Table 3.

Mediation analysis of the effect of empagliflozin 10 mg/day on E/e´ ratio after 12 weeks of treatment in individuals with type 2 diabetes mellitus

	Beta-estimateEmpagliflozin 10 mg/day vs. Placebo [95% CI]	P-value
Basic modela	– 1.18 [– 1.72; – 0.65]	< 0.0001
+ ∆ Hemoglobin	– 0.95 [– 1.54; – 0.36]	0.0021
+ ∆ Hematocrit	– 1.04 [– 1.66; – 0.43]	0.0012
+ ∆ Estimated glomerular filtration rate	– 1.09 [– 1.63; – 0.55]	0.00013
+ ∆ Body mass index	– 1.11 [– 1.69; – 0.52]	0.00031
+ ∆ Systolic blood pressure	– 1.14 [– 1.68; – 0.59]	< 0.0001
+ ∆ Diastolic blood pressure	– 1.16 [– 1.71; – 0.61]	< 0.0001
+ ∆ HbA1c	– 1.17 [– 1.74; – 0.60]	0.0001
+ ∆ NT-proBNP	– 1.19 [– 1.74; – 0.64]	< 0.0001
+ ∆ Uric acid	– 1.19 [– 1.75; – 0.63]	< 0.0001
Full modelb	– 0.77 [– 1.52; – 0.02]	0.046

The table displays linear regression models with E/e´ ratio after 12 weeks of intervention as dependent variable. The beta-estimate provided illustrates the effect of empagliflozin 10 mg/day compared to placebo

In order to decipher the mediation of the effect of empagliflozin on diastolic function via distinct pathways, estimates for empagliflozin versus placebo have been first adjusted for the delta in the following variables between baseline and 12 weeks of intervention in separate models: body mass index (BMI), systolic blood pressure (SBP), diastolic blood pressure (DBP), renal function as estimated by CKD-EPI formula as estimated glomerular filtration rate (eGFR), HbA1c, NT-proBNP, uric acid, hematocrit and hemoglobin

BNP brain natriuretic peptide

^aThe basic model is adjusted for age, sex, and E/e´ ratio at baseline. In the case of effect modification introduced by a single covariate or the sum of covariates, a reduction of the estimate of empagliflozin 10 mg/day versus placebo would be expected

^bThe joint contribution of all factors to the impact of empagliflozin on E/e´ ratio has been evaluated in a full model including all parameters as independent predictors