Pediatric Parapneumonic Empyema, Spain

Increased incidence is principally due to highly invasive nonvaccine serotypes of pneumococci, especially serotype 1.

P leural effusions occur in at least 40% of children hospitalized with bacterial pneumonia. Occasionally, the infectious agent invades the pleura to cause pediatric paraneumonic empyema (PPE) (1), characterized by the pres-ence of pus. Although rarely associated with fatalities in industrialized countries, PPE often results in prolonged hospitalization and surgical intervention, and patients are at risk for serious and long-lasting illness (2,3).
An increasing incidence of PPE has been reported in several countries since the mid-1990s (2-6), but it is not clear why. Streptococcus pneumoniae is the most frequently found microorganism in most recent reports. However, conventional microbiologic culture methods have low sensitivity, usually because of antimicrobial pretreatment before sterile-site sampling. Consequently, the contribution of antimicrobial drug-susceptible serotypes might be higher than reported estimates. Molecular and antigen detection-based techniques, including direct molecular typing of culture-negative pleural fl uid (PF) samples (7), can be useful adjuncts in defi ning the contributory role of different microorganisms and pneumococcal serotypes to PPE etiology (4,8).
Our study's goal was to prospectively investigate the molecular epidemiology of pneumococcal PPE among children admitted to 3 of the largest tertiary-care pediatric hospitals in Spain. There were 4 objectives: 1) identify the serotypes and multilocus sequence typing (MLST) genotypes causing PPE and determine whether a temporal change in the circulating genotypes could explain the recent increase; 2) determine whether the causal genotypes were only associated with PPE or also caused other invasive pneumococcal disease (IPD) in the same population, or were carried by healthy children; 3) compare serotypes and genotypes recovered from northern and southern Spain in the context of regional differences in 7-valent pneumococcal conjugate vaccine (PCV7) uptake; and 4) identify any differences between highly invasive serotypes and more opportunistic serotypes with respect to epidemiology and infl ammatory markers.

Prospective and Retrospective Identifi cation of PPE Cases
PPE cases involving all children <18 years of age at Sant Joan de Deu Hospital Barcelona, Spain, were prospectively enrolled beginning October 1, 2003 PPE was defi ned according to published criteria (6). Requirements for thoracocentesis, decortication, administration of fi brinolytics, and antimicrobial drug therapy were determined according to usual clinical practice. Patients with PPE not requiring thoracocentesis or surgical decortication were excluded, as were cases with PF analysis consistent with a transudate. Thoracocentesis was performed by pediatric surgeons, except for acutely ill patients who were tapped in the emergency department or intensive-care unit. PF specimens were sent for routine microbiologic culture and biochemical analysis; remaining fl uid was frozen at -80ºC for further molecular testing. Clinical, demographic, and outcome data were collected by using a standardized case report form.
To identify temporal trends, we also retrospectively identifi ed PPE cases at VRCH and CHCH during 1998-2004 using International Classifi cation of Diseases, 9th Revision, codes for empyema (510.0 or 510.9) and chart review. PPE defi nition and patient exclusion criteria were the same as those used in the prospective study.

Nasopharyngeal Carriage Study
From January 2005 through June 2006, nasopharyngeal (NP) swab specimens were obtained from 635 children 6 months to 6 years of age attending 4 primary healthcare centers for well-child visits and 2 hospital emergency rooms in Seville. Exclusion criteria were chronic medical condition, moderate to severe acute process including fever >39ºC, lower respiratory tract infection, vomiting, dehydration, or other ill appearance. The study was conducted according to World Health Organization recommendations (9)

Informed Consent
Written informed consent was obtained from parents or legal guardians of participating children before thoraco-centesis or nasopharyngeal swabbing. Hospital ethics committees approved the studies.

Retrospective Analysis of IPD Cases
IPD cases from patients at VRCH and CHCH during 2001-2006 were retrospectively ascertained from microbiology department databases of both centers and confi rmed by chart review. Viable pneumococcal isolates were serotyped (70% of cases) by the Spanish Reference Laboratory of Pneumococci and genotyped by MLST (61% of cases).

Testing of PF Samples
Pneumococci were identifi ed by using microbiologic and molecular genotyping methods; susceptibility testing was performed by agar dilution (10), and Clinical Laboratory Standards Institute interpretive criteria were used to defi ne susceptibility (10). Culture-negative PFs were assayed for the presence of the pneumolysin (ply) gene, by using a real-time PCR in Barcelona adapted from Corless et al. (11) and a published assay (12) in Seville. Conventional serotyping using the Quellung method was performed where possible; culture-negative or incompletely genotyped PFs were serotyped by using a real-time PCR that targets different capsular locus genes (13). DNA was extracted from PFs by using 20% wt/vol Chelex-100 resin (Bio-Rad Laboratories, Hercules, CA, USA) in Barcelona and the Nucleospin kit (Clontech Laboratories, Inc., Mountain View, CA, USA) in Seville.

Molecular Genotyping
MLST was performed by using standard methods (14), with the exception of a change in PCR primers for the gdh, recP, xpt genes when genotyping PFs directly; in the PCR amplifi cation step, fi rst-round primers from a nested PCR (15) were substituted for standard MLST primers to increase sensitivity. Allele and sequence type (ST) designations were made by using the MLST website (www.mlst.net).

Statistical Analyses
Statistical analyses were performed by using SPSS for Windows version 14.0 (SPSS, Inc., Chicago, IL, USA). Reported p values were 2-tailed, and the level of signifi cance was set at 0.05. Analysis of categorical variables was performed with the χ 2 test and Fisher exact test, as appropriate. Continuous variables were compared by analysis of variance followed by a Bonferroni test for multiple comparisons. When data were not normally distributed, we used the Kruskal-Wallis test and conducted posterior comparison between individual groups using the Mann-Whitney U test with the Bonferroni correction. IPD potential was estimated by using a standard odds ratio and Mantel-Haenszel confi dence intervals (16).

Overall PPE Trends
In Seville and Malaga, the annual number of PPE cases increased 13-fold (5 to 66 cases) during 1998-2006 ( Figure  1). In Barcelona, the annual number of PPE cases increased from 11 cases in 2004 to 62 cases in 2006 (data before October 2003 were not available). Over these study periods, no obvious changes in referral patterns, overall pediatric population, guidelines for evaluating children with fever, pneumonia or PPE, or recommendations for performing diagnostic thorachocentesis in children with PPE were found. Table 1 describes the demographic characteristics of the 208 PPE patients prospectively enrolled during the molecular analysis study period (n = 98, Seville and Malaga; n = 110, Barcelona). There were no deaths.

Microbiologic Evaluation
Sixty-seven (32%) patients had positive blood and/or PF cultures for any pathogen, and S. pneumoniae was isolated from 53 (79%) of these cases ( Figure 2). In 51 of these, a pneumococcal serotype could be identifi ed via the conventional Quellung reaction. Evidence of pneumococcal infection in 99 (84%) of 118 culture-negative PF samples was found on the basis of ply or wzg gene detection. PPE cases, diagnosed only by ply or wzg PCR, were signifi cantly more likely to have received antimicrobial drug therapy before PF aspiration than patients with culture-positive pneumococcal PPE (92% vs. 53%; p<0.0001). Of the 99 culturenegative PF samples, 67 (65 ply-positive/wzg-positive and 2 ply-negative/wzg-positive) had a suffi cient sample to enable serotype testing by PCR. In 52 of these samples, a serotype could be identifi ed. Thus, a pneumococcal serotype was identifi ed in 103 PF samples ( Figure 2).
In addition, a predicted serotype based on MLST genotyping was established for 2 cases with negative PCR results and 6 cases for which neither conventional nor PCR-based serotyping was possible ( Figure 2). Such predictions were possible because there is a strong relationship between serotype and MLST genotype for most genotypes (16-18; www.mlst.net), with the exception of a small number of well-known genotypes that are associated with different serotype variations.
Eighty-one PF samples were fully genotyped, and 18 were partially genotyped (>4 alleles), by MLST. An ST was identifi ed for 31 of the 99 culture-negative/ply-positive PPE. Among these 31 cases, there was full concordance between MLST data and PCR results for confi rmation of predicted serotypes (Figure 2). Eighteen PF samples were partially genotyped by MLST: 2 were presumptive serotype 1 pneumococci based on 5-6 loci matching ST228; 1 was a presumptive serotype 5 based on 5 loci matching ST1223; 7 were genotyped at >4 loci and serotyped by PCR (serotype 1, n = 5; serotype 7F and 19A, n = 1 each); and 8 samples were partially genotyped at >4 loci (indicating presence of a pneumococcus), but PCR serotyping was either negative or not performed. Samples with predicted serotypes based on incomplete genotyping data were not included in further analyses.

Serotype Distribution
Ten serotypes were identifi ed among the 111 PPE cases with tentatively assigned or confi rmed serotyping information ( than in Barcelona, the contribution of other serotypes by region was not signifi cantly different (Table 2). PCV7 uptake among PPE patients was signifi cantly higher in Seville and Malaga than Barcelona (40% vs. 22%, p = 0.005), but there were no signifi cant regional differences in vaccination status among children infected with serotype 1 (28% vs. 22%, p = 0.63).
Six of 7 serotype 14-positive PF samples were ST156 (Spain 9V -3). Genotypic diversity among the serotypes in this study was greatest for serotype 19A; 5 unrelated STs were detected, including ST81 (Spain 23F -1). Such variants of ST81 have also previously been detected.

IPD and Nasopharyngeal Carriage in Seville and Malaga
During 2001-2006, 180 cases of IPD involving children <14 years of age were diagnosed with IPD at Seville and Malaga hospitals, and 126 isolates were available for serotyping; 110 of these isolates were also genotyped. Twenty-three percent (29/126) of all IPD was due to serotype 1. Over this period, there was a statistically nonsignificant increase in the proportion of IPD cases due to serotype   (Table 4). Ten (8%) cases of culture-positive IPD were due to serotype 7F, 9 of which were detected after 2004. ST191 was the only serotype 7F genotype in IPD and NP carriage.

Discussion
In this study, we used molecular techniques to sensitively evaluate PPE epidemiology among a large number of patients in geographically diverse locations of Spain. There was evidence of pneumococcal infection in most of the culture-positive and culture-negative cases of PPE, which was mainly associated with nonvaccine serotype 1 followed by 3, 5, 7F, and 19A, as well as vaccine serotype 14. Serotypes 1, 3, and 14 in particular are well-known PPE-associated serotypes (2,4,7,20,21). Antimicrobial drug-susceptible serotypes 1, 3, 5, and 7F were overrepresented in culture-negative PF samples, pointing to an important potential bias in PPE surveillance when surveillance is based solely on conventional microbiologic culture methods. Infection with serotype 3 was a risk factor independently associated with PPE complications, a fi nding also seen in a US study (22). Serotype 1 has also been the most prevalent IPD serotype among Spanish children <14 years of age, representing 5%, 11%, and 27% of all culture-positive pediatric IPD isolates sent to the Pneumococcal Reference Laboratory in 1997, 2003, and 2006, respectively (23). However, the increase in serotype 1 disease cannot easily be explained by a vaccine effect, in part because PCV7 coverage was relatively low in both regions for much of the study period. Registered in Spain in June 2001, PCV7 had a low initial uptake that increased to a reported coverage of 34%-45% in 2004-2005 (24,25).
In addition, increased PPE incidence largely caused by serotype 1 was reported in the United States and the United Kingdom in the decades before PCV7 introduction in 2000 and late 2006, respectively (4,6,20). Previous studies have suggested that the high year-to-year variability of serotype 1 and 5 disease may represent large-scale outbreaks of a cyclical nature (26)(27)(28).
However, the observation in this study that 2 of the 3 MLST genotypes of serotype 1 (ST228 and ST304) had been "resident" in Spain at least since 1990 indicates that serotype 1 PPE increases in Spain were likely not due to a recent introduction of a specifi c clone. In general, MLST analyses demonstrated that the recent increase in PPE was mainly due to pneumococcal STs previously described to be present in Spain and other European countries for some years (18,27,(29)(30)(31). 1394 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 14, No. 9, September 2008 Our study has several limitations. First, the limited study period did not enable a longer-term analysis of PPE epidemiology. Second, our analyses relied exclusively on serotype identifi cation and MLST genotyping, neither of which detects differences in virulence factors apart from the serotype. Genetic factors independent of the capsule have been associated with invasiveness and disease severity (17,32,33). Third, other factors that may also modulate the epidemiology of PPE (e.g., differences in case ascertainment between the retrospective and prospective studies, viral infections, or climatic patterns [34]) were not evaluated. Fourth, it remains diffi cult to evaluate the PCV7 impact because reliable written immunization registries detailing the number of administered doses are lacking, and thus vaccine coverage fi gures mainly come from parent reporting. Finally, the results obtained here may not apply to less severe pneumonia cases, whose etiology may be qualitatively different.
Unfortunately, PCV7 has a serotype coverage of only 11%-14% (including the cross-reactive 6A) in the population of PPE patients. However, conjugate vaccines containing serotypes 1, 5, and 7F, such as the newly developed 10-valent pneumococcal Haemophilus infl uenzae protein D conjugate vaccine candidate (35), could increase the serotype coverage for PPE up to 80%; the subsequent addition of serotypes 3 and 19A in vaccine candidates currently in development would add an additional 18% of coverage (35). Finally, continued epidemiologic surveillance with molecular diagnostic techniques will be crucial to understanding this serious pediatric disease.  (16,17,19). All results shown were statistically significant (p<0.05). There were no significant differences between groups for the following variables: median days febrile preadmission; preadmission antimicrobial therapy; intensive care unit admission; mean leukocyte count; mean C-reactive protein; mean pleural fluid glucose; mean pleural fluid pH; mean lactate dehydrogenase; median days to thorachocentesis; referral; primary fibrinolytics or thoracoscopy; or oxygen requirement >4 d. †HIDP was compared with LIDP by post hoc analysis. ‡Since being admitted to first center. §No significant differences between individual groups by post hoc analysis (p = 0.023 for comparison between serotype 3 and LIDP) ¶Complications included (no. patients): bronchopleural fistula (3), pyopneumothorax (2), pneumatoceles (4), lung abscess (1), mechanical ventilation >48 h (2), severe anemia requiring blood transfusion (2), severe hypoalbuminemia requiring seroalbumin replacement (1). #Serotype 3 compared withHIDP and LIDP groups combined.