The EMPA-REG OUTCOME™ trial (clinicaltrials.gov identifier: NCT01131676) is an ongoing, multicenter, randomized, double-blind, placebo-controlled trial. It was designed to assess the effect of empagliflozin (10 mg or 25 mg once daily) compared with placebo, in addition to standard of care, on CV events in adults with T2DM at high risk of CV events and with less than optimized glycemic control.
The study protocol was approved by the respective Institutional Review Boards, Independent Ethics Committees and Competent Authorities according to national and international regulations.
Randomization and follow-up
Randomization was undertaken using a computer-generated random sequence and an interactive voice and web response system. Patients were stratified by HbA
1c at screening (<8.5%, ≥8.5%), BMI at randomization (<30 kg/m
2, ≥30 kg/m
2), region (North America [plus Australia and New Zealand], Latin America, Europe, Africa, Asia), and renal function (eGFR using the Modification of Diet in Renal Disease [MDRD] equation) at screening (Chronic Kidney Disease [CKD] stage 1: ≥90 ml/min/1.73 m
2; CKD stage 2: 60–89 ml/min/1.73 m
2; CKD stage 3: 30–59 ml/min/1.73 m
2). Patients are instructed to attend the clinic at pre-specified times over the duration of the study including a follow-up visit 30 days after the end of the treatment period (Figure
1). Patients who prematurely discontinue study medication are asked to attend all visits as originally planned.
Outcomes and outcome adjudication
The primary outcome of the study is time to first occurrence of CV death, non-fatal myocardial infarction (MI, excluding silent MI), or non-fatal stroke
i.e., 3-point major adverse cardiovascular events (3P-MACE). The key secondary outcome expands the primary composite outcome to include time to first occurrence of hospitalization for unstable angina (4P-MACE). Further CV outcomes are the individual components of the 4P-MACE, as well as individual occurrence of and time to silent MI, heart failure requiring hospitalization, all-cause mortality, transient ischemic attack (TIA) and coronary revascularization procedures. All CV outcome events and deaths are being prospectively adjudicated by the Clinical Events Committee (one for cardiac events and one for neurological events), as recommended in FDA guidelines (FDA [
9]). Definitions of the major clinical outcomes are presented in Additional file
3 and a non-exhaustive list of further CV outcomes (secondary, tertiary and exploratory) in Additional file
4.
Additional secondary outcomes include the occurrence of and time to new onset albuminuria (urinary albumin:creatinine ratio ≥30 mg/g) and new onset of macroalbuminuria (urinary albumin:creatinine ratio ≥300 mg/g). Other outcomes include the occurrence of and time to a composite microvascular outcome comprising the initiation of laser therapy for retinopathy, vitreous hemorrhage, diabetes-related blindness, and new or worsening nephropathy (new onset macroalbuminuria [albumin:creatinine ratio ≥300 mg/g]; doubling of serum creatinine accompanied by eGFR ≤45 mL/min/1.73 m2; initiation of renal replacement therapy; or death due to renal disease) as well as the individual components of this composite.
The short (12 weeks), medium (52 weeks), and long-term (annually, at end of study, and at follow-up) effects of the two doses of empagliflozin on HbA1c, fasting plasma glucose (FPG), body weight, waist circumference, and BP will be assessed, as well as the proportion of patients who meet the composite outcome of HbA1c reduction ≥0.5%, systolic BP reduction >3 mmHg, and body weight reduction >2%.
The prognostic impact of, and the modulating potential of empagliflozin on, the renal biomarker cystatin C and the CV biomarkers high-sensitivity C-reactive protein and high sensitivity troponin T will be assessed in sub-studies, as will potential associations between genetic variations and drug response.
Safety will be assessed based on adverse events (AEs) reported throughout the study and up to 7 days after the last dose of study medication (coded using the Medical Dictionary for Drug Regulatory Activities [MedDRA]), clinical laboratory tests, vital signs, 12-lead electrocardiogram (ECG), physical examination, and the use of rescue medication. Pre-specified AEs of special interest (AESI) include confirmed hypoglycemic adverse events (plasma glucose ≤70 mg/dL (3.9 mmol/L) and/or requiring assistance), those reflecting volume depletion, bone fracture, hepatic events, malignancies, urinary tract infection (UTI), and genital infection. Events may be defined by either abnormal laboratory values and/or relevant adverse events identified using prospectively defined search categories or both. For qualifying events, relevant source documentation will be requested including lab values, histological analysis, results of ultrasound, CT, MRI, scintigraphy, hospital discharge letters, and medical reports from other physicians. All evaluations will be performed in a blinded fashion.
A list of efficacy and safety outcomes is presented in Additional file
4.
Statistical considerations
Sample size and power calculations
The primary hypothesis aims to show non-inferiority on 3P-MACE for empagliflozin versus placebo based on a non-inferiority margin of < 1.3 (upper limit of the adjusted 95% confidence interval (CI)) for the hazard ratio. The upper limit of the adjusted 95% CI for the HR of <1.3 was based on FDA guidance for CV trials evaluating new anti-hyperglycemic therapies for T2DM [
9]. Patients who receive either 10 mg or 25 mg of empagliflozin will be pooled into a common treatment group for the purposes of the primary analysis. A 4-step hierarchical testing strategy will be followed: 1) non-inferiority test of the primary outcome (3P-MACE), 2) non-inferiority test of the key secondary outcome (4P-MACE), 3) superiority test of the primary outcome (3P-MACE) and 4) superiority test of the key secondary outcome (4P-MACE). A minimum of 691 confirmed primary outcome events are required to provide 90% power with a one-sided α level of 0.025, assuming equal risk between the placebo and empagliflozin groups. With a minimum of 691 events, the trial will also have at least 80% power to detect a hazard ratio of 0.785 (corresponding to a 21.5% risk reduction in CV outcome events) for the primary outcome.
Interim analysis
In order to support a CV meta-analysis of all CV events occurring in the phase III trials involving empagliflozin, as required for all New Drug Applications to be submitted to the FDA [
9], CV outcome data from the ongoing EMPA-REG OUTCOME™ was extracted. The cut off for the data extraction was preplanned and ~ 150 4P-MACE were included in the project level CV meta-analysis. This resulted in addition of a Haybittle-Peto correction for the interim analysis (i.e., 0.0001 of the α was spent on the data extraction for the interim analysis), and subsequent reduction of the final α level to 0.0249 (in order to maintain the experiment-wise α level of 0.025).
The need to prevent the release of any data from the data extraction or interim analysis that could define the effects of empagliflozin on CV outcomes was fundamental to the study design [
23]. Accordingly, procedures, including restricted access to electronic systems, were put in place to ensure that the effect estimate remained blinded and data review by the regulators would not require premature disclosure of the effects of empagliflozin on CV outcome. The data extraction, interim analysis and the following phase III CV meta-analysis were performed by a group independent from the EMPA-REG OUTCOME™ trial team, so that the trial’s operational team and the academic Steering Committee remained blinded to the results. The DMC is the sole group with access to unblinded results beyond the strictly firewalled “CV meta-analysis group” of the sponsor.
Analysis plan
Three analysis populations are defined for this trial: 1) The treated set (TS), consisting of all patients who were treated with at least one dose of study drug, 2) The on-treatment set (OS), consisting of patients who received the drug for at least 30 days (cumulative) in whom events will be considered that occurred within 30 days of the off-treatment period or until the end of the entire trial, whichever will be earlier (patients who did not experience the primary outcome will be censored at the end of the treatment period, if the patient completes treatment as planned, or at the end of the 30 day period) and 3) the Full Analysis Set (FAS), consisting of all patients randomized, treated with at least one dose of study drug and with a baseline HbA1c value.
The primary analysis will be based on a Cox proportional hazards model with treatment (with empagliflozin 10 mg and 25 mg pooled into a single group), age, gender, baseline BMI (<30 kg/m2, ≥30 kg/m2), baseline HbA1c (<8.5%, ≥8.5%), baseline eGFR as well as geographical region (classified as North America, Latin America, Europe, Africa and Asia) as factors. The same Cox proportional hazards model as for the primary outcome will be employed in all steps of the hierarchical testing strategy (3P-MACE and 4P-MACE).
The time to the occurrence of the primary outcome and the key secondary outcomes event will be computed as (event date - randomization date) +1. Patients who do not have the event during the trial period will be censored at the individual day of trial completion. The time to censoring will be computed as (individual day of trial completion – randomization date) + 1. For patients who have more than one primary outcome event during the trial, the time to the first occurrence of the primary outcome event will be considered for the primary analysis. All adjudicated and confirmed events will be used for the primary analysis.
The TS is the basis for the primary analysis and the FAS is the basis for the intention-to-treat (ITT) analysis for efficacy analyses. As sensitivity analyses the primary analysis will also be performed for the OS. Secondary analyses of the primary analysis with pooled active treatment arms will be performed by comparing the active treatments individually versus the placebo arm. Sensitivity analyses of the primary and key secondary outcomes will be performed that include the additional factor of naïve/experienced drug status in the Cox model. The secondary and tertiary further cardiovascular outcomes will be analyzed in a Cox proportional hazards model similar to the primary analysis for the treated set. Of note is that also other sensitivity analysis will be conducted employing other statistical methods, as well as an assessment of outcomes per individual dosages (i.e., empagliflozin 10 mg and empagliflozin 25 mg).
Subgroups to be considered in the analyses will be defined based on, but not limited to, age, HbA1c, BMI, weight, geographical region, race, gender, ethnicity, time since diagnosis of T2DM, renal function, BP, eGFR, glucose-lowering and CV prophylactic medication, CV complications and cohort, all defined at study baseline or screening. In addition, outcomes in patients experiencing severe hypoglycaemia vs those not experiencing severe hypoglycaemia will be assessed. Further details as to the specific categories to be employed for each subgroup factor are provided in Additional file
6.
Patient recruitment and baseline characteristics
Recruitment into the EMPA-REG OUTCOME™ trial began in September 2010 and was completed in April 2013. In total 11507 patients were screened and 7042 participants were randomized to receive study treatment at 592 clinical sites in 42 countries. The main reason for screen-failure was that the HbA1c fell outside protocol specifications. Of those randomized, 7034 participants were treated. The baseline characteristics of treated participants are shown in Table
2. Most came from Europe (41%) or North-America (20%) with 19% from Asia, 15.4% from Latin America and 4% from South-Africa. The mean age of participants was 63 years, with 9% aged ≥75 years. Seventy-two per cent are male, and 72% are white. Time since diagnosis of T2DM was ≤5 years in 18% of participants and >10 years in 57%. At baseline, mean HbA
1c was 8.1% (Table
3) with 68% of participants having HbA
1c <8.5%. Only 2% of participants were drug-naïve; 29% were receiving monotherapy, and 45% were receiving dual therapy. Insulin was used by 36% of participants (as monotherapy or part of dual therapy). A history of CV complications or CV events was demonstrated in 99% of participants and in total 47% had a history of MI and 23% a history of stroke. Fifty-two per cent of participants had an eGFR ≥ 60 and <90 mL/min/1.73 m
2 (i.e., stage 2 CKD) and 26% had an eGFR ≥ 30 and <60 mL/min/1.73 m
2 (i.e., stage 3 CKD). Albuminuria (UACR ≥30 mg/g) was present in 40% of participants. At baseline, 77% of patients were receiving a statin, 9% were receiving a fibrate, 85% were being treated with an acetylsalicylic acid agent, and 94% were receiving any drug for BP reduction (80% on blockers of the renin-angiotensin system).
Table 2
Baseline characteristics (treated set; n = 7034)
Age (years), mean (SD) | 63.1 (8.6) |
≥ 75 years of age, n (%) | 652 (9) |
Male, n (%) | 5026 (72) |
Race, n (%) | |
White | 5089 (72) |
Asian | 1518 (22) |
Black/African American | 357 (5) |
Other* | 70 (1) |
Ethnicity, n (%) | |
Hispanic or Latino | 1268 (18) |
Smoking history, n (%) Current/Ex-smoker | 930 (13)/3216 (46) |
Time since diagnosis, n (%) | |
≤5 years | 1265 (18) |
>5-10 years | 1754 (25) |
>10 years | 4015 (57) |
Region, n (%) | |
Europe | 2885 (41) |
North America/Australia/New Zealand | 1408 (20) |
Latin America | 1081 (15) |
Africa | 313 (4) |
Asia | 1347 (19) |
Northeast Asia | 586 (8) |
South/South-East Asia | 761 (11) |
CV risk factors, any of the below, n (%) | 6978 (99) |
History of MI | 3275 (47) |
Single-vessel CAD | 743 (11) |
Multi-vessel CAD | 3285(47) |
CABG | 1738(25) |
History of stroke | 1631 (23) |
Peripheral occlusive arterial disease | 1449 (21) |
Glucose-lowering therapy at baseline, n (%) | |
None | 128 (2) |
Monotherapy | 2055 (29) |
Metformin (% of monotherapy) | 745 (36) |
Insulin (% of monotherapy) | 954 (46) |
Dual therapy | 3188 (45) |
Metformin + sulfonylurea (% of dual therapy) | 1383 (43) |
Metformin + insulin (% of dual therapy) | 1420 (45) |
Other therapies (n, %) | |
Acetylsalicylic acid | 5990 (85) |
Statins | 5387 (77) |
Fibrates | 630 (9) |
Any antihypertensive therapy (n, %) | 6641 (94) |
Blockers of the renin-angiotensin system | 5651 (80) |
Beta-blockers | 4537 (64) |
Calcium channel blockers | 2114 (30) |
Table 3
Key baseline laboratory data (treated set; n = 7034)
HbA1c (%), mean (SD) | 8.1 (0.8) |
HbA1c <8.5%, n (%) | 4811 (68) |
Fasting plasma glucose (mmol/L), mean (SD) | 8.5 (2.4) |
Body mass index (kg/m2), mean (SD) | 30.6 (5.3) |
≥ 35 kg/m2, n (%) | 1426 (20) |
Weight (kg), mean (SD) | 86.4 (18.9) |
Waist circumference (cm), mean (SD) | 105 (14) |
Systolic/diastolic blood pressure (mmHg), mean (SD) | 135 (17)/77 (10) |
Lipids (mmol/L), mean (SD) | |
Total cholesterol | 4.2 (1.1) |
LDL-cholesterol | 2.2 (0.9) |
HDL-cholesterol | 1.2 (0.3) |
Triglycerides | 1.9 (1.4) |
eGFR according to MDRD (mL/min/1.73 m2), mean (SD) | 74 (21) |
eGFR according to MDRD (mL/min/1.73 m2), n (%) | |
≥90 | 1534 (22) |
60 to <90 | 3671 (52) |
30 to <60 | 1796 (26) |
ACR albumin ratio (mg/g), median (Q1, Q3) | 17.7 (7.1, 72.5) |
ACR ratio (mg/g), n (%) | |
≥ 30 – 300 | 2011 (29) |
≥ 300 | 771 (11) |