PURPOSE The study was undertaken to define the relationship between tumor response and carboplatin area under the curve (AUC) in patients with ovarian cancer; to study the relationship between carboplatin AUC and myelosuppression in the same population; to establish the true impact of carboplatin AUC, prior therapy, and pretreatment platelet and WBC counts on toxicity; and to define an optimal carboplatin exposure for treating patients with ovarian cancer. METHODS With the equation AUC = dose/(glomerular filtration rate [GFR]+25), carboplatin AUC (course 1) was calculated for 1,028 patients (450 previously untreated) who received single-agent carboplatin (40 to 1,000 mg/m2) for advanced ovarian cancer. GFR was measured (chromium-51-edathamil [51Cr-EDTA] or creatinine clearance) in all patients. RESULTS Regression analysis showed that carboplatin AUC, prior treatment, and Eastern Cooperative Oncology Group grade performance status (PS) are predictors of tumor response, thrombocytopenia, and leukopenia. Pretreatment platelet and WBC counts are additional predictors of thrombocytopenia and leukopenia, respectively. Although the likelihood of tumor response increased with increasing carboplatin AUC, this relationship was nonlinear. In all patient subsets, the likelihood of complete response (CR) or overall response did not increase significantly above a carboplatin AUC of 5 to 7 mg/mL x minutes. At any given carboplatin AUC, thrombocytopenia occurred more frequently than leukopenia, although both approached 100% as carboplatin AUC increased. Both thrombocytopenia and leukopenia were more frequent in pretreated than in untreated patients regardless of pretreatment count. At any carboplatin AUC, the influence of PS on likelihood of response and toxicity was profound. CONCLUSION Carboplatin dosing by AUC will lead to more predictable toxicity, and increasing carboplatin AUC above 5 to 7 mg/mL x minutes does not improve the likelihood of response but does increase myelotoxicity. Therefore, careful evaluation of high-dose carboplatin therapy in a prospective, randomized trial is needed before such treatment becomes accepted practice.