The accuracy of creatinine clearance (CLcr) estimates generated for elderly patients with a low serum creatinine (SCr) concentration when the SCr concentration was rounded to 1.0 mg/dL was evaluated, and daily aminoglycoside dosages calculated with CLcr estimates based on the actual and the rounded SCr concentrations were compared with the actual dosage required as determined by aminoglycoside concentration measurements. Twenty-four-hour urine collections were obtained from elderly hospitalized patients with stable renal function, and CLcr was measured. SCr concentrations were measured at baseline, during urine collection, and after urine collection. CLcr was calculated twice with each of seven methods, once by using the actual SCr concentration and once by rounding the SCr concentration to 1.0 mg/dL. Rounding the SCr concentration to 1.0 mg/dL resulted in significant underestimation of CLcr by all seven methods. The Cockcroft-Gault equation used with ideal body weight and the rounded SCr concentration resulted in the largest bias. For six of the seven methods, bias was greater and precision less when SCr concentrations of < 0.7 mg/dL were rounded than when SCr concentrations of > or = 0.7 mg/dL were rounded. The Bjornsson method used with the actual SCr concentration appeared to offer the best combination of low bias and high precision. The required aminoglycoside dosage was significantly underestimated when the rounded SCr concentration was used. In elderly patients with low SCr concentrations, rounding the SCr concentration to 1.0 mg/dL led to underestimates of both CLcr and the correct aminoglycoside dosage.