Polymerase Chain Reaction for Pathogen Identification in Persistent Pediatric Cervical Lymphadenitis

Objective  To study routine culture-negative persistent cervical lymphadenitis in children treated surgically during a 10-year period (December 26, 1997, to October 1, 2007) at a single institution.

Design  Retrospective case series.

Setting  Tertiary university-based pediatric referral center.

Patients  Patients 18 years or younger with cervical lymphadenitis managed surgically (incision and drainage, curettage, and/or excisional lymphadenectomy) and medically (antibiotic therapy), culture-negative after 48 hours, and subsequently diagnosed using the polymerase chain reaction, extended culture incubation, and/or histopathologic evaluation.

Main Outcome Measures  Number of surgical interventions, causative organisms, histopathologic features, and resolution of lymphadenitis.

Results  Ninety surgical procedures were performed in 60 patients. The cure rate was 23% (approximately 14 patients) with incision and drainage, 58% (approximately 35 patients) with curettage, and 95% (57 patients) with excisional lymphadenectomy. Nontuberculous mycobacteria were the most prevalent causative organisms, followed by Bartonella and Legionella organisms. Four of 6 patients with Bartonella infection had a history of cat exposure, and 4 of 6 patients with Legionella infection had a history of hot tub exposure.

Conclusions  Excisional lymphadenectomy is the preferred treatment of mycobacterial persistent cervical lymphadenitis in children. Sufficient data are lacking for similar recommendations in patients with disease caused by Bartonella organisms, whereas for neck disease caused by Legionella organisms, excisional lymphadenectomy may be superior to incision and drainage. The polymerase chain reaction is useful for pathogen identification in pediatric cervical lymphadenitis, although it is less sensitive in identification of mycobacteria. To our knowledge, our study is the first to report multiple cases of legionellosis in otherwise healthy children. Legionella seems to be a previously unrecognized but relatively common pathogen in culture-negative persistent cervical lymphadenitis in children.

Lymphadenitis is a common cause of neck masses in children. Diagnosis is frequently uncertain, and antibiotic therapy is often determined empirically without microbiologic culture data. Staphylococcus aureus or Streptococcus pyogenes (group A) are the most common pathogenic organisms involved¹; however, various other microbes have been identified.^2,3 Clinical management and prognosis differ according to the pathogenic organism involved; therefore, sensitive and accurate microbiologic diagnosis is essential. Although incision and drainage is effective in the management of suppurative lymphadenitis caused by S aureus and S pyogenes, granulomatous lymphadenitis is a more heterogeneous disease and requires determination of specific etiologies. This information also determines the type and duration of antibiotic or medical therapy.

Identifying the infective agent in routine culture-negative tissue samples is a major challenge. Molecular diagnosis in culture-negative persistent cervical lymphadenitis (PCL) via polymerase chain reaction (PCR) of lymph node contents helps identify uncommon organisms in cases in which standard microbiology cultures and histopathology are ineffective.^4-6 However, PCR is not universally used for diagnostic purposes in clinical microbiology laboratories. We present a molecular diagnostic approach using PCR with broad-range primers to amplify part of the eubacterial 16S ribosomal RNA gene followed by amplicon sequencing and database query for organism identification. Amplification and sequencing of a second target specific for the identified organism, and serology when possible, serve to confirm the PCR results.

Otherwise healthy pediatric patients (age ≤18 years) with cervical lymphadenitis treated surgically and with negative routine cultures (aerobic, anaerobic, fungal, and mycobacterial) after 48 hours of incubation were included in the study. All patients had cervical lymphadenitis as defined by tenderness to palpation, red or violaceous discoloration of overlying skin, skin breakdown with purulent drainage, and clinical or radiographic evidence of abscess, with or without constitutional signs or symptoms such as fever and malaise. Data collected retrospectively from the medical record included patient age and sex, size and duration of the neck mass, associated signs and symptoms, environmental exposures, perioperative antibiotic agents used, type of surgery performed (incision and drainage, curettage, or excisional lymphadenectomy) and any associated complications, microbiologic staining and culture results, histopathologic findings, and PCR results. All data were collected with institutional review board approval by the Human Subjects Division at the University of Washington and the Seattle Children's Hospital, Seattle.

Differences in cure rates among the 3 types of surgery performed (incision and drainage, curettage, and excisional lymphadenectomy) were examined for all cases combined and for each set of microbe-specific cases. Statistical variance was calculated using the Kruskal-Wallis nonparametric test with the method of Siegal and Castellan⁷ to determine minimum significant differences in rank. Differences in patient age and in size and duration of neck masses in the various causative groups were examined using 1-way analysis of variance followed by the Tukey post hoc least significant difference test.⁸

Surgically excised tissues or abscess contents were collected in sterile containers, immediately delivered to the laboratory, and split into separate samples. In addition to large portions of tissue ground or minced for bacterial and fungal cultures, 1- to 5-mm³ equivalent aliquots from all areas of the biopsy specimen were split into at least 3 sterile vials. Tissue materials from 2 of the split vials were used for initial PCR using pan-16S ribosomal DNA primers and mycobacterial genus primers. Based on the initial findings, materials in the remaining vials were used for confirmatory PCR using independent signature gene targets. Extraction of DNA was carried out using overnight proteinase K treatment followed by sonication for 15 minutes using the QIAamp DNA Mini Kit (QIAGEN GmbH, Hilden, Germany).

The PCR primers, including 2 sets of 16S ribosomal DNA primers and a number of bacterial genus- or species-specific primers, are given in Table 1. The PCR reactions contained the standard 2-minute pretreatment at 95°C followed by 35 cycles consisting of 30 seconds at 94°C, 30 seconds at 45°C to 58°C (temperatures vary depending on each specific primer pair given in Table 1), and 30 seconds at 72°C, and concluded by the end treatment at 72°C for 5 minutes, then refrigeration at 4°C. The DNA sequencing was carried out using a BigDye sequencing kit and the ABI310 genetic analyzer (both from Applied Biosystems Inc, Foster City, California). On the basis of bacterial identities determined by the primary 16S ribosomal DNA PCR and sequencing, a secondary chromosomal target for the confirmatory PCR was tailored for each specific organism in question. Confirmatory serologic results were sent to reference laboratories.

Lymph nodes, received fresh from the operating room, were subsequently fixed in a 10% formalin solution, embedded in paraffin, sectioned, and stained with hematoxylin-eosin per standard protocols. After initial histopathologic examination, special staining (auramine O or Fite carbolfuchsin acid-fast stains or Gomori methenamine silver stain) for suspected organisms was performed.

Sixty patients treated between December 26, 1997, and October 1, 2007, met the inclusion criteria. Mean (SD) patient age was 4.7 (3.7) years (age range, 10 months to 16 years). There was a slight female predominance (32 of 60 patients [53%]). The mean (SD) size of the neck mass in greatest diameter was 3.2 (1.5) cm (range, 1-8 cm). By far, most masses were located in the superior cervical chain and submandibular nodes, although all sites from perifacial to levels I through V were involved in at least 1 case. The mean (SD) duration of the neck mass was 2.1 (3.4) months (range, 1 week to 24 months).

Preoperative antibiotic usage varied widely in both duration of treatment and type of drug used. Most patients were seen at our clinic after failed antibiotic treatment with amoxicillin/clavulanate potassium, clindamycin hydrochloride, a cephalosporin, or a macrolide antibiotic. Almost all patients received a macrolide antibiotic (clarithromycin or azithromycin) without, or more commonly with, a rifamycin (rifampin or rifabutin) postoperatively for several months. One case of Mycobacterium tuberculosis was treated with a standard 5-drug antituberculosis (TB) regimen. Most patients were referred to a pediatric infectious disease clinic for antibiotic management postoperatively.

Ninety surgical procedures were performed in 60 patients (Table 2). Most were performed at Seattle Children's Hospital, although 3 patients (5%) were treated at our clinic after failed surgical management (incision and drainage) performed elsewhere. Surgical results are given as a total overview of all cases combined and cases categorized by causative organism (Table 2). The differences in cure rates in each group were significant using the Kruskal-Wallis test for total cases combined and for mycobacterial cases (Table 2). For cases caused by Legionella organisms, there was a significant difference in cure rate between incision and drainage vs excisional lymphadenectomy. No surgical complications occurred; however, disease recurred in 32 patients.

Results of microbiologic testing are given in Table 3. Three genera of bacteria, Mycobacterium, Bartonella, and Legionella, were identified at histologic staining, delayed culture, or PCR. In 11 of 60 patients (18%), a causative organism was never identified. In 1 of these patients, sinus histiocytosis (Rosai-Dorfman disease) was ultimately diagnosed at histopathology.

Mycobacterium, the most common infectious organism identified, accounted for 37 of 49 positive diagnoses. Most mycobacterial infections (27 of 37 [73%]) were caused by Mycobacterium avium-intracellulare complex. Other species included M tuberculosis, Mycobacterium lentiflavum, Mycobacterium malmoense, and Mycobacterium szulgai in 1 patient each. In 6 patients, no definitive species was identified. In 26 of 37 patients with Mycobacterium infection, pathologic or microbiologic examination identified acid-fast bacilli. In 32 of 37 Mycobacterium infections, bacterial culture eventually yielded organisms, commonly after 2 weeks or more of incubation. In 30 of 37 Mycobacterium infections, the diagnosis was made within 2 or 3 days using PCR and sequencing. The methods of diagnosis of Mycobacterium infections are given in Table 4.

Conversely, all infections caused by Bartonella (n = 6) and Legionella (n = 6) organisms were diagnosed using only PCR and were negative at staining and culture (2 of 6 patients with Bartonella infections were positive for Bartonella IgG testing). Of 6 patients with diagnosed Bartonella infection, 4 had previous exposure to cats, 1 had previous exposure to a dog, and 1 had no documented history of exposure. Of 6 patients with diagnosed Legionella infection, 4 had previous exposure to whirlpool hot tubs. All Bartonella infections were caused by Bartonella henselae. Of the Legionella infections, 2 were identified as Legionella pneumophila and 2 as Legionella micdadei. The last 2 cases positive for L micdadei were confirmed with serology specific for L micdadei with known cross-reactivity to Legionella longbeachae.

One-way analysis of variance for patient age in the 4 patient groups (Mycobacterium, Bartonella, Legionella, and no organism identified) revealed a significant difference at the level of P = .01. Mean (SD) results of the Tukey post hoc test showed that the group with Mycobacterium infection (3.68 [3.26]) was significantly younger than the group with no organism identified (7.68 [4.15]) at the level of P < .01. Neck mass size (P = .45) and duration of neck mass (P = .28) were not significantly different among groups at 1-way analysis of variance.

Although 27% of patients with non-TB mycobacterial PCL in our series had associated pain or fever, 50% of the patients with Legionella infection exhibited one or both of these symptoms, and half of those with Bartonella infection had pain or malaise associated with lymphadenitis. These differences, however, were not significantly different (P = .12).

Results of histopathologic analysis are given in Table 5. All but 1 patient with Mycobacterium infection, all patients with Bartonella infection, and 3 of 5 patients with Legionella infection exhibited necrotizing granulomas. Of the 11 patients in whom no causative organism was identified, 4 exhibited reactive lymphadenitis (including 1 with Rosai-Dorfman disease), 4 had necrotizing granulomas, and 3 demonstrated the inflammatory granulation tissue of an abscess wall.

Persistent cervical lymphadenitis in children presents a diagnostic and therapeutic challenge. In our experience, non-TB mycobacteria are the most common infectious cause of PCL, accounting for 73% of cases, followed by Bartonella and Legionella organisms, each accounting for 12% of cases. In the present series, cervical tuberculosis was diagnosed in a 16-year-old girl who had recently immigrated to Seattle, Washington, from Ethiopia. In 11 patients (18%), no causative organism was identified. In 1 of these patients, sinus histiocytosis (Rosai-Dorfman disease) was diagnosed at histopathology; however, in the remaining patients, no definitive diagnosis was determined. Some of these patients, especially the 6 with granulomas noted at histopathology, could still be caused by non-TB mycobacterial infection.

Continuous growth monitoring systems such as the BacT/Alert (BioMérieux, Marcy l’Etoile, France) and the BacTec MGIT (Becton Dickinson Microbiology Systems, Cockeysville, Maryland) combined with molecular identification by 16S ribosomal RNA gene partial sequencing have made rapid microbiologic diagnosis of Mycobacterium infection possible; however, molecular approaches in direct tissue specimens have limitations. Mycobacteria are characterized by hydrophobic waxy cell walls rich in mycolic acids and are often difficult to culture. Extraction of DNA from mycobacteria poses a challenge for the same reason. Mycobacterial DNA extraction and specimen sampling problems are associated both with capturing the foci of infection and with minute input DNA material in a 50- to 100-μL PCR reaction.

Surgical excision of involved lymph nodes is the recommended treatment for the management of non-TB mycobacterial PCL.^3,15-17 Our experience confirms that excisional lymphadenectomy results in the highest cure rate. Although data suggest that simple observation is as effective as more invasive therapies,¹⁸ observation typically results in a protracted disease course. However, in cases in which excisional lymphadenectomy may put the facial nerve at excessive risk or in those that involve cosmetically sensitive areas, prudence may dictate a less aggressive approach.

The predominance of M avium-intracellulare complex in our series agrees with published data.¹⁹ This represents a change that occurred in the 1970s, before which the predominant causative species was Mycobacterium scrofulaceum. Antibiotic recommendations for the treatment of non-TB mycobacterial lymphadenitis involve 2 agents, a macrolide such as clarithromycin or azithromycin and a rifamycin such as rifampin or rifabutin.^20,21

Regional lymphadenitis caused by B henselae is the hallmark of cat-scratch disease (CSD).^22,23 Cats are a natural reservoir for B henselae, and the bacteria can persist for a year or longer in some cats.²⁴ Most cases of CSD are associated with a previous cat bite or scratch injury, although flea bites and exposure to dogs have also been linked to CSD. Determining such exposure history in patients with PCL may be helpful in making the diagnosis.

There are no specific criteria for the diagnosis of CSD. Most often, diagnosis is based on clinical history, including that of cat exposure, and findings at physical examination. Laboratory diagnosis can be made with serology yielding a positive B henselae antibody titer; however, current immunoassays used in clinical laboratories are hampered by low sensitivity and specificity. Definitive diagnosis is by growth in culture; however, B henselae is difficult to isolate from tissue specimens, is slow-growing, and requires specific nonstandard culture conditions for optimal growth. Our results demonstrate that PCR is a quick and reliable method for diagnosing Bartonella infection.

Antibiotic recommendations for the treatment of CSD range from no therapy for mild to moderate illness²⁵ to a 5-day course of azithromycin.²⁶ Surgical management of CSD is not typically recommended in the general population because most cases are self-limited.²⁵ Of our 6 patients, 2 underwent incision and drainage, 2 underwent curettage, and 2 underwent excisional lymphadenectomy. One patient had persistent adenopathy for several months after curettage that ultimately resolved with long-term antibiotic therapy; in all other patients, the disease resolved after surgery and antibiotic therapy. Whether surgery substantially affects the natural disease progression in this population of patients is unknown. Our 6 patients did not exhibit substantially different surgical outcomes; however, the sample size is likely too small to detect a true difference if one exists. Nevertheless, the presence of abscess in each case necessitated surgical drainage. A suppuration rate of approximately 10% to 20% in CSD has been reported.^25,27,28

The most common clinical manifestation of legionellosis is legionnaires disease, a necrotizing lobar pneumonia primarily caused by L pneumophila. To our knowledge, the present study is the first case series of PCL caused by Legionella organisms. We previously reported Legionella infection in an immunocompetent child with known hot tub exposure and a 1-week history of an enlarging neck mass and fever in whom L micdadei lymphadenitis was ultimately diagnosed.⁵ Four of 6 patients with Legionella infection in the present series also had a history of hot tub exposure. Water is the primary habitat for Legionella species, and water systems linked to outbreaks of legionnaires disease include whirlpool hot tubs, respiratory therapy equipment, industrial cooling towers, and aerosolized misting devices used in produce sections of grocery stores.²⁹Legionella organisms are never found as normal-colonizing flora in human beings. Legionella organisms are also the cause of Pontiac fever, a febrile, self-limited, flulike illness requiring no specific therapy. Other extrapulmonary manifestations of legionellosis are rare but frequently serious and most often develop in patients with compromised immune systems.³⁰Legionella organisms have been implicated in sinusitis, endocarditis, pericarditis, pancreatitis, and peritonitis.

Although diagnosis is primarily based on growth in culture, Legionella organisms do not grow on routine microbiologic medium, requiring culture on charcoal-buffered yeast-extract agar. In our series, only routine culture medium was used, which explains why Legionella was diagnosed only at PCR.

Most antibiotic trials for treatment of Legionella infection are based on legionnaires disease. Currently recommended agents include levofloxacin, moxifloxacin, or azithromycin. Rifampin may be used as an additional agent in patients who exhibit suboptimal response to a single agent. Among our 6 patients with Legionella infection, 7 incision and drainage procedures were performed, and in all but 1, there was persistence or recurrence of disease despite administration of postoperative antibiotic therapy. Three patients underwent excisional lymphadenectomy with resultant cure. Although our analysis did not reveal significant differences in cure rates between all surgical methods, possibly because of a relatively small sample of 6 patients, excisional lymphadenectomy was significantly better than incision and drainage for resolution of lymphadenitis in patients with Legionella infection. Legionella organisms seem to be a previously unrecognized but relatively common cause of PCL in children. Obtaining a history of hot tub exposure is recommended in patients with PCL. Additional studies are warranted to determine the prevalence of pediatric cervical legionellosis.

All Bartonella and Legionella infections in our series were diagnosed using PCR only, and mycobacteria that eventually grew in culture required longer than 2 weeks in most cases. Organisms such as Mycobacterium, Bartonella, and Legionella are fastidious and difficult to isolate in culture. The histopathologic appearance of infection caused by these 3 microbes is similar. Delay in diagnosis may lead to inappropriate antibiotic or surgical management. Polymerase chain reaction enables rapid and accurate identification of several organisms that are difficult to culture, including Mycobacterium, Bartonella, and Legionella, from surgical specimens and has the potential to improve substantially the clinical course and outcome in children with PCL.

Correspondence: Scott C. Manning, MD, Division of Pediatric Otolaryngology, Seattle Children's Hospital, Mailstop W-7729, 4800 Sand Point Way NE, Seattle, WA 98105 (scott.manning@seattlechildrens.org).

Submitted for Publication: April 5, 2008; final revision received July 7, 2008; accepted August 7, 2008.

Author Contributions: Dr Choi had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Choi, Qin, Perkins, and Manning. Acquisition of data: Choi, Qin, Inglis, Ou, Perkins, Patterson, Berry, and Manning. Analysis and interpretation of data: Choi, Qin, Chen, Inglis, Sie, Patterson, Berry, and Manning. Drafting of the manuscript: Choi, Qin, Inglis, Ou, Perkins, and Manning. Critical revision of the manuscript for important intellectual content: Choi, Qin, Chen, Inglis, Ou, Perkins, Sie, Patterson, Berry, and Manning. Statistical analysis: Choi. Obtained funding: Manning. Administrative, technical, and material support: Qin, Sie, Patterson, and Manning. Study supervision: Inglis, Ou, Perkins, Sie, and Manning.

Previous Presentation: This study was presented in part as an abstract at the annual meeting of the Society for Pediatric Pathology; February 11, 2006; Atlanta, Georgia.

Financial Disclosure: None reported.

Funding/Support: This study was funded by the University of Washington Department of Otolaryngology Research Fund (Dr Manning).