Pleural infection is a common complication in pneumonia and linked with high mortality [18, 19]. Early and reliable detection of the causative organism is beneficial in establishing the correct pathogen-specific treatment, which may lead to decreased morbidity and mortality. However, due to the low sensitivity of conventional cultures , the diagnosis and management of pleural infection remains challenging [2, 3].
In this study we assessed the diagnostic accuracy of a commercial multiplex bacterial PCR for the detection of pathogens in pleural effusion compared with conventional methods in a large group of patients with pleural effusion. We found that the multiplex PCR had a better sensitivity and specificity than conventional methods if the analysis is restricted to microorganisms included on the panel. Indeed, when using the clinical final diagnosis of empyema as a reference, the multiplex PCR-based assay had sensitivity six-times higher than that of the conventional microbiologic methods. However, only a minority of the samples and a small percentage of the pathogens diagnosed by conventional methods were included in this panel including bacteria commonly involved in pneumonia. Considering that empyema is a common complication of pneumonia, it would be reasonable to suggest that the pathogens involved in empyema are similar to those found in pneumonia. Based on pathogens identified for pleural fluid in various other publications [2, 3, 5, 6, 13, 20,21,22], at least 50% of the known pathogens were available on the multiplex PCR-based assay. Yet, only 7 (24%) of the 29 pathogens identified by either conventional or molecular methods in this study were included on the panel of this multiplex bacterial PCR for pneumonia.
Most of the pathogens detected in the patients diagnosed with empyema were anaerobes (69%) followed by gram-positive cocci (22%). These figures were unexpected and differ from previous reports suggesting that organisms related to community-acquired pneumonia, such as S. pneumonia, H. influenza, S. aureus, K.Pneumoniae, are the most common causes of empyema [2, 3, 5, 13, 20, 23]. A more recent paper by Johansson et al. , however, had similar results to our study, in that they found an anaerobe presence of 40%. Thus, empyema might require anaerobe coverage more often than commonly reported.
Streptococcus pneumoniae, the most common pathogen causing parapneumonic effusions [2, 3, 10, 25] was identified in 5 samples using the multiplex PCR-based assay and only once when using conventional methods. Conventional methods identified only one Stenotrophomonas maltophilia sample compared to the multiplex PCR-based assay that identified it in five samples. Although Stenotrophomonas maltophilia is a gram-negative pathogen known to cause many severe infections and can result in mortality in approximately 14.6 to 42.5% of patients [26,27,28], pleural infection associated with this organism is rare . Stenotrophomonas maltophilia is thought to cause nosocomial infections mainly in immunocompromised patients , but in our study, only one patient identified with this pathogen was immunocompromised and four of the five cases positive for Stenotrophomonas maltophilia had a final diagnosis of empyema. It is plausible that the molecular diagnostic technique could improve the recognition of this pathogen, which is associated with a particularly high mortality rate.
In 85% (52/61) of the cases diagnosed with empyema, antibiotics had been administered before sampling. As found previously , we found no association between the multiplex PCR-based assay and prior administration of antibiotics or duration of antibiotic use before sampling in the cases diagnosed with empyema. However, antibiotic use before sampling and the duration of antibiotic use before sampling was associated with more bacterial negative results when using the conventional method in the cases diagnosed with empyema.
When using the final clinical diagnosis of empyema (yes or no) as reference-standard, the multiplex PCR-based assay had a sensitivity of 18.6% and a specificity of 97.3%. When restricting the analysis to pathogens included in the panel, conventional methods had a sensitivity of 3.3%. Twenty-two pathogens were detected by conventional methods and were not on the multiplex PCR panel and when these pathogens were included in the analysis, conventional methods had a sensitivity of 32.8%.
The diagnostic accuracy of this multiplex bacterial panel focusing on pneumonia pathogens for pleural effusion was disappointingly low and highlighted the fact that before introduction in clinical practice, any diagnostic approach should be evaluated in a clinical study. It was apparent that a dedicated panel including anaerobes would be necessary to increase the diagnostic yield of a molecular technique aimed at diagnosing pleural infection.
In this study, we evaluated the diagnostic yield of conventional culture and a molecular technique in pleural effusions in a well-characterized, considerable sized cohort of patients. It is important to take into consideration that a bacterial multiplex PCR is limited its diagnostic yield due to the finite number of in-panel organisms. In addition, specificity issues regarding Enterobacter spp., H. influenzae and S. pneumoniae in comparison to culture-based methods have been previously reported for the molecular diagnostic test prototype . Unlike others  with relevant technical failure rates when analysing respiratory secretions, the multiplex PCR assay had a 2.3% (5/214) failure rate in the current study when analysing pleural fluid.
Amongst its limitations, one has to underline the fact that the molecular diagnostics were performed on frozen material which may have affected the outcome of the analysis. Fernández-Soto P et al. found that less deoxyribonucleic acid (DNA) is extracted from urine frozen and stored for an extended time period compared to fresh urine . Cushwa and Medrano found differences in DNA extraction from blood, based on length of storage time and storage temperature . Thus, the sensitivity of the method may have been negatively impacted by the storage of the samples. However, the main obstacle for the detection of the pathogens by the PCR was the fact that they were not included in the spectrum of the panel for pneumonia. Thus, an increase in DNA amount was not expected to positively increase the sensitivity of the assay. Further, we had not used any further method to confirm the presence or absence of bacteria in the pleural effusion. This was a natural limitation as currently no diagnostic technique would be able to identify infection in this setting . Nevertheless, the clinical final diagnosis had been additionally used as a reference-standard to identify infection. Recently, next-generation sequencing (NGS) was used to identify pathogens in empyema and 385 bacterial detections were made, compared to 38 by culture and 87 by 16S rRNA gene sequencing . The authors concluded that a subgroup of patients with empyema have pathogens more resembling a brain abscess than pneumonia, ie. anaerobic pathogens and that this subgroup should be seen as primary empyema. With this new technique many more bacterial detections were made, but the conclusion that anaerobes are important in empyema remains the same as the conclusion we have drawn in our study. It would be interesting to compare sensitivity and specificity as well as cost-effectiveness between a multiplex PCR-based assay and NGS.
We were unable to determine the clinical implications of the commercial multiplex PCR-based assay on the treatment and outcome of the patients. Thus, further studies are necessary to evaluate whether a molecular diagnostic technique can be helpful in optimizing treatment of pleural infections. We believe that our results are generalizable, in that anaerobes may be important in empyema, but the exact species may differ between regions.