[Diagnosis of fever of unknown origin: use of Bayes theorem]

Sophisticated tests and time are needed to establish the diagnosis of fever of unknown origin (FUO). Bayes theorem can guide the clinician to early probabilistic diagnosis. The authors had estimated the a priori probabilities in a previous study yet. In this paper the diagnostic probabilities of four clinical signs (arthralgias, myalgias, splenomegaly, multiple micro-lymphoadenopathies) and the probabilities of abnormalities of some routine laboratory tests (neutrophils count, lymphocytes count, platelets count, hemoglobin, lactic-dehydrogenase, C3 and C4 complement subunits and fibrinogen plasma concentrations, erythrocyte sedimentation rate) and of the positive bacterial cultures, were estimated in the main groups of FUO disease (indefinite, aspecific bacteria infections, non-bacterial or specific bacteria infections, neoplasias, connective tissue diseases, miscellaneous group). The data suggest that the a priori probability is not affected by the clinical signs. Arthralgias increase the probability of connective tissue diseases. Normal erythrocyte sedimentation rate and plasma fibrinogen concentrations increase the probability of paraneoplastic and indefinite fever. Thrombocytosis increases the probability of paraneoplastic fever. Over 1200 U/l serum lactic-dehydrogenase serum concentrations guide to non-bacterial or specific bacteria infectious diseases and to malignancies. Neutrophilia decreases the probability of indefinite diagnosis, even though it is not specific. The heterogeneity of the miscellaneous group makes the interpretations of the data very hard.