Bacterial Causes of Empyema in Children, Australia, 2007–2009

Most infections were caused by non–7-valent pneumococcal conjugate vaccine serotypes.

An increase in the incidence of empyema worldwide could be related to invasive pneumococcal disease caused by emergent nonvaccine replacement serotypes. To determine bacterial pathogens and pneumococcal serotypes that cause empyema in children in Australia, we conducted a 2-year study of 174 children with empyema. Blood and pleural fl uid samples were cultured, and pleural fl uid was tested by PCR. Thirty-two (21.0%) of 152 blood and 53 (33.1%) of 160 pleural fl uid cultures were positive for bacteria; Streptococcus pneumoniae was the most common organism identifi ed. PCR identifi ed S. pneumoniae in 74 (51.7%) and other bacteria in 19 (13.1%) of 145 pleural fl uid specimens. Of 53 samples in which S. pneumoniae serotypes were identifi ed, 2 (3.8%) had vaccine-related serotypes and 51 (96.2%) had nonvaccine serotypes; 19A (n = 20; 36.4%), 3 (n = 18; 32.7%), and 1 (n = 8; 14.5%) were the most common. High proportions of nonvaccine serotypes suggest the need to broaden vaccine coverage. E mpyema in children is a relatively uncommon disease that occurs in 0.7% of children with pneumonia (1). Many organisms cause empyema in children; Streptococcus pneumoniae is the most common (2)(3)(4)(5)(6). Other important causes, which are becoming increasingly frequent in several countries, are methicillin-sensitive Staphylococcus aureus (MSSA) (2,7,8) and methicillin-resistant S. aureus (MRSA). The latter is particularly problematic in indigenous communities (9). Other commonly identifi ed organisms include S. pyogenes, Haemophilus infl uenzae, Mycoplasma pneumoniae, Pseudomonas aeruginosa, and other Streptococcus spp. (10). The identifi cation of causative organisms is usually determined by standard blood or pleural fl uid cultures. Cultures are limited in that the yield can be as low as 8% (11), possibly because of prior antimicrobial drug treatment. Molecular techniques, such as PCR, are more sensitive in detecting causative organisms than standard culture (11) but are not routinely employed in laboratories for clinical use.
The 7-valent pneumococcal conjugate vaccine (PCV7) (Prevenar; Wyeth, Philadelphia, PA, USA) was introduced in Australia for immunocompromised and indigenous children <2 years of age in 2001 and was added to the national immunization schedule for all children <2 years in 2005 (www.medicareaustralia.gov.au/public/services/ acir/index.jsp). Of >90 pneumococcal serotypes, the 7 included in the vaccine were responsible for 50%-70% of invasive pneumococcal disease (IPD) in children in most populations at the time of its development (12).
The aims of this study were to identify the bacterial causes of empyema in children by using molecular techniques and to assess the effi cacy of PCV7 by using molecular typing of invasive pneumococcal disease serotypes. This information may be helpful in deciding which of the newer conjugate pneumococcal vaccines should be introduced into national vaccination programs.

Methods
The Australian Research Network in Empyema was established in April 2007 and comprises all 13 major tertiary pediatric hospitals from all states and territories. Children with empyema were prospectively recruited over a 2-year period until April 2009.

Patients
A case of childhood empyema was defi ned by the principal site investigators as the presence of pus cells in the pleural fl uid or bacteria isolated from the pleural fl uid of a child with fever, respiratory symptoms, raised serologic infl ammatory markers, and pleural fl uid present on ultrasound image, chest radiograph, or computed tomography scan. Children with postoperative effusions were excluded.
Clinical data collected included age, sex, indigenous status, area of residence, risk factors, congenital or chromosomal abnormality, anatomic or functional asplenia, immunocompromise, and chronic illness. Vaccination status of recruited patients was obtained by either review of the child's hand-held health records (Blue Book), contacting the Australian Childhood Immunisation Registry (with parental permission) for patients <7 years of age (www.medicareaustralia.gov.au/public/services/acir/ index.jsp), or, if these validated sources were unavailable, parental recall.

Microbiologic Investigations
Blood and pleural fl uid specimens were cultured at local hospital microbiology laboratories by standard culture method. If growth was detected, Gram staining was performed and liquid media were subcultured onto horse blood agar. Isolates were identifi ed by using conventional methods. Colonies resembling S. pneumoniae that contained gram-positive diplococci were identifi ed by optochin-susceptibility and bile-solubility testing.
A separate aliquot of pleural fl uid (in ideal circumstances, 10 mL) was collected, labeled according to the central coordinator's de-identifi cation and specimentracking database, and stored at −20°C. Pleural fl uid specimens were transported in batches on dry ice by a commercial transport company to the Centre for Infectious Disease and Microbiology Laboratory, Westmead Hospital, Westmead, New South Wales, Australia, for processing.

Streptococcus pneumoniae PCR
Total nucleic acid was extracted from pleural fl uid specimens by using either NucliSENS easyMAG Total Nucleic Acid Extractor (bioMérieux Australia Pty Ltd, Sydney, NSW, Australia) with enzymes and lysis buffer provided, or SIGMA GenElute Mammalian Genomic DNA Miniprep Kit (Sigma-Aldrich, Sydney, NSW, Australia) with lysis buffer provided plus proteinase K, following the manufacturer's instructions.
S. pneumoniae PCR targeting the autolysin gene (lytA) was performed by using a TaqMan probe and primers as described by McAvin et al. (23), except that the result was read by spectrofl uorometry and interpreted as described by Poddar et al. (24). Briefl y, a 25-μL PCR containing 1.5 mmol/L MgCl 2 , 200 μmol/L dNTPs, 200 nmol/L of each primer, 120 nmol/L probe, 1.23U HotStarTaq DNA polymerase, and 10 μL of total nucleic acid yielded a 101bp product. The PCR cycling conditions included initial denaturation at 95°C for 15 min, followed by 45 cycles at 96°C for 10 s, 63°C for 1 min, and a fi nal extension step of 72°C for 2 min. The endpoint results were analyzed by calculating the postread to preread ratio. Samples with ratios of >2.78 were reported as positive and confi rmed by using pulsed-fi eld gel electrophoresis on a 2% gel, at 200 V for 40 min. Samples with ratios of <1.21 were reported as negative. The limit of detection of the assay was 6 CFU/mL.

Pneumococcal Serotype Identifi cation
All samples in which S. pneumoniae was detected by PCR were examined by multiplex PCR reverse line blot (mPCR/reverse line blot [RLB]) to identify serotypes individually or in small groups of related serotypes (25,26). If serogroup 6 was identifi ed, serotype-specifi c PCRs targeting wciN (to distinguish serotypes 6A and 6C) and the wciP single-nucleotide polymorphism, which distinguishes serotypes 6A and 6C from 6B (27), were performed. Samples that gave no signals in mPCR/RLB (result recorded as below detection level) and those in which only the S. pneumoniae positive control probes targeting ply or lytA produced signals (nontypeable) were further tested when suffi cient DNA remained, by PCR and sequencing of the cpsA-B region of the capsular polysaccharide synthesis (cps) gene cluster, as described and validated (28).

PCR for Other Pathogens
All pleural specimens were tested by separate in-house Testing for S. aureus DNA was undertaken by using a commercial multiplex-tandem PCR targeting the S. aureus nuclease gene nuc, and methicillin-resistance gene mecA (MRSA Easy-plex assay kit; AusDiagnostics, Sydney, New South Wales, Australia), as recommended by the manufacturer.

Assessment of Data Accuracy
To assess the completeness of case ascertainment, we contacted the coding departments of all participating hospitals at the end of the study and asked them to provide data on the number of children 0-18 years of age with empyema (classifi ed according to International Classifi cation of Diseases [ICD] codes J86.0 [pyothorax with fi stula] or J86.9 [pyothorax without fi stula]) who were discharged from each hospital from April 1, 2008, through April 30, 2009. This period was chosen because it represented a time when all hospitals were actively recruiting. These data were compared with our own.
Descriptive statistics were used for all analyses. No power calculation was required because this was an epidemiologic study aiming to capture all cases of empyema.
This study was approved by the local human research ethics committee at each site, and registered with The Australian and New Zealand Clinical Trial Registry (ACTRN12607000476437). Informed parental consent was obtained for each patient before blood and pleural fl uid samples were collected.
Of the 174 children recruited, culture results were available for 172; 140 had blood and pleural fl uid cultured, 20 had only pleural fl uid cultured, and 12 had only blood cultured. Of 152 blood and 160 pleural fl uid cultures, 120 (78.9%) and 107 (66.9%), respectively, were negative. The bacteria isolated are shown in Table 2.
Vaccination status was available for 45 (86.5%) of 52 children who had a pneumococcal serotype detected on PCR. The effect of vaccination status on the acquisition of specifi c serotypes, in relation to age and vaccination status, was assessed ( Figure 2).

Discussion
This study supports previous reports from different countries that have identifi ed S. pneumoniae, S. aureus, and S. pyogenes as notable causes of childhood empyema (2)(3)(4)8,17,21). Most (96.4%) of identifi ed pneumococcal serotypes were nonvaccine related, refl ecting the effectiveness of the PCV7. Furthermore, this study highlights that PCR is more sensitive than culture for identifying pathogens.
Data on bacterial causes of childhood pneumonia in this geographic region are lacking (30). Before this study, the most comprehensive data relating to pneumococcal serotypes causing disease in children were from routine typing of sterile site isolates from cases of IPD reported to the Nationally Notifi able Diseases Surveillance System. However, surveillance of IPD does not specifi cally report on empyema.
By far, the most common organism identifi ed was S. pneumoniae. A variety of organisms other than S. pneumoniae were detected by culture and PCR. After S. pneumoniae, S. aureus was the next most common pathogen, which was identifi ed by culture in 17 pleural fl uid and 2 blood specimens (12 MSSA, 7 MRSA) and by PCR in 13 specimens ( Table 2). The occurrence of MRSA as a cause of community-acquired pneumonia and empyema is of particular concern because it is associated with more severe disease and a higher rate of complications than MSSA (9).
H. infl uenzae was detected by PCR in 4 children, 3 of whom also had positive PCR results for S. pneumoniae. S. pyogenes, Mycobacterium tuberculosis, and P. aeruginosa were isolated only by culture (individual PCR assays for these organisms were not available), and M. pneumoniae was detected by PCR in 1 child (Figure 1). All of these organisms are recognized causes of empyema in children (10). Although C. pneumoniae is a recognized cause of lower respiratory tract infection in children (31,32), its contribution to empyema has not been investigated previously. It was detected by PCR in only 1 child, with S. pneumoniae in the same specimen, which suggests that C. pneumoniae is not a major cause of empyema in children.
This study confi rmed the fi ndings of others, demonstrating enhanced sensitivity of molecular techniques (11,(33)(34)(35). PCR detected many more S. pneumoniae isolates in pleural fl uid than in cultures (51.0% vs. 7.5%), and thus routine use of PCR-based serotype identifi cation in children with empyema may improve the accuracy of pneumococcal disease surveillance, which is   essential for development of new vaccines with broader range of pneumococcal serotypes. In contrast, however, for 4 patients who had a culture positive for S. aureus, PCR results were negative. We are unclear why this occurred, but this PCR was performed last in the sequence, and results may have been due to insuffi cient DNA. Although PCR did not increase the yield of S. aureus, it can detect it more rapidly than culture, enabling rapid change to appropriate therapy, especially when MRSA is found. Reports on childhood empyema pneumococcal serotype distribution from Europe and the United States show differences. Studies in Spain (8,17), the United Kingdom (16,21,36), and the United States (14) have reported a predominance of serotype 1, while in other US studies, 19A is the most common serotype (20,22); both are non-PCV7 serotypes. Bekri et al. (13) identifi ed serotypes 1 and 19A as emerging serotypes in France and also showed that serotype 1 was predominant in children >5 years of age; serotype 19A appeared to only affect children <5 years of age. This age distribution was similarly reported in another study (37). The serotype distribution in our study was similar; serotypes 1, 3, and 19A were predominant ( Figure  2), and most serotypes 3 and 19A were identifi ed in children <5 years of age, similar to results previously reported (37). Although serotype 1 infections were identifi ed across all age groups, most were in children >5 years of age.
We were reassured that only 2 children in our study had serotypes covered by PCV7; 1 child was covered partially, and the other had not been vaccinated. Overall, this study suggests the effi cacy of the PCV7, as previously confi rmed (38). However, public health authorities should be concerned that most pneumococcal infections were caused by nonvaccine serotypes, possibly related to replacement disease after the introduction of PCV7 onto the national vaccination schedule in 2005. We do not have serotype data specifi c to empyema prior to 2005, but this has occurred in other countries and affects all IPD, including meningitis (39). Most studies that compare pre-and post-pneumococcal vaccine effects have shown near extinction of PCV7 serotypes, along with dramatic increases in nonvaccine serotypes, predominantly 1, 3, and 19A (14,15,17,20).
Although the reasons behind serotype changes have not been determined fully, ongoing enhanced surveillance may help clarify them over time, enabling us to predict future serotype trends and tailor new vaccines accordingly. This ability is particularly relevant as 2 new vaccines with broader coverage of pneumococcal serotypes-10-valent pneumococcal conjugate vaccine, with additional serotypes 1, 5, and 7F, and 13-valent pneumococcal conjugate vaccine, with additional serotypes 1, 3, 5, 6A, 19A, and 7F-are being are being added to national vaccination schedules. The 10-valent vaccine offers protection against nontypeable H. infl uenzae through the use of an H. infl uenzae conjugate protein.
Our study has several limitations, nevertheless. First, we cannot know whether the number of children recruited in this study is an accurate snapshot of Australia's true empyema prevalence in children. However, after comparing ICD empyema codes with study recruitment rates over a 1-year period, we determined that we recruited a median of 51.5% (range 0%-200%) of empyema patients admitted to all major pediatric tertiary hospitals recorded by ICD. A limitation of this approach is that we were not able to verify the coding accuracy in each of the 13 hospitals. One likely reason why we did not capture all the cases may be because some patients received treatment from physicians at participating centers who were unaware of the study. Also, some children may have received treatment at smaller rural hospitals where the study was not conducted, even though we have recently shown that most patients are treated in 1 of the tertiary pediatric hospitals included in this study (1). We recruited patients from all states and territories in Australia, which is the strength of the study. A limitation of the PCR data is that the bacteria assessed were restricted to H. infl uenzae, M. pneumoniae, C. pneumoniae, and S. aureus. They are all potentially important bacterial pathogens in empyema in children, however (10). The use of broader PCR, such as 16sPCR, may have detected more organisms, but it is an expensive test and our previous experience suggests that the yield of notable pathogens is poor and that positive results often refl ect contamination.
In conclusion, we have demonstrated a wide variety of bacterial causes for empyema in children. Most infections were caused by non-PCV7 pneumococcal serotypes, which suggests that coverage of pneumococcal serotypes by vaccines needs to be broadened. Ongoing enhanced molecular surveillance is required, particularly to assess the effects of newer vaccines, such as 10-valent and 13-valent pneumococcal conjugate vaccines.