eGFR baseline. Additionally, the combined use of SGLT2i with RASi is not only nephroprotective as shown in CRE-DENCE, DAPA-CKD and subgroup analyses of CVOTs, but is associated with reduced rates of AKI, an effect seen in both the large trials and multiple observational cohorts. 28,29 Attention to volume status, may prevent the few AKIs due to volume depletion that are occasionally observed. Finally the incidence of genital infections may be reduced nearly 10 fold by simple personal hygiene advice that involves washing the urogenital area with water after each void and before going to bed. 30 Armed with the knowledge of benefits, risks, and some clinical sense about risk reduction, the four key specialties should no longer hesitate to discuss SGLT2i for cardiorenal protection with their patients.

| CONCLUSIONS
SGLT2i reduce all cause, cardiovascular mortality, heart failure hospitalizations, need for dialysis and acute kidney injury as a class effect across a broad range of populations with diabetes Type 2 at risk for cardiovascular disease, patients with HFrEF or CKD with or without diabetes. Overall, SGLT2is appear to have a favorable safety profile and thus should be used by the four "key specialties" (cardiology, nephrology, endocrinology, primary care) first and foremost as agents that reduce cardiorenal and heart failure risk and only secondarily as antiglycemics.

ACKNOWLEDGMENTS
Christos Argyropoulos would like to thank multiple individuals who reached out to him and convinced him to convert his two tweetorials about SGLT2i (https://twitter.com/ChristosArgyrop/ status/1301706984379482113?s=20 and https://twitter.com/ ChristosArgyrop/status/1301736105688014849?s=20 to a paper format and Professor Christopher Paul Cannon who invited me to submit to write up to Clinical Cardiology and the reviewers who suggested the editorial be converted to a formal meta-analysis.

CONFLICT OF INTEREST
Michael E Johansen has nothing to disclose. Christos Argyropoulos has received consulting fees from Bayer, Baxter Healthcare, Health Services Advisory Group and research support from Dialysis Clinic, Inc.

DATA AVAILABILITY STATEMENT
The data and software code that support the findings of this study are openly available at repository https://bitbucket.org/chrisarg/ sglt2imetanalysis.