Empagliflozin in Patients With Heart Failure: A Systematic Review and Meta-Analysis of Randomized Controlled Trials

Purpose: The purpose of the study is to evaluate the effect of empagliflozin in patients with heart failure (HF). Method: We performed a systematic search of PubMed, EMBASE, and the Cochrane Library database through January 20, 2021. Randomized controlled trials (RCTs) were included that compared empagliflozin and placebo in patients with HF. Dichotomous variables were expressed as risk ratios (RRs) with 95% confidence intervals (CIs). Continuous variables were calculated and expressed as mean differences (MD) and standard deviation (SD). Meta-analysis was conducted using a random-effects model on outcomes with high heterogeneity. Results: Seven studies were included in our meta-analysis (n = 5,150). Significant differences were observed in a composite of cardiovascular death or hospitalization for worsening heart failure [RR: 0.77 (95% CI 0.68–0.87); I2 = 18%; P < 0.0001), hospitalization for worsening heart failure [RR: 0.71 (95% CI 0.61–0.82); I2 = 0%; P < 0.00001], changes in Kansas City Cardiomyopathy Questionnaire (KCCQ) score [MD: 1.70 (95% CI 1.67–1.73); I2 = 0%; P < 0.00001], and changes in body weight [MD: −1.43 (95% CI −2.15 to −0.72); I2 = 84%; P < 0.0001) from baseline. However, empagliflozin did not show a better change in the 6-min walk test (6MWT) [MD: 34.06 (95% CI −29.75–97.88); I2 = 97%; P = 0.30] or NT-proBNP [MD: −98.36 (95% CI, −225.83–29.11); I2 = 68%; P = 0.13] from baseline. Conclusion: The findings suggest that empagliflozin was effective in reducing a composite of cardiovascular death or hospitalization for worsening heart failure. Further well-designed RCTs are needed to evaluate the long-term effect of empagliflozin in patients with HF. PROSPERO: CRD42021231712.


INTRODUCTION
Heart failure (HF) leads to a high economic burden, and the prevalence of HF is also increasing gradually (1,2). It has been estimated that over 26 million people around the world are suffering from heart failure (3). According to the guidelines (4) for HF patients, it is recommended to apply angiotensinconverting enzyme inhibitors, beta-blockers, mineralocorticoids, or aldosterone receptor antagonists. Currently, new drugs such as sodium-glucose cotransporter 2 inhibitors (SGLT2i) and vericiguat also show positive effects across the spectrum of HF (5).
SGLT2i has been used to treat type 2 diabetes mellitus (6,7). In recent years, SGLT2i has shown a better effect than placebo on the outcomes of all-cause mortality and heart failure hospitalization in patients with type 2 diabetes mellitus in clinical trials, regardless of the presence or absence of heart failure (8,9). In patients with diabetes mellitus, SGLT2i reduced the risk of all-cause mortality and heart failure hospitalizations by 23% (10). Empagliflozin, one of the frequently used SGLT2i, has been proven to be effective in reducing HF hospitalizations, cardiovascular deaths, and biomarkers in patients with HF (11). However, the results in related articles showed heterogeneity in some outcomes, for example, N-terminal pro-brain natriuretic peptide (NT-proBNP) (12,13) and the 6-min walk test (6MWT) (14,15). The effect of empagliflozin in patients with HF has not been evaluated specifically. Previous studies mostly concentrated on SGLT2i rather than empagliflozin. According to previous large-sample trials, empagliflozin also showed different results of a composite cardiovascular endpoint (cardiovascular deaths, non-fatal myocardial infarction, or non-fatal stroke) when compared with dapagliflozin, which means that different drugs may present different results, although all of them belong to SGLT2i (9,16). Thus, we aim to perform a systematic review and meta-analysis of randomized controlled trials (RCTs) to evaluate the effect of empagliflozin on patients with HF.

METHODS
Our systematic review and meta-analysis were conducted based on preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Ethical approval was not required. The study protocol was registered in the PROSPERO international prospective register of systematic review (CRD42021231712).

Search Strategy
Our systematic review and meta-analysis were conducted according to the PRISMA guidelines (17). We performed a systematic search from the following electronic databases: PubMed, EMBASE, and Cochrane Library, from their inception to January 20, 2021. The search terms were as follows: [1-chloro-4-(glucopyranos-1-yl)-2-(4-(tetrahydrofuran-3-yloxy)benzyl] benzene OR BI 10773 OR BI10773 OR BI-10773 OR Jardiance) AND (Cardiac Failure OR Heart Decompensation OR Decompensation, Heart OR Heart Failure, Right-Sided OR Heart Failure, Right Sided OR Right-Sided Heart Failure OR Right Sided Heart Failure OR Myocardial Failure OR Congestive Heart Failure OR Heart Failure, Congestive OR Heart Failure, Left-Sided OR Heart Failure, Left Sided OR Left-Sided Heart Failure OR Left Sided Heart Failure). No restrictions on language, publication date, or publication status were set in our review. We did not include conference abstracts. In addition, we reviewed the reference lists of eligible studies, previous review articles, and registered clinical trials.

Study Selection and Eligibility Data
All searched articles were imported into EndNote software, and the title and abstract were screened by two reviewers (DP and PC). We entirely included clinical trials that met the following criteria for final analysis: (1) RCTs, (2) the target population was patients with HF, and (3) studies that included the comparison between empagliflozin and placebo.
The exclusion criteria were (1) observation, cohort, case control, case series, qualitative studies, uncontrolled trials, and laboratory studies and (2) duplicate studies with the same population (only the study with the largest participants was included in the meta-analysis if multiple studies included overlapping groups of patients).
If disagreements on study selection were identified, another author (MG) was consulted to solve them.

Data Extraction
Two investigators (DP and PC) extracted the included trial data by using a predesigned form independently. The retrieved study characteristics were as follows: (1) first author's name and year of study; (2) study site; (3) intervention, dose of intervention, and comparison; (4) total sample size, and sample size of intervention and control group; (5) primary outcome; and (6) other outcomes.

Outcomes
The primary outcome was determined as a composite of cardiovascular deaths and hospitalization for worsening heart failure. Other outcomes included hospitalization for worsening heart failure, changes in the NT-proBNP, 6WMT, or KCCQ score, and changes in body weight before and after the intervention. The endpoint definitions were those used in the individual trials.

Risk of Bias Assessment
Assessment of the risk of bias for all of the included studies was performed independently by two review authors (DP and PC) through the Cochrane Risk of bias assessment tool. The assessment was performed across the following domains: selection bias (random sequence generation and allocation concealment), performance bias (blinding of patients and investigators), detection bias (blinding of outcome assessors), attrition bias (flawed outcome data), and reporting bias (selective reporting). According to the influence on material biases, the risk of bias was then adjudicated, with low, high, or unclear levels. To evaluate the certainty of the evidence for each outcome, we applied the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Framework. All disagreements were resolved by the other two reviewers (MG and DS).

Data Synthesis
We pooled the study results in the meta-analysis using Review Manager software (version 5.4.1; Nordic Cochrane Center, Cochrane Collaboration).
For dichotomous variables, risk ratios (RRs) with 95% confidence intervals (CIs) were calculated. Continuous variables were calculated and expressed in terms of mean differences (MD) and standard deviation (SD). For the articles that expressed continuous outcomes by median and interquartile range (IQR), the mean value and SD value were by the method of Luo et al. and Wan et al. (18,19). For the articles that reported only the results before and after the intervention, the corresponding change from baseline was calculated according to the Cochrane Handbook for Systematic Reviews of Interventions. In the articles that reported mean and SD before intervention (M baseline , SD baseline ), mean and SD after intervention (M final , SD final ), and mean and SD of the change before and after intervention (M change , SD change ), the correlation coefficient (Corr) was calculated as SD 2 baseline + SD 2 final -SD 2 change /2 * SD baseline * SD final . For the articles that did not report the values of M change and SD change , M change and SD change were calculated by the following equations: baseline + SD 2 final -(2 * Corr * SD final * SD baseline )] 1/2 . We assessed heterogeneity FIGURE 1 | Flow chart of study selection and identification. by using the X 2 test and I 2 statistic. Significant heterogeneity was defined as I 2 > 50% with a P < 0.10. We used the random-effects model to estimate the pooled effect size from the included data if significant heterogeneity was observed across the trials, and the fixed-effects model was used otherwise. By conducting sensitivity and subgroup analysis, we probed sources of heterogeneity. Sensitivity analyses were conducted by one-by-one elimination. Publication bias was assessed using funnel plots if the included trial number was reasonable (10 or more). The quality of evidence for the main outcomes was assessed by the GRADE (20).

Study Selection
We conducted the last search on January 20, 2021, and the results of our literature search are shown in Figure 1. A total of 454 publications were identified. After the removal of duplications and screening, 16 studies were selected for eligibility by full-text screening, and seven trials were included in our systematic review and meta-analysis (11-15, 21, 22). Table 1 reports the characteristics of the included studies. Of the selected seven trials, 5,150 participants were included in total, 2,682 of them were in the empagliflozin group, and 2,468 of them were in the placebo group. Among them, six studies (11)(12)(13)(14)(15)22) were two-arm RCTs. The last was a three-arm RCT (21) comparing two different doses (10 and 25 mg) of empagliflozin with placebo, and it was also the only substudy of a previous trial (9). In the selected trials, Packer's RCT (11) was the study with the largest sample size (n = 3,730). The age ranged from 59 to 70 years old, and the proportion of males was higher than that of females in all of the trials, ranging from 63% to 87%. Except for the three-arm RCT (21) and Mordi et al. (13), the dose of empagliflozin in the remaining trials was 10 mg once daily. The median follow-up time ranged from 6 weeks to 3.1 years.

Assessment of Risk of Bias
All seven included studies reported the generation of random sequences, and all of them also provided generation methods. However, two studies (12,14) did not report the method of allocation concealment, so their risk level remained unclear. One study (22) did not report the blinding of both treatment providers and participants, while the other studies adopted a double-blind design. Two studies (11,12) reported blinding of the outcome assessment. All of the studies were evaluated as having a low risk of attrition bias with no loss to follow-up or low and balanced loss to follow-up. There was no selective bias in any of the seven studies. We also detected no other risk in the seven included studies.

Primary Outcome
Three studies (11,12,21) reported the result of a composite of cardiovascular death or hospitalization for worsening heart failure. Empagliflozin significantly reduced the risk of a composite of cardiovascular death or hospitalization for worsening heart failure compared with the placebo group [RR: 0.77 (95% CI 0.68-0.87]; I 2 = 18%; P < 0.0001; Figure 2A]. We also conducted a meta-analysis by a random-effects model, and the result was consistent [RR: 0.74 (95% CI 0.56-0.96); I 2 = 18%; P = 0.003; Figure 2B].

Six-Minute Walk Test
The outcome of the 6MWT was reported in three studies (14,15,22), and we did not observe a significant difference between the two groups with regard to the distance change in the 6MWT before and after intervention [mean difference (MD): 34.06 (95% CI, −29.75-97.88); I 2 = 97%; P = 0.30; Figure 4].

Sensitivity Analysis
We performed sensitivity analysis to evaluate the influence of any individual study on the overall effect. For NT-proBNP,     Table 2 shows the GRADE assessment of the certainty of the effect of empagliflozin in patients with HF. Because the participants in one study were non-diabetic patients (15), whereas another study focused on patients with type 2 diabetes mellitus (13), the certainty level of body weight was downgraded to moderate. The certainty level of the 6MWT was downgraded Pan et al.    to moderate for one article focused on non-diabetic patients (15), and another focused on patients with type 2 diabetes mellitus or prediabetes (22). The other outcomes, a composite of cardiovascular death or hospitalization for worsening heart failure, hospitalization for worsening heart failure, and changes in NT-proBNP and the KCCQ score, were evaluated as high certainty, with no evidence for downgrading.

Summary of the Evidence
In this systematic review and meta-analysis, we included seven RCTs with 5,150 participants in total, and with regard to the primary outcome of a composite of cardiovascular death or hospitalization for worsening heart failure, the empagliflozin group showed significant superiority to the placebo group. In terms of hospitalization for worsening heart failure, the change in KCCQ score, and the change in body weight, significant differences were observed between the two groups and favored the empagliflozin group. For the results of the changes in 6MWT and NT-proBNP, we did not observe notable differences between the two groups except when studies that caused high heterogeneity were removed, when 6MWT favored the placebo, and NT-proBNP favored empagliflozin.

Results in Relation to Other Studies
To our knowledge, this is the first systematic review and metaanalysis that concentrated on the effect of empagliflozini in patients with HF. Instead of assuming the parameter Corr to be 0.5, we calculated Corr, which made the results more convincing. We also included four new high-quality RCTs in our study and pooled two new outcomes (6MWT and body weight). Hence, our results provided more robust and comprehensive evidence for evaluating the effect of empagliflozin on HF. We observed significant improvements in heart failure hospitalizations and cardiovascular deaths, which is in line with a previous study (32) focused on the effect of SGLT2i in heart failure patients. However, regarding the two outcomes of a composite of cardiovascular death or hospitalization for worsening heart failure and hospitalization for worsening heart failure, our study showed greater RR reduction than the previous meta-analysis (32), which may indicate that empagliflozin is superior in patients with reduced ejection fraction. It is also worth noting that the results of our study and a previous trial with dapagliflozin intervention (33) demonstrated no significant difference in the change in NT-proBNP. However, some trials showed that SGLT2i, especially empagliflozin or dapagliflozin, were associated with a reduction in plasma NT-proBNP levels (11,22,30), which is consistent with the result of the HFrEF subgroup in our meta-analysis. Therefore, the results on NT-proBNP were controversial. However, with pooled results including new high-quality trials, we believe that our results are more convincing.

Potential Mechanism
SGLT2i localize at the brush border of the early proximal tubule and function by reabsorbing nearly all of the filtered glucose (23,24). Because the excretion of glucose was promoted, the blood glucose level was decreased. With lower blood glucose, lower heart failure mortality was observed in a cohort trial (25). However, the outcome cannot be simply explained by lower blood glucose, since other antidiabetic drugs with stronger effects did not show the same cardiovascular protective results (26). In addition to the aforementioned mechanism, SGLT2i prevents the reabsorption of glucose in the proximal tubule, further causing secondary osmotic effects and then leading to natriuretic and diuretic effects (23). This mechanism could account for the reduction in body weight, which was found in our study. Moreover, it has been proven that empagliflozin significantly improved left ventricular diastolic function and reduced mortality in mice, probably by reducing spontaneous diastolic sarcoplasmic reticulum calcium release. Leakage was thought to be the mechanism of diastolic dysfunction (27). In addition, empagliflozin also promoted cardiac function in non-diabetic rats, which may be associated with improved cardiac metabolism and cardiac ATP production (28). Autophagy plays a crucial role in relieving cellular stress caused by different metabolites, such as glucose and lipids (29,34), and clearing accumulated metabolites and dysfunctional organelles, thus, preventing cells from dysfunction or death. In heart failure patients, the autophagy capacity of cardiomyocytes is markedly impaired (35). Autophagic vacuoles, which reflect the extent of autophagy, were found to be associated with a better prognosis of heart failure (36). SGLT2i promotes urinary caloric loss, which mimics nutrient deprivation, in which autophagy is activated, leading to cellular survival, and homeostasis instead of growth (37). In non-diabetic mice, SGLT2i upregulated SIRT1 and AMPK, which promote autophagy, and downregulated the Akt-mTOR pathway, which has the opposite function (35,37). Through the mechanisms mentioned above, autophagy was promoted, thereby protecting cardiomyocytes from dysfunction and death.

LIMITATIONS
There are some limitations that should be noted. First, the doses and durations of follow-up in the included trials were not uniform. Consequently, we conducted sensitivity analysis to evaluate heterogeneity. The results of the 6MWT and NT-proBNP changed after sensitivity analysis, indicating that the two results should be interpreted with great caution. Second, the forms of reported results varied across trials, and the transformation to a consistent form may have introduced some inaccuracy. Nonetheless, we transformed the results according to relevant articles (18,19), which minimized the error as much as possible. Third, we obtained trial-level data instead of individual-level data; thus, we could not conduct subgroup analysis based on type 2 diabetes mellitus, different drug therapies, New York Heart Association (NYHA) functional class, estimated glomerular filtration rate (eGFR), and so on.

CONCLUSION
In conclusion, compared with placebo, empagliflozin significantly reduced a composite of cardiovascular deaths or hospitalizations for worsening heart failure but showed no statistically significant change in NT-proBNP. However, in HFrEF patients, empagliflozin showed a significant reduction in NT-proBNP. More highquality, large-scale RCTs are needed to comprehensively evaluate the long-term effect of empagliflozin in patients with HF.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author/s.