To be cost-effective, PET must be diagnostically accurate and effective in improving management without increasing treatment cost. To evaluate diagnostic accuracy, we performed prospective evaluations of whole-body PET imaging in staging of non-small-cell lung cancer (99 patients), detection of recurrent colorectal cancer (57 patients), diagnosis of metastatic melanoma (36 patients), and staging of advanced head and neck cancer (29 patients). In each case, PET was more accurate than anatomic imaging for determination of the presence and extent of tumor and demonstration of nonresectable disease. PET was also more accurate than conventional imaging in staging Hodgkin's disease (30 patients). We evaluated the management impact of PET retrospectively, by reviewing the treatment records of 72 patients with solitary pulmonary nodules or non-small-cell lung cancer, 68 patients with known or suspected recurrent colorectal cancer, 45 patients with known or suspected metastatic melanoma, and 29 patients with advanced head and neck tumors. PET improved patient management by avoiding surgery for nonresectable tumor and for CT abnormalities that proved to be benign by PET imaging. For determining cost impact, the costs of surgical procedures were determined from Medicare reimbursement rates, and the cost of a PET study was taken to be $1800. The savings from contraindicated surgical procedures exceeded the cost of PET imaging by ratios of 2:1 to 4:1, depending on the indication. PET was decisively more accurate and cost-effective than anatomic imaging by CT, combining improved patient care with reduced cost of management.