Dapagliflozin vs empagliflozin in patients with chronic heart failure: a registry analysis

Abstract

Aim: To assess the relative efficacy of dapagliflozin and empagliflozin in routinely treated chronic heart failure (CHF) patients.

Methods: Data from a registry of prevalent and incident CHF patients were used to set up cohorts (new-user design) of patients started on dapagliflozin or empagliflozin in addition to other guideline-directed therapy. Cohorts were mutually balanced on a range of characteristics, and were assessed for the incidence of a composite of all-cause death/major adverse cardiac events (primary outcome) over the initial 6 months of treatment, and for New York Heart Association (NYHA) functional class at 6 months (secondary outcome). Frequentist and Bayes estimates were generated for the dapagliflozin vs empagliflozin comparison.

Results: In both prevalent (dapagliflozin n=393, empagliflozin n=328) and incident (dapagliflozin n=124, empagliflozin n=116) patients, those prescribed dapagliflozin had somewhat higher incidence of the primary outcome, but the confidence intervals were wide (RR=1.385, 95%CI 0.882-2.173 [prevalent], RR=2.192, 95%CI 0.765-6.282 [incident]), and were more likely to present with a worse NYHA class at 6 months (OR=1.552, 95%CI 1.142-2.108 [prevalent], OR=1.503, 95%CI 0.844-2.676 [incident]). In the pooled data, primary events (n=102) were more common in dapagliflozin-prescribed patients (frequentist estimate RR=1.519, 95%CI 1.239-1.861; Bayes RR=1.380, 95%CrI 0.981-1.944). Dapagliflozin-prescribed patients were also more likely to have a worse NYHA class at 6 months (OR=1.540, 95%CI 1.208-1.962; Bayes OR=1.425, 95%CrI 1.098-1.781).

Conclusion: CHF patients prescribed with dapagliflozin had poorer outcomes than their empagliflozin-prescribed peers over the initial 6 months of treatment. Data emphasize a need for a direct randomized comparison of the two treatments in this setting.

Figure 1

Outline of the present analysis. Included were chronic heart failure (CHF) patients embraced in an institutional registry, started on sodium-glucose co-transporter type 2 inhibitors (SGLT2i) dapagliflozin or empagliflozin, either with a delay vs the CHF diagnosis (prevalent patients, Study 1), or immediately after the CHF diagnosis (incident patients, Study 2). At the start of SGLT2i treatment (baseline), patients prescribed the two SGLT2i were mutually balanced (by optimization-based weighting) on a number of covariates, and were assessed at a control visit 6 months later. MACE – major adverse cardiovascular events; NT-proBNP – N-terminal pro-brain natriuretic polypeptide; NYHA – New York Heart Association.

Figure 2

Summary of the adjusted (weighted) primary analysis (each study separately) of the primary (all-cause mortality/major adverse cardiovascular events, MACE) and secondary (New York Heart Association, NYHA, class at the control visit) outcomes. Effect estimates are for the dapagliflozin (DAPA) vs empagliflozin (EMPA) contrast: relative risks (RR) for the primary and odds ratios (OR) for the secondary outcome (OR>1.0 indicates higher odds of a worse NYHA class). Bayesian credible intervals are the highest posterior density intervals.

Figure 3

Summary of the adjusted (weighted) secondary analysis (pooled data) of the primary (all-cause mortality/major adverse cardiovascular events, MACE) and secondary (New York Heart Association, NYHA, class at the control visit) outcomes with sensitivity to unmeasured confounding. Effect estimates are for the dapagliflozin (DAPA) vs empagliflozin (EMPA) contrast: relative risks (RR) for the primary and odds ratios (OR) for the secondary outcome (OR>1.0 indicates higher odds of a worse NYHA class). Bayesian credible intervals are the highest posterior density intervals. Black squares (bars) denote observed effects. Gray squares (bars) and font denote estimates corrected for a hypothetical bias disfavoring dapagliflozin: (i) we assumed a strong residual confounding bias with an effect of RR = 1.40 for the primary outcome and OR = 1.50 for the secondary outcome, where the overall prevalence of the “biasing variables” is 35%, with a higher prevalence in dapagliflozin-treated than in empagliflozin-treated patients (ratio 1.5:1.0), ie, dapagliflozin 41.4% vs empagliflozin 27.5%. The observed effects were corrected for this hypothetical bias. Gray diamonds (bars) and font denote estimates corrected for the same hypothetical bias but under the assumption that it disfavored empagliflozin: “biasing variables” were assumed more prevalent in empagliflozin vs dapagliflozin-treated patients (ratio 1.5:1.0), ie, empagliflozin 42.3% vs dapagliflozin 28.6%. The E-values denote (on a risk ratio scale) the size of a confounding effect needed to explain away the largest part of the observed point-estimates, ie, to “push” them to RR/OR = 1.10.

Figure 4

Summary of the analysis of chronic kidney disease (CKD), and of left ventricular ejection fraction (mildly reduced/preserved [HFm/pEF] or reduced [HFrEF]) as moderators of the dapagliflozin (DAPA) vs empagliflozin (EMPA) differences in the primary (incidence of all-cause mortality/major adverse cardiovascular events, MACE) and secondary (New York Heart Association, NYHA, class at the control visit) outcomes. Effect estimates are for the DAPA vs EMPA contrast: relative risks (RR) for the primary and odds ratios (OR) for the secondary outcomes (pooled data on prevalent and incident patients). Black squares (bars) denote observed effects. Gray squares and diamonds (bars) and font denote estimates corrected for a hypothetical bias disfavoring dapagliflozin or empagliflozin, respectively (see footnote to Figure 3 for details).