Population Pharmacokinetics of Digoxin in Nonagenarian Patients: Optimization of the Dosing Regimen.

Abstract

Objective: The aim of this study was to develop a population pharmacokinetic model of digoxin in patients over 90 years old and to propose an equation for adjusting digoxin dose in this population.

Methods: We included 326 nonagenarian patients admitted to Severo Ochoa University Hospital (Spain) who received digoxin and were under therapeutic drug monitoring. All data were retrospectively collected, and population modeling was performed with non-linear mixed-effect modeling software (NONMEM^®). One- and two-compartment models were tested to calculate digoxin clearance (Cl), volume of distribution (V_d), absorption rate constant (K_a), and bioavailability (bioavailable fraction, F). The covariates were evaluated by stepwise covariate model building, and the final model was internally validated by bootstrap analysis with 1000 resamples. External validation was performed with another population of 95 patients with the same characteristics as the modeling group.

Results: The population was 26% males, with a mean age of 93.2 years (90-103 years), mean creatinine 1.11 mg/dL (0.42-3.81 mg/dL), and mean total body weight 61.2 kg (40-100 kg). The pharmacokinetics of digoxin were best described by a one-compartment model (ADVAN2 TRANS2), with first-order conditional estimation with interaction. The covariates with influence on our model were creatinine clearance based on the Cockcroft-Gault equation (CG), serum potassium (K), co-administration of loop diuretics, and sex: Cl/F = 4.55 · (CG/36.4)^0.468 · 0.83^LD · 1.21^SEX; V_d/F = 355 · (K/4.3)^-0.849; K_a = 1.22 h^-1 [where LD indicates loop diuretics (1 for administered, 0 for otherwise) and SEX indicates patient sex (1 for male, 0 for female)]. Based on our results, we proposed an equation to adjust the digoxin dosing regimen in nonagenarian patients: dose (mg) = 0.144 · (CG/36.4)^0.468 · 0.83^LD · 1.21^SEX.

Conclusions: The greatest influence on digoxin clearance came from renal function calculated by the Cockcroft-Gault equation. V_d was decreased by K. The model developed showed a precise predictive performance to be applied for therapeutic drug monitoring.