figure a


The global epidemic of type 2 diabetes is expected to affect 642 million individuals by 2040 and impacts women and men equally [1]. It has been postulated that when women develop diabetes, they lose their generally protective effect against cardiovascular (CV) complications. Thus, the relative risk for CV complications is greater in women with diabetes than men when compared with peers without diabetes. This observation may relate to differences in CV risk factors in men and women, influence of sex hormones on CV risk, or sex disparities in the management of diabetes and/or CV risk factors [2]. Moreover and disappointingly, women typically only constitute 25–40% of the study populations in major CV outcome trials [3,4,5]. Thus, there may be less certainty about efficacy and safety of CV preventive strategies in women with diabetes. The EMPA-REG OUTCOME® trial, which included 28.5% women, found that empagliflozin, given in addition to standard of care, reduced the risk of CV death by 38% (HR 0.62 [95% CI 0.49, 0.77]; p < 0.001), heart failure (HF) hospitalisation by 35% (HR 0.68 [0.57, 0.82]; p < 0.001) and a composite endpoint for incident or worsening nephropathy by 39% (HR 0.61 [0.53, 0.70]; p < 0.001) vs placebo [5,6,7]. As a result of this trial, approval for a CV death prevention indication in individuals with type 2 diabetes and established CV disease was granted by the US Food and Drug Administration and other regulatory authorities. Here we report a secondary prespecified analysis [5] of the trial to determine the relative effects of empagliflozin in women vs men.


Description of the trial design (NCT01131676) and methodology used has previously been reported [5,6,7]. In brief, the population studied was individuals with type 2 diabetes (HbA1c 53–86 mmol/mol [7–10%] and eGFR >30 ml min−1 [1.73 m]−2) who had established atherosclerotic CV disease. Individuals were randomised to receive empagliflozin 10 mg or 25 mg, or placebo once daily in addition to standard of care. Throughout the trial, investigators were encouraged to treat CV risk factors (including dyslipidaemia and hypertension) to achieve the best available standard of care according to local guidelines. All CV outcome events were prospectively adjudicated by independent and blinded clinical events committees. For the renal outcomes analysed, serum creatinine and urinary albumin in spot urine samples obtained during regular study visits were measured in central laboratories with the use of standardised procedures. We used the modification of diet in renal disease (MDRD) formula to estimate the eGFR. Safety was assessed on the basis of adverse events that occurred during treatment or within 7 days after the last dose of study drug, and collected as previously reported [5,6,7].

Cox proportional hazards models were used to assess between-group differences in the risk of an outcome after adjustment for study group, age, sex, baseline BMI, HbA1c, eGFR and region. We present overall HR according to sex and the interaction p value considering the interaction of treatment × sex for CV death, HF hospitalisation and the predefined composite renal outcome. Kaplan–Meier plots are presented to display events over time; adverse events are expressed as proportions. Post hoc power considerations were derived using ADDPLAN (version 6.1.1, 2014, ICON, All study participants gave informed consent prior to enrolment, and the investigations were approved by ethics committees or institutional review boards.


The profile at baseline (Table 1, electronic supplementary material [ESM] Table 1) was largely similar between the 2004 women in the trial (age [mean±SD], 63.6 ± 8.8 years; BMI 31.2 ± 5.7 kg/m2; systolic/diastolic blood pressure [SBP/DBP] 136.0 ± 17.6/76.0 ± 10.0 mmHg; eGFR 72.7 ± 22.2 ml min−1 [1.73 m]−2; 64% normoalbuminuria), and the 5016 men (age 63.0 ± 8.6; BMI 30.4 ± 5.1; SBP/DBP 135.2 ± 16.8/76.9 ± 9.8; eGFR 74.6 ± 21.1; 57.6% normoalbuminuria). The geographic representation in the trial was also similar between the sexes (women: 41.0% Europe; 17.8% North America; 18.9% Latin America;18.1% Asia; 4.2% South Africa vs men: 41.1% Europe; 20.7% North America; 14.0% Latin America;19.6% Asia; 4.5% South Africa). In addition, type 2 diabetes disease characteristics were also largely similar: HbA1c 65.5 ± 9.7 mmol/mol (8.1% ± 0.9%) vs 64.4 ± 9.0 mmol/mol (8.0% ± 0.8%); type 2 diabetes duration > 10 years: 57.8% vs 56.9%; and any use of metformin (73.5% vs 74.2%), sulfonylurea (41.0% vs 43.5%) or insulin (51.0% vs 47.1%).

Table 1 Baseline characteristics by sex in EMPA-REG OUTCOME

LDL-cholesterol was, however, numerically higher in women (2.5 ± 1.0 vs 2.1 ± 0.9 mmol/l), likely reflecting lower rates of lipid-lowering therapies (75.4% vs 83.2%), in particular statin use (71.1% vs 79.3%). Women also had lower prevalent use of anti-platelet therapies (80.9% vs 87.7%) and were less likely to have smoked (31.5% vs 69.9%). Use of antihypertensive drugs were similar (75.3% vs 77.0% used ≥ two antihypertensive agents), including use of angiotensin converting enzyme inhibitors (ACEis) / angiotensin receptor blockers (ARBs) (79.7% vs 81.1%), but slightly more women used diuretics (47.6% vs 41.5%) and slightly fewer women used beta blockers (58.9% vs 67.3%). Throughout the trial (ESM Table 2), a higher percentage of patients in the placebo group, for both men and women, received additional antihypertensive therapies (women: 49.8% [placebo] vs 42.9% [empagliflozin]; men: 51.5% vs 45.2%) and acetylsalicylic acid, with the latter occurring numerically more frequent in women (women: 21.7% vs 17.7%; men: 18.6% vs 18.0%), whereas new introduction of statin therapy was similar (women: 25.7% vs 24.9%; men: 25.8% vs 24.2%).

Prior coronary artery disease was less common in women (62.8% vs 80.7%), but women had a more frequent history of stroke (31.6% vs 20.0%). History of HF was similar (10.5% vs 9.9%).

Effects of empagliflozin on reducing BP, HbA1c, weight and waist circumference were of similar magnitude regardless of sex (data not shown). The annualised incidence rate for women in the placebo group was numerically lower than that in men for CV death (1.58% vs 2.19%), numerically higher for HF hospitalisation (1.75% vs 1.33%), and similar for renal events (7.22% vs 7.75%). Empagliflozin reduced the primary composite endpoint (CV death, non-fatal stroke, non-fatal myocardial infarction) relatively by 14% (Fig. 1a,b) by a similar degree, irrespective of sex (interaction p value 0.8114). We did not detect any effect modification by sex within the statistical power restrictions of the analysis for CV death, HF hospitalisation and incident or worsening nephropathy (Fig. 1a, c–e; interaction p values 0.32, 0.20 and 0.85, respectively). Separation of the incidence curves for these events were generally early and risk reduction with empagliflozin vs placebo persisted for the trial’s duration (Fig. 1b–e). Post hoc power considerations are provided in ESM Table 3.

Fig. 1
figure 1

(a) Forest plot of 3-point Major Adverse Cardiovascular Events (MACE [composite of CV death, nonfatal myocardial infarction or nonfatal stroke]), CV death, HF hospitalisation and nephropathy by sex, and interaction p values. Values were determined by Cox regression analysis in patients treated with ≥1 dose of study drug. (be) Kaplan–Meier estimates in patients treated with ≥1 dose of study drug. Solid grey line, women in the empagliflozin group; dashed grey line, men in the empagliflozin group; solid black line, women in the placebo group; dashed black line, men in the placebo group. (b) 3-point MACE over time in women and men with empagliflozin vs placebo. HR (95% CI): women, 0.83 (0.62, 1.11); men, 0.87 (0.73, 1.02). p = 0.8114 for treatment by sex interaction. (c) CV death over time in women and men with empagliflozin vs placebo. HR (95% CI): women, 0.76 (0.48, 1.20); men, 0.58 (0.45, 0.75). p = 0.3219 for treatment by sex interaction. (d) HF hospitalisation over time in women and men with empagliflozin vs placebo. HR (95% CI): women, 0.50 (0.31, 0.81); men, 0.73 (0.53, 1.01). p = 0.1981 for treatment by sex interaction. (e) Occurrence of worsening of nephropathy (progression to macroalbuminuria, doubling of serum creatinine, initiation of renal-replacement therapy, or death from renal disease) over time in women and men with empagliflozin vs placebo). HR (95% CI): women, 0.62 (0.48, 0.80); men, 0.61 (0.52, 0.71). p = 0.8543 for treatment by sex interaction

Empagliflozin was well tolerated by women, with no remarkable difference between empagliflozin and placebo for most adverse events assessed, including urinary tract infections. Genital infections, which were more common in women than in men in both treatment groups, were increased with empagliflozin treatment. In women this was observed in 2.5% in the placebo group and 10.0% in the empagliflozin group while in men, 1.5% and 2.6%, respectively, consistent with the known safety profile of sodium–glucose co-transporter-2 (SGLT2) inhibition. Nevertheless, discontinuation of the study drug due to this adverse event was infrequent. For every 1000 women treated with empagliflozin for 3.1 years, there occurred 12 fewer CV deaths, 27 fewer hospitalised HFs and 83 fewer incident or worsening nephropathy events, with 101 more genital infections, in comparison with placebo.


Women with type 2 diabetes and established CV disease who were enrolled in the EMPA-REG OUTCOME® trial experienced high CV and renal event rates, in particular hospitalisation for HF that occurred numerically more frequently than in men, emphasising the high impact of diabetes on complications in this group. The numerically higher event rate for HF hospitalisation in women than in men is interesting in light of a similar between-sexes a priori 5 year estimated HF risk using the Health ABC HF risk score [8], bearing in mind the limitation that our trial was not powered to address sex differences. This observation could be related both to underlying factors specific to women [2] or, as suggested by others [9], because evidence-based HF therapies are used less often by women. In EMPA-REG OUTCOME®, use of glucose-lowering medications, ACEis/ARBs and overall antihypertensive therapy use (albeit characterised by somewhat more diuretic and somewhat less beta-blocker use), was similar in men and women. However, both women and men allocated to placebo had a higher proportion of new prescriptions of such drugs during the trial than those allocated to empagliflozin. Women did have lower use of lipid-lowering and anti-platelet therapies, but neither of these have been associated with HF outcomes. Also, as fewer women than men were current or former smokers, a habit usually associated with increased HF risk [10], our observation could lend support to the hypothesis of female-specific factors playing a role. Empagliflozin was generally well tolerated, with a higher frequency of genital infections in both sexes, more so both in relative and absolute terms in women. This is congruent with results from previous reports on the use of empagliflozin, which has increased the risk of yeast vaginitis in earlier trials involving women with type 2 diabetes at lower CV risk [11].

In conclusion, CV death, HF hospitalisation and incident or worsening nephropathy rate reductions induced by empagliflozin were not different between women and men.