One hundred fifty-four 24-hour urine collections from 16 patients with stable renal function from whom careful urine collections were made were used to compare the "error" in determining renal function (creatinine clearance) when it is estimated using measured 24-hour creatinine excretion versus when creatinine excretion is predicted from serum creatinine, sex, age, and weight, using a simple estimation scheme. These "errors" were comparable (the coefficients of variation were 10% and 13%, respectively). Thus, the prediction scheme is as acceptable as direct measurement of creatinine excretion for purposes of estimating creatinine clearance in stable patients, and is simpler to perform. Two hundred twenty-two 24-hour urine collections from 86 patients having urine collections performed in a routine fashion were also analyzed in this manner. The routinely assessed patients demonstrated markedly greater intraindividual variability (coefficient of variation, 27%) in measured creatinine excretion than did the carefully studied patients (10%). The authors suspect that the excess variability was due to inaccurate urine collection. If so, the prediction scheme may also be acceptable in the routine clinical setting.