Digoxin associates with mortality in ESRD

Abstract

The safety of prescribing digoxin in ESRD is unknown. Hypokalemia, which frequently occurs among dialysis patients, may enhance the toxicity of digoxin. Here, we analyzed the association between digoxin prescription and survival in a retrospective cohort using covariate- and propensity score-adjusted Cox models to minimize the potential for confounding by indication. Among 120,864 incident hemodialysis patients, digoxin use associated with a 28% increased risk for death (hazard ratio [HR] 1.28; 95% confidence interval 1.25 to 1.31). Increasing serum digoxin level was also significantly associated with mortality (HR 1.19 per ng/ml increase; 95% confidence interval 1.05 to 1.35). This increased mortality risk with level was most pronounced in patients with lower predialysis serum potassium (K) levels (HR 2.53 [P = 0.01] for K <4.3 mEq/L versus HR 0.86 [P = 0.35] for K >4.6 mEq/L). In conclusion, digoxin use among patients who are on hemodialysis associates with increased mortality, especially among those with low predialysis K concentrations.

Figure 1.

Crude survival curves show decreased survival with digoxin use. Digoxin users had a statistically significant increased mortality risk when compared with patients who were not prescribed digoxin (P < 0.0001). Among digoxin users, higher serum digoxin levels in the first 90 days of long-term HD were associated with an increased risk for death (P = 0.01, <0.6 versus 0.6 to 0.9 versus 1.0 to 2.2 ng/ml groups).

Figure 2.

Mortality HRs per 1-ng/ml increase in serum digoxin level stratified by patient characteristics demonstrated a persistent mortality trend in digoxin users. A statistically significant interaction effect existed between K and digoxin level. Models were covariate and propensity score adjusted. ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker.

Figure 3.

The mortality effect associated with a higher serum digoxin level is magnified with decreasing serum K level. Mortality HRs were covariate and propensity score adjusted.