|Home | About | Journals | Submit | Contact Us | Français|
Cat scratch disease (CSD) is mostly due to Bartonella henselae after inoculation of the organism through a skin injury. Since the causative bacteria cannot be easily cultured from human lymph node samples, the diagnosis usually relies on epidemiological, clinical, histological, and serological criteria (classical criteria). A study was performed to determine the diagnostic value of PCR analysis for the detection of B. henselae for the diagnosis of CSD and its place in the diagnostic strategy alongside the classical criteria. Over a 7-year period, lymph node biopsy specimens or cytopunctures from 70 patients were systematically tested by PCR for the presence of B. henselae DNA (htrA gene) in the Bacteriology Laboratory of the Hôpitaux Universitaires de Strasbourg. Serological testing by an immunofluorescence assay for B. henselae antibodies was also performed for each patient, and clinical, epidemiological, and histological data were collected. The patients were then divided into two groups according to the number of positive diagnostic criteria for CSD: 29 patients with definite CSD (two or more classical criteria) and 15 patients with possible CSD (less than two classical criteria). The remaining 26 patients for whom another diagnosis was retained were used as a control group. Among all criteria, PCR analysis had the best specificity (100%). The PCR assay for B. henselae was positive for 22 (76%; 95% confidence interval [CI95], 56.5 to 89.7%) of the 29 definite CSD patients and 3 (20%; CI95, 4.3 to 48.1%) of the 15 possible CSD patients. We then studied combinations of diagnostic criteria, including B. henselae PCR analysis. The best diagnostic performance was observed if at least two criteria were present among serologic, epidemiologic, histological, and molecular criteria.
Cat scratch disease (CSD) is the most frequent clinical manifestation of Bartonella infections in immunocompetent patients (8, 9, 18, 19, 23, 28). Bartonella henselae, the main causative agent of CSD, can be detected in the blood of healthy cats (15), and cats can transmit Bartonella to humans after a skin injury caused by a scratch or bite (19). The disease was first described by Debré et al. in 1950 on the basis of epidemiological and clinical data (8), and the causality of B. henselae in CSD has since been demonstrated by serological and molecular assays (4, 23, 24, 25, 29).
CSD appears as regional lymph node enlargement after a cat scratch or bite in the same area. The clinical manifestations include inflammatory lymphadenopathy, which appears 1 to 7 weeks after the injury, and a papular lesion of the skin, which develops at the site of the injury. The diagnostic challenge for the physician is to prove or invalidate the CSD etiology in the face of a patient with lymph node enlargement. In most cases, the diagnosis is based on a combination of clinical, epidemiological, serological, and histological data. According to Bergmans et al. (5), a diagnosis of CSD usually requires three of the following four criteria: (i) a history of contact with a cat and the presence of a scratch or primary lesion of the skin, eye, or mucous membrane; (ii) a positive cat scratch skin test reaction; (iii) negative laboratory testing for other causes of lymphadenopathy; and (iv) characteristic histopathological findings in a lymph node biopsy specimen or at a site of systemic involvement. However, none of these criteria are sufficiently specific to establish a diagnosis of CSD. In addition, the CSD intradermal skin test (8) is no longer available, and a history of a cat injury is sometimes not reported by the patient. Another possibility for diagnosis is histological examination of the lymph node involved; but this cannot differentiate between several infectious etiologies, including tularemia, bartonellosis, and other inoculation diseases. A typical CSD histology showing a granuloma with central necrosis, multinucleated giant cells, and microabscesses may also be absent. Thus, histological examination at an early stage of the disease in fact shows only lymphoid hyperplasia and arteriolar proliferation. Conversely, in the presence of granuloma, a differential diagnosis with respect to tuberculosis or other infectious diseases that display granulomas can be very difficult. Serology fails in terms of specificity and/or sensitivity (5, 10), while culture of B. henselae from lymph node tissue samples is difficult and has been reported in only a very limited number of cases (18). The detection of specific DNA fragments by PCR has been proposed as a novel method for demonstration of the presence of B. henselae in CSD (3, 4, 5, 18, 21, 22, 25). Its utility nevertheless remains to be assessed, and there is still no established “gold standard” for the diagnosis of cat scratch disease.
The aim of our study was to determine the diagnostic value of B. henselae detection by PCR in CSD and the place of PCR among the other usual diagnostic tools. An observational study was conducted on the basis of data collected prospectively from patients with inflammatory lymph node enlargement requiring biopsy or adenectomy.
Over the period from 1993 to 2000, we prospectively collected the following general and clinical data from every patient consulting at the Hôpitaux Universitaires de Strasbourg for local superficial lymphadenopathy and for whom a cytopuncture or biopsy of the lymph node was performed: age, gender, medical history, localization of the lymph nodes affected, contact with domestic or wild animals, and the presence of a scratch or bite by a domestic or wild animal and its site. Part of the lymphoid tissue was sent to the Bacteriology Laboratory for testing for the usual bacteria and PCR assay for B. henselae. In the case of biopsy samples, another portion was subjected to a histological examination. For each patient, a standard serodiagnostic test for B. henselae was performed in the Bacteriology Laboratory of the Hôpitaux Universitaires de Strasbourg. All patients were reexamined after 2 to 6 months to record the evolution of the adenopathy and determine the validity of the CSD diagnosis. Finally, we focused our study on 44 patients who had provided complete data concerning their medical history and contact with animals and for whom serology testing and PCR analysis B. henselae had been carried out in our laboratory.
For these 44 patients with lymphadenopathy, the final diagnosis of CSD was made or not on the basis of the presence or the absence of the following three additional criteria: (i) close contact with cats or a scratch or bite from a cat, (ii) a typical CSD histology, i.e., granuloma with a central pyogenic abscess (lymphoid hyperplasia not being sufficiently specific to establish a diagnosis of CSD), and (iii) positive serology by an immunofluorescence assay for antibodies against B. henselae. These 44 patients were thus divided into two groups, as follows: the first group of 29 patients was classified as definitely having CSD, according to the presence of at least two of the above three criteria, and the second group of 15 patients had possible CSD and presented with only one or none of the criteria for CSD given above.
Twenty-six lymph node samples from patients for whom a diagnosis other than CSD had been established on the basis of histological criteria or bacteriological tests (positive serology or bacterial or mycobacterial cultures) were used as negative controls.
The lymph node specimens were cut into small pieces by using a sterile scalpel blade. Approximately 40 mg of tissue was then washed twice in 0.5 ml of sterile phosphate-buffered saline (PBS); and the tissue was treated with 500 μg/ml of proteinase K (Sigma) in 1 ml of 10 mM Tris HCl (pH 8.0) containing 0.5% Nonidet P-40 (Sigma), 0.5% Tween 20, 50 mM KCl, and 50 mM MgCl2 at 55°C and with 30 s of vigorous shaking every 15 min for 2 h or until the tissue was entirely digested. The DNA was extracted with phenol-chloroform, ethanol precipitated, air dried, resuspended in 40 μl of TE buffer (10 mM Tris HCl, pH 8.0, 1 mM EDTA), and heated to 95°C for 10 min. A 3-μl aliquot of this suspension was amplified by PCR.
The detection of B. henselae in CSD lymph nodes with the primers and internal hybridization probe employed in this study has previously been described by Anderson et al. (3). These primers target a 414-bp fragment in the htrA gene of B. henselae.
A 3-μl aliquot of the DNA suspension extracted from a lymph node tissue sample was used as the template for 40 cycles of DNA amplification. PCR amplification was performed in 20 μM Tris HCl (pH 8.4) containing 50 mM KCl, 3 mM MgCl2, 1.5 units of Taq polymerase (Invitrogen, Cergy Pontoise, France), 0.2 μM of each primer, and 0.2 μM of each of the four deoxyribonucleotides in a final volume of 100 μl.
To avoid DNA contamination of the samples, the precautions recommended by Kwok and Higuchi (17) were taken. Sample preparation, PCR amplification, and electrophoresis were performed with separate sets of pipettes and the wearing of protective laboratory coats and caps in three different closed rooms where B. henselae had never been cultured. At each step of sample preparation, each tube was carefully and separately uncovered. Gloves were changed between the handling of each sample, and all solutions were manipulated by using pipettes with hydrophobic filter tips (Multiguard; Sorenson).
In each run of four coded tissue samples, three negative controls were added. The first consisted of the reaction mixture without any DNA template, while the second contained DNA from a strain of Bartonella other than B. henselae. In order to detect sample-to-sample contamination during DNA preparation, a third control consisting of a 0.5-ml aliquot of a tissue sample from a patient with a noninfectious disease was blindly and simultaneously processed and amplified with the four other samples.
To monitor the DNA amplification efficiency, a positive control containing 1 pg of purified B. henselae DNA (ATCC 49882) was included in each run. All positive samples were checked by processing and amplification of another frozen aliquot of the same tissue specimen. All negative samples were amplified again after addition of 1 pg of purified B. henselae DNA, in order to detect a possible inhibitor of the amplification reaction.
PCR amplification was performed in an Applied Biosystems 9700 thermal cycler. After predenaturation for 3 min at 94°C, samples were amplified through 40 cycles of 93°C for 30 s, 55°C for 30 s, and 72°C for 60 s, followed by a final extension step of 8 min at 72°C.
A 10-μl aliquot from each PCR tube was electrophoresed through a 3% agarose NuSieve containing 1% SeaKem agarose gel (FMC Bioproducts) for 1.5 h at 120 V. DNA was transferred onto a positively charged nylon membrane (Roche, Meylan, France) and fixed for Southern blotting according to the manufacturer's recommendations. The membranes were prehybridized for 30 min at 55°C in 6× SSPE buffer (1× SSPE buffer is 0.18 M NaCl, 10 mM NaH2PO4, and 1 mM EDTA [pH 7.7]) supplemented with 0.02% bovine serum albumin, 0.02% Ficoll 400, and 0.02% polyvinylpyrrolidone and then transferred into fresh hybridization buffer containing 0.5 pmol/ml of the internal probe 5′ labeled with [γ-32P]ATP. Hybridization was performed for 2 h at 45°C and was followed by two washes at 30°C for 10 min in 2× SSPE buffer containing 0.1% sodium dodecyl sulfate. After the membranes were air dried, they were exposed overnight at-70°C to an X-ray film (Fuji) with two intensifying screens.
B. henselae (ATCC 49882) grown on Vero cells (ATCC CCL-81) was used to prepare the antigen. The ATCC 49882 strain was previously compared to three other B. henselae strains isolated from stray cats (13) as the antigen source for indirect immunofluorescence assays and was shown to be superior or equivalent to the other strains tested for both specificity and sensitivity (data not shown). The cells were first cultured in 25-cm2 flasks (Corning-Costar, Brumath, France) as a confluent unicellular layer in the presence of 89% Dulbecco's modified Eagle's medium (Invitrogen, Cergy Pontoise, France), 1% 0.2 M glutamine solution (Merck, Nogent sur Marne, France), and 10% fetal calf serum (Seromed, Berlin, Germany). The cultures were incubated at 35°C under 5% CO2 and the confluent cellular layer was dissociated with 0.05% trypsin in 0.53 M EDTA solution.
The cells were then infected by addition of 5 ml of a 0.5 to 1 McFarland suspension of B. henselae previously grown on Columbia agar (Becton Dickinson, Meylan, France) enriched with 5% rabbit blood for 6 days at 35°C under 5% CO2. After 3 days, the infected cells were washed in prewarmed sterile PBS. Uninfected Vero cells were cultured in parallel for use as controls. Infected and uninfected cells were dissociated with 0.05% trypsin in 0.53 M EDTA. The cells were resuspended in fresh culture medium and centrifuged at 200 × g for 10 min. This washing step was repeated once, after which the pellets were resuspended in fresh culture medium and the cell concentration was adjusted to 500 to 700 × 103 cells/ml. These cell suspensions were layered onto immunofluorescence assay slide wells and incubated at 35°C under 5% CO2 for 16 to 18 h. The cells were then fixed on the slides with cold acetone for 15 min.
In the indirect immunofluorescence assays, the human sera to be tested for the presence of B. henselae antibodies were diluted in PBS and incubated on the fixed bacterial antigens for 30 min at 37°C. After the slides were washed for 10 min in PBS containing 1% (wt/vol) bovine serum albumin (Sigma), the slides were incubated with fluorescein isothiocyanate-conjugated goat anti-total human immunoglobulins (Fluoline H; BioMérieux, Marcy l'Etoile, France) previously diluted to 1/50 in PBS. The slides were mounted in buffered glycerol (Fluoprep; BioMérieux) and read under an Olympus fluorescence microscope at ×400 magnification. The titer of a serum sample was defined as the highest dilution that still showed for 50% of the infected cells a fluorescence intensity equal to the highest intensity displayed by the positive control serum. Titers of <1/32 were considered negative, titers of ≥1/64 were considered positive (sensitivity, 0.70; specificity, 0.95), and titers of 1/32 were considered uncertain.
For a proportion, the 95% exact confidence interval (CI95) was computed by using the binomial distribution. The comparison of localization of lymphadenopathy for each diagnostic group was done by the Fisher-Freeman-Halton test, which extends the Fisher exact test for tables with more than two rows and/or columns. The alpha level was set at 5%, and the test was bilateral.
General data were obtained for all patients tested for Bartonella by PCR analysis of lymph node tissue samples (Table (Table1).1). The patients were predominantly children or young, and there was no significant difference in the sex ratio between the three diagnostic groups.
The lymph node enlargements were mainly axillar in the definite CSD group (51.7% of patients; CI95, 32.5 to 70.6%) and, less frequently, inguinal or cervical (24.1% of patients in both cases; CI95, 10.3 to 43.5%) (Table (Table1).1). However, inguinal lymph node enlargement was observed more often in individuals in the definite CSD group than in those in the other two groups. In the possible CSD group, axillar and cervical lymph node enlargements were found at approximately the same frequency (40 to 50%). Cervical lymph node enlargement was most commonly observed in the negative control group (61.5%; CI95, 40.6 to 79.8%) (Table (Table11).
In the definite CSD group, analysis of the number of classical criteria for CSD (Table (Table2)2) revealed that 26 of the 29 patients (89.7%; CI95, 72.6 to 97.8%) had experienced a cat scratch or had been in contact with cats, one had been injured by a monkey, and the two remaining patients did not recall any animal contact. Among the 19 patients in whom a histological examination of the lymph node was performed, 16 (84.2%; CI95, 60.4 to 96.6%) presented with a pyogenic granuloma and the three presented with other nonspecific lymphocytic inflammation. Serological testing for B. henselae antibodies was positive for 25 of these 29 individuals (86.21%; CI95, 68.3 to 96.1%). PCR assay for Bartonella was positive for 22 patients (sensitivity, 0.76; CI95, 56.5 to 89.7%) (Table (Table2),2), while all 7 patients from this group with a negative PCR result (6 of them had a histological examination) were positive for only two of the three classical criteria for CSD.
The possible CSD group, which comprised 15 patients who presented with only one or no criteria for CSD (Table (Table2),2), was heterogeneous, with 4 individuals (26.7%; CI95, 7.8 to 55.1%) having a history of cat contact. A histological examination was performed for nine of these patients, and the result was never compatible with CSD. B. henselae serology was positive for five patients (33.33%; CI95, 11.8 to 61.6%), while only three patients (20%; CI95, 4.3 to 48.1%) had a positive PCR assay result. The latter set of patients always displayed at least one classical criterion for CSD: two patients had a history of contact with cats, and one was positive for CSD serology. A histological analysis of the lymph node was not available for any of these three individuals because these three samples were pus samples. Six patients in whom no other final diagnosis had been retained presented no CSD criteria. All patients in this group had good evolution of their lymph node enlargement 2 to 6 month after diagnosis.
For the 26 negative controls for whom another diagnosis had been established, the causes of adenopathy are shown in Table Table3.3. These individuals mainly presented with noninfectious adenopathy (17 patients), including 8 cases of lymphoma, 4 cases of benign tumor, and 2 cases of carcinoma. Among the nine patients with infectious adenopathy, three had tuberculosis, as revealed by Mycobacterium tuberculosis-positive tissue cultures and/or histological necrosis with caseum; three had pyogenic adenitis due to Staphylococcus aureus with histologically evident, acute purulent inflammation of the lymph node; and three had serologically confirmed tularemia.
Analysis of the distribution of the classical criteria (Table (Table2)2) for the 26 patients in the negative control group revealed that 9 mentioned contact with cats but only 1 had experienced a cat scratch before the appearance of adenopathy. The B. henselae serology was tested in 21 patients in this group, and 3 of them were positive with a titer of ≥1/64, but the final diagnosis for these patients was staphylococcal lymphadenitis. PCR assay for B. henselae was negative for all 26 individuals in this group (0%; CI95, 0 to 13.2%). A histological examination was performed for 23 patients, and 7 cases showed signs of granuloma, 3 of which were histologically compatible with CSD. However, the clinical data together with a positive specific serology led to a diagnosis of tularemia in all three cases.
Culture of B. henselae on chocolate agar (Becton Dickinson) enriched with IsoVitaleX was done in our study and was always negative.
The sensitivity and specificity of the PCR assay were calculated by comparing the results of Bartonella DNA testing for patients in the definite CSD and control groups. The sensitivity was 76% (CI95, 56.5 to 89.7%), and the specificity was 100% (CI95, 86.7 to 100%). Moreover, good sensitivity was maintained whatever the type of sample analyzed. Thus, among the five cases in the definite CSD group for whom only pus and not tissue samples were tested for B. henselae DNA, the PCR assay was positive for four of them. The positive predictive value is 100% (CI95, 84.6 to 100%) if the control group is the patients with a diagnosis of CSD by another means. The predictive positive value is 88% (CI95, 68.8 to 97.5%) if the analysis is done with the group with a possible diagnosis of CSD as the control group.
In our study, 38 patients in the definite and possible CSD groups had at least one of the previously defined CSD criteria (Table (Table4).4). Among these, 29 displayed at least two of these classical criteria and could be diagnosed as having definite CSD. Among the 15 other possible CSD patients, the criteria were insufficient to establish a diagnosis of CSD. On the other hand, the good specificity and sensitivity observed for PCR diagnosis of CSD allowed us to evaluate another combination of criteria (enhanced diagnostic criteria) that included the PCR assay as an additional factor (Table (Table4).4). By consideration of all individuals positive for two of the enhanced diagnostic criteria, we defined a group of 32 patients that comprised the previous 29 in the initial definite CSD group and 3 others. The PCR result was always associated with at least one other criterion.
In this new group of 32 CSD patients, B. henselae PCR testing was positive for 25 (78.1%; CI95, 60.0 to 90.7%) cases, and hence, the sensitivity was 78%. By application of the same enhanced criteria to the group of 12 patients for whom a diagnosis of CSD had been excluded, 6 of these individuals had one positive criterion for CSD, but this was never the PCR result (CI95, 0 to 26.5%) (Table (Table44).
In this work, we used an efficient and specific PCR method to detect the B. henselae htrA gene in lymph node tissues. We chose to study patients selected on the basis of clinical symptoms compatible with a diagnosis of CSD. This allowed us to determine the diagnostic value of the PCR assay in different groups of patients classified according to the number of criteria for CSD so as to be able to predict the sensitivity of the PCR method with patients presenting with more or fewer (sometimes no) CSD criteria. Such an approach has, to our knowledge, never been used before and is close to that used by a physician who must make a diagnosis for a patient with lymphadenopathy with no etiological indication. Compared to the classical diagnostic criteria for CSD, PCR analysis displayed excellent specificity, since no false-positive results were observed in our control group. The classical criteria nevertheless remain useful, because the PCR result can sometimes be negative for patients with authentic CSD (7 of 29 patients in the definite CSD group in our study).
The diagnosis of this disease relies on several criteria similar to those originally described by Debré et al. (8). However, the intradermal skin test is no longer available in several countries, and in addition, none of the criteria initially used by Debré et al. are etiologic markers of the disease (8). Thus, among the classical criteria, neither a history of contact with cats nor a clinical or histological examination alone is sufficient for the diagnosis of CSD. A small minority of patients with cat scratches develop CSD, and many cases of possible CSD-related adenopathy can be attributed to other causes. Similarly, a histology picture compatible with CSD may be seen in other conditions, such as tularemia, Nicolas Favre disease, or even mycobacteriosis. New criteria which include serology and PCR diagnosis should be of value for the diagnosis of an infection due to B. henselae.
Serological testing for B. henselae antibodies was the first microbiological test available but currently has a variable positive predictive value. It is an indirect diagnostic method which can be negative in the early stage of the disease. In some studies (7, 11, 28), the positive predictive value of the indirect immunofluorescence assay for B. henselae was reported to be high (≥91.4%). Conversely, Bergmans et al. (6) and Dupon et al. (10) found a lack of sensitivity of the serological test among patients with CSD.
On the other hand, both CSD serology and PCR assays specific for B. henselae have been reported to be negative (1, 3, 4, 5, 6, 12, 19, 28) in cases of authentic CSD, and the sensitivity of PCR detection is often less than 80%. Among studies that have tested well-defined cases of CSD, none have shown that one PCR assay of a lymph node sample is sufficient for the diagnosis of CSD. Avidor et al. (4) reported a sensitivity of 100%, using three different PCR assays with three different targets, but this is not current practice in routine diagnosis.
Several groups have already assessed the diagnostic value of PCR analysis for CSD (1, 3, 4, 5, 6, 12, 20, 27). A comparison of these studies is, however, difficult due to differences in the PCR target, the sample type, and the criteria used to define CSD. Thus, several primer pairs have been used to detect B. henselae by PCR amplification (3, 4, 5, 6, 14). The 16S rRNA target first employed by Bergmans et al. (5) gave sensitivities of 96% among patients with a positive skin test result for CSD and 60% among patients with a negative skin test result. In a second study, the same authors (6) found that the sensitivities were 86.4 and 100% for patients with more than two or more than three criteria for CSD, respectively. The htrA gene used in our study has frequently been employed to test clinical samples among patients with suspected CSD. Anderson et al. (3) and Goldenberger et al. (12) obtained sensitivities of 84 and 61%, respectively, so that our result (sensitivity of 76%) is close to the best for this target (3, 4). A comparison of the 16S rRNA and htrA targets showed a better sensitivity of the former (60 versus 43%) (26). Avidor et al. (4) compared the gltA gene (which encodes citrate synthase) with the 16S rRNA and htrA genes and found the first two targets to be more sensitive (100 and 94%, respectively) than the htrA sequence (69%). Other PCR targets were not tested in our work. However, the specificity of the results was ensured by processing and amplifying a second aliquot for all the positive samples.
False-negative results can be explained either by a lack of sensitivity, as suggested by the comparative studies of Avidor et al. (4) and Sander et al. (25), or by the presence of other species of Bartonella in CSD (13, 16, 21). A poor quality of clinical samples without lymph node tissue or samples taken after a long period of antibiotic therapy could also explain some of these false-negative results. In most of the studies, the samples were fresh lymph node biopsy specimens or pus drawn from the lymph nodes (3, 4, 5, 6). Two other groups used fixed paraffin-embedded lymph nodes (26, 27) and obtained sensitivities of 40 to 70%, according to the amplification target and the criteria used to define CSD.
A diagnosis of CSD must rely on the presence of a combination of epidemiological, histological, and bacteriological criteria, since no single criterion may be considered the gold standard. The criteria used to define CSD are hence of great importance for estimation of the sensitivities and the specificities of the biological tests used for its diagnosis, as has been pointed out by several authors (6, 26, 27). Anderson et al. (3) and Avidor et al. (4) selected patients with lymphadenopathy with only contact with cats as the criterion for CSD. In our study, the latter criteria misclassified one of our patients as having CSD, although the patient in fact had pyogenic adenopathy. The sensitivity of the PCR assay of Sander and Penno (26) was 65% by the use of only histological criteria for case definition and increased to 87% when serological results were also considered, illustrating the low specificity of histological criteria. In the study of Scott et al. (27), the patients were selected because they fulfilled histopathological conditions and were then analyzed according to different criteria. The sensitivity of the PCR assay in that work was 68% (27). In our study, histological evidence was present in 84% of the patients displaying classical criteria but in only 72% of patients when the enhanced criteria, including the PCR results, were used. There were histological manifestations compatible with CSD in three patients for whom this diagnosis was finally not retained. In our study, we employed precisely defined clinical, serological, epidemiological, and histological criteria. Our patients were selected not only among those with a previously established diagnosis of CSD but also among all patients presenting with lymphadenopathy and were divided into different groups, according to the classical diagnostic criteria. This allowed us to obtain a good estimation of the sensitivity of the PCR assay.
Goldenberger et al. (12) classified their patients into four categories (certain CSD, possible CSD, unknown diagnosis, and a control group) and tested miscellaneous samples, not all of which were derived from cases of lymphadenopathy, and obtained a sensitivity of 61% and a specificity of 100%. To estimate the diagnostic value of our assay, especially for patients with uncertain CSD, we preferred to focus blindly on cases of lymphadenopathy and to collect the data prospectively, so as to define the different groups using the usual criteria for CSD. We therefore determined the diagnostic value of htrA PCR detection of B. henselae as an additional criterion for CSD and that of the expanded criteria that included the PCR result. On the basis of our findings, only a positive PCR assay result may be considered to be sufficiently specific for the diagnosis of CSD, since no patient in the control group had a positive PCR test result, in contrast to the results of serology (three false-positive results) and histology (two false-positive results).
Adopting a clinical approach, we first determined the diagnostic value of PCR analysis for a group of patients fulfilling the classical criteria for CSD. For such patients, the diagnosis is generally easy to make. More interesting are patients who do not fulfill all the criteria for CSD, for whom the diagnosis can be very difficult and PCR assay of B. henselae is very helpful. This situation is frequent in clinical practice: absent or nonspecific histolopathology, negative serology, or contact with cats without any scratch, giving several combinations of criteria. However, in our possible CSD patients who presented with only one or none of the classical criteria but for whom no other diagnosis could be retained, the B. henselae PCR assay was positive in three cases. Insofar as these three patients always displayed one of the classical criteria for CSD, we tested the diagnostic value of the enhanced criteria (at least two criteria, including the PCR result). By using these new criteria, a diagnosis of CSD was established for an additional 10% of patients. Thus, by using the PCR assay as an additional criterion, the sensitivity of CSD diagnosis could be improved without any decrease in specificity, especially for patients with incomplete diagnostic criteria. In our study, PCR detection of B. henselae had a specificity of 100%. Hence, a PCR analysis could be sufficient for the diagnosis of CSD in patients with lymphadenopathy in the presence of only one other diagnostic criterion. Interestingly, a lymph node biopsy could be avoided because PCR amplification can be performed with pus samples drawn from lymph nodes with good sensitivity (four of the five pus samples from the group with definite CSD tested were PCR positive) and specificity (pus samples were obtained from three patients in the control group, and all were PCR negative). Because of the low number of patients, this aspect should be confirmed in further studies.
Other direct methods for detection of Bartonella infections, like immunohistochemical staining or culture, have been reported for CSD diagnosis. These methods have not been used in our work because of their lack of sensitivity and specificity. Culture on chocolate agar enriched with IsoVitaleX was done in our study and was always negative.
To establish a diagnosis of CSD in patients presenting with superficial lymphadenopathy in one isolated area, we propose the use of an etiological approach which consists of looking first for the presence of B. henselae DNA by PCR analysis. In the case of PCR positivity, CSD may be retained on account of the excellent specificity. In the case of a negative PCR result, the diagnosis could rely on the presence of at least two of the following criteria: (i) positive serology, (ii) histology compatible with CSD (pyogenic granuloma), or (iii) contact with cats during the days or weeks preceding lymphadenopathy, together with elimination of any other cause of lymph node enlargement (Fig. (Fig.11).
We thank C. Barthel and E. Collin for their excellent technical assistance.