The applicability of a kinetic model for the prediction of steady-state blood levels, based on body composition as assessed by bioelectric impedance analysis (BIA), was applied to a population of elderly patients, candidates for digoxin therapy. Elderlies, all >70 years of age, underwent standard laboratory and clinical evaluation but no further characterization of liver or renal function. These 72 patients were given 0.125 mg digoxin for 5 days, in order to reach steady-state levels. Treatment was then interrupted and samples were collected 2 and 48 h after the last administration. Plasma digoxin levels were determined both by the immunochemical method with TDX and according to the BIA method described in the accompanying paper. Plasma levels calculated and measured in 2 h samples did not differ statistically, but levels were about 15% higher in the directly measured samples. There was a similar underestimation, i.e. about 15%, for the 48 h calculated levels. However, only approximately 5% of the levels were outside of the 95% confidence intervals as determined from the directly measured levels. These findings indicate that digoxin levels, calculated based on a BIA evaluation, may be sufficiently reliable, in the majority of patients, to allow direct determination of the more appropriate doses of digoxin.