Objectives: To evaluate the use of multifrequency bioelectrical impedance analysis to predict creatinine/urea clearance based on 24 hours urine collection. A practical formula was developed, and its performance was compared with that of established formulas such as Cockcroft-Gault, Modification of Diet in Renal Disease, and Jelliffe’s. Design: An open-label prospective observational cohort study. Setting: A 12-bed ICU at a nonuniversity major teaching hospital (Gelre ziekenhuizen Apeldoorn/Zutphen, The Netherlands). Patients: Adult critical care patients with an expected ICU length of stay at admission of at least 48 hours. Interventions: Each patient’s body composition was assessed using a validated Quadscan 4000 analyzer (Bodystat, Isle of Man, British Isles). Twenty-four hours urine was collected, and laboratory variables in serum including creatinine, urea, and albumin were obtained at the beginning and end of the collection period. Measurements and Main Results: A total of 151 patients, stratified to an acute and nonacute ICU-group, were enrolled in the study over a 2-year period. A formula to predict creatinine/urea clearance based on 24 hours urine collection was developed using stepwise linear regression using a training data set of 75 patients. This formula was subsequently tested and compared with other relevant predictive equations using a validation data set of 76 patients. Serum creatinine values ranged from 40 to 446 µmol/L. With the predictive model based on estimated body cell mass and a “prediction marker” more than 71% of the observed variance in creatinine/urea clearance based on 24 hours urine collection could be explained. Predictive performance was superior to the other eight evaluated models (R 2 = 0.39–0.55) and demonstrated to be constant over the whole range of creatinine/urea clearance based on 24 hours urine collection values. Conclusions: Multifrequency bioelectrical impedance analysis measurements can be used to predict creatinine/urea clearance based on 24 hours urine collection with superior performance than currently established prediction models. This rapid, noninvasive method enables correction for influences of a patient’s actual body composition and may prove valuable in daily clinical practice.