Assessment of renal function prior to cisplatin chemotherapy has long been based on measurement of creatinine clearance by 24-hour urine collection (CrCmeas). Estimated creatinine clearance (CrCest) as calculated from the patient's age, weight, and serum creatinine level has been suggested as an adequate surrogate for CrCmeas, as it provides advantages of improved convenience, decreased cost, and possibly increased accuracy. We studied 847 patients receiving cisplatin-based chemotherapy on Cancer and Leukemia Group B (CALGB) protocols to determine whether the CrCmeas, CrCest, or serum creatinine value or the age of the patient would predict the subsequent genitourinary (GU) toxicity. Both CrCmeas (P = 0.001) and CrCest (P = 0.02) were predictive of subsequent grade 2+ GU toxicity, with CrCmeas being a slightly better predictor. Patient age also influenced subsequent GU toxicity, with the risk increasing with age (P = 0.0008). When patients were classified by age group and by CrCmeas, distinct subgroups were identified, with differences in the risk for grade 2+ GU toxicity ranging from 14% to 32%. Using a logistic model to assess the probability of grade 2+ GU toxicity, we found that an age of greater than or equal to 60 years (P = 0.005), a CrCmeas value of less than 75 ml/min (P = 0.004), and the risk characteristics of the individual cisplatin trial were important, whereas CrCest was not. Furthermore, CrCest proved to be a poor predictor of a CrCmeas value of less than 75 ml/min, "misclassifying" nearly half of the patients to a "lower-risk" subgroup. In summary, both CrCmeas and the patient's age independently provided predictive information concerning cisplatin GU toxicity. Our data support the continued clinical usefulness of determining the CrCmeas value prior to the administration of cisplatin-based chemotherapy to most patients.