To identify the current etiologies of blood culture-negative infective endocarditis and to describe the epidemiologic, clinical, laboratory, and echocardiographic characteristics associated with each etiology, as well as with unexplained cases, we tested samples from 348 patients suspected of having blood culture-negative infective endocarditis in our diagnostic center, the French National Reference Center for Rickettsial Diseases, between 1983 and 2001. Serology tests for Coxiella burnettii, Bartonella species, Chlamydia species, Legionella species, and Aspergillus species; blood culture on shell vial; and, when available, analysis of valve specimens through culture, microscopic examination, and direct PCR amplification were performed. Physicians were asked to complete a questionnaire, which was computerized. Only cases of definite infective endocarditis, as defined by the modified Duke criteria, were included. A total of 348 cases were recorded-to our knowledge, the largest series reported to date. Of those, 167 cases (48%) were associated with C. burnetii, 99 (28%) with Bartonella species, and 5 (1%) with rare, fastidious bacterial agents of endocarditis (Tropheryma whipplei, Abiotrophia elegans, Mycoplasma hominis, Legionella pneumophila). Among 73 cases without etiology, 58 received antibiotic drugs before the blood cultures. Six cases were right-sided endocarditis and 4 occurred in patients who had a permanent pacemaker. Finally, no explanatory factor was found for 5 remaining cases (1%), despite all investigations.Q fever endocarditis affected males in 75% of cases, between 40 and 70 years of age. Ninety-one percent of patients had a previous valvulopathy, 32% were immunocompromised, and 70% had been exposed to animals. Our study confirms the improved clinical presentation and prognosis of the disease observed during the last decades. Such an evolution could be related to earlier diagnosis due to better physician awareness and more sensitive diagnostic techniques. As for Bartonella species, B. quintana was recorded more frequently than B. henselae (53 vs 17 cases). For 18 patients with Bartonella endocarditis, the responsible species was not identified. Species determination was achieved through culture and/or PCR in 49 cases and through Western immunoblotting in 22. Comparison of B. quintana and B. henselae endocarditis revealed distinct epidemiologic patterns. The 2 cases due to T. whipplei reflect the emerging role of this agent as a cause of infective endocarditis. Because identification of the bacterium was possible only through analysis of excised valves by histologic examination, PCR, and culture on shell vial, the prevalence of the disease might be underestimated. Among patients who received antibiotic drugs before blood cultures, 4 cases (7%) were found to be associated with Streptococcus species (2 S. bovis and 2 S. mutans) through 16S rDNA gene amplification directly from the valve, which shows the usefulness of this technique in overcoming the limitations of previous antibiotic treatment. Right-sided endocarditis occurred classically in young patients (mean age, 36 yr), intravenous drug users in 50% of cases, and suffering more often from embolic complications. Finally, 5 cases without etiology or explaining factors were all immunocompetent male patients with previous aortic valvular lesions, and 3 of the 5 presented with an aortic abscess. Further investigations should be focused on this group to identify new agents of infective endocarditis.