Sterile blood cultures are noted in one third of patients with infectious endocarditis. Although in half of cases this is due to previous antibiotic therapy, in the other half, the aetiology of culture-negative endocarditis is intracellular bacteria such as Coxiella burnetii or fastidious growing bacteria. Although it was previously considered that the prevalence of such organisms was identical throughout the world, recent investigations on Bartonella endocarditis clearly showed that the aetiology of culture-negative endocarditis is likely to be strongly related to epidemiology of the agent in each country. During the past decade the use of molecular techniques such as PCR with subsequent sequencing to detect or to identify bacteria in valves from patients with infectious endocarditis have considerably improved the aetiological diagnosis. This is especially true in the case of culture-negative endocarditis following earlier antibiotic therapy. However, the fact that DNA remnants of past endocarditis can be detected some time after the acute episode, when the patient has been cured, suggests that the predictive value of these techniques along with the traditional histology and culture need to be evaluated closely.
Infective endocarditis can be evoked in a patient by fever and a new or changing cardiac murmur. The diagnosis is most often based upon the detection of vegetation on the cardiac valves using echocardiography and positive blood culture. However, numerous situations in which blood culture or echocardiography are not able to confirm the diagnosis, lead to a high degree of suspicion of endocarditis.
Fever, the most common finding in endocarditis, may not be present in the elderly or in patients given antibiotic therapy before presentation, or in Whipple's disease (Richardson et al., 2003) and it may be low-grade or intermittent in Q fever endocarditis (Brouqui et al., 1993). Cardiac murmur is the second most frequent finding in endocarditis but may be absent in the initial stage of right side endocarditis. A new or changing murmur is detected in 40% of patients with endocarditis only (Stamboulian & Carbone, 1997). Echocardiography has assumed a central role in the diagnosis of suspected endocarditis. Transeosophagal echocardiography is more invasive and expensive than transthoracic echocardiography but reportedly is more efficient in detecting smaller lesions. Its role in routine screening for endocarditis is controversial. The sensitivity of echocardiography depends upon the causative organism, the operator and the quality of the echocardiograph. In a recent study, vegetation was detected in 87% and 75% of cases of Bartonella and Whipple's disease endocarditis, respectively, and in only 13% of Q fever cases (Fenollar et al., 2001), while it was detected in 39% of positive blood culture endocarditis (Watanakunakorn & Burkert, 1993).
However, with the use of modern diagnosis techniques that will be reviewed below, especially PCR and culture of infected valves, the number of cases without a detected aetiology dropped from 27% to 9% and 1.4% in the last published series (Hoen & Alla, 2002; Lamas & Eykyn, 2003; Werner et al., 2003a; Houpikian & Raoult, 2004). Antibiotic treatment preceded blood culture in 45–60% of cases of culture-negative endocarditis (CNE) (Lamas & Eykyn, 2003; Werner et al., 2003a). Another cause of CNE is subacute right side endocarditis and mural endocarditis (Brouqui & Raoult, 2001). Slow-growing and fastidious organisms represent approximately 5–15% of infective endocarditis and 50% of CNE. The most frequently reported of these are Coxiella burnetii, Brucella spp., Abiotrophia spp., HACEK group endocarditis and Listeria monocytogenes (Brouqui & Raoult, 2001). More recently discovered, an increase in endocarditis due to the fastidious Bartonella spp. has been reported (Fournier et al., 2001). Some of these slow-growing bacteria require specific media such as l-cysteine-enriched medium for Abiotrophia spp., special culture conditions for anaerobes or intracellular bacteria and incubation times as long as 6 weeks (Maurin & Raoult, 1996). The lack of systematic serological testing for Bartonella spp. and C. burnetii in CNE, as well as the length of incubation of blood cultures, are important limiting factors in the aetiological diagnosis of CNE and may explain the variation in the proportion of CNE in patients with infectious endocarditis reported in the literature (Werner et al., 2003a). In such situations, infective endocarditis remains a diagnostic challenge.
This review is based on a previous review (Brouqui & Raoult, 2001) completed using a Medline® search with the following keywords from 1998–2004: culture-negative endocarditis (n=161), diagnosis (n=130), echocardiography (n=41), PCR (n=40). Articles have been selected on the basis of the scope of the study, the impact factor and the half life citation of the journal, as well as the accessibility of the journal.
Endocarditis with a negative blood culture account for 2.5–48% of cases (Tariq et al., 2004). Developing countries have the highest rate of CNE, but it is usually admitted that one major explanation for this is the lack of efficient microbiology. However, in a comparative study made by the same research group on the aetiology of endocarditis in Slovakia, CNE accounted for 26.7% in the 1991–1997 period and for 53.3% of cases in the 1998–2001 period, despite use of better bacteriological techniques (Krcmery et al., 2003). Interestingly, no serological testing was performed for Coxiella burnetii and Bartonella in that study and the respective part these two major infectious agents play in CNE is unknown. This suggests that epidemiological factors interfere with the aetiology of CNE. In two recent studies of our laboratory, we demonstrated that the prevalence of Bartonella spp. as an aetiological agent of CNE varies greatly depending on the country. In Tunisia and Algeria, Bartonella endocarditis accounts for 9.8% and 15% of all endocarditis, respectively (Znazen et al., 2005; Benslimani et al., 2005). The poor living conditions in some countries such as Algeria, where louse-borne typhus has re-emerged, may explain the high prevalence of Bartonella quintana in CNE.
Several studies on Bartonella endocarditis have previously been published, reporting prevalences of 0–4.5% of all infective endocarditis. In France, Bartonella species account for 3% of infective endocarditis in Marseilles and for 4.5% in Lyon (Raoult et al., 1996). In Germany, 3% of all endocarditis cases investigated retrospectively by PCR on valve tissues were attributed to Bartonella (with 2.6% attributed to Bartonella henselae and 0.4% attributed to Bartonella quintana). In Sweden, Bartonella endocarditis was reported in only one case in 1997 and a study on 334 patients with infective endocarditis did not find any cases of Bartonella (Werner et al., 2003b). In the UK, a recent study showed that Bartonella species accounts for 1.1% of infective endocarditis (Lamas & Eykyn, 2003). These data suggest a north to south gradient in the distribution of Bartonella infections (Fig. 1). One may also suggest that the prevalence of C. burnetii as an aetiological agent of CNE may vary depending on the level of exposure to cattle. As a consequence, the prevalence of the aetiological agent in CNE is likely to be dependent upon environmental exposure and may vary greatly between countries.
Prevalence (%) of Bartonella quintana endocarditis in Europe and North Africa.
Diagnosis of infectious endocarditis
Role of echocardiography in the diagnosis of infective endocarditis
Transeosophagal echocardiography (TEE) has been reported to have a better sensitivity than transthoracic echocardiography (TTE) (Jacob & Tong, 2002), most studies reporting sensitivities of 30–50% and 85–100% for TTE and transeosophagal echocardiography, respectively. Transeosophagal echocardiography appears to be especially useful for the diagnostic evaluation of patient with suspected prosthetic valve endocarditis (Roe et al., 2000). Transeosophagal echocardiography is also superior to TTE in detecting mechanical complications such as valve perforation and chordal rupture (Jacob & Tong, 2002). As a general rule, TTE should be performed in anyone with suspected endocarditis (Robles, 2003). Transeosophagal echocardiography should be performed first in patients with possible endocarditis with clinical criteria, or in cases of prosthetic valve endocarditis or in cases of complicated infective endocarditis (Li et al., 2000).
New approaches of diagnostic scores
To both assist physicians in establishing the final diagnosis of endocarditis and to allow comparison of published cases, diagnostic scores have been defined. For many years, the Beth Israel criteria were the only recognized diagnostic criteria (Von Reyn et al., 1981). The use of echocardiography has led to the inclusion of echocardiographic findings in the criteria of the Duke endocarditis service (Durack et al., 1994). According to the Duke endocarditis service, the diagnosis of infective endocarditis is definite: (1) when a microorganism is demonstrated by culture or histological testing of a vegetation, an embolism, or an intracardiac abscess; (2) when active endocarditis is confirmed by histological examination of vegetation or intracardiac abscess; or (3) when two major criteria and three minor criteria or five minor criteria are met (Durack et al., 1994). Evaluation of these criteria in patients with pathologically proven endocarditis showed that 24% of patients were misclassified as ‘possible’ despite the use of Duke criteria especially in case of CNE or Q fever infective endocarditis (Habib et al., 1999). The overall sensitivity of the Beth Israel and Duke criteria was evaluated to be 60% and 80%, respectively (Lamas & Eykyn, 1997; Habib et al., 1999). Consequently, several researchers have tried to improve the value of these criteria and have suggested modifications that take into account several clinical or microbiological criteria. Among these proposed modifications are Coxiella burnetii serology or the molecular detection of the aetiological agent in the removed valve as major criteria (Fournier et al., 1996; Li et al., 2000). In a study of 20 patients with infective endocarditis confirmed at pathological examination, the Duke endocarditis criteria misclassified four of them, all Q fever endocarditis with C. burnetii IgG Phase I >800. The authors conclude that including C. burnetii IgG Phase I >800 or a single positive blood culture for C. burnetii as a major criterion in the Duke criteria would improved their sensitivity (Fournier et al., 1996). In another study of 100 cases of proven native valve endocarditis, the addition of some minor criteria increased the sensitivity from 83% to 94% (Lamas & Eykyn, 1997). An often-heard criticism of the Duke criteria is the over broad categorization of the group ‘possible infective endocarditis’. In 2000, a redefinition of this group was proposed, based upon the review of the Duke University data on more than 800 patients. This category has been proposed to include patients with one major criterion and one minor criterion or three minor criteria (Li et al., 2000). Several issues remained including the relative risk of infective endocarditis in the cases of Staphylococcus aureus bacteraemia and the relative role of transeosophagal echocardiography. Previous studies have shown that S. aureus infective endocarditis is unlikely when the bacteraemia is nosocomially acquired and a primary focus, such as intravascular device, is present at the time of bacteraemia (Li et al., 2000). As a result, the original Duke criteria considered blood cultures that were positive for S. aureus to be a major criterion only if community acquired in the absence of primary focus (Durack et al., 1994). In a review of the Duke endocarditis service database, 13% of patients with nosocomially acquired S. aureus bacteraemia developed infective endocarditis whether this bacteraemia was catheter-related or not. The authors suggest that S. aureus bacteraemia should be considered a major criterion regardless of whether the infection is nosocomially acquired or a removable source of infection is present (Li et al., 2000). The Duke University Medical Center maintains prospective databases on all echocardiograms performed since 1994. Using these databases they suggest that transthoracic echocardiography should be used at first in all patients except those for whom transeosophagal echocardiography is recommended (patients with prosthetic valves, those rated as at least ‘possible infective endocarditis’ by clinical criteria, or those with complicated infective endocarditis such as paravalvular abscess). In some cases of afebrile chronic endocarditis caused by slow growing fastidious organisms such as Whipple's disease bacilli, the diagnosis remains a challenge and is made only at surgery or autopsy (Gubler et al., 1999). The revised Duke criteria are summarized in Tables 1 and 2 (Li et al., 2000).
Definition of terms used in the proposed modified Duke criteria for the diagnosis of infective endocarditis (infective endocarditis)
• Blood culture positive for infective endocarditis:
• Typical microorganisms consistent with infective endocarditis from two separate blood cultures: Viridans streptococci, Streptococcus bovis, HACEK group, Staphylococcus aureus; or community-acquired enterococci, in the absence of a primary focus; or
• Microorganisms consistent with infective endocarditis from persistently positive blood cultures, defined as follows:
At least two positive cultures of blood samples drawn >12 h apart; or
All of three or a majority of ≥4 separate cultures of blood (with first and last samples drawn at least 1 h apart)
• Single positive blood culture for Coxiella burnetii or antiphase I IgG antibody titre >1 : 800
• Evidence of endocardial involvement
• Echocardiogram positive for infective endocarditis [TEE recommended in patients with prosthetic valves, rated at least ‘possible infective endocarditis’ by clinical criteria, or complicated IE (paravalvular abscess); TTE as first test in other patients], defined as follows:
Oscillating intracardiac mass on valve or supporting structures, in the path of regurgitant jets, or on implanted material in the absence of an alternative anatomic explanation; or Abscess; or new partial dehiscence of prosthetic valve
• New valvular regurgitation (worsening or changing of pre-existing murmur not sufficient)
• Predisposition, predisposing heart condition or injection drug use
• Fever, temperature >38°C
• Vascular phenomena, major arterial emboli, septic pulmonary infarcts, mycotic aneurysm, intracranial haemorrhage, conjunctival haemorrhages, and Janeway's lesions
• Microbiological evidence: positive blood culture but dues not meet a major criterion as noted above or serological evidence of active infection with organism consistent with infective endocarditis
• Echocardiographic minor criteria eliminated
TEE, transesophageal echocardiography; TTE, transthoracic echocardiography; IE, infective endocarditis.
↵Excludes single positive cultures for coagulase negative staphylococci and organisms that do not cause endocarditis.
Conditions leading to negative blood culture
Several intracellular agents have been reported to be involved in infective endocarditis (Table 3). The most frequently detected is Coxiella burnetii, the agent of Q fever, but others have been reported, such as Tropheryma whippelii and Chlamydia pneumoniae (Fenollar et al., 2001; Gdoura et al., 2002).
Q fever has been found in all the countries where it has been investigated. More than 400 cases of Q fever endocarditis have been found, 359 in the literature and more than 60 new cases diagnosed in our laboratory (Houpikian & Raoult, 2004). Q fever endocarditis is often a severe disease associated with a long diagnostic delay. Q fever represents 5% of endocarditis cases in France. It occurs almost exclusively in patients with previous a cardiac defect or in immunocompromised patients, and Q fever endocarditis is being recognized increasingly all over the world. The clinical presentation has changed over the last 30 years. With faster diagnoses, the prevalence of heart failure, hepatomegaly, inflammatory syndrome, anaemia and leucopaenia and abnormal liver function tests have decreased significantly (Houpikian et al., 2002). Diagnosis based on serum antibody response to C. burnetii phase I and II should be carried out systematically in confronting CNE as it represents the main aetiological diagnosis.
Whipple's disease endocarditis is a specific entity. A previous valvular disease is found in only 13% of patients compared to 52% and 88% in Bartonella and Q fever endocarditis, respectively (Fenollar et al., 2001). The absence of fever is a characteristic of Whipple's endocarditis and is reported in only 26% of patients. In the majority of cases the endocarditis is associated with other signs of Whipple's diseases such as diarrhoea, arthralgia, abdominal pain and lymphadenopathy (48%). Death occurs in 57% of cases. Vegetation is detected by echocardiography in 75% of cases. Anaemia (80%) and hypereosinophilia (40%) are specifically associated with Whipple's disease (Fenollar et al., 2001).
Chlamydia spp. have often been suggested to cause CNE but this has rarely been proven. Most reported cases were due to serological cross-reaction with Bartonella spp. (Brouqui & Raoult, 2001). To our knowledge there are two documented cases of chlamydial endocarditis reported, one due to Chlamydia psittaci in which the organism has been cultured from the blood and another recent case of Ch. pneumoniae in which it as been molecularly characterized (Gdoura et al., 2002).
Isolation of these agents requires tissue cell culture facilities and a laboratory staff experienced in the growth of intracellular organisms. The availability of sequenced genomes now provides the opportunity to define culture media for the growth of fastidious pathogens. This has been applied successfully to the growth of T. whippelii in cell-free culture medium, creating the opportunity to transfer the technology to routine laboratories (Renesto et al., 2003). In the meantime, infective endocarditis caused by these organisms remains the true CNE. Etiological diagnosis of these intracellular agents relies mostly on serology and molecular techniques, which will be described below.
Several other agents are classified as causing CNE because the blood cultures are still negative after 48 h of incubation. This is due to the slow growth of these agents. The most prevalent organisms in this category are Bartonella spp. and Brucella spp.
Bartonella endocarditis has been reported in 120 cases in the literature. Most cases are caused by B. quintana or B. henselae but other Bartonella spp. have occasionally been reported. Bartonella quintana has been associated with body-louse infested alcoholic homeless persons and B. henselae with patients with a previous valvulopathy and contact with cats or their fleas (Fournier et al., 2001). Valvular surgery is needed in 96% of cases. Bartonella spp. are isolated from blood by prolonged incubation in the Bactec® system (Becton Dickinson), inoculation on rabbit blood agar or on tissue cell culture (Brouqui & Raoult, 2001).
Brucella endocarditis represent 1.1% of brucellosis and 3.5% of endocarditis cases in Spain (Reguera et al., 2003). An occupational exposure is found in almost all patients and an underlying heart disease is reported in 45% of patients. The mean duration of symptoms before diagnosis is 3 months. The aortic valve is involved in 80% of cases. Although some systems such as the lysis concentration technique lead to shorter isolation time, Brucella spp. growth takes at least 4–10 days. However, despite the fact that Brucella endocarditis is generally associated with CNE, in a recent study, 64% of Brucella endocarditis cases were culture positive in patients without previous antibiotic therapy if processed correctly (Reguera et al., 2003). Brucella endocarditis is severe and has cited as responsible for death in 80% of patients who died of brucellosis.
Other agents are Mycobacterium spp., Francisella tularensis, HACEK group and Legionella spp. Francisella tularensis was detected in blood after 9 days of incubation, whereas the mean duration time for HACEK group is 3–5 days, and Actinobacillus actinomycetemcomitans may require up to 30 days for growth (Houpikian & Raoult, 2003a). In some situations, Brucella spp. as well as Bartonella spp. were recovered from tissue cell culture only (Rovery et al., 2003).
Endocarditis caused by anaerobes is uncommon and usually associated with CNE. Propionibacterium spp. and the Bacteroides fragilis group are the cause of most cases of anaerobic endocarditis (Bisharat et al., 2001). Although anaerobes are usually detected in anaerobic blood culture media, subculture and isolation in nonselective culture media under anaerobic conditions is slow and requires an incubation time of at least 5 days.
Nutritionally deficient bacteria
Abiotrophia species formerly known as nutritionally deficient streptococci can be detected in routine blood culture in 2 or 3 days. However, subculture usually requires supplementation of blood agar or broth with pyridoxal hydrochloride or l-cysteine at 1–1000 µg mL−1. Legionella species required buffered charcoal yeast extract (BCYE) agar. Mycoplasma spp. grow better in media such as SP4 glucose at pH 4.5. Mycobacterium spp. have been isolated in conventional blood culture systems with prolonged incubation time, but special media such as Middlebrook 7H13 broth should be considered especially for Mycobacterium tuberculosis. Most Haemophilus spp. grow well on conventional chocolate agar, but require either exogenous haemin (X factor) or NAD (V factor).
Previous antibiotic therapy
Previous antibiotic therapy is noted in two thirds of patients with CNE (Brouqui & Raoult, 2001; Lamas & Eykyn, 2003). The duration of previous antimicrobial therapy is an important factor. If antibiotics are given for only 2–3 days, blood cultures that were initially negative rapidly become positive. However, after longer noncurative courses of therapy, some blood cultures remained negative for a number of weeks (Tunkel & Kaye, 1992).
Other conditions (right heart)
Blood cultures are frequently negative in patients with right-side endocarditis and mural endocarditis (Brouqui & Raoult, 2001). Of 11 nondrug addict patients with tricuspid valve endocarditis diagnosed with two-dimensional echocardiography, seven had S. aureus positive blood culture and four were culture negative (Naidoo, 1993).
Strategies developed for the diagnosis of CNE
From blood specimens
Quantitative culture techniques show that blood from patients with infective endocarditis contains 1–10 bacteria per mL, and this quantity remains constant during the course of the disease. Because of the approximate correlation between the yield of the bacteria from blood and the volume of blood drawn, it has been recommended that at least 10 mL of blood be obtained for each culture. The use of more than three blood cultures will not improve diagnosis, and therefore culture of three sets (anaerobic and aerobic) of blood drawn with an interval of at least 1 h within a 24–48 h period are normally sufficient to establish the diagnosis of culture positive endocarditis and, conversely, to indicate a possible diagnosis of culture negative endocarditis (Brouqui & Raoult, 2001; Houpikian & Raoult, 2003a). Incubation time is one of the major limiting factors for recovery of fastidious organisms. Several of them may require a number of weeks to grow and in patients with suspected endocarditis with negative blood culture the incubation time can be as long as 30–42 days. If antibiotics have not been previously administered and the blood cultures are negative, a fastidious growing organism should be suspected.
Tissue cell culture
Tissue cell culture was developed initially for isolation of strict intracellular bacteria. Use of the ‘shell vial’ technique, adapted from a technique used for isolation of cytomegalovirus, has resulted in the isolation of C. burnetii, recognized today as one of the most common aetiological agents of CNE (Marrero & Raoult, 1989; Brouqui & Raoult, 2001). Coxiella burnetii can be isolated from blood in 53% of untreated patients (Musso & Raoult, 1995) but this should be restricted to biosafety level 3 equipped laboratories. A large number of cell lines can be used. L929 cells, Vero, and human embryonic lung fibroblast (HEL cells) are used for C. burnetii, L929 for Chlamydia spp., and ECV cells for Bartonella spp. (Brouqui & Raoult, 2001). This technique has been applied successfully for the isolation of other agents such as Bartonella spp. and Whipple's disease bacilli in CNE (Drancourt et al., 1995; Raoult et al., 2000; Fenollar et al., 2001). In patients with Bartonella spp. endocarditis the sensitivity of the shell vial assay when inoculated with blood was 28%, compared with only 5% when cultured onto agar plates. The most efficient method for recovering Bartonella spp. from blood in the case of endocarditis was the subculture in shell vial of the aerobic Bactec Plus® (Becton Dickinson) blood culture broth on day 7 (LaScola & Raoult, 1999).
Of the fastidious organisms causing CNE, serum antibody testing is available for Chlamydia spp., Legionella, Brucella, Bartonella and C. burnetii. The serological tests are included as a part of the Duke criteria for the diagnosis of infective endocarditis but their predictive value differs. The good predictive value of positive serology to C. burnetii has led authors to propose that a single titre of C. burnetii antibody IgG phase 1–800 be a major criterion for infective endocarditis (Fournier et al., 1996; Li et al., 2000). For C. burnetii, the most reliable and commonly used methods are indirect immunofluorescence and the complement fixation test (Houpikian & Raoult, 2003a). At present, the reference technique is indirect immunofluorescence assay. Both C. burnetii Nine-Mile strain Phase I and Phase II are used as antigens. Antibodies to Phase I and II can be determined in the IgG, IgM and IgA classes. Q fever endocarditis is characterized by a very high titre of anti-Phase I antibodies; an IgG anti-Phase I antibody titre of: 1600 is considered to be highly predictive and sensitive, with a 98% positive predictive value, whereas anti-Phase I IgA and IgM titres do not contribute usefully to diagnosis (Dupont et al., 1994). A single serum sample is sufficient for the diagnosis of Q fever endocarditis. Cross-reactions, however, may be a source of confusion when interpreting serological results. These vary according the serological technique employed. They have been described between C. burnetii and either Legionella or Bartonella species, but a differential diagnosis is easily established when quantitative antibody titres against both Phase I and II of C. burnetii antigens are determined.
Currently, both indirect immunofluorescent assay and enzyme-linked immunosorbent assay (ELISA) are used in the diagnosis of Bartonella infection to detect specific antibodies. Current serological tests may not distinguish reliably between antibody responses to B. quintana and B. henselae, although antibody cross-absorption and Western immunoblotting allow differentiation of the serological responses to these two species (Fig. 2) (Houpikian & Raoult, 2003b). Furthermore, cross-reactions may occur at a low level with C. burnetii (La Scola & Raoult, 1996) and significantly with Chlamydia spp. (Maurin et al., 1997). Patients with proven B. quintana endocarditis have been reported with IgG titres of >1 : 256 against Ch. pneumoniae and titres of 1 : 64 against Chlamydia trachomatis and Ch. psittaci (Drancourt et al., 1995). Absorption of the sera with Ch. pneumoniae did not reduce the high antibody titres against B. quintana, but absorption with B. quintana eliminated reactivity with the Ch. pneumoniae antigen. This cross-reactivity was confirmed using immunoblotting (Drancourt et al., 1995). Taken altogether, an indirect immunofluorescence assay antibody titre toward Bartonella spp. >1 : 800 has a predictive value of 95% to detect Bartonella infection in patients with endocarditis, leading to the suggestion that Bartonella serology should be included as a major criterion in the Duke criteria (Fournier et al., 2002).
Western immunoblot with cross adsorption for the species-specific diagnosis of Bartonella endocarditis. Antigens used are Bartonella quintana (L1), Bartonella henselae Houston (L2), Bartonella elizabethae (L3) Bartonella vinsonii spp. berkofii (L4) and Bartonella vinsonii spp. arupensi (L5). The patient's serum was not adsorbed, adsorbed with B. quintana or adsorbed with B. henselae. Note that all reactions disappeared when adsorbed with B. quintana, whereas when adsorbed with B. henselae a reaction appears with B. quintana, indicating that the specific antibodies contained in the patient's sera are those toward B. quintana.
In a series of 10 patients reported to have Chlamydia endocarditis, eight were finally diagnosed with Bartonella endocarditis after testing their sera by cross-absorption procedures and western immunoblotting. Because of the serological cross-reaction describe below, an elevated titre of antibody to Chlamydia spp. in a patient with CNE should prompt Bartonella antibody testing.
Although several methods such as indirect immunofluorescent assay, ELISA and Western blot analysis have been developed to detect specific antibodies to Brucella spp., the tube agglutination test is still the reference test. It can be used to make a presumptive diagnosis in the absence of bacteriological confirmation because most cases of active infection will be associated with titres >1 : 160 (Young, 1991). The physician should be aware of serological cross-reactions that exist between Brucella, Yersinia and Francisella species that can lead to confusion in the aetiological diagnosis.
Considerable efforts have been recently made to establish the aetiological diagnosis of CNE. Molecular techniques by using PCR with subsequent sequencing of the amplicon and gene analysis and comparison in the data bank have allow to a considerable number of new aetiological diagnoses of CNE (Mueller et al., 1999; Millar et al., 2001; Casalta et al., 2002; Gauduchon et al., 2003; Lang et al., 2004). In our series, PCR amplification using universal primers in valves removed at surgery allowed 100% aetiological diagnosis in patients classified with possible or definite endocarditis, including those who had blood or valve material that cultured negative (Millar et al., 2001). Although the sensitivity and specificity of PCR is well established in resected valves (see below), the usefulness of this technique in blood is still debated (Bosshard et al., 2003). The low specificity of PCR in blood is likely due to pre-PCR contamination. When universal 16S rRNA gene PCR is coupled with a simple pre-PCR decontamination step by enzymatic digestion, the specificity of the PCR was 100% in patients with culture positive endocarditis, providing a promising diagnostic in CNE (Rothman et al., 2002). Molecular methods seem particularly indicated in CNE, both due to previous antibiotic therapy and due to fastidious organisms (Podglajen et al., 2003). In their study, Millar (2001) recovered DNA of possible causal agents in 10 of the 25 CNE blood cultures they tested. Three were Streptococcus spp. in three patients who had received previous antibiotic therapy (Millar et al., 2001). Blood samples from two patients who had been treated previously with antibiotics were negative with PCR, but Streptococcus spp. was detected in the valve (Khulordava et al., 2003). While waiting for more data, broad range PCR of blood in cases of CNE should be restricted to CNE patients who have previously received antibiotics and who are Bartonella and C. burnetii seronegative. Recently, Light Cycler Nested PCR (LCN-PCR) in sera from patients with a proven diagnosis of either Bartonella or C. burnetii endocarditis has been reported to be more sensitive and specific than other traditional methods such as culturing or PCR of EDTA-treated blood (Zeaiter et al., 2003; Fenollar et al., 2004).
From cardiac valves
Because histopathology can confirm the diagnosis by revealing valvular inflammation, the vegetation, the organisms or other changes consistent with infective endocarditis, the histology of the resected valve remains the gold standard for the diagnosis of infective endocarditis and is a major criterion of the Duke classification.
Endocarditis can be identified histologically with haematoxylin–eosin (H&E) staining by demonstration of an inflammatory reaction in valvular tissue and vegetation. The presence of vegetation and a significant inflammation with up to 2% polymorphonuclear neutrophil leucocytes is a key in the diagnosis of endocarditis and should be considered a major criterion (Lepidi et al., 2005). Eight nonspecific stains can be used to detect bacteria and fungi in paraffin sections of CNE valve specimens: Giemsa, Brown–Brenn and Brown–Hopps tissue Gram stains, periodic acid-Schiff, Grocott–Gomori methenamine silver, Warthin–Starry, Gimenez, and Ziehl–Neelsen stains. Although the Giemsa stain appears to be the most sensitive, the most popular and widely used histological method for detection of bacteria is the tissue Gram stain. In CNE due to previous antibiotic therapy, the Gram stain appears to be helpful for further identification. The diagnosis of T. whippelii endocarditis can usually be made with periodic acid-Schiff staining, which reveals numerous characteristic periodic acid-Schiff positive granules. Silver impregnation using the Warthin–Starry stain is among the most sensitive methods for detection of bacteria, including those that stain weakly with a tissue Gram stain, such as Bartonella sp. Silver impregnation is not specific, staining virtually all bacteria, spirochetes, and fungi. The Gimenez stain is a good method for the Legionella species. The Ziehl–Neelsen stain is used for detection of acid-fast bacteria, especially mycobacteria. Sometimes, antibiotics induce changes in bacterial morphology and staining properties which lead to an erroneous diagnosis of a yeast or fungal infection. Histological evaluation in infective endocarditis has recently been reviewed (Lepidi et al., 2002).
The successful isolation and cultivation of C. burnetii, B. quintana, B. henselae and T. whippelii has made it possible to generate polyclonal rabbit or monoclonal mouse antibodies to these bacteria. Specific detection of these microorganisms in tissues may now be achieved by using these antibodies for immunohistology. Intracellular bacteria such as C. burnetii or T whippelii are demonstrated in large numbers within swollen histiocytes, whereas Bartonella spp. are found in extracellular locations without inflammatory infiltrates (Brouqui et al., 1994; Lepidi et al., 2000, 2003) (Fig. 3).
Immunohistological demonstration of Whipple's disease bacillus in cardiac valve using species-specific monoclonal antibody. Magnification × 400. Courtesy of Dr Hubert Lepidi, CNRS UMR 6020.
Pathogens can also be isolated from resected valves or biopsy specimens by inoculation onto agar or tissue culture. C. burnetii, Mycobacterium spp., Brucella, Legionella, Bartonella and Whipple's disease bacteria have been isolated in this manner. In Bartonella endocarditis, the sensitivity of the shell vial culture of valve biopsy is greater than that of blood culture (44% vs. 28%) (LaScola & Raoult, 1999). No patient with prior antibiotic therapy yielded a positive blood culture; this does not affect isolation of Bartonella from the resected valve, although one must not allow the strain to become established in culture (LaScola & Raoult, 1999). Among 35 valves resected in patients with Whipple's disease endocarditis, 34 were positive at histological examination, eight using PCR and two in culture (Fenollar et al., 2001). Thus surgically removed material should be systematically cultured in appropriate medium when possible. Serological testing may help to indicate the appropriate media (Brouqui & Raoult, 2001).
Molecular detection in heart valves is certainly the most sensitive tool today for the diagnosis of CNE. PCR is especially helpful in the aetiological diagnosis of CNE with prior antibiotic therapy. In a study of 49 patients with CNE for whom the infected valve was available for diagnostic tests, the sensitivity, specificity and predictive positive and negative values were 17.6%, 88.9%, 75% and 36%, respectively, with valve culture compared with 82.6%, 100%, 100% and 76.5% with PCR (Bosshard et al., 2003). The most common situation is that of an organism identified in the valve by histology or culture and characterized by PCR. Recently, the detection of S. pneumoniae rpoB DNA in a histologically normal heart valve of a man, who had presented with pneumococcal endocarditis 7 years earlier, raised the question of the persistence of DNA in the absence of any evidence of infection (Branger et al., 2003). In a recent series, we report that bacterial DNA is detected in 60% of patients with infective endocarditis while on antibiotic therapy, and that this DNA was still detected in 37% of patients who had completed their treatment (Rovery et al., 2005).
Perspective in the diagnosis of CNE
The most important new trend in the diagnosis of CNE in the last decade is the use of molecular techniques. Molecular detection of microorganisms either by broad range PCR or specific PCR has led to detection or characterization of a considerable number of new organisms responsible for fastidious endocarditis. This technique is by especially by in CNE with previous antibiotic treatment. PCR has been helpful for the diagnosis of streptococcal endocarditis in patients receiving antibiotic therapy, for fastidious streptococci, Tropheryma whippelii endocarditis, Mycoplasma spp. and Mycobacterium spp. endocarditis and Bartonella spp. and C. burnetii when serology was not tested (Table 4).
Usefulness of polymerase chain reaction in culture-negative endocarditis
Microorganism identified by PCR on valve
In patient receiving antibiotic
When serology is not performed
When serology is not performed
The sensitivity of this technique is much greater in the resected valves than in blood. A number of authors recently proposed modifying the Duke criteria to include molecular results as a major criterion (Millar et al., 2001; Millar & Moore, 2004). However, the recently published detection of remnant DNA in patients with previously cured infectious endocarditis (Branger et al., 2003; Rovery et al., 2004) raised the question of the specificity of such technique. This specificity may be raised to 100% if interpreted following a predefined procedure (Fig. 4) (Greub et al., 2005). Histological examination of the valve remains the gold standard for the diagnosis of infective endocarditis, and the molecular diagnosis of CNE should interpreted carefully as molecular detection cannot be considered a major criterion.
(1997) Suggested modifications to the Duke criteria for the clinical diagnosis of native valve and prosthetic valve endocarditis: analysis of 118 pathologically proven cases. Clin Infect Dis 25: 713–719.
(2002) Detection of bacteremia in emergency department patients at risk for infective endocarditis using universal 16S rRNA primers in a decontaminated polymerase chain reaction assay. J Infect Dis 186: 1677–1681.