EMPA-REG OUTCOME Subgroup Analyses: Who Responded Most and Least

At a glance

Parameter	Detail
Trial	EMPA-REG OUTCOME
N	7,020 (empagliflozin 10 mg: 2,345; 25 mg: 2,342; placebo: 2,333)
Intervention	Empagliflozin 10 mg or 25 mg once daily
Comparator	Placebo (added to standard of care)
Duration	Median 3.1 years
Primary endpoint	3-point MACE (CV death, nonfatal MI, nonfatal stroke)
Key result	HR 0.86 (95% CI 0.74-0.99; p = 0.04 for superiority)
CV death	HR 0.62 (95% CI 0.49-0.77)
All-cause mortality	HR 0.68 (95% CI 0.57-0.82)

Why Subgroup Analyses Matter Here

The primary EMPA-REG OUTCOME publication demonstrated a 14% relative risk reduction in 3-point MACE and a striking 38% reduction in cardiovascular death with empagliflozin versus placebo. Those topline numbers changed cardiology and endocrinology practice overnight. But a pooled hazard ratio tells you what happened on average across 7,020 patients. It does not tell a clinician whether the 72-year-old woman with stage 3a CKD sitting in front of them will see the same benefit as the 55-year-old man with preserved kidney function.

That is the purpose of subgroup analyses. EMPA-REG OUTCOME included 16 pre-specified subgroups in the statistical analysis plan. Several post-hoc analyses followed over the next two years. Together, they form the most granular look at which type 2 diabetes patients gain the most (and least) from SGLT2 inhibition for cardiovascular protection.

Pre-Specified Subgroup Design

The trial's statistical analysis plan defined subgroups before unblinding. Each subgroup analysis tested interaction between treatment assignment and the subgroup variable using a Cox proportional-hazards model. The threshold for a significant interaction was p < 0.05 (two-sided), though the investigators acknowledged that multiple comparisons increased the risk of false positives.

Pre-specified subgroups included:

• Age (<65 vs. ≥65 years)
• Sex (male vs. female)
• Race (White, Asian, Black/African American)
• BMI (<30 vs. ≥30 kg/m²)
• HbA1c (<8.5% vs. ≥8.5%)
• eGFR (<60, 60 to <90, ≥90 mL/min/1.73 m²)
• Region (North America, Latin America, Europe, Asia, Africa)
• Time since T2D diagnosis (<5, 5-10, >10 years)
• Baseline insulin use (yes vs. no)
• Baseline metformin use (yes vs. no)
• Baseline SU use (yes vs. no)

Results by Subgroup: The Full Picture

Age

Interaction p-value for MACE: 0.74. Both age strata showed consistent direction of benefit. The cardiovascular death reduction was remarkably similar in both groups, and the point estimates overlapped almost completely. This result is clinically reassuring: older patients with more comorbidities did not lose the cardioprotective signal. The 2019 ESC/EASD diabetes guidelines later cited this consistency when recommending SGLT2 inhibitors without an upper age cutoff.

Sex

Interaction p-value: 0.50. Women made up only 28.5% of the trial population, a limitation that reduced statistical power in the female subgroup. The wider confidence interval for women reflects that smaller sample size, not a biological difference in drug effect. The point estimate for women still favored empagliflozin. Underpowered female enrollment is a systemic problem in cardiovascular outcome trials. The FDA label for empagliflozin does not restrict by sex based on these data.

BMI

Interaction p-value: 0.71. Empagliflozin's cardiovascular benefit did not depend on obesity status. This finding is worth noting because SGLT2 inhibitors produce modest weight loss (typically 2-3 kg), and some clinicians assumed the cardiovascular benefit might track with body composition changes. The subgroup data suggest the cardioprotection operates through mechanisms beyond weight reduction, likely involving hemodynamic and natriuretic effects.

Baseline eGFR (Kidney Function)

Interaction p-value for MACE: 0.56. The kidney function subgroup is among the most clinically consequential. Patients with eGFR <60 had numerically less MACE benefit (HR 0.95) compared to those with eGFR 60-90 (HR 0.81), but the interaction was not statistically significant. The CV death reduction trended in the same protective direction across all three strata.

This matters because clinicians frequently hesitate to prescribe SGLT2 inhibitors in CKD. The EMPA-KIDNEY trial later confirmed renal benefit down to eGFR 20, but at the time of EMPA-REG OUTCOME, the eGFR <60 subgroup analysis was the earliest signal that SGLT2 inhibitors were not only safe in moderate CKD but retained cardiovascular benefit.

HbA1c at Baseline

Interaction p-value: 0.73. The cardiovascular benefit was independent of glycemic control at enrollment. Patients already near target (HbA1c <8.5%) saw similar protection to those with poor control (≥8.5%). This finding was an early indicator that empagliflozin's cardiovascular mechanism is not primarily glucose-mediated. It later supported the "pleiotropic effects" hypothesis that guided the 2018 ADA/EASD consensus report on T2D management, which repositioned SGLT2 inhibitors as cardiovascular drugs that also lower glucose, rather than the reverse.

Race and Ethnicity

Interaction p-value: 0.86. Directional consistency across all racial groups. The Black/African American subgroup was too small (5.1% of enrollment) for meaningful statistical interpretation. Trial sites spanned 42 countries, but recruitment in Sub-Saharan Africa was minimal. This limits the generalizability of race-based conclusions.

Baseline Insulin Use

Interaction p-value: 0.60. Patients on insulin at baseline, a marker of more advanced or difficult-to-control T2D, showed a numerically stronger MACE signal than those not on insulin. The CV death benefit was also preserved in insulin users (HR 0.63, 95% CI 0.47-0.84). This was clinically important in 2015, when there was debate about whether adding an SGLT2 inhibitor to insulin regimens was worth the complexity.

Post-Hoc Analyses That Shaped Practice

Several post-hoc analyses published between 2016 and 2018 extended the subgroup picture:

Heart failure at baseline. A 2016 post-hoc analysis found that empagliflozin reduced heart failure hospitalization by 35% (HR 0.65, 95% CI 0.50-0.85) regardless of whether patients had heart failure at enrollment. Among the approximately 10% of patients with baseline HF, the absolute risk reduction was larger, though the relative reduction was consistent. This analysis seeded the hypothesis that led to EMPEROR-Reduced and EMPEROR-Preserved.

By number of cardiovascular risk factors. Patients with a greater burden of CV risk (prior MI, prior stroke, peripheral artery disease) showed consistent empagliflozin benefit. The multi-vessel disease subgroup had a numerically larger absolute risk reduction because of higher baseline event rates, not a different relative effect.

Geographic region. European and North American sites showed nearly identical hazard ratios. Latin American sites had a numerically larger point estimate for benefit, though confidence intervals were wide.

Limitations of the Subgroup Data

The trial's own investigators flagged several constraints in the primary publication and supplemental materials:

1. Multiple comparisons. Sixteen pre-specified subgroups tested at p < 0.05 means roughly one false-positive interaction is expected by chance alone. No correction for multiplicity was applied to subgroup interaction tests.

2. Underpowered strata. The Black/African American subgroup (n = 358), the eGFR <60 subgroup (n = 1,447), and the female subgroup (n = 2,002) each had limited power to detect a 14% relative risk reduction.

3. No dose-response subgroup analysis. The 10 mg and 25 mg arms were pooled for all subgroup analyses. Whether specific subgroups respond differently to different doses remains unknown from this trial.

4. Post-hoc status. The heart failure and multi-vessel disease subgroup analyses were not pre-specified. They generated hypotheses rather than confirmed them. Dedicated trials (EMPEROR program) were needed to validate these signals.

5. Baseline therapy confounding. Patients on insulin had different disease durations, comorbidity profiles, and concomitant medications than those not on insulin. Subgroup analyses adjust for the subgroup variable but not for all correlated confounders.

What This Means for Real-World Prescribing

The consistency of empagliflozin's benefit across subgroups is the main takeaway, not the point estimate in any single stratum. No pre-specified subgroup showed significant heterogeneity of treatment effect (all interaction p-values > 0.05). The 2023 ADA Standards of Care reflect this by recommending SGLT2 inhibitors for cardiovascular risk reduction across a broad T2D population with established atherosclerotic CVD, without restricting by age, sex, BMI, or baseline HbA1c.

Practical implications for clinicians:

• Do not withhold empagliflozin from older patients. The ≥65 subgroup showed equivalent or slightly better point estimates for CV death reduction.
• Women benefit too. The wider confidence interval in women reflects sample size, not lack of efficacy. Guideline recommendations apply equally.
• eGFR ≥30 is the threshold. While the EMPA-REG eGFR <60 subgroup was underpowered, subsequent data from EMPA-KIDNEY extended the evidence down to eGFR 20.
• Glycemic control is not the gatekeeper. Patients with HbA1c already near target still benefit cardiovascularly from empagliflozin. This argues against deprioritizing the drug in well-controlled patients.
• Insulin co-use is fine. The subgroup data and safety profile support adding empagliflozin to insulin regimens, with appropriate dose adjustment to avoid hypoglycemia.

Frequently asked questions

›Did any EMPA-REG OUTCOME subgroup show harm from empagliflozin?

No. Every pre-specified subgroup showed a point estimate favoring empagliflozin over placebo for the primary MACE endpoint. No subgroup crossed an HR of 1.0 in a direction suggesting harm, though some confidence intervals included 1.0 due to limited sample size within those strata.

›Was the cardiovascular benefit of empagliflozin different in men versus women?

The point estimate was numerically better in men (HR 0.84) than in women (HR 0.93), but the interaction p-value was 0.50, meaning this difference was not statistically significant. Women comprised only 28.5% of trial participants, limiting statistical power in the female subgroup.

›Does empagliflozin work less well in patients with CKD?

In the eGFR <60 subgroup, the MACE hazard ratio was 0.95 (not statistically significant on its own), compared to 0.81 in the eGFR 60-90 group. The interaction p-value was 0.56, meaning the difference was not significant. Later trials like EMPA-KIDNEY confirmed benefit in CKD patients down to eGFR 20.

›Is empagliflozin's cardiovascular benefit explained by glucose lowering?

The subgroup analysis by baseline HbA1c strongly suggests no. Patients with HbA1c <8.5% and those ≥8.5% had nearly identical hazard ratios (0.85 vs. 0.89, interaction p = 0.73). This supports a mechanism independent of glycemic control, likely involving hemodynamic, natriuretic, and metabolic effects.

›Were EMPA-REG OUTCOME subgroup analyses pre-specified or post-hoc?

Sixteen subgroup analyses were pre-specified in the statistical analysis plan before unblinding. Additional analyses (heart failure at baseline, number of CV risk factors, geographic region) were post-hoc. Pre-specified analyses carry greater statistical credibility, while post-hoc analyses are considered hypothesis-generating.

›Did obese patients respond differently to empagliflozin?

No. The BMI <30 group (HR 0.84) and the BMI ≥30 group (HR 0.88) showed nearly identical benefit (interaction p = 0.71). This suggests the cardioprotective effect is not mediated primarily through weight loss, despite empagliflozin causing modest weight reduction.

›How did race and ethnicity affect EMPA-REG OUTCOME results?

White, Asian, and Black/African American subgroups all showed directional benefit with empagliflozin (interaction p = 0.86). The Black/African American subgroup was very small (n = 358), limiting statistical conclusions. Broader enrollment diversity remains a gap in SGLT2 inhibitor evidence.

›Should empagliflozin be avoided in patients already on insulin?

No. The insulin-user subgroup (n = 3,375) showed a MACE hazard ratio of 0.84, consistent with the overall trial result. The 2023 ADA Standards of Care support adding SGLT2 inhibitors to insulin therapy for cardiovascular benefit, with appropriate glucose monitoring.

›Were both empagliflozin doses equally effective across subgroups?

The published subgroup analyses pooled the 10 mg and 25 mg arms. In the overall trial, both doses showed similar cardiovascular benefit (HR 0.85 for 10 mg, HR 0.86 for 25 mg). Dose-specific subgroup breakdowns were not published.

›How do EMPA-REG OUTCOME subgroups compare to DAPA-HF or CANVAS subgroups?

DAPA-HF (dapagliflozin) and CANVAS (canagliflozin) showed similarly consistent subgroup results. Across the SGLT2 inhibitor class, no large cardiovascular outcome trial has identified a subgroup that clearly does not benefit. This class-wide consistency strengthened guideline recommendations for SGLT2 inhibitors in T2D with established CVD.

References

1. Zinman B, Wanner C, Lachin JM, et al. Empagliflozin, Cardiovascular Outcomes, and Mortality in Type 2 Diabetes. N Engl J Med. 2015;373(22):2117-2128. PubMed
2. Fitchett D, Zinman B, Wanner C, et al. Heart failure outcomes with empagliflozin in patients with type 2 diabetes at high cardiovascular risk: results of the EMPA-REG OUTCOME trial. Eur Heart J. 2016;37(19):1526-1534. PubMed
3. Wanner C, Inzucchi SE, Lachin JM, et al. Empagliflozin and Progression of Kidney Disease in Type 2 Diabetes. N Engl J Med. 2016;375(4):323-334. PubMed
4. Cosentino F, Grant PJ, Aboyans V, et al. 2019 ESC Guidelines on diabetes, pre-diabetes, and cardiovascular diseases developed in collaboration with the EASD. Eur Heart J. 2020;41(2):255-323. PubMed
5. The EMPA-KIDNEY Collaborative Group. Empagliflozin in Patients with Chronic Kidney Disease. N Engl J Med. 2023;388(2):117-127. PubMed
6. American Diabetes Association Professional Practice Committee. Standards of Care in Diabetes, 2023. Diabetes Care. 2023;46(Suppl 1). PubMed