Infective Endocarditis in Patients with Negative Blood Cultures: Analysis of 88 Cases from a One-Year Nationwide Survey in France

Transcription

1 501 CLINICAL ARTICLES Infective Endocarditis in Patients with Negative Blood Cultures: Analysis of 88 Cases from a One-Year Nationwide Survey in France Bruno Hoen, Christine Selton-Suty, Flore Lacassin, Jerome Etienne, Serge Briancon, Catherine Leport, and Philippe Canton From the Departments of Infectious and Tropical Diseases. Cardiology and Epidemiology, Centre Hospitalier Universitaire de Nancy, Vandoeuvre-les-Nancy; the Labomtoly of Bacteriology, HOPital Cardiologique, Lyon; and the Division of Infectious and Tropical Diseases, Centre Hospitalier Universitaire Bichat, Paris, France Blood cultures were negative in 88 (14%) of 620 cases of infective endocarditis (IE) documented in France during a 1-year nationwide survey. In 15 of these 88 cases, the causative microorganism was identified: seven cases of Q fever endocarditis and two cases of chlamydial endocarditis were diagnosed by serological and/or immunohistologic techniques, and a pathogen was cultured from five surgically removed valves and one arterial septic embolus. Forty-two (48%) of the 88 cases involved patients who had received antibiotics before the first blood sample was taken for culture. Mortality was lower in this group than among patients who had not previously received antibiotics (7% vs. 22%, P =.05). Comparison of blood culture negative cases of IE with blood culture positive cases revealed that the former tended to occur more often on prosthetic valves (32% vs. 22%, P =.16), were more often left-sided (97% vs. 83%, P =.0009), less often included extracardiac symptoms at presentation (52% vs. 63%, P =.06), and were more often surgically treated (53% vs. 34%, P =.001). Mortality was similar regardless of the results of blood culture (15% vs. 21%, P =.18). This study showed that more than 10% of all cases of IE in France are still associated with negative blood cultures and confirmed that a search for pathogens such as Coxiella burnetii and Chlamydia species is worthwhile in this situation. Cases of infective endocarditis (IE) that are associated with negative blood cultures and are not preceded by antibiotic treatment are presumably caused by nonbacterial organisms or fastidious slow-growing bacteria [ 1 ]. In large series, 2.5%-3 1% of cases of IE are blood culture negative [2, 3]. This wide range of proportions is attributable mainly to variations in the criteria used for the definition of IE and in the sensitivity of the bacteriologic techniques used for the detection of pathogens [3]. Rates of blood culture negativity were highest in the oldest series (cases collected between 1944 and 1966). In most recent series, fewer than 5% of all cases of IE have been associated with negative blood cultures [3-6]. However, in a recent epidemiological survey, we found that patients with negative blood cultures accounted for more than 10% of all cases of IE collected during a 1-year period [6a]. The purposes of the present study were to describe the Received 28 January 1994; revised 7 September This work was presented in part at the 32nd Interscience Conference on Antimicrobial Agents and Chemotherapy held on October 1992 in Anaheim, California. (abstract no. 3). Reprints or correspondence: Dr. Bruno Hoen, Service de Maladies Infectieuses et Tropicales, CHU de Nancy, F Vandoeuvre Cedex, France. Clinical Infectious Diseases 1995;20: by The University of Chicago. All rights reserved /95/ $02.00 epidemiological and clinical features of blood culture negative IE and to examine the current reasons for negative blood cultures in IE. Patients and Methods A nationwide survey on IE was conducted in France between 1 November 1990 and 31 October During this period information was collected on patients with IE at all public or private hospitals in France. These hospitals were located in three regions with a population of 18.5 million. Hospitals located outside these regions volunteered to participate in the study. All physicians and microbiologists involved in the care of patients with IE were asked to participate. Physicians were informed about and periodically reminded of the study by their professional societies (cardiology, cardiothoracic surgery, internal medicine, infectious diseases, and/or intensive care). They were asked to report each case of IE by calling a toll-free answering machine. A physician reporting a case was subsequently sent a questionnaire to fill out and return to the coordinating investigator after the patient's discharge from the hospital. The questionnaire included information on the patient's gender, date of birth, history of heart disease in general and of IE in particular, prosthetic valves, and current episode of IE (date of hospitalization, location of endocarditis, clinical signs and

2 502 Hoen et al. CID 1995;20 (March) Table 1. Criteria for the definition of infective endocarditis (IE), based on von Reyn's criteria with modifications to incorporate echocardiographic results and observations made macroscopically at the time of surgery. Definite IE Direct evidence of IE on macroscopic examination and/or histologic study of samples obtained at surgery or autopsy or on bacteriologic study (gram staining or culture) of valvular vegetation or peripheral embolus Probable IE A. Persistently positive blood cultures* plus one of the following: 1. New regurgitant murmur 2. Predisposing heart disease and vascular phenomena* 3. Predisposing heart disease and echocardiographicall). demonstrated vegetation 4. Vascular phenomena and echocardiographicallr demonstrated vegetation B. Negative or intermittently positive blood cultures* plus either of the following sets of findings: 1. Fever, new regurgitant murmur, and vascular phenomena 2. Fever, predisposing heart disease, vascular phenomena, and echocardiographicallr demonstrated vegetation Possible IE A. Persistently positive blood cultures plus one of the following: 1. Predisposing heart disease 2. Vascular phenomena B. Negative or intermittently positive blood cultures with all three of the following: 1. Fever 2. Predisposing heart disease 3. Vascular phenomena C. For viridans streptococcal cases only: at least two positive blood cultures without an extracardiac source plus fever Rejected IE A. Endocarditis unlikely (alternative diagnosis generally apparent) B. Endocarditis likely (empirical antibiotic therapy warranted) C. Culture-negative endocarditis diagnosed clinically but excluded by findings at surgery or postmortem examination NOTE. Table is adapted from [4]. Modifications of von Reyn's criteria are in italic type. * At least two blood cultures obtained; both of two, all of three, or at least 70% of four or more cultures positive. t Definite valvular or congenital heart disease or the presence of a prosthetic valve (other than a permanent pacemaker). * Petechiae; splinter hemorrhages; conjunctival hemorrhages; Roth's spots; Osler's nodes; Janeway lesions; aseptic meningitis; glomerulonephritis; and pulmonary, CNS, coronary, or peripheral emboli. Any rate of blood-culture positivity not meeting the definition for persistent positivity. symptoms, echocardiographic findings, number of positive blood cultures, presumed portal of entry of the pathogen, valve surgery, and final outcome). If data were missing, the information was requested again by the coordinating investigator. Once completed, the questionnaire was coded by two independent investigators and the data were entered in a computerized database by two independent operators. The two database files were then electronically matched, and discrepancies were resolved by a consensus of two investigators. Each case of IE was defined as definite, probable, possible, or rejected, according to von Reyn's criteria modified to incorporate echocardiographic results and observations made macroscopically at the time of surgery (table 1). After the exclusion of cases that had occurred outside the study period or that did not fulfill the criteria defining IE, 620 cases remained in the database. The figures for these cases, which are being published elsewhere [6a], indicated a rate of blood culture negative IE of nearly 15%. This higherthan-expected rate prompted us to review the charts of all patients from the whole database who had negative cultures. A case of IE was defined as blood culture negative when a minimum of two blood cultures had been performed and all blood cultures were negative. The additional information scrutinized during the chart review included the duration of symptoms before hospitalization, the history (if any) of injection drug use, the total number of blood cultures performed, the antibiotics (if any) administered before blood was drawn for culture, the number of blood cultures performed before the administration of any antibiotics, and the efforts made to identify fastidious organisms (e.g., Legionella, Chlamydia, and Mycoplasnia species and Coxiella burnetii). At least one blood culture was positive in 506 (82%) of 620 cases. All blood cultures were negative in 88 cases (14%). The remaining 26 cases (4%) were excluded either because no blood cultures had been performed or because it was impossible to ascertain that at least two blood cultures had been performed. The 88 cases of blood culture negative IE were characterized with regard to a variety of parameters and were compared with cases of blood culture positive IE. Values for quantitative variables were expressed as means ± standard deviations. Mann-Whitney U test, Pearson's x 2 test, and Fisher's exact test were used as appropriate. Results Characteristics of blood culture negative IE. The mean age (±SD) of patients with blood culture negative IE was 53 ± 17 years (range, 6-89 years). There were 66 males and 22 females. Twenty-eight patients (32%) had had a prosthetic valve for 4.8 ± 4.9 years (range, 0-15 years), 35 (40%) had native-valve disease, and 25 (28%) had no previously recognized cardiac disease. Ten (11%) of 88 patients had previously had IE. Three patients had a history of injection drug use; two were still injection drug users at the time of IE. When two cases diagnosed >1 year after the onset of symptoms were excluded from the calculations, the mean duration of symptoms before admission to the hospital was found to be 32 ± 40 days (range, days, with one patient admitted on the first day of acute cardiac failure and discovered to have endocarditis). High-grade fever (>39 C) was recorded for 13 patients, medium-grade fever for 57, and low-grade or no fever (<38 C) for 17. Data for one patient were not available. More than half of the patients (46 of 88,

3 CID 1995;20 (March) IE with Negative Blood Cultures 503 or 52%) presented with at least one extracardiac manifestation; 20 patients (23%) experienced an embolic accident. The great majority of cases were left-sided. The tricuspid valve was involved in only three cases, two of which were bilateral. None of these three cases involved injection drug users. Echocardiographic findings included an oscillating intracardiac mass in 62 cases, an intracardiac abscess in four, and a dehiscent prosthetic valve in six. In eight other cases echocardiography yielded findings consistent with IE but failed to reveal any of the three major abnormalities just mentioned; the diagnosis of IE was confirmed at surgery in all of six patients who underwent surgery. In the remaining eight cases, echocardiography either showed no evidence of IE (six cases) or was not performed (two cases). The diagnosis of IE was confirmed by surgery in all of three patients who underwent surgery; two patients had serological evidence of C. burnetii endocarditis, and one patient had a culture-positive surgically removed embolus. Eventually, the diagnosis of IE remained questionable in only four cases. The mean number of blood cultures performed was 9 ± 4 (range, 2-26) for all 88 patients. However, 42 patients (48%) had received antibiotics before blood was drawn for culture (either before or after admission to the hospital). Of these patients, 31 (74%) had either a prosthetic valve or known native-valve disease. Serological tests for C. burnetii had been conducted in 35 cases (40%). Although blood cultures remained negative, a bacterial etiology was suspected in 15 cases. A microorganism was cultured from a surgically removed valve in five cases and from an arterial septic embolus removed by percutaneous embolectomy in one case. The isolated pathogen was a coagulasenegative Staphylococcus in three cases, a Streptococcus species in one case, a Corynebacterium species in one, and Propionibacterium acnes in one. However, because a pathological examination was performed (and contributed to the diagnosis of IE) in only one of these six cases, contamination of the removed valve could not be ruled out. A specific bacteriologic diagnosis was made in nine other cases. In seven cases, patients' sera contained high titers of immunofluorescent IgG (>1,000) and IgA (>800) antibodies to phase I antigens of C. burnetii. Q fever endocarditis was diagnosed and treated in these instances. Three of the seven patients underwent surgery and had C. burnetii recovered from a valve culture. In the other two cases, chlamydial antigens were detected from the removed valve in immunofluorescent studies using genus-specific antibodies. Forty-seven patients (53%) underwent surgical valve replacement as part of the treatment of IE. A pathological and/ or bacteriologic examination of surgically removed valves was completed in 36 cases (77%). One or both of these examinations contributed to the diagnosis of IE in 18 cases: specifically, 12 cases were confirmed pathologically, five bacteriologically, and one by both techniques. In the other 18 cases, Table 2. Characteristics of 88 patients with blood culture negative infective endocarditis (IE), by prior antibiotic therapy status. No. (%) in indicated group* No prior Prior antibiotics antibiotics Characteristic (n = 46) (n = 42) Male gender Age, y (mean ± SD) Underlying cardiac condition No known valve disease Native-valve disease Prosthetic valve Diagnostic classification Definite Probable Possible Clinical symptoms Extracardiac symptoms Embolic event Echocardiographic findings 34 (74) 54.3 ± (30) 17 (37) 15 (33) 24 (52) 13 (28) 9 (20) 28 (61) 14 (30) 33 (79) ± (26) 18 (43).84 13(31) 24 (57) 1 10 (24).87 8 (19) 18 (43).09 6 (14).07 Major* 37 (80) 35 (83) Minorg 5 (11) 3 (7).83 None or not evaluated 4 (9) 4(10) Surgical treatment 23 (50) 24 (57).50 Death 10 (22) 3 (7).05 * Results are expressed as number (percentage) of patients except as otherwise specified under "Characteristic." Each case was categorized according to whether or not antibiotics were administered before blood was drawn for culture. t Mann-Whitney U test, Pearson's x 2 test, or Fisher's exact test. * Vegetation, abscess, or dehiscent prosthesis. Abnormalities consistent with IE but not meeting criteria for major findings. only a bacteriologic examination was conducted and no organisms were detected. In these 18 cases and the 11 cases in which neither a pathological nor a bacteriologic examination was conducted (a total of 29 cases, or 62%), the diagnosis of IE was based on macroscopic findings alone. Thirteen of 88 patients died. Thus the overall mortality was 15%. The 88 cases of blood culture negative IE fell into two subgroups: 42 cases in which antibiotics had been administered before blood was drawn for culture and 46 cases in which no such therapy had preceded the drawing of blood (table 2). The latter subgroup included the seven cases in which C. burnetii and/or its antibodies were detected, the two cases in which chlamydial antigens were found, and three of the six cases of culture-proven endocarditis. Only the cases in this subgroup can be regarded as strictly defined blood culture negative IE (authentic blood culture negative endocarditis). Although the two subgroups were comparable in terms of the criteria used for case definition, age, gender, underlying cardiac condition, clinical symptoms, echocardiographic findings, and rate of surgery, mortality was three

4 504 Hoen et al. CID 1995;20 (March) Table 3. Comparison of characteristics of 88 patients with blood culture negative infective endocarditis (IE) and 506 patients with blood culture positive IE from a 1-year nationwide survey in France. No. (%) in indicated group* Characteristic Negative blood cultures = 88t) Positive blood cultures (n = 506t) P$ Male gender 67 (76) 335 (66).07 Age, y (mean ± SD) 53.2 ± ± Underlying cardiac condition No known valve disease 25 (28) 177 (35) Native-valve disease 35 (40) 214 (42).16 Prosthetic valve 28 (32) 115 (23) Previous IE 10(11) 50 (10).68 Diagnostic classification Definite 48 (55) 169 (33) Probable 23 (26) 264 (52) Possible 17 (19) 73 (14) Left-sided location of current IE 85 (97) 420 (83).0009 Clinical symptoms Temperature of 38 C 71 (81) 482 (95) 6 X 10-7 Extracardiac symptoms 46 (52) 318 (63).06 Embolic event 20 (23) 138 (27).37 Suspected portal of entry 31/79 (39) 323/459 (70) 10-7 No. of blood cultures (mean ± SD) 8.8 ± ± Surgical treatment 47 (53) 163/485 (34).001 Death 13 (15) 100/477 (21).18 * Results are expressed as number (percentage) of patients except as otherwise specified under "Characteristic." t Unless a different denominator (total number of patients) is specified. Mann-Whitney U test or Pearson's x 2 test. times lower among patients who had received antibiotics before having blood drawn for cultures than among those with authentic blood culture-negative endocarditis (P =.05). Six of the 10 deaths in the former subgroup were due to heart failure, two were due to cerebral emboli, and two were not explained. The cause of the three deaths in the latter subgroup remained unknown. Comparison of blood culture-positive and blood culturenegative IE. Patients with blood culture-negative IE did not differ significantly from those with blood culture-positive IE with respect to age, gender, or prior IE (table 3). Although the difference was not statistically significant, there was a trend towards a higher frequency of prosthetic valves among patients with negative blood cultures. Moreover, the proportion of cases classified as definite was higher among patients with negative blood cultures. Most cases of blood culture-negative IE were left-sided, and the proportion of left-sided cases was significantly higher among blood culture-negative than among blood culture-positive cases. Almost 20% of patients with negative blood cultures had mild fever or none at all; the figure was 5% for patients with positive blood cultures. On the whole, extracardiac symptoms tended to be less frequent in blood culture-negative cases, although the difference was not statistically significant. The same trend was observed for each of the specific symptoms studied: hemorrhagic, renal, or meningeal symptoms; splenomegaly; Roth's spots; Janeway lesions; Osler's nodes; and embolic events. Each variety of embolus (cerebral, coronary, pulmonary, and limb) tended to be less frequent among patients with negative blood cultures. A portal of entry was suspected or identified almost twice as often in blood culture-positive cases. Although suspected in some instances, a specific portal of entry was not identified for any patient with negative blood cultures. The rate of surgery was significantly higher in blood culture-negative than in blood culture-positive cases. Among patients with prosthetic valves, 19 (68%) of the 28 with negative blood cultures, but only 41 (36%) of the 115 with positive blood cultures underwent surgical valve replacement (P =.002). The nine prosthetic-valve patients with negative blood cultures who did not undergo valve replacement included five with probable IE and four with possible IE; none of these patients died, and three had Q fever endocarditis. Mortality did not differ significantly with blood culture results. Discussion In our series of cases of IE, the rate of negative blood cultures (14%) was higher than in other recent series. Recruitment biases may be a major factor in the large variations observed. For example, the 1% rate of blood culture-nega-

5 CID 1995;20 (March) IE with Negative Blood Cultures 505 tive IE recorded by Van der Meer et al. is obviously related to their study's strict case-recruitment design, which was based on the reporting of positive blood cultures by microbiologists [5]. Conversely, exceedingly flexible diagnostic criteria have been cited as an explanation for high rates of blood culture negative endocarditis in some series [3]. We used von Reyn's criteria modified to incorporate echocardiographic findings and macroscopic observations at surgery; the fact that these modified criteria are less strict than the original criteria may account for our high proportion of definite cases. The original von Reyn criteria, which limit definite diagnoses to pathologically proven cases, are now regarded as too strict. Their use means that most cases of IE are classified as probable or possible even when the diagnosis seems certain on clinical grounds [6]. We do not regard our criteria as too flexible. Of our 29 patients whose diagnosis of definite IE relied on macroscopic findings at surgery, 22 would have been classified as probable according to the original von Reyn criteria. While seven cases would have been categorized as "rejected," all seven of the patients involved had a new or altered cardiac murmur in addition to echocardiographically documented and surgically confirmed vegetations. After our study had been completed, new criteria for the diagnosis of IE were proposed by Durack et al. [7]. Application of these criteria revealed that 18 of our 88 cases of blood culture negative IE were pathologically definite, 27 were clinically definite, 43 were possible, and none were "rejected." We cannot rule out the possibility that some of our blood culture negative cases were actually instances of noninfective inflammatory (marantic) endocarditis; however, we can assume that these cases were very few. In our series, only four patients had neither firm echocardiographic evidence of IE nor pathologically or bacteriologically demonstrated IE. Although the diagnosis of IE was debatable in these four cases, the patients were treated by their attending physicians as though they had IE, as were all the patients in our series. Recruitment may also influence the proportion of cases of endocarditis in a given series that are blood culture negative, especially when all cases are recruited from a single center. In our nationwide multicenter survey, such a bias was likely to be minor. In our series, cases with negative blood cultures did not differ remarkably from cases with positive blood cultures in terms of patients' characteristics, clinical features, and outcome. The same was true in the comparative study by Pesanti and Smith of 52 blood culture negative cases of IE and 84 blood culture positive cases [8]. The most striking difference that both we and they observed was the significantly higher frequency of antibiotic administration preceding the drawing of blood for culture among culture negative than among culture positive cases. It is noteworthy that in 31 (35%) of our 88 patients negative blood cultures were probably attributable to such treatment, which was administered even though these patients had cardiac conditions that put them at risk for IE. No clinical feature distinguished blood culture negative from blood culture positive cases. Fewer patients with negative blood cultures had fever; this finding has not been reported previously. In contrast with other investigators, we found neither a longer duration of symptoms before admission [9] nor a higher rate of embolic events [8-10] or congestive heart failure [8] among patients with negative blood cultures. Blood culture negative cases of IE were more often left-sided than blood culture positive cases. This difference was attributable to the fact that most cases of tricuspid-valve IE were caused by staphylococci and were associated with positive blood cultures. A portal of entry was more often suspected in blood culture positive cases, obviously because knowledge of the causative microorganism facilitated identification of its portal of entry. The high rate of surgery among our blood culture negative patients shows that the treatment of IE is more difficult when blood cultures yield no bacteria. In patients with prosthetic valves, valve replacement often becomes unavoidable when no pathogen can be identified by blood culture. In our series such patients underwent valve surgery about twice as often as patients with prosthetic-valve endocarditis and positive blood cultures. In spite of or perhaps because of this high proportion of patients undergoing surgery, the overall outcome in this group was still good: mortality was lower (albeit not significantly lower) among blood culture negative than among blood culture positive patients. Some evidence suggests that surgical valve replacement is worthwhile early in the course of IE, even during its active phase, with an improved final outcome [11]. The high rate of favorable outcome in blood culture negative IE, especially when fever resolves soon after the initiation of antibiotic therapy [8], has been described previously [1, 2]. Our 15% overall mortality rate was lower than the figures for previous series [8]. However, in-hospital mortality was higher among the 46 patients with authentic blood culture negative endocarditis than among the 42 patients whose blood cultures may have been rendered negative by prior antibiotic administration. Organisms such as C. burnetii or chlamydiae appear to be often responsible for authentic blood culture negative endocarditis. A fl-lactam/aminoglycoside combination usually recommended for the treatment of blood culture negative IE is likely to be ineffective against these organisms. The use of such regimens may best explain the poorer prognosis in the group of patients with no prior antibiotic therapy than in those receiving such treatment. In addition to antibiotic treatment preceding blood cultures, inadequate bacteriologic and diagnostic techniques may increase the proportion of cases of IE associated with negative blood cultures. Because of the multicenter design of our study, we cannot comment further on the adequacy of the blood culture techniques used. We found some evidence

6 506 Hoen et al. CID 1995;20 (March) of deficiencies in serological testing and in examination of valves removed at surgery. Antibodies to C. burnetii were sought in serum from only 40% of our patients. Since 8% of our blood culture-negative cases were diagnosed as Q fever endocarditis, we advocate that this specific serological test, along with a specific blood culture, be undertaken for patients with apparent blood culture-negative endocarditis, especially for those with prosthetic valves [12, 13]. To a lesser degree, this recommendation for specific serological testing might also apply to other bacteria that display fastidious growth characteristics, such as Legionella pneumophila [ 14], Chlamydia psittaci [15, 16], Chlamydia pneuinoniae [17], Mvcoplas#na species [18, 19], Brucella species [20], and Bartottella (Rochalimaea) species [21, 22]. Finally, we speculate that some of our cases were due to unknown organisms that cannot be cultured by routine laboratory procedures. The routine identification of these organisms by nucleic acid amplification technology may become possible in the near future [23]. Recently, for example, ribosomal RNA amplification led to the identification of Trophervma whippelii, the "unculturable" bacillus of Whipple's disease [24]. Thus it may be that valves and vegetations removed from patients with blood culture-negative endocarditis should be frozen and stored for future etiologic studies based on nucleic acid amplification techniques. The above-mentioned recommendations imply that "the blood culture is the single most important laboratory test in the diagnosis of endocarditis" [3]. From a clinical point of view, two practical conclusions may be drawn from this statement. First, we advocate that cardiac patients at risk for IE who present with unexplained fever-even of short duration, not be given antibiotics before blood is obtained for culture. Second, when a patient has symptoms consistent with IE and has received outpatient antibiotic therapy, large volumes of blood should be collected for culture over a long interval-i.e., until the end of the potential period of suppression of bacteremia [25, 26]. Acknowledgments The authors thank all the physicians who collaborated on this study, especially Caroline Oudot, M.D., for her review of patients' charts; Francois Delahaye, M.D., M.P.H., and Veronique Goulet, M.D., for their participation in the French Study Group on Infective Endocarditis; and the following colleagues for their granting of permission to study patients' charts: M. C. Aumont, M. C. Cajot, F. Cartier, P. Charon, C. Daubert, F. De Bourayne, P. Deleuze, M. Gatfosse, B. Godeau, P. Godeau, F. Guerin, F. Janbon, F. Leca, P. Longuet, F. Pernot, T. Petitclerc, R. Philipo, C. Polidori, G. Remy, D. Rondepierre, A. Ros, F. Spatz, B. Veyre, and H. Warembourg. The authors are also indebted to Christine Dobrovolny for her assistance in reviewing the manuscript. References 1. Van Scoy RE. Culture-negative endocarditis. Mayo Clin Proc 1982; 57: Cannady PB Jr, Sanford JP. Negative blood cultures in infective endocarditis: a review. South Med J 1976;69: Tunkel AR, Kaye D. Endocarditis with negative blood cultures [editorial]. N Engl J Med 1992;326: von Reyn CF, Levy BS, Arbeit RD, Friedland G, Crumpacker CS. Infective endocarditis: an analysis based on strict case definitions. Ann Intern Med 1981;94: Van der Meer JTM, Thompson J, Valkenburg HA, Michel MF. Epidemiology of bacterial endocarditis in the Netherlands. I. Patient characteristics. Arch Intern Med 1992; 152: Lukes AS, Bright DK, Durack DT. Diagnosis of infective endocarditis. Infect Dis Clin North Am 1993;7:1-8. 6a.Delahaye F, Goulet V, Lacassin F, et al. Characteristics of infective endocarditis in France in 1991: a one-year longitudinal survey. Eur Heart J 1995 (in press). 7. Durack DT, Lukes AS, Bright DK, et al. New criteria for diagnosis of infective endocarditis: utilization of specific echocardiographic findings. Am J Med 1994; 96: Pesanti EL, Smith IM. Infective endocarditis with negative blood cultures: an analysis of 52 cases. Am J Med 1979;66: Hampton JR, Harrison MJG. Sterile blood cultures in bacterial endocarditis. Q J Med 1967;36: Lerner PI. Infective endocarditis; a review of selected topics. Med Clin North Am 1974;58: Middlemost S, Wisenbaugh T, Meyerowitz C, et al. A case for early surgery in native left-sided endocarditis complicated by heart failure: results in 203 patients. J Am Coll Cardiol 1991;18: Kimbrough RC, Ormsbee RA, Peacock M, et al. Q fever endocarditis in the United States. Ann Intern Med 1979;91: Brouqui P, Dupont HT, Drancourt M, et al. Chronic Q fever; ninetytwo cases from France, including 27 cases without endocarditis. Arch Intern Med 1993;153: Tompkins LS, Roessler BJ, Redd SC, Markowitz LE, Cohen ML. Legionella prosthetic-valve endocarditis. N Engl J Med 1988;318: Levison DA, Guthrie W, Ward C, Green DM, Robertson PCG. Infective endocarditis as part of psittacosis. Lancet 1971;2: Shapiro DS, Kenney SC, Johnson M, Davis CH, Knight ST, Wyrick PB. Brief report: Chlamydia psittaci endocarditis diagnosed by blood culture. N Engl J Med 1992;326: Etienne J, Ory D, Thouvenot D, et al. Cffiamydia/ endocarditis: a report on ten cases. Eur Heart J 1992; 13: Cohen JI, Sloss LJ, Kundsin R, Golightly L. Prosthetic valve endocarditis caused by Mycoplasma hominis. Am J Med 1989; 86: Popat K, Barnardo D, Webb-Peploe M. Mycoplasma pneumoniae endocarditis. Br Heart J 1980;44: Flugelman MY, Galun E, Ben-Chetrit E, Caraco J, Rubinow A. Brucellosis in patients with heart disease: when should endocarditis be diagnosed? Cardiology 1990; 77: Daly JS, Worthington MG, Brenner DJ, et al. Rochalimaea elizabethae sp. nov. isolated from a patient with endocarditis. J Clin Microbiol 1993;31: Hadfield TL, Warren R, Kass M, Brun E, Levy C. Endocarditis caused by Rochalimaea henselae. Hum Pathol 1993;24: Relman DA. The identification of uncultured microbial pathogens. J Infect Dis 1993;168: Relman DA, Schmidt TM, MacDermott RP, Falkow S. Identification of the uncultured bacillus of Whipple's disease. N Engl J Med 1992;327: Mermel LA, Maki DG. Detection of bacteremia in adults: consequences of culturing an inadequate volume of blood. Ann Intern Med 1993; 119: Pazin GJ, Saul S, Thompson ME. Blood culture positivity; suppression by outpatient antibiotic therapy in patients with bacterial endocarditis. Arch Intern Med 1982; 142:263-8.