We previously correlated both renal function and thrombocytopenia, the dose limiting toxicity of carboplatin, with the plasma pharmacokinetics of carboplatin. From these correlations, we developed equations to calculate carboplatin dosage for any patient based on that patient's creatinine clearance, body surface area, pretreatment platelet count, desired platelet nadir, and status of prior chemotherapy. We prospectively applied these equations in 44 courses of carboplatin given to 24 patients. There were 13 males and 11 females with median age 53 (range, 33-77), median Karnofsky performance status 80 (range, 50-100), and creatinine clearance 32 to 118 ml/min. Ten patients had creatinine clearances less than 60 ml/min. Precision of the equations used for dose calculation was evaluable in 38 courses administered to 23 patients. In 23 courses of carboplatin administered to 12 patients without extensive prior chemotherapy, the observed change in platelets = 1.04 X predicted change -48,000 (r = 0.96). In the 15 courses of carboplatin administered to 11 heavily pretreated patients, the observed change in platelets = 1.13 X predicted change +6,600 (r = 0.97). For the overall combined population, the observed change in platelets = 0.96 X predicted change -7,000 (r = 0.94). These relationships which nearly define the line of identity (observed = expected) validate our initial observations. Only 2 patients developed WBC less than 2,000, but 12 patients developed hematocrit less than or equal to 29% and 8 required RBC transfusions. Fifteen patients had nausea and vomiting greater than or equal to grade 2. There were no other nonhematological toxicities observed. In view of continuing documentation of the antitumor activity of carboplatin, these equations allow safe and rational drug dosing of patients with potentially platinum-responsive tumors but with renal function too poor to receive cisplatin. Among the 9 patients in this study evaluable for response, there was 1 partial response in a patient with malignant melanoma and 1 objective response (less than partial response) in a patient with adenocarcinoma of the cervix.