25 patients with cancer of unknown primary (CUP) were studied using PET-FDG. Whole-body-PET scans were performed based on previous morphological information. All patients underwent conventional diagnostic modalities prior to and after the PET-study. True positive results were obtained in 12/25 patients (48%) and the primary identified by PET was confirmed by further procedures. True negative results were obtained in 8/25 patients (32%) and no primary tumor was found by PET or other clinical investigations. False positive results were obtained in 5/25 patients (20%). No false negative results were obtained. In 11 patients having CT compared with PET the primary tumor was exclusively localized by PET in 8 patients (73%) and also by CT in 3 patients (27%); CT was unable to identify the tumor in 8/11 patients detected by PET. PET affected management and directed treatment in 11/25 patients (44%). The sensitivity of PET-FDG was 100%, specificity 61%. These observations demonstrate that PET-FDG is a highly sensitive method to detect CUP. Correctly classified patients in an early stage of disease may lead to benefit from a more targeted therapy with prolonged survival time. In cases of advanced disease PET might not decisively influence survival time.