From January 1992 until January 1994, we used a standardized diagnostic protocol for the 167 immunocompetent patients with fever of unknown origin (FUO) admitted on the internal medicine wards in all 8 university hospitals in the Netherlands. This protocol consisted of a standardized coded history and standardized physical examination for all 167 patients. A number of additional obligatory investigations had to be performed in the first week of admission for all patients, and all potentially diagnostic clues (PDCs) thus retrieved had to be registered. In the presence of PDCs, specific investigations had to be performed based on the differential diagnosis. In the absence of PDCs or in the presence of only misleading PDCs, patients underwent a screening 2-staged diagnostic protocol. In 162 (97%) patients, PDCs were present after 1 week of admission. In 61 patients these PDCs were all misleading. The likelihood of reaching a diagnosis in patients with PDCs was not significantly higher than that in patients without PDCs, probably because of the high proportion of misleading PDCs. The likelihood of establishing a diagnosis was significantly lower (< 10%) only for patients with recurrent fever, normal erythrocyte sedimentation rate (ESR), and normal hemoglobin. All other PDCs were not significantly different in patients with a diagnosis compared with patients without a diagnosis. The screening 2-staged diagnostic protocol proved useful in 10 of 43 patients in whom it was used. The screening value of immunologic and microbiologic serology and endocrine investigations was nil; these investigations probably should be performed only when PDCs for the disease searched for are present. Scintigraphic techniques, echocardiography, and other imaging procedures were never helpful in our population in the absence of PDCs. Many patients with FUO had nonspecific anemia and disturbed liver chemistry. In the presence of these findings alone, without other more specific PDCs, the likelihood of reaching a diagnosis with help of bone marrow aspiration was nil, and with help of liver biopsy, it was low. Enteric biopsy was never helpful. If lymphadenopathy was confined to the cervical or inguinal region (with negative chest X-ray and abdominal ultrasound), lymph node biopsy was not helpful, in contrast to patients having generalized lymphadenopathy, in whom the technique had a yield of 79%. As shown in this study, the search for PDCs remains an important tool for establishing the diagnosis in patients with FUO, although in many cases these PDCs appear to be misleading. Directed diagnostic workup--using the PDCs retrieved by repeated, meticulous history taking and physical examination--remains the most efficient and intellectually satisfactory way to solve the problem of FUO in the individual patient. A standard protocol in patients with FUO in whom the obligatory investigations, as used by us, do not lead to the diagnosis can be limited to the tests that proved to be of some use as screening procedure: temporal biopsy in patients older than 55 years; fundoscopy; serology (Western blot) for Yersinia enterocolitica; serum for cryoglobulin at an early stage of the diagnostic process; and bone biopsy, liver biopsy, abdominal computed tomography (CT), and chest CT at a later stage. Repeating a thorough history-taking, physical examination, and obligatory investigations and waiting for PDCs to appear probably is better than ordering more screening investigations in the hope that something abnormal will come up. Supportive treatment with nonsteroidal anti-inflammatory drugs (NSAIDs) can be helpful at this stage. Only rarely do patients deteriorate while using NSAIDs without presenting new PDCs. In these rare patients, further diagnostic workup should be performed or a therapeutic trial with, for example, antibiotics, steroids, or antituberculous agents started.