BACKGROUND The anaerobic bacteremia incidence is decreasing during the last years, and although it carries a high mortality rate there are studies that ask about the possibility of avoid anaerobic blood cultures thinking about the clinical prediction capacity of them. The objective of this study is the analysis of incidence and clinical characteristics of anaerobic bacteremias in two periods with 10 years of difference and empiric antibiotic treatment received, and if it was changed when microbiological results were received . METHODS Prospective study of anaerobic bacteremias during 1985-86 and 1996-97, at university hospital analyzing clinical characteristics, incidence, analysis, evolution and empiric antibiotic treatment received. The statistical analysis was performed with the chi square test or exact Fisher test with statistical significance with p<0.05, talking about variables with p<0.10. RESULTS The incidence of anaerobic bacteremias was higher in 1996-97 with 24 cases (5.08%), that during 1985-86 with 22 cases (4.24%). The variables with statistical significative differences with p<0.10 were: intrahospitalary adquisition (p<0.10); genitourinary and vascular manipulation (p=0.02 and p=0.06), and hypotension (p=0.034) more frequent during the first period than during the second one. There isn t statistical difference in evolution to cure although the percentage was higher during the second period (62.5%), that during the first one (54.6%), being the empiric treatment very high in both periods. Bacteroides fragilis was the most frequent microorganism and abdominal origin the most frequent one in both periods. CONCLUSION Although there is a low anaerobic bacteremia incidence this one hasn't diminished in ten years. Anaerobic bacteremias have a high mortality index although the most part of empirical antibiotic treatments are correct. It could ask about the possibility of avoid anaerobic blood cultures or ask about them only in clinic suspicion of them.