This investigation compares the accuracy of calculating gentamicin pharmacokinetic parameters by a noninvasive body composition technique (bioelectrical impedance analysis; BIA) with an empiric method, against the two-point method as the criterion standard. A prospective concurrent open label design was used. The 32 medical and surgical intensive care unit beds at Henry Ford Hospital, a not-for-profit, university-affiliated teaching hospital, served as the setting. Twenty critical ill adults, Therapeutic Index Scoring System (TISS) = 4, who required gentamicin as part of their normal course of therapy for gram-negative bacillary infections, were evaluated. Gentamicin Vd and k were calculated by three methods. After measurement of body composition parameters by BIA, previously derived gentamicin dosing equations were used to predict gentamicin volume of distribution (Vd) and elimination rate constant (k) (BIA method). Empiric estimates of these parameters (Vd = 0.3L/kg and k derived from creatinine clearance) were compared with the BIA parameters against a criterion standard Vd and k determined from a two-point sampling of gentamicin serum concentrations. Measurements of BIA parameters and gentamicin serum concentrations were made in duplicate with coefficients of variation, < or = 2% and < or = 3%, respectively. The BIA and empiric methods produced resultant pharmacokinetic parameters (Vd and k) not different than those measured by the two-point method. There were no statistically significant differences in mean error (bias), or mean squared error (precision) for both Vd and k assessed by the empiric or BIA methods.(ABSTRACT TRUNCATED AT 250 WORDS)