|Home | About | Journals | Submit | Contact Us | Français|
Based on a single placebo-controlled randomized clinical trial, empagliflozin is licensed to reduce cardiovascular death in diabetes and comorbid cardiovascular disease.
We examined the comparative effectiveness of empagliflozin on mortality and cardiovascular morbidity in type 2 diabetes. We conducted random-effects direct frequentist meta-analyses of aggregate data and appraised the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology. Our search in PubMed, EMBASE, the Cochrane Library, clinicaltrials.gov, and PharmaPendium up to May 2017 identified 11 meta-analyses, multiple publications, and unpublished data from 29 randomized controlled trials (RCTs).
Empagliflozin reduces all-cause mortality [relative risk (RR) of death, 0.69; 95% confidence interval (CI): 0.58–0.82; number needed to treat (NNT) to postpone mortality in one patient, 39; 95% CI: 26–79; 1 RCT of 7,020 patients) in patients with but not without (RR, 0.90; 95% CI: 0.36–2.23; 14 RCTs of 7,707 patients) established cardiovascular disease when compared with placebo. Empagliflozin reduces cardiovascular mortality (RR, 0.62; 95% CI: 0.50–0.78; NNT, 45; 95% CI: 30–90; 1 RCT of 7,020 patients) in patients with but not without (RR, 0.98; 95% CI: 0.29–3.33; 10 RCTs of 5,429 patients) established cardiovascular disease when compared with placebo. There are no differences in cardiovascular morbidity and mortality and all-cause mortality between empagliflozin and metformin (4 RCTs of 1,344 patients), glimepiride (1 RCT of 1,549 patients), linagliptin (2 RCTs of 1,348 patients), or sitagliptin (3 RCTs of 1,483 patients). Two network meta-analyses concluded that sodium-glucose cotransporter 2 (SGLT2) inhibitors, mostly due to empagliflozin, decrease all-cause and cardiovascular mortality but increase the risk of nonfatal stroke, genital infection, and volume depletion.
We conclude that empagliflozin reduces all-cause and cardiovascular mortality in patients with established cardiovascular disease and type 2 diabetes. Sparse direct evidence suggests no difference in mortality between empagliflozin and metformin, glimepiride, linagliptin, or sitagliptin. Long-term comparative safety needs to be established.
One of the main goals in managing type 2 diabetes in adults is prevention of cardiovascular morbidity and mortality (1,2). Only a few of the available diabetes medications have shown benefits in reducing cardiovascular risks; most available drug classes such as thiazolidinediones, sulfonylureas, dipeptidyl peptidase-4 (DPP-4) inhibitors, and sodium-glucose cotransporter 2 (SGLT2) inhibitors have been approved based on their ability to decrease glycosylated hemoglobin A1c (HbA1c) rather than their ability to prevent morbidity and mortality (3,4). Network meta-analyses of SGLT2 inhibitors suggest lower risk of all-cause and cardiovascular mortality from 3 oral SGLT2 inhibitors combined, at the expense of higher risk of nonfatal stroke [pooled relative risk (RR) 1.30; 95% confidence interval (CI): 1–1.68], genital infection (pooled RR 4.75; 95% CI: 4.00–5.63), and volume depletion (pooled RR 1.53; 95% CI: 1.27–1.83) (5,6). However, the reduction in the risk of mortality and morbidity is mostly attributable to one drug, empagliflozin, the only drug approved by the FDA in 2016 to reduce the risk of cardiovascular death in adult patients with type 2 diabetes mellitus and established cardiovascular disease (5,6).
Older meta-analyses focused on intermediate outcomes of empagliflozin when compared with placebo; e.g., HbA1c, blood pressure, and body weight (7-15). The most recent high-quality meta-analyses included 13 (6) and 16 (5) randomized controlled trials (RCTs) that reported mortality and morbidity in adults with type 2 diabetes treated with empagliflozin but did not examine the comparative effectiveness of empagliflozin and other specific antidiabetic drugs (5,6). Clinicians have to select specific drugs for individual patients rather than relying on drug class benefits and harms.
To support clinical decisions at point of care with all available evidence, we conducted a rapid review of the published and unpublished data from the recently completed RCTs, meta-analyses of RCTs, and primary observational studies that compared the effects of empagliflozin with those of other antidiabetic drugs on all-cause and cardiovascular mortality and morbidity.
We used a standard recommended methodology in conducting systematic literature reviews and meta-analyses from the Cochrane Collaboration and the Agency for Healthcare Research and Quality (16,17). We developed a priori protocol for a systematic literature review to answer the clinical question about the efficacy and comparative effectiveness of empagliflozin and other antidiabetic medications against mortality and cardiovascular morbidity in adults with type 2 diabetes.
We defined the target population as adults with type 2 diabetes. Eligible interventions included SGLT2 inhibitor empagliflozin when compared with placebo or other antidiabetic medications. Eligible outcomes included all-cause and underlying cause-specific mortality, myocardial infarction (MI), stroke, incidence or progression of heart failure, and hospitalizations for major cardiovascular events. Intermediate outcomes included diabetes control as HbA1c <7% or as defined in the primary studies. We reviewed the frequency and severity of hypoglycemia as well as any harms from examined treatments.
We conducted a comprehensive search in PubMed, EMBASE, the Cochrane Library, www.clinicaltrials.gov and PharmaPendium (www.pharmapendium.com) up to May 2017 to find systematic reviews, published and unpublished RCTs, and nationally representative controlled observational studies that reported adjusted effect estimates (16,17). All of the authors determined the studies’ eligibility. All citations found during the searches are stored in a reference database.
The data was extracted from the Clinical Trials Transformation Initiative (CTTI) (https://www.ctti-clinicaltrials.org/aact-database), checked for quality, and stored in the HPCC platform (High-Performance Computing Cluster, https://hpccsystems.com/).
We performed direct frequentist meta-analyses of aggregate data when definitions of the active and control intervention and patient outcomes were deemed similar for pooling (18). We used random effects models to address inevitable differences in patient characteristics across primary RCTs. For each abstracted hypothesis, we calculated absolute risk difference and RR with 95% CI. We calculated number needed to treat (NNT) and number of attributable events per 1,000 treated with 95% CI based on statistically significant differences in absolute risks of the outcomes. We examined consistency in results across studies with chi-square tests and I2 statistics and concluded statistically significant heterogeneity if I2 was >50% (16). Statistically significant heterogeneity did not preclude statistical pooling (18). However, we planned exploring heterogeneity with a priori defined patient characteristics, drug doses, and study quality if this information was available in the studies (18).
We used consensus method guidelines for systematic review and meta-analyses that do not recommend conducting post hoc analyses of statistical power (19-22). Instead, we downgraded our confidence in true treatment effects based on calculated optimal information size as the number of patients required for an adequately powered individual trial (23). Since power is more closely related to number of events than to sample size, we concluded imprecision in treatment effects if fewer than 250 patients experienced the event (23).
We used Statistics/Data Analysis, STATA software (StataCorp LP, College Station, Texas). Statistical significance was evaluated at a 95% confidence level.
We evaluated the quality of systematic reviews using the Assessment of Multiple Systematic Reviews (AMSTAR) (24). For primary RCTs, we used the Cochrane risk of bias tool on a 3-point scale: high bias, low bias, and unclear (25,26). A low risk of bias was assumed when RCTs met all the risk-of-bias criteria, a medium risk of bias if at least 1 of the risk-of-bias criteria was not met, and a high risk of bias if 2 or more risk-of-bias criteria were not met. An unknown risk of bias was assigned for the studies with poorly reported risk-of-bias criteria. We assigned high risk of bias to all observational studies.
The authors assigned the quality of evidence ratings as high, moderate, low, or very low, according to risk of bias in the body of evidence, directness of comparisons, precision and consistency in treatment effects, and the evidence of reporting bias, using Grading of Recommendations Assessment, Development and Evaluation (GRADE) methodology (27).
A high quality of evidence was assigned to well-designed RCTs with consistent findings. The quality of evidence was downgraded to moderate if at least 1 of 4 quality of evidence criteria was not met; for example, moderate quality of evidence was assigned if there was a high risk of bias in the body of evidence or if the results were not consistent or precise. The quality of evidence was downgraded to low if 2 or more criteria were not met. We concluded a high risk of bias in the body of evidence if at least one RCT had high risk of bias. We downgraded the quality of evidence when we suspected high risk of publication bias due to unavailability of the results in clinicaltrials.gov or journal articles.
A low quality of evidence was assigned to nonrandomized studies, but the rating was upgraded if there was a strong or dose-response association (28). Evidence was defined as insufficient when no studies provided valid information about treatment effects. This approach was applied regardless of whether the results were statistically significant.
Our comprehensive search in PubMed, EMBASE, the Cochrane Library, and clinicaltrials.gov up to May 2017 identified 11 meta-analyses, multiple publications as well as unpublished data from 29 RCTs, and one non-randomized study that examined the benefits and harms of empagliflozin in people with type 2 diabetes (5-14,29). We also identified 2 high-quality meta-analyses and multiple publications as well as unpublished data from 9 RCTs that directly compared empagliflozin with other antidiabetic drugs in people with type 2 diabetes (5-7,30-48).
Primary studies enrolled adults with type 2 diabetes and various baseline degrees of cardiovascular risk, permitted administration of metformin and other antidiabetic drugs, and aimed mostly at diabetes control and drug safety. Only one large non-inferiority trial, the EMPA-REG OUTCOME trial, was designed to examine difference in a composite outcome defined as the first occurrence of cardiovascular death, nonfatal MI, or nonfatal stroke in adults with type 2 diabetes and high cardiovascular risk (49).
Moderate-quality evidence suggests that empagliflozin reduces all-cause mortality (7,11,40,44,46,47,49-61) and increases the rates of diabetes control without increasing the risk of serious adverse effects and hypoglycemia when compared with placebo in adults with type 2 diabetes (Table 1) (50,51,70). The increase in rates of glycemic improvement starts at the dose of 10 mg/day (150 attributable events per 1,000 treated, Figure 1) and increases to 210 attributable events per 1,000 treated after the larger dose of empagliflozin (25 mg/day, Figure 1).
Low-quality evidence suggests that empagliflozin reduces cardiovascular mortality, the risk of hospitalization for any cause (73), and hospitalizations for heart failure (59), as well as the risk of developing heart failure (73), developing or worsening of nephropathy (58), and the risk of treatment discontinuation due to lack of efficacy, at the expense of higher risk of adverse effects (Table 1) (7,11,40,44,46,47,49-61,70,75-77). The observed improvement in patient outcomes is attributable to the largest RCT, EMPA-REG OUTCOME, which enrolled patients with established cardiovascular diseases (59). Sensitivity analyses excluding this RCT demonstrate no protective effects from empagliflozin against all-cause and cardiovascular mortality in all other RCTs combined (Table 1). There are no differences in the risk of stroke or coronary events between empagliflozin and placebo (Table 1) (40,44,49-59,61,70).
Subgroup analysis of the EMPA-REG OUTCOME trial demonstrates that empagliflozin is better than placebo in reducing the risk of major cardiovascular events in older patients and adults with HbA1c 7.0–8.5% (P value for interaction <0.05, Table S1). Baseline heart failure does not modify empagliflozin effects on cardiovascular mortality and the risk of major cardiovascular effects (data not shown) (73). Empagliflozin is not better than placebo in adults with BMI ≥30 kg/m2 (P value for interaction 0.06, Table S1).
Safety analyses demonstrate that empagliflozin increases the risk of genital infection, thirst, and polyuria and reduces the risk of acute renal injury and failure, hypertension, and worsening of heart failure (Table S2) (7,11,31,40,50,51,53-56,58,59,62-67,77,78,81-83). Empagliflozin’s safety profile is similar in patients with normal and impaired baseline renal function (Table S2).
Post-marketing surveillance suggests more than 1900 case reports of various adverse effects, including fungal and urinary tract infections, diabetic ketoacidosis, unintentional weight loss, pollakiuria, dizziness, dehydration, nausea, and vomiting reported by patients taking empagliflozin among other drugs for type 2 diabetes (Table S3). In addition, the European Medicines Agency recently requested that information on potential risk of toe amputation be included in prescribing information for all SGLT2 inhibitors (84).
Low-quality evidence suggests that there are no differences in mortality, morbidity, diabetes control, and serious adverse effects between empagliflozin and metformin (Table 2) (5,31,33,34,37,39,43,46-48,81). Empagliflozin reduces the risk of total non-serious adverse effects when compared with metformin, with 51 avoided adverse events per 1,000 treated (Table 2) (5,31,33,34,37,39,43,46-48,81).
Very low-quality evidence from a single RCT suggests that empagliflozin decreases HbA1c, the risk of hypoglycemia, and total non-serious adverse effects when compared with glimepiride, at the expense of higher cumulative risk of total combined serious adverse effects (Table 3) (36,42,45). There are no differences in any specific serious harms between empagliflozin and glimepiride (data not shown).
Low-quality evidence suggests that there are no differences in mortality, morbidity, and serious adverse effects between empagliflozin and linagliptin (Table 4) (5,34,35,46,47). Empagliflozin decreases HbA1c and the risk of total non-serious adverse effects (Table 4) (5,34,35,46,47). Low-quality evidence suggests that there are no differences in mortality, morbidity, and total adverse effects between empagliflozin and sitagliptin (Table 5) (5,32,33,38,40,43,44).
Specific adverse effects differ among examined drugs, according to the labeling information (Table S4). These differences should be taken into account when selecting specific drugs for patients with higher baseline risk of specific harms. Combined drug formulations would likely result in a cumulative increase in the risk of specific adverse effects.
Post-marketing surveillance suggests that lactic acidosis (6,754 cases), diarrhea (3,774 cases), and acute renal failure (3,754 cases) are the most common adverse effects reported in patients taking metformin among other drugs for type 2 diabetes (Table S3). Hypoglycemia (675 cases), hypoglycemic coma (142 cases), and acute renal failure (126 cases) are the most common adverse effects reported in patients taking glimepiride among other drugs for type 2 diabetes (Table S3). Pancreatitis (328 cases), nausea (216 cases), and rash (178 cases) are the most common adverse effects reported in patients taking linagliptin among other drugs for type 2 diabetes (Table S3). Pancreatitis (2,459 cases), pancreatic carcinoma (1,604 cases), diarrhea (1,175 cases), nausea (1,175 cases), and hypoglycemia (1,163 cases) are the most common adverse effects reported in patients taking sitagliptin among other drugs for type 2 diabetes (Table S3).
Our findings that empagliflozin decreases overall and cardiovascular mortality are in concordance with high-quality meta-analyses (5,6). The results are applicable to predominantly white adults with HbA1c 7–10% and established cardiovascular disease. Although the tests for statistical interaction were not significant, ethnic and gender differences in drug benefits need further investigation. Previously published network meta-analyses also suggest that empagliflozin has a favorable benefits-to-harm profile, because it decreases HbA1c and arterial blood pressure without increased risk of hypoglycemia or weight gain (9,85). The evidence regarding the effects of empagliflozin on quality of life and the long-term safety of empagliflozin is insufficient.
Our review also found low-quality evidence that all-cause and cardiovascular mortality and morbidity are comparable after empagliflozin when compared with metformin, glimepiride, linagliptin, or sitagliptin in adults with type 2 diabetes. We downgraded the quality of evidence due to risk of bias in the body of evidence and small number of events in RCTs. We also concluded reporting bias, because reporting of patient morbidity and specific adverse effects was inconsistent across studies, and the results of several completed studies were not available for the analysis. None of the head-to-head RCTs were powered to detect differences in mortality and morbidity. A single RCT suggested reduction in mortality and cardiovascular morbidity after empagliflozin when compared with placebo in adults with established cardiovascular disorder (58,59,68). This pivotal RCT provided only indirect comparative evidence that empagliflozin may be a drug of choice in people with type 2 diabetes and comorbid cardiovascular disorder (58,59,68).
The direct evidence regarding the comparative effectiveness of empagliflozin and other antidiabetic drugs including other SGLT2 inhibitors is insufficient. A recent single RCT demonstrated that injectable liraglutide reduces the risk of major cardiovascular events [hazard ratio (HR) 0.87; 95% CI: 0.78–0.97] and all-cause mortality (HR 0.85; 95% CI: 0.74–0.97) in adults with type 2 diabetes and high cardiovascular risk when compared with placebo (86). Although the RR reduction is larger with empagliflozin, well-designed direct RCTs are needed to conclude the comparative effectiveness of the 2 drugs. Previously published network meta-analyses of intermediate outcomes also suggest that empagliflozin has a favorable benefits-to-harms profile, because it decreases HbA1c without increased risk of hypoglycemia or weight gain (9,15,85).
Our rapid review has several limitations. We did not contact drug manufacturers or principal investigators regarding unpublished or missing data. We do not know how many unregistered, unpublished studies have been conducted. We found no observational studies that provide adjusted for confounding estimates of the comparative effectiveness and safety of empagliflozin and other antidiabetic drugs.
Evidence-based guidelines recommend SGLT2 inhibitors among other available drug classes for adults with type 2 diabetes who could not control diabetes with behavioral changes and metformin (1,2,4). A British guideline based on comprehensive evidence specifies that empagliflozin in combination with metformin should be recommended only to patients who cannot tolerate sulfonylureas or have a high risk of hypoglycemia or its consequences (3).
Future research should examine the long-term comparative benefits and harms of empagliflozin and other drug choices in patient subpopulations by demographics, comorbidities, and concomitant treatments.
We would like to thank David R. Goldmann, MD for his contribution to the refinement of the clinical question and review protocol.
This work is supported by Elsevier Evidence-based Medicine Center.
Conflicts of Interest: The authors have no conflicts of interest to declare.