MAJOR ARTICLE

Comprehensive Diagnostic Strategy for Blood

Culture–Negative Endocarditis: A Prospective Study
of 819 New Cases

Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021

Pierre-Edouard Fournier,1,2 Franck Thuny,3 Hervé Richet,1 Hubert Lepidi,1 Jean-Paul Casalta,2 Jean-Pierre Arzouni,2
Max Maurin,5 Marie Célard,6 Jean-Luc Mainardi,7 Thierry Caus,8 Frédéric Collart,3 Gilbert Habib,4
and Didier Raoult1,2
1
Unité de Recherche sur les Maladies Infectieuses et Tropicales Emergentes, Centre National de la Recherche Scientifique–Institut de Recherche
pour le Développement, Unité Mixte de Recherche 6236, Faculté de Médecine, Université de la Méditerranée, and 2Pôle de Maladies
Infectieuses, 7Service de Chirurgie Cardiaque, and 8Service de Cardiologie, Hôpital de la Timone, Marseille, 3Laboratoire de Bactériologie, Centre
Hospitalo-Universitaire de Grenoble, Grenoble, 4Laboratoire de Bactériologie, Centre de Biologie Est, Hospices Civils de Lyon, Lyon, 5Service de
Microbiologie, Hôpital Européen Georges Pompidou, Paris, and 6Service de Chirurgie Cardiaque, Centre Hospitalier Universitaire, Amiens, France

(See the editorial commentary by Lamas, on pages 141–142.)

Background. Blood culture–negative endocarditis (BCNE) may account for up to 31% of all cases of
endocarditis.
Methods. We used a prospective, multimodal strategy incorporating serological, molecular, and histopatho-
logical assays to investigate specimens from 819 patients suspected of having BCNE.
Results. Diagnosis of endocarditis was first ruled out for 60 patients. Among 759 patients with BCNE, a
causative microorganism was identified in 62.7%, and a noninfective etiology in 2.5%. Blood was the most useful
specimen, providing a diagnosis for 47.7% of patients by serological analysis (mainly Q fever and Bartonella
infections). Broad-range polymerase chain reaction (PCR) of blood and Bartonella–specific Western blot methods
diagnosed 7 additional cases. PCR of valvular biopsies identified 109 more etiologies, mostly streptococci, Tro-
pheryma whipplei, Bartonella species, and fungi. Primer extension enrichment reaction and autoimmunohisto-
chemistry identified a microorganism in 5 additional patients. No virus or Chlamydia species were detected. A
noninfective cause of endocarditis, particularly neoplasic or autoimmune disease, was determined by histological
analysis or by searching for antinuclear antibodies in 19 (2.5%) of the patients. Our diagnostic strategy proved
useful and sensitive for BCNE workup.
Conclusions. We highlight the major role of zoonotic agents and the underestimated role of noninfective
diseases in BCNE. We propose serological analysis for Coxiella burnetii and Bartonella species, detection of anti-
nuclear antibodies and rheumatoid factor as first-line tests, followed by specific PCR assays for T. whipplei, Bartonella
species, and fungi in blood. Broad-spectrum 16S and 18S ribosomal RNA PCR may be performed on valvular
biopsies, when available.

Blood culture-negative endocarditis (BCNE)—that is,
endocarditis in which no causative microorganism can
be grown in a culture taken from a blood sample by
using the usual laboratory methods—accounts for
Received 18 December 2009; accepted 20 March 2010; electronically published
11 June 2010.
Reprints or correspondence: Dr. Didier Raoult, Unité de Recherche sur les
Maladies Infectieuses et Tropicales Emergentes, CNRS-IRD UMR 6236, Faculté de
médecine, Université de la Méditerranée, 27 Blvd. Jean Moulin, 13385 Marseille
cedex 05, France (didier.raoult@gmail.com).
Clinical Infectious Diseases 2010; 51(2):131–140
 2010 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2010/5102-0002$15.00
DOI: 10.1086/653675
2.5%–31% of all cases of endocarditis [1]. This varia-
tion in incidence may be explained by several factors,
including (i) differences in the diagnostic criteria used;
(ii) specific epidemiological factors, as for fastidious
zoonotic agents; (iii) variations in the early use of an-
tibiotics prior to blood sampling; (iv) differences in
sampling strategies [2]; or (v) involvement of unknown
pathogens.
In our laboratory, we have diversified the diagnostic
tests used over the past few years for the diagnosis of
BCNE. In particular, we have demonstrated the use-
fulness of systematic serological testing for the detec-
tion of fastidious agents, especially Coxiella burnetii and
Bartonella species, but also Brucella species, Legionella

Blood Culture–Negative Endocarditis • CID 2010:51 (15 July) • 131

 Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021
Figure 1. Distribution of the 819 patients with suspected blood culture–negative endocarditis (BCNE) studied from 1 June 2001 to 1 September
2009, according to the etiological diagnosis (a) and the year (b). Black columns, Number of patients per year for whom we obtained an etiological
diagnosis (infectious or not). Gray columns, Number of patients without any etiological diagnosis. Values above each column represent percentages
of etiological diagnoses obtained each year. Agents include Tropheryma whipplei.

pneumophila, and Mycoplasma species [2, 3]. Two other meth-
ods that exhibited great potential for the diagnosis of BCNE
were histological examination [4] and broad-range polymerase
chain reaction (PCR), particularly when applied to valvular
biopsies [5]. The few studies using PCR for the diagnosis of
BCNE published to date have highlighted the importance of
streptococci and fastidious bacteria, although their respective
prevalence has not been estimated [3, 5–12]. In addition,
among fastidious bacteria, the role of Chlamydia pneumoniae
in BCNE remains uncertain, as is the case for viruses [13].
Consequently, diagnosis of BCNE remains a challenge, as high-
lighted by the 21% rate of cases without any identified causative
agent in the largest series of BCNE cases reported to date [3].
Since our previous publication of a large series of BCNE
cases [3], our laboratory received an increasing number of spec-
imens from patients with BCNE. Here, we prospectively used

132 • CID 2010:51 (15 July) • Fournier et al

 Table 1. Epidemiological Variables among the 740 Studied Patients Classified as
Definite or Possible According to the Duke Criteria [15]

Variable Definite patients Possible patients

No. of patients 549 191
Male-to-female ratio 2.7 1.9
Age, mean years  SD 58.3  15.7 58.2  17.1
Age range, years 12–92 3–92
Geographic origin
a
France (Marseille) 477 (156) 172 (60)
Europe 32 12
North America 20 4
Middle East 16 1

Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021

Otherb 4 2
Echocardiographic signs of endocarditis, % 84.9 73.8
Left-sided endocarditis, % 91.6 83.8
Right-sided endocarditis,c % 2.9 1.6
Pacemaker, % 5.5 14.6
Antibiotics prior to blood cultures, % 44.3 60.2
a
The main contributing cities were Amiens, Grenoble, Lyon, and Paris.
b
Includes patients from South America, North Africa, Asia, and Australia.
c
Not including pacemakers.

a comprehensive strategy incorporating a battery of laboratory
techniques [3, 5, 14], including several new ones, to increase
the rate of diagnoses in patients with BCNE and to evaluate
the prevalence of various agents missed by standard blood cul-
tures, including fastidious bacteria and viruses. We also esti-
mated the incidence of noninfective causes of endocarditis.
PATIENTS AND METHODS
Patients
From 1 June 2001 to 1 September 2009, we prospectively in-
cluded all patients with suspected BCNE for whom specimens
were referred to our laboratory (Figure 1 and Table 1). For
each studied patient, a questionnaire was completed by the
physician in charge. The questions asked are detailed in Tables
2 and 3. Answers were not obtained for all questions from all
patients. When results of all assays were negative, we recon-
tacted physicians in charge of patients to enquire about any
neoplasic or automimmune disease that would have been di-
agnosed elsewhere. The study was approved by the local ethics
committee under reference 07–015. The study was also ap-
proved by the Commission Nationale Informatique et Libertés
under reference 1223186.
Diagnostic Procedures
Serological analysis. Indirect immunofluorescence assays to
detect significant levels of antibodies to C. burnetii (phase I
immunoglobulin [Ig] G titer, 11:800), Bartonella quintana,
Bartonella henselae (IgG titer, ⭓1:800), and L. pneumophila
(total antibody titer, ⭓1:256) were performed as previously
described [3]. Specific antibodies to Brucella melitensis and My-
coplasma pneumoniae were detected with an immunoenzymatic
antibody test (titer, ⭓1:200) and the Platellia M. pneumoniae
IgM kit (Bio-Rad), respectively. Because of a lack of standard-
ization, antibodies to Mycoplasma hominis were not investi-
gated. When results of first-rank tests were negative, we sys-
tematically performed Western blot using Bartonella species
antigens [16], as described in the Appendix, which appears
only in the electronic version of the journal.
Molecular detection methods. Bacterial DNA was extract-
ed from surgically excised valves, or EDTA blood when no valve
was available, using the QIAmp Tissue kit (QIAGEN) as de-
scribed by the manufacturer. PCR primers and targets are de-
tailed in Table 4, and PCR and sequencing conditions are de-
scribed in the Appendix. Primer extension enrichment reaction
(PEER) was performed on valvular biopsies from patients for
whom results of other tests were negative, as previously de-
scribed [22, 23]. We used the buffy coat obtained from each
patient following antibiotic therapy as control tissue (“driver”).
Cell culture. Homogenized cardiac valve specimens suit-
able for culture—that is, frozen at ⫺80C following surgery
and sent in dry ice—and heparinized blood specimens pro-
cessed in the same way were inoculated onto human endothelial
cells (ECV 304) grown in shell vials, as reported elsewhere [24].
Three weeks after inoculation, bacteria detected by Gimenez
and acridine orange staining, electron microscopy, or immu-
nofluorescence using the patient’s serum, were identified by
amplification and sequencing of the 16S rRNA gene as previ-
ously described [8].

Blood Culture–Negative Endocarditis • CID 2010:51 (15 July) • 133

Table 2. Comparison of the Demographic Features and the Involved Cardiac Valves of the 476 Patients with an Etiological Agent
Identified with Those of the 73 Definite Patients without any Diagnosis and the 191 Possible Patients

Patients
P by univariate Relative risk (95% CI) P by multivariate
with an a a,b
analysis by univariate analysis analysis
identified Definite Possible
agent patients patients Definite Possible Definite Possible Definite Possible
Variable (n p 476) (n p 73) (n p 191) patients patients patients patients patients patients
Male-to-female ratio 3.2 1.4 1.9 !.01 !.01 1.1 (1.03–1.2) 1.2 (1.04–1.3) .4 .09
Age, years
Mean  SD 57.9  15.7 60.7 16.0 58.3 17.1 .15 .8
Median 58 62 59
Known preexisting valvular defect 94.3 94.0 83.4 .6 !.01 1.0 (0.9–1.1) 1.5 (1.1–1.9) ND !.01
Valve involved
Native valve 67.9 63.4 62.6 .5 .3 1.0 (0.9–1.1) 1.0 (0.9–1.2)

Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021

Bioprosthetic valve 16.7 26.8 15.4 .06 .7 0.9 (0.8–1.0) 1.0 (0.8–1.2)
Mechanical valve 8.4 7.0 16.3 .7 .01 1.0 (0.9–1.2) 0.8 (0.6–0.9) ND .4
Pace-maker 7.0 2.8 13.0 .3 .06 1.1 (1.0–1.2) 0.8 (0.6–1.04)
Aortic 45.5 39.4 47.1 .4 .7 1.0 (0.9–1.1) 1.0 (0.9–1.1)
Mitral 33.0 43.7 33.3 .1 .9 0.9 (0.8–1.0) 1.0 (0.9–1.1)
Aortic and mitral 12.3 11.3 3.2 .8 !.01 1.0 (0.9–1.1) 1.3 (1.1–1.4) ND .2
Tricuspid 2.0 2.8 1.6 .6 .9 0.9 (0.6–1.3) 1.0 (0.7–1.5)

NOTE. Data are percentage of patients, unless otherwise indicated. CI, confidence interval; ND, not determined.
a
P values !.05 are in boldface type.
b
Only variables for which the P value obtained in univariate analysis was ⭐.1 were included in the multivariate logistic regression analysis.

Histopathological analysis. Paraffin-embedded heart valves negative, we performed autoimmunohistochemistry as previ-

were examined with hematoxylin-eosin for histopathologic fea-
tures [4]. To detect microorganisms within tissues, the Giemsa,
Gram (Brown-Brenn and Brown-Hopps), periodic-acid Schiff,
Grocott-Gomori, Warthin-Starry, Gimenez, and Ziehl-Nielsen
stains were systematically performed as described elsewhere [25].
For patients for whom results of all other techniques remained
ously described [26].
Detection of autoantibodies. Presence of rheumatoid fac-
tor, antinuclear antibodies, and anti-DNA antibodies was de-
termined using the Rheumatoid Factor IgM kit (Orgentec),
ANA Hep2 kit (BMD), and MuST Connective kit (Inodiag),
respectively.

Table 3. Comparison of the Clinical Features, Laboratory Results, and Outcome of the 476 Patients with an Etiological Agent Identified
to Those of the 73 Definite Patients without any Diagnosis, and the 191 Possible Patients

Patients
P by univariate Relative risk (95% CI) P by multivariate
with an analysis
a
by univariate analysis analysisa,b
identified Definite Possible
agent patients patients Definite Possible Definite Possible Definite Possible
Variable (n p 476) (n p 73) (n p 191) patients patients patients patients patients patients
Previous antibiotic therapy 39.8 63.9 60.6 !.01 !.01 0.9 (0.8–0.9) 0.8 (0.7–0.9) .2 !.01
Clinical symptoms
Fever (temperature, ⭓38.5C) 92.0 93.1 81.4 .8 !.01 1.4 (1.1–1.8) 1.4 (1.1–1.8) ND !.01
Cardiac murmur 72.9 79.3 68.0 .3 .3 1.0 (0.9–1.0) 1.0 (0.9–1.2)
Arterial emboli 13.3 46.5 9.0 !.01 .2 0.7 (0.6–0.8) 1.1 (1.0–1.3) !.01 ND
Digital clubbingc 7.2 8.6 6.2 .7 .7 1.0 (0.8–1.1) 1.0 (0.8–1.3)
Osler nodesc 1.7 1.7 0.7 .9 .4 1.0 (0.7–1.3) 1.2 (0.9–1.6)
Glomerulonephritisc 9.4 6.9 5.5 .5 .1 1.0 (0.9–1.2) 1.1 (1.0–1.3)
Laboratory results
Leukocytosisc 49.0 51.7 47.2 .7 .7 1.0 (0.9–1.1) 1.0 (0.9–1.1)
c
Anemia 51.8 50.0 45.8 .8 .2 1.0 (0.9–1.1) 1.1 (1.0–1.2)
Rheumatoid factorc 6.9 31.0 11.1 !.01 .1 0.6 (0.5–0.8) 0.8 (0.6–1.1) !.01 ND
Death 3.1 6.8 5.8 .2 .2 0.8 (0.6–1.1) 0.8 (0.5–1.1)

NOTE. Data are percentage of patients, unless otherwise indicated. CI, confidence interval; ND, not determined.
a
P values !.05 are in boldface type.
b
Only variables for which the P value obtained in univariate analysis was ⭐.1 were included in the multivariate logistic regression analysis.
c
Data were not available for all patients.

134 • CID 2010:51 (15 July) • Fournier et al

 Table 4. Primers, Probes, and Polymerase Chain Reaction Conditions Used in This Study

Microorganisms
Primer name Nucleotide sequence, 5
r3
detected Molecular target Reference
a
536F CAGCAGCCGCGGTAATAC All bacteria 16S rRNA [8]
RP2b ACGGCTACCTTGTTACGACTT
a
CUF TCCGTAGGTGAACCTGCGG Fungi 18S rRNA [17]
b
CUR GCTGCGTTCTTCATCGATGC
IS1111fa CAAGAAACGTATCGCTGTGGC Coxiella burnetii htpAB-associated element [18]
b
IS1111R CACAGAGCCACCGTATGAATC
c
IS1111probe CCGAGTTCGAAACAATGAGGGCTG
ITS Fa GGGGCCGTAGCTCAGCTG Bartonella species 16S-23S rRNA spacer [19]
ITS Rb TGAATATATCTTCTCTTCACAATTTC

Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021

ITSprobec CGATCCCGTCCGGCTCCACCA
a
199F GGTTTCTCTGTTACATGTATGTC Tropheryma whipplei WISP family protein Present study
492Rb AACCCTGTCCTGCACCCC
c
TWprobe CTTTGTTATGGAGATTACTTTCTCATCTCC
ChlamOmp2da ATGTCCAAACTCATCAGACGAG Chlamydia species ompD2 Present study
ChlamOmp2rb CCTTCTTTAAGAGGTTTTACCCA
CMV Fa GCAGCCACGGGATCGTACT Cytomegalovirus pp65 protein [20]
b
CMV R GGCTTTTACCTCACACGAGCATT
CMVprobec CGCGAGACCGTGGAACTGCG
a
EnteroV F CCCTGAATGCGGCTAATCC Enteroviruses Polyprotein [21]
EnteroV Rb ATTGTCACCATAAGCAGCCA
EnteroVprobec CADGGACACCCAAAGTAGTCGGTTCC

NOTE. rRNA, ribosomal RNA; WISP, WNT1-inducible-signaling pathway.

a
Forward primer.
b
Reverse primer.
c
5
-FAM–3
-TAMRA probe.

Statistical Methods
To identify the variables associated with absence of diagnosis,
we compared the 476 patients with an identified etiological
agent with the 73 definite patients without etiological agents
and with the 191 possible patients, using the Mantel-Haenszel
x2 test. Mean ages were compared using the Anova and
Bartlett’s tests. The variables independently associated with the
negativity of all diagnostic tests were identified using an un-
conditional logistic regression analysis (Tables 2 and 3 and the
Supplementary Results in the Appendix). We also compared
the diagnoses obtained by our strategy with those of other
published BCNE series, using the Mantel-Haenszel x2 test. All
tests were performed using the EPI info software, version 3.3.2
(http://www.cdc.gov/epiinfo/index.htm). Observed differences
were considered significant when P was !.05 for 2-tailed tests.
RESULTS
Patients. Specimens from 819 new patients from France or
abroad were included in this study (Figure 1). By using the
modified Duke criteria [15], 60 patients (7.3%) were excluded
from the diagnosis of endocarditis, including 57 who did not
meet criteria for possible endocarditis, and 3 for whom his-
topathological results identified angiosarcoma (2 patients ) and
a myxoma of the left atrium (1 patient). Among the remaining
759 patients with endocarditis, 19 (2.5%) were classified as
having noninfective endocarditis. Among these, histopathology
identified a marantic endocarditis, Libmann-Sacks endocarditis,
and Behcet disease in 7, 4, and 1 cases, respectively. Subse-
quently, by evaluating the presence of antinuclear antibodies
of 129 of 290 patients for whom results of all assays were
negative and by calling their physicians in charge, we identified
an additional 7 patients in whom a diagnosis of autoimmune
disease had been done elsewhere using diagnostic criteria [27,
28], including 5 patients with Libmann-Sacks endocarditis and
2 with rheumatic arthritis. Of the 740 patients putatively clas-
sified as having infective endocarditis, 549 (74.2%) were clas-
sified as having definite endocarditis using the Duke criteria
[15], including 476 for whom our diagnostic strategy allowed
identification of an infectious agent (Table A1 in the online
Appendix), and 73 for whom no etiological agent could be
found. The remaining 191 patients (25.8%) were classified as
possible patients. The epidemiological data of the 740 patients
with definite or possible endocarditis are presented in Table 1.
Diagnostic procedures. Serological analysis using immu-
nofluorescence assay provided a diagnosis for 356 (47.8%) of
745 tested patients (Figure 2). Chronic Q fever (IgG titer to

Blood Culture–Negative Endocarditis • CID 2010:51 (15 July) • 135


Figure 2. Diagnostic tests applied to clinical specimens for identification of causative agents of blood culture–negative endocarditis. Agents include
Tropheryma whipplei. AIHC, autoimmunohistochemistry; PCR, polymerase chain reaction; PEER, primer extension enrichment reaction; rRNA, ribosomal
RNA.
phase I C. burnetii, 11:800) was diagnosed in 274 patients
(77%). Eighty patients (22.5%) had an IgG titer to B. quintana
and/or B. henselae ⭓1:800 (Table A1). One patient had a titer
of 1:2048 to L. pneumophila and a positive result of L. pneumo-
phila urinary antigen assay. One patient had a titer of 1:256 to
Legionella anisa and was later diagnosed as having L. long-
beachae endocarditis on the basis of both culture and PCR
of a valvular biopsy. Among the patients for whom no diag-
nosis was obtained with other tests, serum was available for
Bartonella Western blot for 148 patients. For these patients, a
diagnosis of B. quintana infection was obtained for 4 patients
who were seronegative using immunofluorescence assay.
Additionally, 16S ribosomal DNA (rDNA) and 18S rDNA
PCR assays performed on EDTA blood provided diagnoses for
35 (13.6%) and 1 (0.4%) of 257 patients, respectively. Also,
16S rDNA and 18S rDNA PCR assays of valvular specimens
were positive for 150 (66.1%) and 7 (3.0%) of 227 patients,
respectively (Table A1). PCR amplification of C. burnetii, Bar-
tonella species, and Tropheryma whipplei was positive for 16,
12, and 5 EDTA blood specimens, respectively, and for 30, 26
and 17 valvular biopsies, respectively. Results for negative con-
trols were in all assays. Overall, PCR identified a causative agent
in 109 seronegative patients, including 106 by PCR of valvular
specimens and 3 by PCR of blood only. PCR assays for cyto-
megalovirus, enteroviruses, and C. pneumoniae were negative
for all tested specimens. PEER, performed on 49 valvular bi-
opsies from patients for whom all other diagnostic methods
failed to identify a causative agent, identified a microorganism
136 • CID 2010:51 (15 July) • Fournier et al
Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021
in 4 patients: Mycobacterium tuberculosis, Micrococcus luteus,
Candida dubliniensis, and Cryptococcus laurentii. Sequences ob-
tained from these microorganisms have been deposited in
GenBank under accession numbers EU139422, EU139419,
EU139420, and EU139421, respectively. Results for negative
controls remained negative.
Cell culture of heparinized blood allowed bacterial recovery
in 7 (4.1%) of 169 patients. When applied to valvular biopsies,
culture was positive for 58 (45.7%) of 127 patients. Culture
did not provide any diagnosis that was not made by another
method. Detailed results are presented in Table A1.
Culture and PCR of valvular biopsies (positive for 58 of 127
patients and 157 of 227 patients, respectively) were significantly
more sensitive than were culture and PCR of blood (positive
for 7 of 169 patients and 36 of 257 patients, respectively; P !
.01 for both). In addition, PCR was significantly more sensitive
than was culture for valvular biopsies (P ! .01 ). The sensitivity
of our diagnostic strategy (478 etiological agents identified
among 740 patients [64.6%]) was significantly higher than that
obtained for a previous series from France (15 [17.0%] of 88;
P ! .01), Great Britain (31 [49.2%] of 63; P p .01), and Algeria
(28 [45.1%] of 62; P ! .01) but was significantly smaller than
that obtained for our previous series from France (271 [77.9%]
of 348; P ! .01) (Table 5). However, in the latter study, we had
considered only patients classified as having definite endocar-
ditis and identified an agent different from C. burnetii and
Bartonella species in only 5 patients, compared with 110 pa-
tients in the present study (P ! .01).
 Presence of rheumatoid factor was detected in 73 patients.
Among 115 tested patients without diagnosis, antinuclear an-
tibodies were found in 10, including 5 who also had anti-DNA
antibodies. By contacting the physicians in charge of these 10
patients, we had confirmation that 5 patients received a di-
agnosis of systemic lupus erythematosus and 2 received a di-
agnosis of rheumatic fever.
Histopathological examination of valvular biopsies enabled
the diagnosis of an angiosarcoma and a myxoma of the left
atrium in 2 and 1 patients, respectively, and a diagnosis of
noninfective endocarditis in 12 patients: marantic endocarditis
in 7, Libman-Sachs endocarditis in 4, and cardiac involvement
detection from valvular biopsies, identified a causative micro-
bial agent in 476 (62.7%) of 759 tested patients with BCNE
and a high incidence of noninfective endocarditis (2.5%). Our
current strategy allowed the identification of a significantly
greater variety of etiological agents than was identifed in a
previous series from our laboratory (P ! .01 ) [3]. However, new
techniques introduced in this study to detect new agents, such
as autoimmunohistochemistry and PEER, were disappointing.
Apart from serological analysis by immunofluorescence as-
say, which provided the greatest number of diagnoses (356
[74.8%] of 476) (Table A1), PCR was the second most efficient
diagnostic method and provided a diagnosis for 109 patients

Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021

of Behcet disease in 1. Valvular and serum specimens suitable
for autoimmunohistochemistry were available from 52 patients
without any identified etiology. Results were negative except
for 1 patient in whom cocci were observed (Figure 3). However,
results of all other tests performed on this surgically excised
valve, including PEER, 16S rDNA PCR, and 18S rDNA PCR,
remained negative.
DISCUSSION
In the present series, in an effort to increase our detection
sensitivity for the diagnosis of BCNE, we used a multimodal
diagnostic strategy, including validated methods for the iden-
tification of endocarditis agents, assays newly applied for this
purpose, and assays for the detection of microorganisms of
uncertain role. Our study, in addition to reporting the largest
diagnostic series of BCNE to date and the largest series of PCR
for whom serological results were negative. However, the PCR
output varied depending on specimens. When applied to blood,
16S rRNA amplification exhibited a poor sensitivity, with 35
positive specimens of 257 tested (13.6%). This low sensitivity
was also recently reported by Casalta et al [32]. In contrast,
dedicated real-time PCR assays targeting small DNA fragments
and using fluorescent probes exhibited a high sensitivity for
Bartonella species, C. burnetii, and T. whipplei, as previously
described [33]. However, in our study, PCR of blood did not
provide any additional diagnosis of Q fever with regard to
serological analysis. Therefore, our data suggest that dedicated
real-time PCR assays, in particular those targeting Bartonella
species and T. whipplei, might be valuable for use on EDTA
blood, in particular for patients for whom valvular biopsies are
not available. In addition, by using a broad-range, 18S rRNA–
based PCR assay targeting fungi, we identified a Penicillium

Table 5. Comparison of Microorganisms Identified in Published Series of Blood Culture–Negative

Endocarditis

Study by location [reference]

a
Present study France [3] France [29] Great Britain [30] Algeria [31]
Microorganism (n p 740) (n p 348) (n p 88) (n p 63) (n p 62)
Bartonella species 12.4 28.4 0 9.5 22.6
Brucella melitensis 0 0 0 0 1.6
Chlamydia species 0 0 2.2 1.6 0
Corynebacterium 0.5 0 1.1 0 1.6
species
Coxiella burnetii 37.0 48 7.9 12.7 3.2
Enterobacteriaceae 0.5 0 0 0 0
HACEK bacteria 0.5 0 0 0 3.2
Staphylococcus 2.0 0 3.4 11.1 6.4
species
Streptococcus 4.4 0 1.1 6.3 3.2
species
Tropheryma whipplei 2.6 0.3 0 0 0
Other bacteria 3.0 1.1 1.1 1.6 1.6
Fungi 1.0 0 0 6.3 1.6
No etiology 36.5 22.1 82.9 50.8 54.8

NOTE. Data are percentages. HACEK, Haemophilus, Actinobacillus, Cardiobacterium, Eikenella, Kingella.
a
Patients classified as excluded were not included in this analysis.

Blood Culture–Negative Endocarditis • CID 2010:51 (15 July) • 137

 Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021
Figure 3. a, Immunohistochemical detection of bacteria from the cardiac valve of a patient with culture-negative endocarditis, using the patient’s
serum. Note the extracellular location of bacteria revealed by the peroxidase method and hematoxylin counterstain (original magnification, 400⫻).
b, Immunohistochemical detection of Tropheryma whipplei in a resected cardiac valve from a patient with a Whipple endocarditis, using a rabbit
polyclonal antibody and hematoxylin counterstain. Note the intracellular location of the bacteria in the macrophage cytoplasm (original magnification,
200⫻). c, Immunohistochemical detection of Coxiella burnetii in a resected cardiac valve from a patient with Q fever endocarditis, using a mouse
monoclonal antibody and hematoxylin counterstain. Note the intracellular location of the bacteria in the macrophage cytoplasm (original magnification,
400⫻). d, Immunohistochemical detection of Bartonella henselae in a resected cardiac valve from a patient with Bartonella endocarditis, using a
rabbit polyclonal antibody and hematoxylin counterstain. Note the extracellular location of the bacteria in the valvular vegetation (original magnification,
400⫻).

species as an agent in a patient. To date, PCR assays targeting
fungi have been used in a limited number of studies [9, 11].
Given the fact that these agents are not covered by most em-
pirical antibiotic therapies used for BCNE, we propose that this
assay should form part of the diagnostic strategy for BCNE.
Valvular biopsies, when available, proved to be of great di-
agnostic value, allowing the detection of a microorganism in
157 patients. The 16S rDNA and 18S rDNA PCR assays of
valves were positive for 150 (66.1%) and 7 (3.0%) of 227 pa-
tients, respectively (Table A1). In recent studies, broad-range
PCR for bacteria was demonstrated to be highly valuable for
valvular specimens [3, 6–11, 14], with the identification of
streptococci and fastidious bacteria as main agents (Table 5).
In our study, the main bacterial agents identified by PCR were
streptococci (mainly Streptococcus gallolyticus), T. whipplei, Bar-
tonella species, and staphylococci. These diagnoses suggest that,
unlike for blood specimens, broad-range PCR may be preferred
as the first-line test for valvular specimens. We detected no viral
agent or Chlamydia species. Among assays newly applied for
this purpose, Western blot for Bartonella species, PEER, and
autoimmunohistochemistry provided a diagnosis for 4, 4, and
1 patients, respectively, for whom results of all other methods
were negative. Such findings suggest that these methods should
be considered only when others fail.
Of the identified agents, zoonotic bacteria were the most
frequent (Supplementary Comments and Table A1 in the Ap-

138 • CID 2010:51 (15 July) • Fournier et al

pendix). C. burnetii was the main identified etiological agent
(57.3%), followed by Bartonella species (19.2%). Because we
acknowledge the fact that this prevalence may be caused by a
sampling bias, we differentiated patients from Marseille and
those from other hospitals (Table A1). In both populations, C.
burnetii and Bartonella species were predominant, accounting
for 46.0% and 15.1% of identified agents in Marseille, respec-
tively, and for 61.4% and 20.7% of identified agents in other
hospitals, respectively. Such results confirm the role of C. bur-
netii as a major agent of BCNE in France and other countries
[29, 30, 34, 35], followed by Bartonella species. For these agents,
the diagnosis was mostly obtained by serological analysis, which
demonstrates the necessity to systematically include antibody
detection for C. burnetii and Bartonella species in the diagno-
sis of BCNE (Figure 2) [36]. Other fastidious bacteria, such as
T. whipplei (4.0%), Legionella species (0.4%), mycobacteria
(0.4%), M. hominis (0.2%), Gemella morbillorum (0.2%), and
Abiotrophia defectiva (0.2%) accounted for 5.4% of identified
agents. Fungi were identified in 8 patients (1.7%). In previous
series of BCNE, fungi accounted for up to 6.3% of patients
[30], which suggests the need to systematically search for these
microorganisms in BCNE. The usual bacteria that may have
been inactivated by early antibiotic therapy, in particular strep-
tococci (6.9%) and staphylococci (3.1%), accounted for 16.5%
of cases. The role of antibiotics in the absence of diagnosis,
well-recognized in previous studies [37–39], was emphasized
by our statistical analysis (P ! .01 ) (Table 3). Among strepto-
cocci, S. gallolyticus was the main species (1.9% of microor-
ganisms), exclusively detected in elderly patients.
An unexpected finding of our study was the identification
of 19 patients with noninfective endocarditis, including 7 cases
of marantic and 9 cases of Libmann-Sacks endocarditis. As a
matter of fact, in a preliminary study, the demonstration of a
significant association between an elevated rheumatoid factor
rate and absence of diagnosis, as observed in our statistical
analysis (P ! .01) (Table 3), motivated us to evaluate the pres-
ence of antinuclear antibodies and to call physicians of patients
without identified etiology and enabled the retrospective iden-
tification of 5 more patients with systemic lupus erythematosus
and 2 with rheumatoid arthritis. These data demonstrate that
the BCNE picture may include cases of nonbacterial thrombotic
endocarditis associated with cancers and autoimmune diseases
[4, 40–42]. To the best of our knowledge, the prevalence of
noninfective etiologies of endocarditis has not previously been
evaluated. In our study, noninfective endocarditis accounted
for 2.5% of cases of BCNE. Our results support the usefulness
of a systematic detection of rheumatoid factor and antinuclear
antibodies, as well as histopathological analysis of all valvular
biopsies taken from patients with suspected endocarditis.
On the basis of our results, we propose a refined strategy for
the diagnosis of BCNE (Figure 2). Serum should be considered
a priority specimen, with Q fever and Bartonella serological
analysis being systematic, followed by specific PCR assays for
Bartonella species, T. whipplei, and fungi. Other serological anal-
yses, Bartonella-specific Western blot, and broad-range PCR of
blood [32] should be reserved for patients with negative results.
We also suggest that detection of antinuclear antibodies and
rheumatoid factor should be systematic [43]. When available,
valvular biopsies should first be tested using broad-range PCR
assays for bacteria and fungi, with autoimmunohistochemistry
being reserved for patients without any identified diagnosis.
Finally, given the poor sensitivity of commercially available PCR
kits for detection in blood [32], we believe that highly sensitive
Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021
PCR assays specifically targeting staphylococci and streptococci,
in particular S. gallolyticus, may be valuable detection tools for
blood specimens in the future.
Acknowledgments
We thank Carine Almani, Karine Puggioni, and Marielle Bedotto for
technical help.
Financial support. Protocole Hospitalier de Recherche Clinique Ré-
gional 2005.
Potential conflicts of interest. D.R. is cofounder of the biotechnology
company Inodiag (http://www.inodiag.com) and is inventor of a patent on
a serological diagnostic method for endocarditis held by the Université de
la Méditerranée, which was not used in this study. All other authors: no
conflicts.
References
1. Brouqui P, Raoult D. Endocarditis due to rare and fastidious bacteria.
Clin Microbiol Rev 2001; 14:177–207.
2. Raoult D, Casalta JP, Richet H, et al. Contribution of systematic se-
rological testing in diagnosis of infective endocarditis. J Clin Microbiol
2005; 43:5238–42.
3. Houpikian P, Raoult D. Blood culture-negative endocarditis in a ref-
erence center: etiologic diagnosis of 348 cases. Medicine (Baltimore)
2005; 84:162–73.
4. Lepidi H, Durack DT, Raoult D. Diagnostic methods current best
practices and guidelines for histologic evaluation in infective endo-
carditis. Infect Dis Clin North Am 2002; 16:339–61.
5. Greub G, Lepidi H, Rovery C, et al. Diagnosis of infectious endocarditis
in patients undergoing valve surgery. Am J Med 2005; 118:230–8.
6. Podglajen I, Bellery F, Poyart C, et al. Comparative molecular and mi-
crobiologic diagnosis of bacterial endocarditis. Emerg Infect Dis 2003; 9:
1543–7.
7. Breitkopf C, Hammel D, Scheld HH, Peters G, Becker K. Impact of a
molecular approach to improve the microbiological diagnosis of in-
fective heart valve endocarditis. Circulation 2005; 111:1415–21.
8. Goldenberger D, Kunzli A, Vogt P, Zbinden R, Altwegg M. Molecular
diagnosis of bacterial endocarditis by broad-range PCR amplification
and direct sequencing. J Clin Microbiol 1997; 35:2733–9.
9. Grijalva M, Horvath R, Dendis M, Erny J, Benedik J. Molecular di-
agnosis of culture negative infective endocarditis: clinical validation in
a group of surgically treated patients. Heart 2003; 89:263–8.
10. Bosshard PP, Kronenberg A, Zbinden R, Ruef C, Bottger EC, Altwegg
M. Etiologic diagnosis of infective endocarditis by broad-range poly-
merase chain reaction: a 3-year experience. Clin Infect Dis 2003; 37:
167–72.
11. Millar B, Moore J, Mallon P, et al. Molecular diagnosis of infective en-
docarditis—a new Duke’s criterion. Scand J Infect Dis 2001; 33:673–80.
12. Gauduchon V, Chalabreysse L, Etienne J, et al. Molecular diagnosis of
Blood Culture–Negative Endocarditis • CID 2010:51 (15 July) • 139
 infective endocarditis by PCR amplification and direct sequencing of
DNA from valve tissue. J Clin Microbiol 2003; 41:763–6.
13. Berlot G, Bussani R. Cytomegalovirus endocarditis: a case report and
a review of the literature. Minerva Anestesiol 2003; 69:801–5.
14. Rovery C, Greub G, Lepidi H, et al. PCR detection of bacteria on cardiac
valves of patients with treated bacterial endocarditis. J Clin Microbiol
2005; 43:163–7.
15. Li JS, Sexton DJ, Mick N, et al. Proposed modifications to the Duke
criteria for the diagnosis of infective endocarditis. Clin Infect Dis 2000;30:
633–8.
16. Houpikian P, Raoult D. Western immunoblotting for Bartonella en-
docarditis. Clin Diagn Lab Immunol 2003; 10:95–102.
17. White TJ, Bruns T, Lee S, Taylor JW. Amplification and direct se-
quencing of fungal ribosomal RNA genes for phylogenetics. In: Innis
MA, Gelfand DH, Sninsky JJ, White TJ, eds. PCR protocols: a guide
to methods and applications. New York: Academic Press, 1990:315–22.
18. Willems H, Thiele D, Frolich-Ritter R, Krauss H. Detection of Coxiella
burnetii in cow’s milk using the polymerase chain reaction. Zentralbl
Veterinarmed B 1994; 41:580–7.
19. Roux V, Raoult D. Inter-and intraspecies identification of Bartonella
(Rochalimea) species. J Clin Microbiol 1995; 33:1573–9.
20. Griscelli F, Barrois M, Chauvin S, Lastere S, Bellet D, Bourhis JH.
Quantification of human cytomegalovirus DNA in bone marrow trans-
plant recipients by real-time PCR. J Clin Microbiol 2001; 39:4362–9.
21. Watkins-Riedel T, Woegerbauer M, Hollemann D, Hufnagl P. Rapid
diagnosis of enterovirus infections by real-time PCR on the LightCycler
using the TaqMan format. Diagn Microbiol Infect Dis 2002; 42:99–105.
22. Ganova-Raeva L, Zhang X, Cao F, Fields H, Khudyakov Y. Primer
extension enrichment reaction (PEER): a new subtraction method for
identification of genetic differences between biological specimens. Nu-
cleic Acids Res 2006; 34:e76.
23. Biagini P, de Lamballerie X, de Micco P. Effective detection of highly
divergent viral genomes in infected cell lines using a new subtraction
strategy (primer extension enrichment reaction—PEER). J Virol Meth-
ods 2007; 139:106–10.
24. Raoult D, Vestris G, Enea M. Isolation of 16 strains of Coxiella burnetii
from patients by using a sensitive centrifugation cell culture system
and establishment of strains in HEL cells. J Clin Microbiol 1990; 28:
2482–4.
25. Bruneval P, Choucair J, Paraf F, et al. Detection of fastidious bacteria
in cardiac valves in cases of blood culture negative endocarditis. J Clin
Pathol 2001; 54:238–40.
26. Lepidi H, Coulibaly B, Casalta JP, Raoult D. Autoimmunohistochem-
istry: a new method for the histologic diagnosis of infective endocar-
ditis. J Infect Dis 2006; 193:1711–7.
27. Hochberg MC. Updating the American College of Rheumatology re-
vised criteria for the classification of systemic lupus erythematosus.
Arthritis Rheum 1997; 40:1725.
28. Arnett FC, Edworthy SM, Bloch DA, et al. The American Rheumatism
Association 1987 revised criteria for the classification of rheumatoid
arthritis. Arthritis Rheum 1988; 31:315–24.
29. Hoen B, Selton-Suty C, Lacassin F, et al. Infective endocarditis in
140 • CID 2010:51 (15 July) • Fournier et al
patients with negative blood cultures: analysis of 88 cases from a one-
year nationwide survey in France. Clin Infect Dis 1995; 20:501–6.
30. Lamas CC, Eykyn SJ. Blood culture negative endocarditis: analysis of
63 cases presenting over 25 years. Heart 2003; 89:258–62.
31. Benslimani A, Fenollar F, Lepidi H, Raoult D. Bacterial zoonoses and
infective endocarditis, Algeria. Emerg Infect Dis 2005; 11:216–24.
32. Casalta JP, Gouriet F, Roux V, Thuny F, Habib G, Raoult D. Evaluation
of the LightCycler SeptiFast test in the rapid etiologic diagnosis of in-
fectious endocarditis. Eur J Clin Microbiol Infect Dis 2009; 28:569–73.
33. Fenollar F, Lepidi H, Raoult D. Whipple’s endocarditis: review of the
literature and comparisons with Q fever, Bartonella infection, and blood
culture–positive endocarditis. Clin Infect Dis 2001; 33:1309–16.
34. Scott JW, Baddour LM, Tleyjeh IM, Moustafa S, Sun YG, Mookadam
F. Q fever endocarditis: the Mayo Clinic experience. Am J Med Sci 2008;
336:53–7.
35. Marin M, Munoz P, Sanchez M, et al. Molecular diagnosis of infective
Downloaded from https://academic.oup.com/cid/article/51/2/131/301131 by guest on 09 January 2021
endocarditis by real-time broad-range polymerase chain reaction (PCR)
and sequencing directly from heart valve tissue. Medicine (Baltimore)
2007; 86:195–202.
36. Raoult D, Abbara S, Jassal DS, Kradin RL. Case records of the Mas-
sachusetts General Hospital. Case 5–2007: a 53-year-old man with a
prosthetic aortic valve and recent onset of fatigue, dyspnea, weight loss,
and sweats. N Engl J Med 2007; 356:715–25.
37. Grace CJ, Lieberman J, Pierce K, Littenberg B. Usefulness of blood
culture for hospitalized patients who are receiving antibiotic therapy.
Clin Infect Dis 2001; 32:1651–5.
38. Branger S, Casalta JP, Habib G, Collard F, Raoult D. Escherichia coli
endocarditis: seven new cases in adults and review of the literature.
Eur J Clin Microbiol Infect Dis 2005; 24:537–41.
39. Werner M, Andersson R, Olaison L, Hogevik H. A 10-year survey of
blood culture negative endocarditis in Sweden: aminoglycoside therapy
is important for survival. Scand J Infect Dis 2008; 40:279–85.
40. Asopa S, Patel A, Khan OA, Sharma R, Ohri SK. Non-bacterial throm-
botic endocarditis. Eur J Cardiothorac Surg 2007; 32:696–701.
41. DeLong CE, Roldan CA. Noninfective endocarditis in rheumatoid ar-
thritis. Am J Med 2007; 120:e1–2.
42. Horino T, Takao T, Taniguchi Y, Terada Y. Non-infectious endocardi-
tis in a patient with cANCA-associated small vessel vasculitis. Rheuma-
tology (Oxford) 2009; 48:592–4.
43. Gouriet F, Bothelo-Nevers E, Coulibaly B, Raoult D, Casalta JP. Eval-
uation of sedimentation rate, rheumatoid factor, C-reactive protein,
and tumor necrosis factor for the diagnosis of infective endocarditis.
Clin Vaccine Immunol 2006; 13:301.
44. Fenollar F, Fournier PE, Carrieri MP, Habib G, Messana T, Raoult D.
Risks factors and prevention of Q fever endocarditis. Clin Infect Dis
2001; 33:312–6.
45. Brouqui P, Raoult D. New insight into the diagnosis of fastidious
bacterial endocarditis. FEMS Immunol Med Microbiol 2006; 47:1–13.
46. Fournier PE, Lelievre H, Eykyn SJ, et al. Epidemiologic and clinical
characteristics of Bartonella quintana and Bartonella henselae endocar-
ditis: a study of 48 patients. Medicine (Baltimore) 2001; 80:245–51.
47. Pierrotti LC, Baddour LM. Fungal endocarditis, 1995–2000. Chest
2002; 122:302–10.