Seven methods using patient's age, weight, gender, and serum creatinine concentration were evaluated for mathematical and clinical predictive performance in estimating the creatinine clearance of Japanese patients. We compared the measured and estimated creatinine clearance in 179 hospitalized heart failure patients. The Gates equation was the most biased (mean prediction error, 17.9 mL/min), and the Yukawa equation was the most precise (mean absolute prediction error, 10.4 mL/min), followed by the Jelliffe, Bjornsson, Cockcroft and Gault, and Mawer equations. For all methods, precision was less when serum creatinine concentrations of <0.6 mg/dL were rounded than when serum creatinine concentrations of >/=0.6 mg/dL were rounded. We conclude that in clinical practice where the renal function can vary widely, any of the published formulas can give satisfactory results. The Yukawa formula is concise and gives better results in Japanese patients.