A prospective clinical trial was performed to evaluate the usefulness of a human tumor cloning system for selecting single-agent chemotherapy for patients with advanced cancers. Six hundred four single-agent trials were performed in the 470 patients whose tumors were submitted for drug sensitivity testing. Only 246 of these 604 trials (41%) could be directed by the cloning system results because of inadequate tumor growth and other difficulties. In these 246 prospective trials, there was a 60% true positive and an 85% true negative rate for predicting for response or lack of response of an individual patient's tumor to the single agent. There was also a relationship between the percentage of decrease in survival of tumor colony-forming units and the probability of a clinical response of the patient's tumor to the same drug used in vivo. Despite these encouraging findings, work to improve tumor growth and additional prospective clinical trials of the system are needed before the system can be recommended for routine clinical use.