We compared agreement between creatinine clearance values in obese, critically ill patients calculated using three common empirically derived formulas and modifications thereof, with creatinine clearance obtained by conventional 24-hour urine collection. We selected the charts of 22 patients in intensive care units (86% medical, 14% surgical) according to the following criteria: actual body weight greater than 150% of ideal body weight; serum creatinine variation of less than 15% from the day of starting 24-hour urine collection to the day before or after the collection; presence of a urinary bladder catheter; no history of renal dialysis; and clinical indication for renal function assessment. Mean measured 24-hour urinary creatinine clearance for all patients was 72 +/- 64 ml/minute (range 8-248 ml/min). The method of estimating creatinine clearance that showed the least mean bias was the equation of Salazar and Corcoran using a corrected serum creatinine concentration (mean bias -2 ml/min); however, the corresponding 95% confidence intervals were wide (-133-129 ml/min). The narrowest range of 95% confidence intervals were seen with Jelliffe's equation (mean bias 25 ml/min, 95% confidence intervals -41-90 ml/min). In this sample, estimated creatinine clearances did not agree acceptably with measured values. Despite low mean bias values, none of the empirically derived equations that we studied had clinically acceptable 95% confidence intervals. We recommend using the 24-hour urine collection method when assessing creatinine clearance in obese, critically ill patients.