Pneumococcal empyema and complicated pneumonias: global trends in incidence, prevalence, and serotype epidemiology

This review evaluates the serotype epidemiology of complicated pneumococcal pneumonia (CPP) during the period 1990–2012. PubMed and EMBASE were searched using the terms “empyema”, “complicated pneumonia”, “pleural infection”, “necrotizing pneumonia”, “pleural effusion”, “parapneumonic effusion”, “pneumatocele”, or “lung abscess”; “pneumococcal” or “Streptococcus pneumoniae”; and “serotype” for studies on the epidemiology of complicated pneumonias published from January 1, 1990 to October 1, 2013. Studies with data on incidence and serotypes were included; reviews, case reports, and conference abstracts were excluded. Of 152 papers, 84 fitted the inclusion criteria. A few pneumococcal serotypes were predominant causes of CPP, particularly serotypes 1, 19A, 3, 14, and 7F. CPP was a more common manifestation of pneumococcal disease among older (>2 years old) than younger children. The data support increases in both reported incidence rates and proportions of CPP in children and adults during the period 1990–2012; specific increases varied by geographic region. The proportions of serotype 3 and, particularly in Asia, serotype 19A CPP have increased, whereas most studies show declines in serotype 14. Serotype 1 has been a predominant cause of CPP since 1990, while antibiotic resistance was infrequent among serotype 1 isolates. The reported incidence and proportions of CPP among pneumonia cases steadily increased from 1990 to 2012. Several factors might account for these increases, including enhanced disease detection due to a higher index of suspicion, more sophisticated diagnostic assays, and changes in the prevalence of serotypes with capacity to invade the pleural space that were not targeted by the 7-valent pneumococcal conjugate vaccine (PCV7).


Introduction
Streptococcus pneumoniae is the most common cause of pneumonia in children and a major cause of pneumonia in adults worldwide [1,2]. Among patients with pneumonia, as many as half may develop pleural effusions (i.e., fluid in the pleural space); of these, 5-10 % may progress to empyema [3]. In general, "complicated pneumonia" refers to pneumonia accompanied by pleural effusion. Empyema is a serious complication characterized by pus and bacteria in the pleural space [3,4], which may progress to necrosis, cavitation, or fistulas in the thoracic cavity. S. pneumoniae is the most common cause of complicated pneumonia in children and a common cause in adults [5,6]. Other bacteria associated with acute complicated pneumonias include S. pyogenes, S. milleri, Staphylococcus aureus, Haemophilus influenzae, Mycoplasma pneumoniae, Pseudomonas aeruginosa, other Streptococcus species, and, less commonly, Klebsiella, Enterobacter, Proteus, Salmonella, and Yersinia species [5].
A limited number of pneumococcal serotypes have been associated with CPP and pneumococcal empyema (PnEmp). Changes in the serotype epidemiology of invasive pneumococcal disease (IPD) and pneumococcal pneumonia have been reported in recent years, with significant declines in incidence and in the proportion of disease caused by PCV7 serotypes and increases in the proportion of non-PCV7 serotype disease [22][23][24][25][26][27]. Although these changes are suggestive of serotype replacement, similar trends have also been reported in some countries prior to the introduction of PCV7 [28,29,68].
The purpose of this review is to evaluate the serotype epidemiology of CPP during the period 1990-2012.

Methods
PubMed and EMBASE were searched for studies on the epidemiology and incidence of CPP published from January 1, 1990 through October 1, 2013 using the terms: "empyema", "complicated pneumonia", "pleural infection", "necrotizing pneumonia", "pleural effusion", "parapneumonic effusion", "pneumatocele", or "lung abscess"; "pneumococcal" or "Streptococcus pneumoniae"; and "serotype". Studies with data on incidence, prevalence (i.e., proportion of cases), or serotypes were included; reviews, case reports, and conference abstracts were excluded. In addition, the references sections of relevant review articles were checked for studies not identified during the online search.

Results
A total of 152 papers were initially identified; 68 were excluded because there were no data on incidence or serotypes, or they were case reports; consequently, 84 were included in this analysis. Table 2 presents data on the incidence and proportion of CPP [7, 9, 10, 12-14, 16-19, 30-47, 49-62]. Trends and agerelated differences in the studies are discussed below.
The incidence for PnEmp tended to increase from the late 1990s to the mid-2000s in children in the USA [10,16] and Spain [12], but remained lower in children aged 5-17 years compared with younger children (  [16]. In Spain from 1997Spain from -2001Spain from to 2002Spain from -2006 the PnEmp incidence increased from 2.2 to 9.2 per 100,000 population in children aged <2 years and from 1.5 to 9.2 per 100,000 population in children aged 2-4 years [12]. Of note, two studies from Spain showed significant declines in the PnEmp incidence following the introduction of PCV13 into the pediatric immunization program: in one, from 6.73 to 4.14        [49].

Proportion of cases with CPP
Outcomes used as numerators for proportions of cases with CPP ranged from more general (e.g., pulmonary complications, complicated pneumonia) to more specific (e.g., empyema, cavitatory disease). Denominators varied, including hospitalized patients, hospitalized patients with CAP, hospitalized patients with pneumococcal CAP, or children with parapneumonic PnEmp. In addition, the assessment methods varied in sensitivity, which may have affected the reported proportions. For example, in hospitalized children with empyema in Australia, only 7.5 % of pleural fluid cultures, but 51.0 % of pleural fluid PCRs, were pneumococcal-positive [58].
Effect of age on the incidence and proportion of cases of CPP The effect of age on CPP was not consistent across studies, and age effects for incidence differed as compared to proportion. In most studies, older children comprised a larger proportion of those with CPP or PnEmp relative to other pneumococcal diseases. In Utah (USA) (1997-2010), children with CPP were significantly older than those with other forms of IPD (37 months vs. 25 months; p<0.001) [18]. Among children with pneumococcal pneumonia in the USA (1993)(1994)(1995)(1996)(1997)(1998)(1999)(2000), the proportion of cases with CPP increased with age from 26.4 % (ages 0-12 months) to 53.0 % (ages >61 months) [34]. In contrast, in a study of children from four Asian countries (Vietnam, China, Korea, and Taiwan), empyema and PE were most common in the younger age groups, particularly those ≤4 years of age [13].

Serotype epidemiology
Serotype 19A also appears to be important, particularly in the Asia-Pacific region. For example, serotype 19A caused 69.2 % and 71.0 % of cases of pneumococcal necrotizing pneumonia and PnEmp, respectively, in children in Taiwan [83], 46.2 % of PnEmp cases in children in Korea [54], and 36.4 % of PnEmp cases in children in Australia [58].
Serotypes varied in their prevalence as complicated versus uncomplicated pneumonia. In children in Utah (USA) (1997-                     PnEmp in Spain (2003Spain ( -2006 identified the same three serotype 1 clones in pleural fluid [87]. MLST types associated with the increased incidence of pediatric PnEmp had been present previously in Spain and elsewhere in Europe, and, therefore, the increase in proportions of PnEmp (predominantly serotype 1) was probably not associated with the emergence of new clones or of capsular switching [87,101]. For adults in Spain (1996Spain ( -2010, an increase in PnEmp incidence in otherwise healthy adults with pneumonia was associated predominantly with serotype 1, in particular, ST306 [47].
Several studies evaluated the association of pneumococcal conjugate vaccine introduction with antibiotic resistance among pneumococcal serotypes associated with CPP or PnEmp. In Israel (1990Israel ( -2002, no penicillin-resistant pneumococci were  [94]. Although in some studies the emergence of serotypes 1, 19A, 3, and 14 in CPP and PnEmp has corresponded to the years following the introduction of PCV7 [17,18,31,35,50,52,86], a few studies demonstrated that their increasing role in CPP and PnEmp began prior to the introduction of PCV7 [28,29,68]. For example, in a study of IPD in Spain (1989Spain ( -2008, the proportion of cases of IPD caused by serotype 1 was increasing prior to the introduction of PCV7, and this trend continued after the introduction of PCV7 (2 %, 8.6 %, 14.9 %, and 23.8 % of cases in 1989-1993, 1994-1998, 1993-2003, and 2004-2008, respectively [p<0.001]) [68]. This suggests that emergence involves more than simply serotype replacement following PCV7 introduction; characteristics particular to certain serotypes, such as differences in antibiotic sensitivity or their propensity to cause pleural infection, may be responsible for these increases. In addition, several studies established a relationship between the proportion of CPP and age that contrasts with the age-specific incidence of IPD [32,40,44,49,87,90,97], which may suggest an interaction between the propensity of given serotypes to infect and the age-dependent susceptibility of patients to infection by these same serotypes.
Recent data suggest that PCV13 may impact on pediatric PnEmp caused by serotypes associated with CPP, such as 1 and 19A. In Spain, where PCV13 was introduced in 2010, the incidence of pediatric PnEmp caused by serotypes targeted by  [49].
Although this review has recapitulated the basic epidemiology of CPP (including PnEmp) over the past several decades, the increased use of more specific diagnostic imaging (e.g., CT and ultrasound) in recent years may affect the reported rate of PE and empyema because of the more accurate identification. Improved diagnostic methodologies, clinical (ultrasound or CT vs. physical examination or chest X-ray) and microbiological (e.g., PCR vs. culture), as well as increased awareness of and vigilance for PE and empyema may have resulted in an apparent increase in the proportion of CPP over the time period included in this analysis, which, given the limitations of the retrospective nature of this analysis, may be a confounding factor in the determination of any true increase in the proportion of CPP.
This review is also limited by the fact that many of the studies used culture to identify pathogens. Culture may be less than ideally sensitive to the presence of specific pathogens because of factors such as prior antibiotic use or other difficulties in culturing pathogens. In recent years, PCR has been applied to identify the pneumococci causing CPP and to identify serotypes in culture-negative specimens [32,59,64,84,87,93]. As antibiotic treatment may reduce the likelihood of detecting bacteria via culture, PCR is useful in detecting pneumococci in culture-negative samples from patients previously treated with antibiotics. For example, in a Spanish study of culture-negative pleural fluid specimens from children with empyema, PCR typing identified eight different serotypes (i.e., serotypes 1, 3, 5, 7F/7A, 8, 14, 19A, and 19F/B/C) in 52 of 67 culture-negative pleural fluid samples from children with PnEmp, with a sensitivity of 96.0 % and a specificity of 98.6 % [93]. Spanish pediatric empyema patients with S. pneumoniae culture-negative/PCR-positive samples were found to be significantly more likely to have received antibiotics than those with culture-positive samples (92 % vs. 53 %, respectively; p<0.0001) [87]. Rapid pneumococcal antigen detection by means of immunochromatography has also been used to detect pneumococci in isolates from patients with empyema [51]. In hospitalized children with parapneumonic effusion, conventional microbiologic culture of pleural fluid samples detected pneumococci in 15 of 55 isolates, real-time PCR detected pneumococci in 13 of 16 culture-negative isolates (81.2 %), and immunochromatographic testing detected pneumococci in 24 of 27 culture-negative isolates (88.9 %) [51]. Such molecular methods may complete the information available on changes in the IPD and CPP serotype epidemiology over time.
Finally, most of the studies reviewed here used retrospective database analyses to identify cases of CPP and PnEmp. The specificity of information in these databases could vary, as cases may have been missed due to misclassification. In recent years, increased awareness of CPP and PnEmp has led to prospective surveillance studies, which may enhance disease identification. In addition, studies used different breakpoints to determine penicillin sensitivity, making it difficult to compare antibiotic resistance results.
In conclusion, the reported proportion of cases of CPP and PnEmp due to non-PCV7 serotypes has increased over the past several decades in countries that introduced PCV7 into the pediatric immunization program. Whether this increase reflects the advent and wider use of more specific diagnostic methods and increased awareness due to research initiatives, or if it, indeed, represents a true increase in disease incidence, is unclear. Several factors may account for these greater proportions, including enhanced disease detection due to a higher index of suspicion and more sophisticated diagnostic assays, as well as the prevalence of certain non-PCV7 serotypes that are capable of invading the pleural space.
It is established that reductions in the proportion of PCV7 serotype CPP have been observed in countries using PCV7.
Most serotypes associated with CPP and PnEmp-particularly serotypes 1, 19A, 3, and 7F-are targeted by PCV13, which was registered for pediatric vaccination from 2009 and for adult vaccination from 2011. Early reports suggest declines in the incidence and proportion of cases of vaccine serotype CPP post-PCV7 introduction, and then further with PCV13, although studies are ongoing.