Background: Antibiotic choices for pleural infection are uncertain as its bacteriology is poorly described.
Methods: Pleural fluid from 434 pleural infections underwent standard culture and a screen for bacteria by amplification and sequencing of bacterial 16S ribosomal RNA gene.
Results: Approximately 50% of community-acquired infections were streptococcal, and 20% included anaerobic bacteria. Approximately 60% of hospital-acquired infections included bacteria frequently resistant to antibiotics (methicillin-resistant Staphylococcus aureus, 25%; Enterobacteriaceae, 18%; Pseudomonas spp., 5%, enterococci, 12%). Mortality was increased in hospital-acquired infection (hospital, 17/36 [47%]; community, 53/304 [17%]; relative risk, 4.24; 95% confidence interval, 2.07–8.69; p < 0.00001; χ2, 1 df = 17.47) and in gram-negative (10/22 [45%]), S. aureus (15/34 [44%]), or mixed aerobic infections (13/28 [46%]), compared with streptococcal infection (23/137 [17%]) and infection including anaerobic bacteria (10/49 [20%]; p < 0.00001, χ2, 4 df = 23.35).
Conclusion: Pleural infection differs bacteriologically from pneumonia and requires different treatment. Antibiotics for community-acquired infection should treat aerobic and anaerobic bacteria. Hospital-acquired, gram-negative S. aureus and mixed aerobic infections have a high mortality rate.
Bacterial pleural infection has been a substantial clinical challenge since ancient times. Descriptions of its treatment date from Hippocrates and perhaps ancient Egypt (1, 2). Today, it affects up to 65,000 patients each year in the United Kingdom and United States (3) and has a 12-mo mortality of 22%, with another 15% of patients requiring surgical abscess drainage (4). Appropriate antibiotic choices are important to minimize this morbidity, and these choices are likely to be different from those of pneumonia because small series suggest these syndromes differ bacteriologically (5–15). However, this difference in antibiotic choice is unclear because these studies were small, biased by case selection, or retrospective (5–15), and do not take into account the increasing prevalence of antibiotic-resistant pathogens. Antibiotic selection is also problematic because about 40% of patients with convincing pleural infection have no pathogens identified on standard laboratory culture (4).
The First Multicenter Intrapleural Sepsis Trial (MIST1) assembled a large, well-characterized patient cohort (4), and this article presents the bacterial microbiology of this cohort, including prognostic significance of different bacterial isolates. The results include a screen for bacterial pathogens from pleural fluid samples by the amplification and sequencing of the bacterial 16S ribosomal RNA gene, which has been shown to improve the identification of pathogens from cases of pleural infection (15).
This analysis provides a detailed description of the microbiology of pleural infection using current techniques, describes how this differs from pneumonia, and determines the influence of different bacterial species on prognosis, allowing clinicians to make logical therapeutic choices.
This article reports the bacteriology from the Medical Research Council/ British Thoracic Society (BTS) MIST1 trial of streptokinase in pleural infection (4). This trial showed no treatment effect from streptokinase and so patient outcome is not confounded by this factor. Briefly, the MIST1 trial recruited 454 patients from 52 centers in the United Kingdom. Entry criteria were macroscopically purulent, or bacterial culture, or Gram stain–positive pleural fluid, or a pleural fluid of pH < 7.2, in the presence of clinical evidence of infection. Apart from trial intrapleural streptokinase, patients received standard clinical care. All patients received antibiotics at the discretion of their local managing physician, but antibiotic guidelines were provided in the trial protocol (available in the online supplement).
Baseline microbiological information included the results of pleural fluid Gram stain, and aerobic and anaerobic pleural fluid culture, performed in the recruiting center. The trial outcomes included mortality, which is used in the survival analyses presented here. At randomization, pleural fluid was collected and transferred to the coordinating center and frozen at −70°C. Samples were frozen within 48 h of being taken. Culture-negative pleural fluid samples were also gathered from 20 patients with pleural effusions due to noninfectious causes to act as control samples. The trial was approved by the Anglia and Oxford Multicentre Research Ethics Committee (MREC) (ref: 98/5/61), the Oxford Local Research Ethics Committee, and the local research ethics committees for each center. All participants gave informed, written consent.
Pleural fluids were thawed from –70°C storage and equilibrated to ambient temperature. Bulk nucleic acid was promptly extracted using the Qiagen QIAmp Mini Kit following the manufacturer's guidelines (Qiagen, Hilden, Germany). The nucleic acid was eluted in 100 μl of buffer.
The polymerase chain reaction (PCR) was modified from Woo and colleagues (17). The PCR reaction mixture consisted of 5 μl Bioline KCl buffer (Bioline, London, UK), 27.3 μl PCR-grade water, and final concentrations of 0.5mM MgCl2, 0.2mM dNTPS (Promega), and 0.5 μM each primer (Sigma Genosys, Poole, UK) and 1 unit per reaction Taq polymerise (Bioline). The PCR reaction mixture was aliquoted and exposed to ultraviolet light for 2 min. Ten microliters of nucleic acid extract were added to give a final volume of 50 μl. The cycling conditions were as follows: 94°C for 5 min, followed by 30 cycles of 94°C for 1 min, 60°C for 1 min, and 72°C for 2 min, with a final elongation at 72°C for 5 min (Progene Thermal Cycler; Techne, Cambridge, UK).
PCR amplimers were cleaned using QIAquick PCR clean-up (QIAquick PCR purification kit; Qiagen), and the size of the amplion was verified using Hyperladder I (Bioline). Sequencing reactions were performed using the BigDye Terminator Cycle Sequencing Ready Reaction DNA sequencing kit (Applied Biosystems, Inc., Foster City, CA) and analyzed on the ABI 377 Genetic Analyser (Applied Biosystems, Warrington, UK). Sequences obtained were in the region of 400–700 base pairs. Sequencing was performed twice in the forward direction only, and the sequences aligned to ensure base pair accuracy using the CLUSTALW algorithm.
Twelve samples that had mixed sequences underwent cloning using the TOPO TA cloning kit (Invitrogen, Paisley, UK). This cloning strategy was limited to 12 samples for reasons of cost. Cloning reactions contained the following: 1 μl fresh PCR product, 1 μl salt solution, 2 μl sterile water, and 1 μl TOPO vector. The mixtures was incubated at room temperature for approximately 20 min and then placed on ice. The 2 μl of the cloning mixture were added to the competent cells (that had been thawed on ice) and heat-shocked at 42°C for 30 s and then incubated with 250 μl media at 37°C on an orbital shaker. The 10- and 50-μl volumes were then spread on Luria-Bertani (LB) agar containing 50 μg/ml ampicillin. Plates were incubated overnight at 37°C.
Twelve colony picks of transformants per sample were transferred directly in M13 PCR mixtures. The 50-μl M13 PCR reaction mixtures contained the following: 0.2 μM each primer, 0.2 mM deoxynucleotide triphosphates (dNTPs), 2 mM MgCl2 (Bioline), 5 μl buffer, and 1 unit Taq polymerase (Bioline). Primer sequences were as follows: M13(f) 5′CAGGAAACAGCTATGA and M13(r) GTAAAACGACGGCCAG. The cycling conditions were as follows: initial denaturation, 94°C for 5 min; then 30 cycles of 94°C for 40 s; 55°C for 40 s; and 72°C for 1 min followed by a final elongation at 72°C for 7 min. Amplimers were resolved through 1% agarose, and successful transformants were assumed to be those with bands at 1,300 bp. Amplimers were then cleaned and sequenced using the methods described above (using 16S primers).
Identification of bacteria from pleural fluids was performed by the comparison of bacterial DNA sequences coding for the 16S ribosomal RNA subunit with the GenBank database available at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov). A positive identification was recorded when the sequence completely matched a database entry and other matches showed significantly less homology. Sequences that showed clear amplification and strong signal but were unreadable (i.e., where base pairs could not be assigned due to two peaks in the same position) were recorded as “mixed.” Twelve of these samples were then cloned and resequenced and the identities of the bacteria present clarified. Details of the identity percentage data for the genetic bacterial identification are available from the corresponding author, if required.
To compare the results of the standard pleural fluid bacterial culture and the nucleic acid amplification, pre hoc rules defining these comparisons were established. These rules were as follows:
Where the bacteria identified by culture and nucleic acid amplification were identical (or both were negative), this was defined as “agreement.”
Where the identified bacterium was consistent, but one technique provided better information, this was defined as “agreement but superior” (e.g., methicillin-resistant Staphylococcus aureus by culture/S. aureus by nucleic acid amplification = “agreement, culture superior”, or S. milleri by culture/S. intermedius by nucleic acid amplification = “agreement, nucleic acid amplification superior”).
Where one technique identified a bacterium that was not detected by the other, this was defined as “superior” (e.g., nucleic acid amplification negative/S. aureus by culture = “culture superior”).
Where the two techniques identified different bacteria, it was considered impossible to assess which test was more informative (e.g., Streptococcus pneumoniae by culture/Haemophilus influenzae by nucleic acid amplification = “indeterminable”).
The survival of patients in the different bacterial pathogen groups was defined by selecting groups of subjects with bacteria of a particular class blind to these subjects' clinical outcome. Mortality was then studied in these groups. The following groups were identified to this analysis:
No pathogen identified (n = 71)
S. pneumoniae (n = 59)
S. intermedius group (n = 55)
Other streptococci (n = 23)
Anaerobic or mixed aerobic/anaerobic infection (n = 49)
S. aureus (n = 34)
Mixed aerobic bacteria (n = 28)
Gram-negative bacteria (n = 22)
Survival was also compared between cases acquired in the community and in hospital.
The χ2 analysis and Fisher exact test were used when comparing proportions and overall mortalities. Cohort survivals were described using Kaplan-Meier survival curves. Statistical analysis was performed with SPSS, version 10 (SPSS, Inc., Chicago, IL).
The clinical and blood culture characteristics of the subjects with pleural infection are shown in Table 1, and elsewhere (4). Of the patients, 82 of 454 (18%) had chronic lung disease. The bacteriology was similar in this group to the whole sample.
Age, yr, mean (SD)
|Male, n (%)||302 (67)|
|Duration of symptoms prior to presentation, median (IQR)||14 (8–28)|
|Comorbidity, n (%)|
|Diabetes mellitus||48/454 (11)|
|Excessive alcohol consumption||43/454 (9)|
|Etiology of pleural infection, n (%)|
|Community-acquired infection||394 (87)|
|Hospital-acquired pneumonia||29 (6)|
|Post–thoracic surgery||11 (2)|
|Post–abdominal surgery||5 (1)|
| Unclassified||11 (2)|
Twenty age-matched control patients (11 male; mean age, 69 yr [SD, 20 yr]), with pleural effusion believed clinically to be of a noninfectious cause, gave samples to act as negative controls. Seven were transudative effusions (by “Light's criteria” ) and 13 had histocytologically proven malignant pleural effusion. These controls were selected on the basis of the security of their diagnosis and negative standard cultures for aerobic and anaerobic bacteria, mycobacteria, and fungi.
Of cases, 434 of 454 (95%) had standard microbiological pleural cultures and Gram's stain performed, of which 250 of 434 (58%) were positive, as follows: 151 (35%) produced a single aerobic growth, 29 (9%) produced a single anaerobic growth, 52 (12%) produced polymicrobial cultures, and 18 (2%) were Gram-stain positive only. In 20 subjects, pleural fluid was not available for culture. Seventy-seven subjects received antibiotics before pleural fluid sampling, and culture was negative in 47 of (61%) of these patients. Pleural fluid was available for molecular microbiological analysis in 404 of 434 (93%) subjects. Seventy (16% of total sample) of culture-negative cases had bacteria identified by subsequent nucleic acid amplification, leaving 114 of 434 (26%) cases still bacteriologically obscure.
Among the 12 of 404 (3%) cases where nucleic acid amplification showed more than one bacterial DNA sequence, cloning suggested the following bacteria to be present. Six samples revealed only one bacterium: Proechimys oris in two cases; and one case each of Staphylococcus intermedius, Streptococcus pneumoniae, Streptococcus pyrogenes, and Filifactor micros. Six samples revealed more than one bacterium: Fusobacterium nucleatum and Peptostreptococcus micros in three cases; and one case each of Bacteriodes fragilis + P. oris; F. nucleatum subsp. vincentii + P. oral clone; Peptostreptococcus micros + Peptostreptococcus oral clone + Fusobacterium oral clone + Bacteroides oral clone. These results were combined with those from nucleic acid amplification.
The summary bacteriology derived from both standard culture and nucleic amplification is presented in Table 2. S. intermedius, S. anginosis, and S. constellatus were grouped together as the “S. intermedius group.”
Community Acquired (no. isolates)
Hospital Acquired (no. isolates)
|Streptococcus intermedius– anginosus–constellatus (“milleri”) group||80||4|
|Other Streptococcus species||16||4|
|Methicillin-resistant S. aureus||7||15|
|Mixed anaerobes, unclassified||8||2|
The bacteriologies of hospital- and community-acquired infection differed substantially (Table 2), and both differ from pneumonia, which is consistent with previous small reports (5–15). Pleural infections acquired in the community were most frequently due to streptococcal infection (S. pneumoniae, 71/336 [21%]; S. intermedius group, 80/336 [24%]; other streptococcal species, 25/336 [7%]), and infections due to S. aureus and enterococci were more prevalent in the infections acquired in hospital. Fifteen of the 60 (25%) isolates in hospital-acquired infection were due to methicillin-resistant S. aureus.
Overall mortality was substantially increased in subjects who acquired their infection in hospital compared with those who acquired their infection in the community (hospital mortality, 17/36 [47%]; community, 53/304 [17%]; relative risk, 4.24; 95% confidence interval [CI], 2.07–8.69; p < 0.00001; χ2, 1 df = 17.47). Survival curves are shown in Figure 1.
One-year mortality outcome data were available in 438 of 440 (99.5%) subjects in the survival analyses. All the streptococcal subsets showed similar mortalities (S. pneumoniae, 10/59 [17%]; S. intermedius group, 9/55 [16%]; other streptococci, 4/23 [17%]; p = 0.92; χ2, 2 df = 0.17; see survival curve in Figure E1 of the online supplement). These groups were combined in later comparisons.
The mortality in the group in whom no pathogen was identified was 9 of 71 (13%), similar to that in streptococcal infection, which was 23 of 137 (17%; relative risk, 1.27; 95% CI, 0.53–3.06; p = 0.60; χ2, 1 df = 0.28; survival curve in Figure E2).
Mortality at 1 yr varied substantially between the culture-positive bacterial groups (p < 0.00001; χ2, 4 df = 23.35; survival curve in Figure 1). This was due to a statistically significant increase in mortality in three groups: those with gram-negative bacteria (10/22 [45%]), S. aureus (15/34 [44%]), or mixed aerobic bacteria (13/28 [46%]), compared with the others; those with streptococcal infection (23/137 [17%]); and those with infection including anaerobic bacteria (10/49 [20%]).
The pathogens causing hospital-acquired infection differ from those causing community-acquired infection and are generally from bacterial classes with a worse outcome. To clarify whether the bacterial mortality effect and the hospital-acquired mortality effect were independent, we have explored this interaction in Table 3. This shows that the place of infection and the type of bacterial infection are independently associated with variations in mortality.
Statistical Significance in Hospital- and Community-acquired Infection (χ2)
|Streptococcal infection||19/131||0/6||p = 0.60 (χ2, 1 df = 1.01)|
|Staphylococcal infection||6/22||8/12||p = 0.04 (χ2, 1 df = 4.98)|
|Infections including anaerobic bacteria||5/44||2/4||p = 0.13 (χ2, 1 df = 3.37)|
|Infections with mixed aerobic bacteria||5/18||7/10||p = 0.05 (χ2, 1 df = 4.68)|
|Infections with gram-negative bacteria||9/20||0/2||p = 0.49 (χ2, 1 df = 1.52)|
|Statistical significance across bacterial groups (χ2)||p = 0.01 (χ2, 4 df = 13.13)||p = 0.03 (χ2, 4 df = 10.93)|
In 140 of 404 (35%) cases, the same organism was found by both nucleic acid amplification (or cloning) and standard culture. Of these, using the rules previously described, in 120 of 140 cases, the standard culture and nucleic acid amplification were in agreement; in 9 of these 140, the standard culture was superior; and in 11 of 140 cases, the nucleic acid amplification was superior.
In 107 cases, the standard culture was superior to the nucleic acid amplification because it identified a bacterium not identified from the DNA studies. In 54 cases, nucleic acid amplification was superior to standard culture because it identified a bacterium not identified by culture. In 50 cases, the standard culture and the nucleic acid amplification identified different bacteria, and it could not be assessed which method was more informative. The results of conventional culture and nucleic acid amplification are compared in Table 4.
Number of Positive Isolates
|Organism||Bacteria Isolated by Conventional Culture and Nucleic Acid Amplification||Bacteria Isolated by Conventional Culture But Not by Nucleic Acid Amplification||Bacteria Isolated by Nucleic Acid Amplification But Not by Culture|
|S. intermedius–anginosus–constellatus group||23||26||10|
|Methicillin-resistant S. aureus||3||5|
|Mixed (anaerobes only)||4||4||1|
|No organism identified (negative)||89|
In 19 of the 20 culture-negative and presumed uninfected control subjects, no bacterial DNA was identified. In one, a Prevotella spp. was identified (from a patient with malignant pleural mesothelioma). Post hoc review of the case notes showed this patient was febrile at the time of pleural fluid sampling, although this fever had been attributed to the tumor.
This report has clarified the bacteriology of pleural infection and may improve antibiotic choices for the 65,000 people who develop this infection each year in the United Kingdom and the United States, helping minimize the 22% mortality associated with pleural infection (4). It is the first study to examine the standard culture and genetic bacteriology of pleural infection in a large generalizable cohort, and to relate this to patient mortality. These data provide a foundation for clinical trials to define whether better bacterial diagnosis and antibiotic choices improve outcome in pleural infection.
The cases studied here were accumulated from 52 centers in the United Kingdom, including both teaching and district hospitals, and so are likely to be representative. This study's results include DNA sequencing and cloning to clarify the bacteriology in detail, a technique that has proved effective in other smaller samples (15). This has reduced the number of cases that are bacteriologically undiagnosed from 42 to 26%, a diagnostic improvement that is likely to be clinically valuable.
This analysis confirms smaller studies (5–15), which suggest the bacteriology of pleural infection differs from that of pneumonia (reviewed in Reference 21), and that these syndromes should be considered separate. Because pleural infection follows bacterial migration from beneath the visceral pleura, it is often termed “complicated parapneumonic effusion,” which could be taken to imply a bacteriologic etiology similar to that of pneumonia. This study shows that this is simplistic and antibiotic choices targeted at the typical range of pneumonic pathogens are not ideal for pyogenic pleural infection. The differences in the bacteriology are probably due to the acidic and hypoxic environment of the infected pleural space favoring selected pathogens. Many of the anaerobic bacteria of pleural infection are strictly anaerobic and cannot tolerate the Po2 of lung parenchyma, whereas streptococci of the “intermedius–anginosus–constellatus” group characteristically flourish in low pH and Po2 tissue environments and favor these conditions in artificial culture (19). The microbiological differences from pneumonia suggest that future studies should try to define which clinical phenotypes are associated with which pattern of pathogens. For example, preliminary data are beginning to suggest that different radiographic patterns of lung disease are associated with different bacterial pathogens (20).
Interestingly, even the genetic bacterial analysis presented here may underestimate the variety of bacteria present in pleural infection. Our limited cloning strategy suggests even greater bacterial diversity than the 16S amplification, and may provide more clinical information in the future.
Of the cases with a confirmed bacteriology, community-acquired pleural infection is caused by penicillin-sensitive streptococci in about 50% of cases, with the other 50% being due to organisms that are usually penicillin resistant, including staphylococci and Enterobacteriaceae. About 25% of community-acquired pleural infections include anaerobic bacteria. Appropriate empiric antibiotic choices for these patients should therefore cover streptococci, penicillin-resistant staphylococci, and Enterobacteriaceae and should usually also include anaerobic bacterial therapy.
By contrast, hospital-acquired infection includes more staphylococcal infection (more than 70% of which is due to methicillin-resistant S. aureus) and Enterobacteriaceae, organisms that are often multiple-drug resistant. This may contribute to the poor prognosis in these patients. Here, empiric antibiotic therapy should be effective against these multidrug-resistant organisms. The differences in the bacteriology and the prognosis between hospital- and community-acquired empyema are sufficiently marked that these syndromes should probably be considered separate clinical entities.
Mycoplasma spp., and Legionella spp. are respiratory pathogens that are difficult to isolate using routine culture methods. The molecular bacterial techniques found no evidence of either of these species, which suggests that these pathogens are not likely causes of pleural infection and that empiric antibiotic regimes targeted at them are probably not required. By contrast, molecular diagnostics increased the yield of anaerobic bacteria in comparison to routine culture methods, which is expected given the difficulties of isolating these fastidious organisms in a clinical microbiological laboratory. Therefore, empiric antibiotics targeted at anaerobic infection should be usual. An exception to this is where routine culture has identified S. pneumoniae (because we found no evidence of any coincidental anaerobic infection in these cases, either by standard culture or nucleic acid amplification). Here, specific pneumococcal therapy would be appropriate provided the prevalence of pneumococcal penicillin resistance is low.
There are substantial differences in mortality in different bacterial subgroups. Patients with streptococcal infection generally have the best prognosis, with 83% of these patients alive at 12 mo (Figure E1). A similar 87% survival is seen in subjects in whom no pathogen can be identified by either culture or nucleic acid amplification (Figure E2). This suggests that the majority of culture- and nucleic acid amplification–negative cases may be of streptococcal origin, with antibiotic therapy having suppressed bacterial numbers to undetectable concentrations.
Most previous reports of the bacteriology of pleural infection have suggested that infections including anaerobic bacteria have a particularly high mortality (7, 8, 11). The data reported here contradict this view, showing that infections including anaerobic bacteria have an 80% survival, which is similar to streptococcal infections (Figure 1). The groups with the highest mortality were those patients with staphylococcal, enterobacterial, and mixed aerobic infections: about 45% at 1 yr (Figure 1). Because standard clinical care (with chest tube drainage and antibiotics) is associated with such a high mortality in this group, rapid, aggressive empyema drainage (e.g., with early surgical intervention) should be considered for these patients.
Our study demonstrates that bacterial nucleic acid amplification allows the identification of the pathogen in up to 75% of cases. This rate could have been improved further if the cloning strategy had been applied to all the cases where mixed sequences were detected. False-positive rates with nucleic acid amplification have been low. Sterile laboratory saline controls have been reliably negative, and only one of the fluids from patients not believed to have pleural infection identified any bacterial DNA. Interestingly, with hindsight, this subject was febrile when the pleural fluid sample was collected, with a fever attributed to his tumor. This result suggests it could have been infectious in origin. Larger studies relating clinical phenotypes with genetic bacteriology in patients with sterile pleural fluid on standard culture are needed to resolve whether such isolates are true or false positives.
The substantial increase in the number of bacteria identified by nucleic acid amplification when compared with normal culture in this study is similar to the advantage seen when this strategy was applied to a small sample of pediatric empyema fluids (15), strengthening the case for its use.
This study presents the largest comprehensive description of the bacteriology of pleural infection using conventional and molecular methods. The results confirm pleural infection is bacteriologically significantly different from pneumonia, and that hospital-acquired pleural infection is a subentity with a high mortality. Over 40% of hospital-acquired infections were due to multiantibiotic-resistant pathogens. Three bacterial subgroups were associated with a substantially increased mortality and should probably be targeted for early surgical abscess drainage. In contrast to previous reports, we found that infections including anaerobic bacteria do not have a poor outcome. This improved description of the bacteriology of this disease allows better antibiotic and therapeutic strategies to reduce the high morbidity and mortality associated with this disease.
The authors thank Prof. John Macfarlane, City Hospital, Nottingham; Dr. Robert Miller, University College, London; and Prof. Tim Peto, Oxford University, Oxford, for their advice on the manuscript.
The samples were provided by the Medical Research Council/BTS MIST1 trial group comprising the following members:
O.A. Afolabi (North Tyneside General Hospital); N. Ali (Kingsmill Hospital, Mansfield); M. Allen (North Staffordshire Hospital); A. G. Arnold (Castle Hill Hospital, Cottingham); D. Baldwin (City Hospital, Nottingham); J. R. Bateman (Derby City General Hospital); J. Bennett (Derby Royal Hospital); A. Bentley (North Manchester Hospital); D. Boldy (Pilgrim Hospital, Boston); M. Bone (South Tyneside District Hospital); M. Britton (St. Peters Hospital, Surrey); S. Burge (Birmingham Heartlands Hospital); R. Butland (Gloucester Royal Hospital); I. Campbell (Llandough Hospital, Cardiff); J. Congleton (Worthing Hospital); D. C. Currie (Dewsbury & District Hospital); R. J. O. Davies (Oxford Radcliffe Hospital); C. W. H. Davies (The Royal Berkshire Hospital, Reading); A. G. Davison (Southend Hospital); O. Dempsey (Aberdeen Royal Infirmary); R. G. Dent (QEII, Welwyn Garden City); A. P. Greening (Western General Hospital, Edinburgh); J. Hadfield (Chesterfield & Derbyshire Royal Hospital); B. D. W. Harrison (Norfolk and Norwich Hospital); J. Harvey (Southmead Hospital, Bristol); M. Hetzel (Bristol Royal Infirmary); H. Hosker (Airedale General Hospital, Keighley); P. Hughes (Derriford Hospital, Plymouth); A. M. Hunter (York District Hospital); M. Jackson (Brighton General Hospital); W. J. M. Kinnear (Queens Medical Centre, Nottingham); C. Laroche (West Suffolk Hospital, Bury St. Edmonds); C. Laroche (Papworth Hospital, Cambridge); A. Leonard (Great Western Hospital, Swindon); M. C. I. Lipman (Royal Free Hospital, London); D. Lomas (Addenbrookes Hospital, Cambridge); A. Malin (Royal United Hospital, Bath); D. McLeod (Sandwell General Hospital, West Bromwich); M. L. D. Morgan (Glenfield Hospital, Leicester); M. F. Muers (The Leeds General Infirmary); N. C. Munro (University Hospital of North Durham); E. Neville (St. Mary's Hospital, Portsmouth); D. J. Newbury (Ashford Hospital); D. Peckham (St. James University Hospital, Leeds); M. Pond (Bridlington & District Hospital); S. Saboor (Chase Farm Hospital, Enfield); D. Seaton (Ipswich Hospital); R. Smith (Horton General Hospital, Banbury); S. Spiro (The Middlesex Hospital, London); B. Stack (West Glasgow Hospital); D. Thickett (Queen Elizabeth Hospital, Birmingham); C. Wathen (Wycombe General Hospital); J. H. Winter (Nine Wells Hospital, Dundee); N. Withers (Royal Devon & Exeter Hospital); M. Woodhead (Manchester Royal Infirmary); F. Zaman (Pendle Community Hospital, Burnley).
|1.||Hippocrates. Aphorism 44. In: Adams LB, editor. The genuine works of Hippocrates: the classics of surgery. Birmingham AL.: Gryphon Editions; 1985. pp. 768–771.|
|2.||Breasted JH. The Edwin Smith surgical papyrus. Chicago, IL: University of Chicago Press; 1980.|
|3.||Colice GL, Curtis A, Deslauriers J, Heffner J, Light R, Littenberg B, Sahn S, Weinstein RA, Yusen RD. Medical and surgical treatment of parapneumonic effusions: an evidence-based guideline. Chest 2000; 118:4–71.|
|4.||Maskell NA, Davies CWH, Nunn AJ, Hedley EL, Gleeson FV, Miller R, Gabe R, Rees GL, Peto TE, Woodhead MA, et al. UK controlled trial of intrapleural streptokinase for pleural infection. N Engl J Med 2005;352:865–874.|
|5.||Alfageme I, Munoz F, Pena N, Umbria S. Empyema of the thorax in adults. Etiology, microbiologic findings, and management. Chest 1993; 103:839–843.|
|6.||Ali I, Unruh H. Management of empyema thoracis. Ann Thorac Surg 1990;50:355–359.|
|7.||Bartlett JG. Anaerobic bacterial infections of the lung and pleural space. Clin Infect Dis 1993;16:S248–S255.|
|8.||Bartlett JG, Gorbach SL, Thadepalli H, Finegold SM. Bacteriology of empyema. Lancet 1974;7853:338–340.|
|9.||Brook I, Frazier EH. Aerobic and anaerobic microbiology of empyema: a retrospective review in two military hospitals. Chest 1993;103:1502–1507.|
|10.||Civen R, Jousimies-Somer H, Marina M, Borenstein L, Shah H, Finegold SM. A retrospective review of cases of anaerobic empyema and update of bacteriology. Clin Infect Dis 1995;20:S224–S229.|
|11.||Ferguson AD, Prescott RJ, Selkon JB, Watson D, Swinburn CR; Empyema Subcommittee of the Research Committee of the British Thoracic Society. The clinical course and management of thoracic empyema. Q J Med 1996;89:285–289.|
|12.||Hernandez B, Alfageme M, Munoz M, Campos R, Pena G, Villagomez C. Thoracic empyema in HIV-infected patients: microbiology, management, and outcome. Chest 1998;113:732–738.|
|13.||Smith JA, Mullerworth MH, Westlake GW, Tatoulis J. Empyema thoracis: 14-year experience in a teaching center. Ann Thorac Surg 1991;51: 39–42.|
|14.||Snider GL, Saleh SS. Empyema of the thorax in adults:review of 105 cases. Dis Chest 1968;54:410–415.|
|15.||Saglani S, Harris KA, Wallis C, Hartley JC. Molecular techniques improve organism identification from pleural fluid in empyema. Arch Dis Child 2005;90:70–73.|
|16.||Martin FE, Nadkarni MA, Jacques NA, Hunter N. Quantitative microbiological study of human carious dentine by culture and real-time PCR: association of anaerobes with histopathological changes in chronic pulpitis. J Clin Microbiol 2002;40:1698–1704.|
|17.||Woo PCY, Cheug EYL, Leung K, Yuen K. Identification by 16S ribosomal RNA gene sequencing of an enterobacteriaceae species with ambiguous biochemical profile from a renal transplant recipient. Diagn Microbiol Infect Dis 2001;39:85–93.|
|18.||Light RW, MacGregor MI, Luchsinger PC, Ball WCJ. Pleural effusions: the diagnostic separation of transudates and exudates. Ann Intern Med 1972;77:507–513.|
|19.||Whiley RA, Beighton D, Winstanley TG, Fraser HY, Hardie JM. Streptococcus intermedius, Streptococcus constellatus, and Streptococcus anginosus (the Streptococcus milleri group): association with different body sites and clinical infections. J Clin Microbiol 1992;30:243–244.|
|20.||Rahman NM, Batt S, Gleeson FV, Hedley EL, Maskell NA, Gillespie S, Davies RJO. The relationship between lung parenchymal CT appearances and pleural bacteriology in pleural infection [abstract]. Am J Respir Crit Care Med 2006:173:A11.|
|21.||Marrie TJ, editor. Community acquired pneumonia. New York: Plenum Press; 2001.|