Comparative cardiovascular benefits of individual SGLT2 inhibitors in type 2 diabetes and heart failure: a systematic review and network meta-analysis of randomized controlled trials

Background In patients with type 2 diabetes (T2D) and a history of heart failure (HF), sodium–glucose cotransporter-2 inhibitors (SGLT2is) have demonstrated cardiovascular (CV) benefits. However, the comparative efficacy of individual SGLT2is remains uncertain. This network meta-analysis (NMA) compared the efficacy and safety of five SGLT2is (canagliflozin, dapagliflozin, empagliflozin, ertugliflozin, and sotagliflozin) on CV outcomes in these patients. Materials and methods PubMed, Embase, and the Cochrane Central Register of Controlled Trials were searched up to September 23, 2022, to identify all randomized controlled trials (RCTs) comparing SGLT2is to placebo in T2D patients with HF. The main outcomes included composite CV death/heart failure hospitalization (HFH), HFH, CV death, all-cause mortality, and adverse events. Pairwise and NMA approaches were applied. Results Our analysis included 11 RCTs with a total of 20,438 patients with T2D and HF. All SGLT2is significantly reduced HFH compared to standard of care (SoC) alone. “Add-on” SGLT2is, except ertugliflozin, significantly reduced composite CV death/HFH relative to SoC alone. Moreover, canagliflozin had lower composite CV death/HFH compared to dapagliflozin. Based on the surface under the cumulative ranking curve (SUCRA), the top-ranked SGLT2is for reducing HFH were canagliflozin (95.5%), sotagliflozin (66.0%), and empagliflozin (57.2%). Head-to-head comparisons found no significant differences between individual SGLT2is in reducing CV death. “Add-on” SGLT2is reduced all-cause mortality compared with SoC alone, although only dapagliflozin was statistically significant. No SGLT2is were significantly associated with serious adverse events. A sensitivity analysis focusing on HF-specific trials found that dapagliflozin, empagliflozin, and sotagliflozin significantly reduced composite CV death/HFH, consistent with the main analysis. However, no significant differences were identified from their head-to-head comparisons in the NMA. The SUCRA indicated that sotagliflozin had the highest probability of reducing composite CV death/HFH (97.6%), followed by empagliflozin (58.4%) and dapagliflozin (44.0%). Conclusion SGLT2is significantly reduce the composite CV death/HFH outcome. Among them, canagliflozin may be considered the preferred treatment for patients with diabetes and a history of heart failure, but it may also be associated with an increased risk of any adverse events compared to other SGLT2is. However, a sensitivity analysis focusing on HF-specific trials identified sotagliflozin as the most likely agent to reduce CV death/HFH, followed by empagliflozin and dapagliflozin. Systematic review registration https://www.crd.york.ac.uk/prospero/, identifier CRD42022353754.


Introduction
Heart failure (HF) is a prevalent and debilitating complication of type 2 diabetes (T2D), contributing to increase morbidity and mortality in affected individuals.Worldwide, more than 26 million people are affected by this condition (1, 2), and T2D is a wellestablished risk factor with approximately 10%-30% of T2D patients aged over 70 years reported to have had HF (3).
Comorbid T2D with cardiovascular disease (CVD) is associated with higher mortality (4), highlighting the importance of reducing the risk of CVD in T2D management.A previous systematic review and meta-analysis (SRMA) indicated that intensive glucose lowering was not significantly associated with CVD risk reduction but conversely increased HF by 47% (5).Novel strategies are therefore necessary to improve prognosis and lower mortality in patients with T2D.
The associated costs of SGLT2is also limit their accessibility, especially in limited-resource settings.The cost for SGLT2i therapy in the United States ranged from $405.98 to $426.27/person/month, with an out-of-pocket cost of $36.76 to $56.64/person/month (22).However, given the variable individual medication pricing, an improved understanding of individual SGLT2i efficacy and safety will inform treatment decisions.Direct head-to-head comparisons of all SGLT2is are unlikely; a network meta-analysis (NMA) may provide indirect comparisons and a ranking of the efficacy and safety of individual SGLT2is.
Although several trials were conducted in patients with T2D, only ~10% had previously reported HF (19,23,24), in contrast to the ~50% of HF patients who had previously reported T2D (14,15,17,20,25).None of these studies were sufficiently powered to evaluate SGLT2i efficacy in HF-T2D patients.Previous SRMAs have evaluated the efficacy of SGLT2is in various patient groups (26)(27)(28)(29), although none have specifically targeted HF-T2D or used an NMA approach.Therefore, this NMA was conducted to compare the CV benefits and adverse events (AEs) associated with individual SGLT2is in HF-T2D patients.Specifically, we aimed to determine which SGLT2i provides the greatest efficacy in reducing cardiovascular events in this patient population.

Materials and methods
This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement (30) and was registered in PROSPERO (CRD42022353754).

Data sources, search strategy, and data extraction
Three electronic databases, PubMed, Embase, and the Cochrane Central Register of Controlled Trials, were searched from inception to August 15, 2022, with an update on September 23, 2022, without language restrictions.The search terms and strategies are provided in Supplementary Table S1.The titles and abstracts were reviewed by two independent reviewers (TK and PH), and disagreements were resolved with a third reviewer (PL).The inclusion criteria included i) randomized controlled trials (RCTs) or their subgroup or post-hoc analyses of SGLT2is in HF-T2D, ii) compared SGLT2is with standard of care (SoC), and iii) included any outcome of interest (i.e., composite CV death/HFH, HFH, CV death, and allcause mortality).
Two reviewers (TK and PH) independently extracted the data; disagreements were adjudicated by PL.The data extractions included i) study characteristics (i.e., study participants and number, study design, follow-up period, age, sex, baseline ejection fraction (EF), HF type (preserved/reduced EF), functional class (New York Heart Association (NYHA) Functional Classification), and other concomitant medications); (ii) interventions (SGLT2i type, dose, and duration); and (iii) outcomes (i.e., composite CV death/HFH, HFH, CV death, all-cause mortality, and AEs).
The primary outcome included composite CV death/HFH originally defined by individual RCTs.Secondary outcomes included HFH, CV death, all-cause mortality, and any AEs (i.e., volume depletion, acute kidney injury (AKI), urinary tract infection, genital tract infection, diabetic ketoacidosis (DKA), and bone fracture) in addition to serious AEs (SAEs).SAEs were defined as i) death or immediate life-threatening event, ii) persistent or clinically significant disability or incapacity, iii) events requiring hospitalization, iv) events related to a congenital anomaly or birth defect, or v) deemed serious for any other reason (14,31,32).

Risk of bias assessment
Two authors (TK and PK) independently assessed the risk of bias (RoB) using the Cochrane Risk of Bias tool version 2 (RoB2) based on five domains: randomization process, deviations from the intended protocol, missing outcome data, measurement of the outcomes, and selection of the reported result.Disagreements were adjudicated by PL.The overall quality was graded as high, with some concern and a low risk of bias (33).

Statistical analysis
Effect sizes (i.e., unstandardized mean difference (USMD) and risk ratio (RR)) along with 95% confidence intervals (CIs) were estimated for continuous data and dichotomous outcomes.Heterogeneity was assessed using the Q-test and I 2 statistics.If heterogeneity was present (Q test <0.1 or I 2 > 50% (34)), a metaanalysis (MA) random-effects model was used; otherwise, a fixedeffects model was considered.A meta-regression investigated the heterogeneity source by fitting each co-variable in the model including age, sex, baseline EF, HF type (reduced or preserved EF), functional class (NYHA), SoC with any HF treatment (i.e., MRA, renin-angiotensin system inhibitor [RASi], and angiotensin receptor/neprilysin inhibitor [ARNI]), treatment duration, and acute/chronic HF.
A two-stage NMA was applied as follows.First, a relative treatment effect (i.e., lnRR and USMD) was estimated with common variance-covariance.Second, treatment effects were pooled across studies using a multivariate MA with a consistency model.Transitivity was evaluated by exploring patient characteristics between comparisons or intervention arms, where appropriate.The inconsistency assumption was assessed using a global design-by-treatment interaction model, if applicable.Relative treatment effects were ranked using a rankogram and surface under the cumulative ranking curve (SUCRA).Publication bias was assessed using Egger's test and adjusted comparison funnel plots, which, if asymmetrical, were evaluated further using a contourenhanced funnel plot.
Subgroup/sensitivity analyses were pre-planned by HF type (preserved and reduced EF) and concomitant use of ARNI, MRA, and RAS blockade as SoC, if data were available.Furthermore, clustered-ranking plots were used to evaluate and rank risks and benefits associated with individual SGLT2is.
STATA version 17 (StataCorp, College Station, TX, USA) was used for all analyses.A significance threshold of p < 0.05 was considered, except for the heterogeneity and Egger's tests, where a p-value <0.10 was used.

Cardiovascular death
Of the nine studies (n = 14,349) that reported CV death as an outcome, an MA approach was applied across two (32,39) The network plot of the included studies.

Canagliflozin
Comparison should be read from right to left.In the upper rectangle, relative risk <1 favors the drug in the column.In the lower rectangle, relative risk <1 favors the drug in the row.*Statistical significance.
Clustered ranking suggested that canagliflozin and sotagliflozin offered the best efficacy in reducing HFH and composite CV death/ HFH, although both had higher risks associated with any AE.Empagliflozin was associated with a high probability of reducing HFH and the lowest probability of any AEs (Figure 4).

Transitivity assessment
Characteristics for each comparison, including age, sex, the strength of the medication, baseline EF, functional class, concurrent medications, and treatment duration, were explored across comparisons (see Supplementary Table S6).Age, percentage

Canagliflozin
Comparison should be read from right to left.In the upper rectangle, a relative risk of <1 favors the drug in the column.In the lower rectangle, a relative risk of <1 favors the drug in the row.*Statistical significance. of females, and EF varied between comparisons.In a network of the composite CV death/HFH outcome, patients in the canagliflozin-SoC comparison were likely to be younger and female compared to patients in other comparisons.In HFH, CV death, and all-cause mortality networks, patients in the canagliflozin-SoC comparison were more likely to be female and have a higher EF compared to other comparisons.Patients in the ertugliflozin-SoC comparison were more likely to be younger, while those in the sotagliflozin-SoC comparison were more likely to have had a shorter follow-up period for CV death and all-cause mortality networks.

Risk of bias assessment
The RoB was assessed by both reviewers (TK, PK) with 72.73% agreement (kappa 0.48, p = 0.03).Of the 11 RCTs, five were considered low risk, and six had some concerns; the major domain driving this tended to be the randomization process.Given that NMA focused on HF and T2D, studies that did not stratify analyses based on diabetes or HF were considered to have some concerns regarding randomization (Supplementary Figure S6).

Publication bias
Comparison-adjusted funnel plots indicated evidence of asymmetry associated with HFH, CV death, and all-cause mortality networks due to small-study effects from a single study (see Supplementary Figures S1B, S2C, S3C, S4C, S5D, S5H).However, this was due to a very small effect size.

Sensitivity analysis
Since the HF diagnostic criteria varied between HF-specific RCTs and post-hoc analyses of CV death/HFH death outcomes, we performed a sensitivity analysis that included only HF-specific trials at baseline.Consequently, only three SGLT2is, i.e., empagliflozin, dapagliflozin, and sotagliflozin, were retained within the analysis.The results showed that dapagliflozin, empagliflozin, and sotagliflozin significantly reduced composite CV death/HFH with corresponding RRs (95% CI) of 0.82 (0.74-0.92), 0.79 (0.71-0.89), and 0.70 (0.62-0.78); these RRs were comparable with the original analysis that included all HF-specific and T2D-specific trials with corresponding RRs of 0.80 (0.72-0.87), 0.79 (0.71-0.88), and 0.74 (0.68-0.81) (see Supplementary Figure S7).A NMA also identified add-on therapies in combination with these three SGLT2is to be significantly associated with reduced risk of composite CV death/ HFH compared to SoC.However, no significant differences in headto-head comparisons were identified (see Supplementary Table S7).Similar to our main analysis, SUCRA identified sotagliflozin with the highest probability of reducing composite CV death/HFH (97.6%), followed by empagliflozin (58.4%) and dapagliflozin (44.0%).
In addition, we performed a sensitivity analysis that excluded a single study with a small-study effect (sample size <100) and a treatment duration of less than 6 months (38) from the main analysis.However, our findings remained unchanged (see Supplementary Figure S8).
The Confidence in Network Meta-Analysis (CiNeMA) for each outcome is shown in Supplementary Table S8.The minimal clinically important differences for each outcome were set according to the Dutch guidelines committee T2D in primary care (42).CiNeMA indicated canagliflozin, sotagliflozin, and empagliflozin had very low confidence ratings for composite CV death/HFH.Within-study bias, reporting bias, and incoherence were the reasons for these downgrades.There significant concern with incoherence given the lack of a closed loop within the network framework.

Discussion
A NMA was conducted and revealed that when added to SoC, SGLT2is significantly reduce the composite outcomes of CV death/ HFH.Notably, canagliflozin was the most effective, followed by sotagliflozin, while dapagliflozin and empagliflozin exhibited comparable efficacy.The addition of SGLT2is beyond SoC reduced CV death by between 8% and 22%.Only dapagliflozin and canagliflozin were associated with lower all-cause mortality compared to SoC.Importantly, we did not find any statistically significant associations between SGLT2is and adverse side effects or SAEs.
Our findings indicate that SGLT2is reduce composite CV death/ HFH outcomes in patients with T2D and previously documented HF by approximately 20%.Although our study encompasses participants from both HF-specific trials and post-hoc analyses, our main findings and sensitivity analyses align with those previously reported in an SRMA that focused exclusively on HF-specific trials (43).Notably, the composite outcome of CV death/HFH was primarily influenced by HFH.In our analysis, canagliflozin and sotagliflozin ranked first and second, respectively, in reducing HFH, while they ranked first and third, respectively, in reducing CV death.This ranking is consistent with previous NMA findings (44), which support the notion that nonselective SGLT2is may offer greater advantages in treating HF compared to selective SGLT2is for reducing HFH (44).It is hypothesized that SGLT1 plays a pivotal role in glucose absorption in the intestines, and concurrent inhibition of SGLT1 and SGLT2 may further enhance renal sodium and glucose reabsorption.Furthermore, SGLT1 receptors are expressed in the human myocardium, and their upregulation has been observed in HF patients (45).However, the understanding of the role of SGLT1 cardiac expression and its interactions with SGLT2 in HF patients remains limited.
This study reveals that despite differences in chemical structure, pharmacokinetic and pharmacodynamic properties, as well as variations in SGLT1/SGLT2 receptor selectivity, all SGLT2is investigated in this study generally reduce the risk of HFH, consistent with previous SRMA results (8).We also observed little disparity in the efficacy of individual SGLT2is, with the exception of dapagliflozin, which exhibited a 36% higher rate of HFH compared to canagliflozin.As such, our findings support the beneficial effects of SGLT2is in reducing HFH as a class effect.Notably, the natriuretic and diuretic effects that lead to increased renal glucose excretion may have beneficial implications for endothelial progenitor cells, weight loss, improved myocardial energetics, adaptive cellular reprogramming, and reductions in both blood pressure and left ventricular hypertrophy (46)(47)(48).
Previous SRMAs have consistently reported a significant reduction of CV death in patients with T2D who were prescribed SGLT2is (HR 0.85, 95% CI 0.78-0.93,I 2 = 64.5%,p = 0.02) (8) as well as in patients with HF with or without T2D (HR 0.87, 95% CI 0.79-0.95,p = 0.94) (43).Our study specifically focused on patients with comorbid HF and T2D, and our findings align with the previously reported evidence.Furthermore, our NMA highlights that canagliflozin and dapagliflozin provide the greatest reduction in the risk of CV death, corroborating earlier research (44).Interestingly, we did not observe any significant differences in the ability to reduce CV death between selective and nonselective SGLT2is.
Our findings demonstrate that SGLT2is can reduce all-cause mortality in patients with HF-T2D by approximately 10%.However, only dapagliflozin reached statistical significance, possibly due to the inclusion of two large-scale placebo-controlled RCTs (DECLARE-TIMI and DAPA-HF).The robust reduction in all-cause mortality observed in our study was predominantly driven by the DAPA-HF trial, which revealed a remarkable 17% reduction in all-cause mortality in patients with HF-prescribed dapagliflozin, with or without T2D.In contrast, while the EMPA-REG RCT demonstrated a significant reduction in all-cause mortality, only 9.9% of the patients had a history of cardiac failure at baseline.Moreover, empagliflozin exhibited no survival benefits in the EMPEROR-reduced and EMPEROR-preserved RCTs.Similarly, the impact of canagliflozin, ertugliflozin, and sotagliflozin on mortality outcomes in patients with T2D and HF at baseline was found to be minimal in the CANVAS, VERTIS-CV, and SOLOIST-WHF RCTs, respectively.
The safety profile of SGLT2is is firmly established, encompassing known risks such as mycotic genital infections, urinary tract infections, diabetic ketoacidosis, volume depletion, kidney impairment (16,19,23), and the risk of amputation (21).Our findings, as corroborated by our NMA, confirm that SAEs were notably absent across all individual SGLT2is analyzed.However, our analysis did reveal an increased risk of any adverse event associated with canagliflozin.
Although the benefit of SGLT2is in reducing HFH appears to be a class effect, our findings highlight variations among individual SGLT2is in reducing CV and all-cause death, and safety profiles, which may be attributed to several factors.First, each SGLT2i exhibits distinct properties including their selectivity for SGLT1/SGLT2 inhibition, particularly within cardiomyocytes, which could influence CV and renal effects.Second, the differences in the characteristics of the study populations, concomitant medications, the duration of treatments, and follow-up time may introduce elements of heterogeneity, potentially confounding the observed outcomes.
Our NMA has several strengths: first, this is the first NMA to address uncertainties regarding the ranking of CV benefits provided by individual SGLT2is for HF-T2D patients.Second, our NMA includes a broader evidence base, incorporating more RCTs and a larger cohort of HF-T2D patients in comparison to the most recent SRMA (43) (20,438 vs. 9,739).Third, we have considered all available SGLT2is (5 SGLT2is vs. 3 SGLT2is) and have included additional CV outcome measures, including HFH, CV death, and all-cause mortality in HF-T2D patients.These efforts enable us to comprehensively rank the clinical efficacy and safety profile of individual SGLT2is across all of the CV outcomes of interest.
We also recognize several limitations in our study.First, we employed aggregated study-level data rather than individual patient data, which limited our ability to explore additional baseline factors that might potentially confound outcomes, including concomitant drug used, EF, and the etiology of HF (ischemic or non-ischemic heart disease).Second, our study outcomes may have been influenced by differences in patient study designs, and trial durations.For instance, the SOLOIST-WHF trial focused on T2D patients with more severe HF, enrolling participants either before or within 3 days of HFH, whereas other studies included T2D patients with chronic HF.Variances in the duration of participant follow-up were also observed with CANONICAL and SOLOIST-WHF, which monitored participants for less than 1 year, while other RCTs had longer follow-up periods.Third, our study encompassed both HF-specific and post-hoc analysis of CVOTs.We observed disparities in the diagnostic criteria for HF between HF-specific RCTs and post-hoc CVOTs.Specifically, all participants enrolled in HF-specific trials exhibited elevated brain natriuretic peptide or NT-proBNP levels, which are established HF diagnostic biomarkers, while diagnostic criteria in CVOTs were less strictly defined.Nevertheless, it is noteworthy that despite these potential confounding factors, the observed heterogeneity in our NMA remained low, and the results from a sensitivity analysis that focused solely on HFspecific trials were consistent with the findings of the overall analysis.Fourth, the efficacy of canagliflozin is primarily derived from the posthoc analysis of CANVAS studies, which did not specifically focus on heart failure at baseline.These results should be interpreted with caution.Fifth, many of the treatment comparisons in our NMA exhibited low confidence levels, as assessed using the six-domain CINeMA tool.These findings underscore the significance of taking into account the uncertainty associated with these comparisons when drawing conclusions from our study.

Conclusions
SGLT2is significantly reduce the composite CV death/HFH outcome.Among them, canagliflozin may be considered the preferred treatment for patients with diabetes and a history of heart failure, but it may also be associated with an increased risk of any adverse events compared to other SGLT2is.However, a sensitivity analysis focusing on HF-specific trials identified sotagliflozin as the most likely agent to reduce CV death/HFH, followed by empagliflozin and dapagliflozin.

FIGURE 1 PRISMA
FIGURE 1 PRISMA flow diagram.PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-Analyses.

FIGURE 2
FIGURE 2Composite cardiovascular death or heart failure hospitalization in type 2 diabetes with heart failure patients receiving SGLT2 inhibitors versus standard of care.SGLT2, sodium-glucose cotransporter-2.
(A) Composite cardiovascular (CV) death or heart failure hospitalization (HFH).(B) Heart failure hospitalization (HFH).(C) Cardiovascular death (CV death).(D) All-cause mortality.The size of the nodes indicates the total sample size of the associated intervention (blue circles).The thickness of each line represents a direct comparison between two therapies and corresponds to the number of trials that examined each comparison.

4
FIGURE 4Clustered ranking plot of standard of care (SoC), canagliflozin, dapagliflozin, empagliflozin, and sotagliflozin showing the surface under the cumulative ranking curves (SUCRAs) for the highest probability of any adverse events (AEs) versus the SUCRAs for the highest probability of improving cardiovascular outcomes.(A) Composite cardiovascular (CV) death or heart failure hospitalization.(B) Heart failure hospitalization.(C) CV death.(D) All-cause mortality.Intervention lying in the upper right corners are associated with the higher-of probability of treatment efficacy and "higher" probability of AEs.

TABLE 1
Characteristics of studies included in the quantitative analysis.

TABLE 2
Relative treatment effect comparison (95% CI) for composite cardiovascular death or heart failure hospitalization (HFH) (upper triangle) and HFH (lower triangle).

TABLE 3
Relative treatment effect comparisons (95% CI) for cardiovascular death (upper triangle) and all-cause mortality (lower triangle).