Eplerenone and New-Onset Diabetes in Mild Heart Failure
Eplerenone and New-Onset Diabetes in Mild Heart Failure
Of the 2737 patients randomized in EMPHASIS-HF, 891 (32.5%) were known to have diabetes at baseline. Analyses in this report are limited to the 1846 patients without known diabetes at baseline. The average age was 69 years and average BMI was 27 kg/m. During the 21 months of follow-up, 69 (3.7%) of the 1846 patients without diabetes at baseline were diagnosed with diabetes, an approximate incidence of 21 cases per 1000 patient-years.
The baseline characteristics of patients without diabetes at baseline were balanced between randomized treatment arms (Supplementary material online, Table S1). Eplerenone had no effect on the incidence of new-onset diabetes; 33 (3.7%) of 894 patients assigned to treatment with eplerenone developed diabetes compared with 36 (3.8%) of 952 patients assigned placebo (Figure 1). This corresponded to a HR of 0.94 [95% confidence interval (CI) 0.59–1.52, P = 0.81]. The eplerenone:placebo HR for the composite of all-cause mortality or new-onset diabetes was 0.80 (95% CI 0.64–1.01); P = 0.06 (Figure 2).
(Enlarge Image)
Figure 1.
Kaplan–Meier plot for new-onset diabetes on eplerenone compared with placebo in EMPHASIS-HF.
(Enlarge Image)
Figure 2.
Kaplan–Meier plot for the composite endpoint of new-onset diabetes or mortality on eplerenone compared with placebo in EMPHASIS-HF.
The baseline characteristics of patients who did and did not develop new diabetes are shown in Table 1. In univariate Cox proportional hazard analysis, the strongest predictors of new-onset diabetes (in decreasing order of magnitude as estimated by the Wald χ statistic) were: higher serum alanine aminotransferase (ALT), use of digitalis, higher serum aspartate aminotransferase (AST), higher waist circumference, higher weight, higher BMI, longer duration of heart failure, current or previous smoking, higher heart rate, hypertension, higher systolic blood pressure, lower age, higher diastolic blood pressure, higher serum potassium, not using a beta-blocker. and the presence of atrial fibrillation (see Table 2).
In multivariable Cox proportional hazard analysis, Model 1 [full model where only variables with P < 0.10 in univariate analysis were included but for the following groups of highly correlated variables only the variable that provided the most information was included: (i) systolic and diastolic blood pressure: systolic blood pressure selected; (ii) weight, waist circumference, and BMI: waist circumference selected; (iii) ALT and AST: ALT selected], only seven variables remained significantly associated with new-onset diabetes (see Table 3). These independent predictors of new-onset diabetes (in decreasing order of magnitude as estimated by the Wald χ statistic) were: use of digitalis, higher serum ALT, longer heart failure duration, current or previous smoking, higher waist circumference, lower age, and higher systolic blood pressure. Variables identified in Model 2 (the reduced model) were similar.
In the ROC curve analysis, the individual c-statistic for variables independently associated with new-onset diabetes varied from 0.53 to 0.65 in Models 1 and 2. When all the relevant variables were combined, the overall c-statistic was 0.74 for Model 1 and 0.76 for Model 2 (see Table 4).
Results
Of the 2737 patients randomized in EMPHASIS-HF, 891 (32.5%) were known to have diabetes at baseline. Analyses in this report are limited to the 1846 patients without known diabetes at baseline. The average age was 69 years and average BMI was 27 kg/m. During the 21 months of follow-up, 69 (3.7%) of the 1846 patients without diabetes at baseline were diagnosed with diabetes, an approximate incidence of 21 cases per 1000 patient-years.
Effect of Eplerenone on New-onset Diabetes
The baseline characteristics of patients without diabetes at baseline were balanced between randomized treatment arms (Supplementary material online, Table S1). Eplerenone had no effect on the incidence of new-onset diabetes; 33 (3.7%) of 894 patients assigned to treatment with eplerenone developed diabetes compared with 36 (3.8%) of 952 patients assigned placebo (Figure 1). This corresponded to a HR of 0.94 [95% confidence interval (CI) 0.59–1.52, P = 0.81]. The eplerenone:placebo HR for the composite of all-cause mortality or new-onset diabetes was 0.80 (95% CI 0.64–1.01); P = 0.06 (Figure 2).
(Enlarge Image)
Figure 1.
Kaplan–Meier plot for new-onset diabetes on eplerenone compared with placebo in EMPHASIS-HF.
(Enlarge Image)
Figure 2.
Kaplan–Meier plot for the composite endpoint of new-onset diabetes or mortality on eplerenone compared with placebo in EMPHASIS-HF.
Predictors of New-onset Diabetes in Mild Heart Failure
The baseline characteristics of patients who did and did not develop new diabetes are shown in Table 1. In univariate Cox proportional hazard analysis, the strongest predictors of new-onset diabetes (in decreasing order of magnitude as estimated by the Wald χ statistic) were: higher serum alanine aminotransferase (ALT), use of digitalis, higher serum aspartate aminotransferase (AST), higher waist circumference, higher weight, higher BMI, longer duration of heart failure, current or previous smoking, higher heart rate, hypertension, higher systolic blood pressure, lower age, higher diastolic blood pressure, higher serum potassium, not using a beta-blocker. and the presence of atrial fibrillation (see Table 2).
In multivariable Cox proportional hazard analysis, Model 1 [full model where only variables with P < 0.10 in univariate analysis were included but for the following groups of highly correlated variables only the variable that provided the most information was included: (i) systolic and diastolic blood pressure: systolic blood pressure selected; (ii) weight, waist circumference, and BMI: waist circumference selected; (iii) ALT and AST: ALT selected], only seven variables remained significantly associated with new-onset diabetes (see Table 3). These independent predictors of new-onset diabetes (in decreasing order of magnitude as estimated by the Wald χ statistic) were: use of digitalis, higher serum ALT, longer heart failure duration, current or previous smoking, higher waist circumference, lower age, and higher systolic blood pressure. Variables identified in Model 2 (the reduced model) were similar.
In the ROC curve analysis, the individual c-statistic for variables independently associated with new-onset diabetes varied from 0.53 to 0.65 in Models 1 and 2. When all the relevant variables were combined, the overall c-statistic was 0.74 for Model 1 and 0.76 for Model 2 (see Table 4).
Source...