Rationale. Pleural infections are common and associated with substantial healthcare cost, morbidity, and mortality. Accurate pleural infection diagnosis remains challenging due to low culture positivity rates, frequent polymicrobial involvement, and non-specific diagnostic biomarkers.
Objective. To undertake a prospective pilot study examining the feasibility and challenges associated with various molecular methods for diagnosing suspected pleural infection.
Methods. We prospectively characterised 26 consecutive, clinically suspected pleural infections, and 10 consecutive control participants with suspected non-infective pleural effusions, using shotgun metagenomics, bacterial 16S ribosomal RNA metataxonomics, panbacterial and panfungal quantitative PCR, and conventional culture.
Results. We demonstrate the feasibility of culture-independent molecular techniques for diagnosing suspected pleural infection. Molecular methods exhibited excellent diagnostic performance, with each method identifying 54% (14/26) positive cases among the pleural infection cohort, vs. just 38% (10/26) with culture. Meta-omics methods unveiled complex polymicrobial infections largely missed by culture. Dominant infecting microbes included streptococci (S. intermedius, S. pyogenes, S. mitis), Prevotella spp. (P. oris, P. pleuritidis), staphylococci (S. aureus, S. saprophyticus), and Klebsiella pneumoniae. Of the three PSI cases with <30-day mortality, two were infected with K. pneumoniae, and two had polymicrobial infections according to metagenomics. In contrast, no patient surviving beyond 30 days (n=23) harboured Klebsiella spp. according to any testing method. Metagenomics detected the non-pathogenic single-stranded DNA virus, adeno-associated dependoparvovirus A, in one control participant. Unexpectedly, pleural fluid from two control participants cultured Candida parapsilosis and Moraxella osloensis; the latter was also observed in the metagenomic data. We encountered challenges that complicated pleural infection interpretation, including: i) uncertainties regarding microbial pathogenicity and the impact of prior antibiotic therapy on diagnostic performance; ii) lack of a clinical diagnostic gold-standard for molecular performance comparisons; iii) potential accidental microbial contamination during specimen collection and processing; and iv) difficulties distinguishing background microbial noise from true microbial signal, particularly in low-biomass specimens.
Conclusions. Our pilot study demonstrates the potential utility and value of molecular methods in diagnosing pleural infection, and highlights key concepts and challenges that should be addressed when designing larger prospective trials.