Abstract
Pleural space infections (PSIs) are common and associated with substantial healthcare cost, morbidity, and mortality. Accurate PSI diagnosis remains challenging due to low culture positivity rates, frequent polymicrobial involvement, and non-specific diagnostic biomarkers. To better understand the comparative diagnostic yield of culture-independent technologies for PSI diagnosis, we prospectively characterised 26 clinically suspected PSIs and 10 control patients with non-infective pleural effusions using shotgun metagenomics, bacterial metataxonomics, quantitative PCR, and conventional culture. We demonstrate that culture-independent molecular techniques have superior sensitivity and negative predictive values for PSI diagnosis compared with conventional culture. Bacterial metataxonomics and metagenomics were both sensitive to the detection of polymicrobial infections, and unveiled microbial heterogeneity. Metagenomics also detected fungi and DNA viruses within pleural fluid. Dominant pathogens in the PSI cohort according to the highest resolution method, metagenomics, included streptococci (S. intermedius, S. pyogenes, S. mitis), Prevotella spp. (P. oris, P. pleuritidis), staphylococci (S. aureus, S. saprophyticus), and Klebsiella pneumoniae. Taken together, these results demonstrate the potential utility of molecular methods in the accurate diagnosis of PSI, which represents an important step towards improving personalised treatment decision-making and antimicrobial stewardship in suspected PSI.
Introduction
Pleural space infection (PSI) is a complex, heterogeneous, and challenging clinical entity associated with prolonged hospitalization, significant healthcare costs, morbidity and mortality 1-4. Frustrations in the clinical management of suspected PSI arise due to limitations in culture-based microbiologic diagnostics, which are notoriously insensitive (culture-positive rates of only 20-40% 5-8), resulting in prolonged and potentially inappropriate broad-spectrum antimicrobial use. Culture-based diagnostic sensitivity in PSI is poor due to multiple factors, including prompt institution of empiric antibiotics prior to pleural sampling, low pathogen concentration in pleural fluid, and difficulties in cultivating fastidious microorganisms 4,9. Limitations in culture-based diagnostics have obscured characterization of the complete pleural space ‘microbiome’ (i.e. the microbial communities present in clinical specimens), hampering clinical progress 10.
Molecular diagnostic approaches enable culture-independent microbial identification from complex niches such as clinical specimens 11. Such molecular techniques include shotgun metagenomics, metataxonomics, and quantitative PCR (qPCR). Shotgun metagenomics involves DNA sequencing of genomes and genes from all microbiota (excluding RNA viruses), followed by mapping to a reference microbial genome database, allowing strain-level detection of all bacteria, archaea, fungi, and DNA viruses in a sample, along with determining their putative functions (e.g. resistance and virulence genes). Metataxonomics involves high-throughput amplicon sequencing of the bacterial 16S ribosomal RNA (16S rRNA) or fungal internal transcriber spacer (ITS) regions to characterize entire bacterial or fungal microbiota, respectively, in a given sample; these methods typically resolve microbiota to the genus level. qPCR is a rapid and inexpensive same-day diagnostic method that specifically amplifies target taxa of interest, with resolution ranging from strain to domain level, the latter of which involves targeting the 16S or ITS regions to quantify only total bacterial or fungal load within a given sample 12,13.
Several recent studies applying culture-independent techniques have revealed improved diagnostic yield, and a previously untapped PSI microbiome, confirming that, in many cases, PSI is a polymicrobial disease with a diverse range of co-aggregating pathogens 7,8,14,15. Complex symbiotic interactions and cross-talk between and among microbes may therefore play an important role in PSI development, natural history, response to treatment, and prognosis 8.
To better understand the comparative diagnostic yield and clinical value of culture-independent technologies in PSI diagnosis, we performed a prospective observational single-center study of 26 consecutive patients presenting with suspected PSI at an Australian hospital, and 10 consecutive control patients lacking clinical suspicion of PSI. We then compared the diagnostic performance and resolution of microbe-enriched shotgun metagenomics, bacterial metataxonomics, panbacterial qPCR, and panfungal qPCR with conventional culture.
Methods
Ethics Statement
The study was approved by The Prince Charles Hospital (TPCH) Human Research Ethics Committee (HREC) HREC/2022/QPCH/81858 and the Sunshine Coast Hospital and Health Service Research Governance Office. All participants provided written informed consent. Patient IDs are coded and only re-identifiable by members of the research clinical team.
Study Participants
All patients ≥18 years of age presenting with suspected PSI (denoted with a ‘-P’ suffix) admitted between March and November 2022 to the Sunshine Coast University Hospital (SCUH), Queensland, Australia, were prospectively invited to participate in this study. Suspected PSI, as determined by a specialist respiratory physician, was based on a supportive clinical history (e.g., fever, pleuritic chest pain), imaging evidence of pleural infection (pleural fluid loculation/septation on ultrasound and/or computed tomography), and/or biochemical evidence of infection (elevated white cell with left shift and elevated C-reactive protein [CRP]). Individuals undergoing a diagnostic and/or therapeutic procedure for suspected PSI were invited to participate. Based on these criteria, 26 consecutive patients with ‘clinically suspected PSI’ (see: ‘Definition of Diagnostic Reference Standard’ below) were recruited to the study.
For the control group, patients with pleural effusion without clinical suspicion for PSI (e.g., malignancy or heart failure) at the time of pleural sampling (denoted with a ‘-C’ suffix) were enrolled. These individuals were approached if they had an alternative effusion cause, absence of clinical history to support active infection, and simple effusion appearance on ultrasound and/or computed tomography. Patients were excluded if they had prior sampling of the pleural space, any clinical concern for current or recent infection, concomitant abdominal ascites, hepatic hydrothorax, or antibiotic administration within 3 weeks of pleural sampling. Ten participants meeting these criteria were enrolled.
Specimen Collection, Storage, and Processing
For the PSI cohort, specimens were collected as soon as practicable from the time of admission to hospital. Antibiotic commencement ranged from 21 days prior to pleural sampling through to the time immediately post-pleural aspiration (Table S1). Between 10 and 60 mL pleural fluid was collected into two sterile specimen pots at the time of diagnostic pleural sampling using aseptic conditions. One pot was immediately mixed with 7mL RNA stabilization agent 16 for total DNA extraction (bacterial metataxonomics and qPCR); the second pot contained only pleural fluid for culture and host DNA depletion and extraction (metagenomics). Specimens were immediately stored at 4°C until processed. In parallel, specimens were also collected for routine microscopy (cell count, white blood cell differential, Gram stain), and biochemistry analyses (pleural fluid pH, protein, glucose, lactate dehydrogenase).
Microbial culturing was undertaken by both SCUH’s onsite Pathology laboratory and by our Research laboratory, with matrix-assisted laser desorption ionization time-of-flight mass spectrometry (bioMérieux, North Ryde, NSW, Australia) or Illumina NovaSeq 6000 whole-genome sequencing
(WGS; Australian Centre for Ecogenomics, St Lucia, QLD, Australia) used respectively to confirm species identity. 16S rRNA (i.e. bacterial) metataxonomic sequencing was carried out using V3-V4 amplification 17, followed by Illumina MiSeq 300bp paired-end sequencing (Ramaciotti Centre for Genomics, Sydney, NSW, Australia). Metagenomic sequencing on host-depleted DNA was performed using Illumina NovaSeq 6000 150bp paired-end reads (17-27 million reads/sample; Azenta Life Sciences, Suzhou, China). For qPCR analyses, total genomic DNA was tested against panbacterial (16S rRNA) 18, panfungal (28S rRNA) 19, and human (β-globin) 20 qPCR assays to quantify microbial-to-host load. For the panbacterial qPCR assay, a cycles-to-threshold (CT) cut-off value of 31 was used to delineate PSI-positive vs. PSI-negative cases. Further methodological details are described in Supplementary Material – ‘Methods’.
Definition of Diagnostic Reference Standard
Due to the lack of an optimal ‘gold-standard’ test for PSI confirmation, ‘clinical likelihood of PSI’ was objectively stratified using predefined clinical criteria:
Definite/Probable PSI: (i) clinical evidence of PSI as assessed by the recruiting physician based on fever, elevated white cell count, presence of pleurisy, raised inflammatory markers, and suitable clinical presentation; AND (ii) either direct evidence of PSI defined as the presence of Gram stain and/or culture positive on pleural fluid; OR indirect evidence of PSI based upon pleural fluid pH < 7.2 as measured on a blood gas analyser and/or glucose <3.0 mmol/L AND (iii) accompanied by at least two of the following criteria: serum CRP > 100 mg/L; complex pleural fluid on imaging; fever (> 38.0°C). 2. Possible PSI: (i) clinical evidence of PSI (as defined in (i) above), AND (ii) negative Gram stain and culture negative on pleural fluid AND pH ≥ 7.2 and glucose >3.0 mmol/L) AND (iii) accompanied by at least 2 of the following criteria: CRP > 100 mg/L; complex (i.e. loculated) pleural fluid collection on imaging; fever (≥ 38.0 °C). 3. Unlikely PSI: (i) No clinical evidence of PSI, AND (ii) all of the following criteria are met: pleural fluid Gram stain and culture negative with pH ≥ 7.2, glucose >3.0 mmol/L, serum CRP < 100 mg/L, uncomplicated effusion on CT and or ultrasonographic imaging, and an alternate diagnosis confirmed (i.e. heart failure, malignancy).
Statistical Analyses
To determine diagnostic test performance, ‘Definite/Probable PSI’ was used as the PSI-positive reference. Specificity, sensitivity, positive predictive value (PPV), and negative predictive value (NPV) were calculated for microbial culture and the three molecular methods (qPCR, bacterial metataxonomics, metagenomics). Additionally, Spearman’s Rho was used to determine PSI/non-PSI diagnostic correlations between the four testing methods.
Results
Study participants and clinical outcomes
Twenty-six patients (n=7 female; median age of 67 years [range 20 – 95]) with suspected PSI had an exudative pleural effusion as defined by Light’s criteria 21. Twenty-four (92%) had loculated (complex) pleural effusions on ultrasound and/or computed tomography imaging, with 20 patients (77%) also having associated lung consolidation at the time of cross-sectional imaging. Among the 10 control subjects, 9 were diagnosed with malignant pleural effusion (MPE), and one with decompensated congestive heart failure (Table 1).
Twenty-five patients (96%) with suspected PSI underwent intercostal catheter insertion, whereas one underwent diagnostic/therapeutic needle aspiration alone. All suspected PSI patients received broad-spectrum antimicrobial therapy (Table S1). Intrapleural enzyme therapy (5mg DNase and 10 mg tissue plasminogen activator) was administered in 17 patients (65%) with suspected PSI, and four patients (15%) required surgical decortication (Table 1). Thirty-day mortality within the suspected PSI cohort was 12% (n=3).
Test Performance across diagnostic methods
Conventional culture
Pathology lab testing returned 10 (38%) microbial culture-positive results from the PSI cohort (Table 2). Six (23%) were culture-positive from Research lab testing, with 100% overlap with the Pathology results.
qPCR
Positive bacterial detection was achieved with panbacterial qPCR in 14/26 (54%) suspect PSI cases (Table 2). All panbacterial qPCR-positive results were also PSI-positive by metagenomics and/or bacterial metataxonomics (Table 2). Panfungal qPCR did not identify fungi in any PSI or control specimens.
Bacterial metataxonomics and shotgun metagenomics
Together, bacterial metataxonomics and metagenomics identified 16/26 (62%) suspect PSI cases as PSI-positive, with 14/26 (54%) cases determined by each individual method. Metagenomics identified 7/14 (50%) cases as polymicrobial, whereas bacterial metataxonomics identified 5/14 (36%) cases as polymicrobial (Table 2 & Figure 1). Despite both methods identifying 14 PSI cases each, only 12 cases were identified as PSI by both methods, of which three were polymicrobial by both methods (Table 2). Far fewer taxa were identified by bacterial metataxonomics in two polymicrobial cases (SCHI0131-P, SCHI0172-P) (Table 2).
*Conventional culture positive; ^16S ribosomal RNA (panbacterial) quantitative PCR positive; #Whole-genome sequencing-confirmed (from culture). Dominant taxa are labelled on each bar.
Molecular diagnostics vs microbial culture
When compared against our ‘PSI-positive’ reference classification (Table 1), all molecular diagnostic methods demonstrated superior sensitivity and NPVs compared with microbial culture (Table 3). As expected, specificity and PPVs remained high across all diagnostic methods.
Using Spearman’s rank correlation (rs) to examine the relationship between PSI diagnostic methods (Figure 3), conventional culture was less strongly correlated with molecular methods (metagenomics: rs=0.56 (p<0.001); bacterial metataxonomics: rs=0.39 (p=0.05); panbacterial qPCR: rs=0.39 (p=0.05)). Metagenomics demonstrated the strongest correlation across methods (bacterial metataxonomics: rs=0.69 (p<0.001); panbacterial qPCR: rs=0.69 (p<0.001); conventional culture: rs=0.56 (p<0.01)). The strongest individual correlation was between panbacterial qPCR and bacterial metataxonomics (rs=0.85; p<0.001; Figure 3); this result was expected as both methods target the same gene.
PSI microbial aetiology
A diverse array of naso-oropharyngeal-, respiratory tract-, and gut-associated microbes were detected in the 14 metagenomic PSI-positive specimens (Table 2 & Figure 1). The most common dominant organisms were Streptococcus spp. (n=5; 2 x S. intermedius, 2 x S. pyogenes, 1 x S. mitis), followed by Prevotella spp. (n=2; P. oris, P. pleuritidis), Staphylococcus spp. (n=2; S. aureus, S. saprophyticus), and Klebsiella pneumoniae (n=2) (Table 2). Seven (50%) metagenomic PSI-confirmed cases were identified as monomicrobial, of which two harboured S. intermedius, along with one each of K. pneumoniae, S. saprophyticus, S. aureus, Enterococcus faecium, and Escherichia coli (Table 2 & Figure 2). In contrast, bacterial metataxonomics identified nine (35%) monomicrobial PSI cases, of which four were Streptococcus sp., vs. one each of Escherichia-Shigella sp., Fusobacterium sp., Haemophilus sp., Klebsiella sp., and Prevotella sp. (Table 2 & Figure 2). Five monomicrobial cases overlapped with the two meta-omics methods; however, bacterial metataxonomics erroneously assigned two monomicrobial PSI cases, SCHI0132-P and SCHI0143-P, as ‘Streptococcus’, despite the SCHI0132-P isolate being confirmed as coagulase-negative Staphylococcus sp. by MALDI-TOF testing (Table 2), indicating a probable bacterial metataxonomic database error for this genus.
*Conventional culture positive; ^16S ribosomal RNA (panbacterial) quantitative PCR positive; #Whole-genome sequencing result from culture; ‡Likely incorrect assignment of Staphylococcus as Streptococcus (1 case confirmed as a coagulase-negative Staphylococcus by MALDI-TOF analysis of culture). Dominant taxa are labelled on each bar.
As expected, bacterial metataxonomics failed to identify fungi or viruses in the PSI cases, whereas conventional culture (followed by WGS) and metagenomics identified P. kudriavzevii in SCHI0131-P. Metagenomics further identified Rhizoctonia solani, Wallemia mellicola, Meyerozyma guilliermondii, Truncatella angustata, Cokeromyces recurvatus, Bipolaris maydis, Candida orthopsilosis, and Gilbertella persicaria in SCHI0131-P (subject suffered from confirmed Boorhaaves syndrome), Candida dubliniensis in SCHI0140-P, and Candida tropicalis, Fonsecaea pedrosoi, and Pyricularia pennisetigena in SCHI0125-P. In all instances, fungal taxa were in relatively low abundance in the metagenomic data. Metagenomics also identified bacteriophages in 5/26 PSI-positive cases, with Staphylococcus phage Andhra seen in SCHI0123-P, SCHI0125-P, SCHI0127-P, and SCHI0128-P, and Streptococcus phage SM1 seen in SCHI0131-P. Staphylococcus phage Andhra was among the most dominant taxa seen in SCHI0125-P and SCHI0128-P, indicating very high phage levels in pleural fluid. Both of these specimens harboured S. aureus according to metagenomics, and SCHI0125-P also had S. saprophyticus according to culture and WGS (Table 2 & Figure 2).
Pleural fluid from two non-PSI control participants, SCHI0152-C and SCHI0175-C, unexpectedly cultured Candida parapsilosis and Moraxella osloensis, respectively, in the Research lab. These isolates were confirmed by WGS. Metagenomics also unexpectedly found microbes in 2/10 control cases (Table 2). The non-pathogenic single-stranded DNA virus, adeno-associated dependoparvovirus A, was identified in SCHI0173-C, and M. osloensis was confirmed in SCHI0175-C. The culture- and WGS-confirmed C. parapsilosis isolate retrieved from SCHI0152-C was not detected by metagenomics, suggesting possible operator contamination during specimen collection or processing. Bacterial metataxonomics failed to identify the culture-confirmed M. osloensis in SCHI0175-C, but identified Escherichia-Shigella in SCHI0152-C (1453 reads) and Corynebacterium in SCHI0173-C (190 reads), findings that were not supported by the other methods, again likely due to contamination.
Microbiology and clinical outcomes
Of the three PSI cases with a < 30-day mortality, two (SCHI0128-P and SCHI0134-P) were infected with K. pneumoniae, and two (SCHI0131-P and SCHI0128-P) had polymicrobial infections according to metagenomics. In contrast, no patient surviving beyond 30 days (n=23) harbored Klebsiella spp. according to any testing method.
Antimicrobial resistance gene detection
In silico antimicrobial resistance gene detection only identified innate drug resistance genes, with no evidence of acquired drug resistance development in any case (Table S1).
Discussion
Clinical decision-making in suspected PSI cases is challenging, predominantly due to low culture positivity rates 5-8, resulting in frequent diagnostic uncertainty and antimicrobial overuse. Patients with culture-positive PSI exhibit more severe disease and higher mortality 3; understanding PSI microbiome composition thus plays a critically important role in improving patient prognosis. Recent work 7,8,14,15 has highlighted the potential advantages of applying culture-independent molecular methods in PSI diagnostics. To this end, we performed a pragmatic, single-centre, prospective study of individuals with clinician-suspected PSI to better understand the comparative diagnostic yield and clinical value of culture-independent technologies. Confirming these prior studies, we demonstrate that all three culture-independent molecular techniques are superior to currently used conventional culture for PSI diagnosis and aetiologic confirmation. Furthermore, metagenomics and bacterial metataxonomics unveiled greater microbial heterogeneity than could be detected by the other methods.
The advent of metataxonomics and shotgun metagenomics has facilitated comprehensive microbiome characterisation in myriad human diseases. In line with recent PSI molecular studies 8,15, we confirm that PSIs can be caused by either single or multiple pathogens. The pleural microbiome in our cohort was frequently characterised by naso-oropharyngeal-colonising bacteria including streptococci (S. intermedius, S. pyogenes, S. mitis), staphylococci (S. aureus, S. saprophyticus), and the anaerobic Prevotella spp. (P. oris, P. pleuritidis), with occasional occurrence of lower respiratory or gut tract flora (E. coli, E. faecium, K. pneumoniae) (Table 2). Similar to recent work published from China 15, we observed that polymicrobial bacterial infection was also occasionally associated with fungal coinfection using metagenomics, confirming the extremely challenging nature of treating complex polymicrobial PSIs across microbial domains. Streptococcal and staphylococcal phages were also seen in five PSI cases, with their host species not always detected, suggesting possible lytic activity towards their host within the pleural space, or incidental phage transit from other body sites 22. Although not examined here, others have shown that phage presence in clinical specimens can negatively impact culture rates 23. Finally, we found that K. pneumoniae presence was associated with mortality, consistent with recent bacterial metataxonomic data showing an association between Enterobacteriaceae and PSI mortality 8.
Prior studies 7,8,14 of the pleural space microbiome have identified 100% PSI positivity using molecular testing. In contrast, only 62% PSI-positive cases were identified in our study. Our lower positivity rate likely reflects our study design, which, unlike most earlier studies 7,8,14, prospectively recruited consecutive patients with suspected PSI without prior knowledge of pleural fluid biochemistry or microbiology. This study design thus aimed to capture a more accurate representation of the diagnostic uncertainty facing clinicians in real-world practice. To our knowledge, only one other study has not used prior knowledge of pleural fluid biochemistry or microbiology for molecular PSI identification and characterisation; their PSI positivity rate was 70% 15. Lower ‘real world’ PSI-positive rates suggest that a subset of patients with clinically suspected PSI and/or para-pneumonic effusions have no pleural microbial biomass at time of sampling (38% in this study), even with highly sensitive deep sequencing methodologies. We suspect that many, or all, of these patients may truly have non-infective parapneumonic effusions – a conclusion further supported by their more benign pleural biochemistry using recognised cut-off values (i.e. pH < 7.2; glucose <3.0 mmol/L) 1. These collective data suggest that molecular testing could play an important role in clinical practice, particularly as an adjunct to current PSI diagnostics, thereby reducing empiric antimicrobial burden, duration, and spectrum in pleural effusion cases. In particular, we and others 24 have shown that panbacterial qPCR stands out as a cost-effective, simple, and rapid (same-day) diagnostic screening tool that could be readily implemented in clinical practice; its routine use on pleural fluid specimens would assist with both antimicrobial stewardship efforts and treatment decision-making in pleural effusion cases. In support of its clinical utility, panbacterial qPCR showed excellent diagnostic performance (sensitivity 82%, specificity 95%, PPV 93%, NPV 86%) and strong correlation with shotgun metagenomics, bacterial metataxonomics, and conventional culture (Figure 3).
1=perfect correlation; -1=no correlation. Conventional culture showed the weakest correlation when compared with each of the three molecular methods, an expected result due to known PSI underdiagnosis using this method 5-8.
Many patients are possibly over-treated for suspected PSI, usually with broad-spectrum antimicrobials, without definitive microbial confirmation or knowledge of the underlying infecting pathogen(s). This practice can result in treatment-related adverse events, and may contribute to AMR emergence. However, in this study, we did not identify acquired AMR genes amongst our cohort (Table S1), suggesting that a shift towards narrow-spectrum antimicrobials 8 may be achievable with the routine implementation of bacterial metataxonomics and metagenomics in PSI diagnostics. Such a shift would also contribute towards improved antimicrobial stewardship practices.
Finally, we also compared the performance of conventional culture with molecular tests in a negative control group of patients with pleural effusions from known non-infectious causes (predominantly MPEs). Although most controls were indeed found to have a sterile pleural space, one MPE case cultured Moraxella osloensis, which was confirmed by WGS and metagenomics. Furthermore, adeno-associated dependoparvovirus A (considered a non-pathogenic virus) and Corynebacterium was also seen in another control patient with MPE by metagenomics and bacterial metataxonomics, respectively. Although specimen or sampling contamination represents the most plausible explanation for these results, a bacterial metataxonomic study suggested that the pleural space of patients with non-infective pleural disease may not be sterile, instead potentially containing unique microbial signatures 25. Further work in this area is needed.
This study has several recognised limitations. Although the first study to systematically compare performance across these diagnostic methods, we acknowledge the limited sample size. Expanding this proof-of-principle work on a larger scale, and across multiple testing sites, will be important for understanding diagnostic performance and clinical utility of these tests among broader Australian and international cohorts. Secondly, immediate antimicrobial administration upon suspicion of PSI, which often occurs prior to pleural sampling, is required to prevent potential sepsis development; however, this practice may have negatively affected our culture and molecular PSI diagnostic rates. In particular, our implementation of human DNA depletion prior to metagenomic sequencing may have negatively impacted our ability to detect extracellular microbial DNA 26. One way to overcome this issue is to perform deeper metagenomic sequencing on total DNA; however, such an approach is more costly, and may increase the risk of missing low-abundance microbes due to the overwhelming human signal 26,27. Third, increasingly sensitive molecular assays come with the trade-off that microorganism presence might represent either a contaminant or a bona fide pathogen; such false-positive results can be mitigated by exercising strict aseptic practice during specimen collection and processing. Fourth, index hopping and ‘kitome’ contamination are known drawbacks of next-generation sequencing, especially in low-biomass specimens 27. However, overcoming these issues by subtracting background signal resulted in ‘real’ E. coli signal (as determined by WGS-confirmed culture) being inadvertently omitted in SCHI0125-P due to higher E. coli metagenomic/Shigella-Escherichia bacterial metataxonomic reads in the reagent-only controls. One way to reduce this issue might be to employ ultraclean, certified microbe-DNA free reagents 28, although using such reagents represents an added cost. Fifth, ‘incorporation bias’ 29 should be considered when interpreting the sensitivity/specificity of conventional culture results, given that in vitro culture data contribute to the PSI diagnostic ‘reference standard’, thereby inflating estimates of culture performance in this study design. Finally, we did not assess RNA viruses in this study using e.g. viral-specific qPCRs or metatranscriptomics; their contribution to PSI aetiology in our cohort is therefore unknown.
Conclusion
This real-world prospective study demonstrates the superiority of molecular methods (panbacterial qPCR, bacterial metataxonomics, and shotgun metagenomics) for both PSI diagnosis and aetiology. Based on our findings, we postulate that routine incorporation of molecular tests alongside conventional culture would result in more accurate PSI diagnoses. An important next step will be to determine, via randomised clinical trials, whether same-day panbacterial qPCR testing, followed by high-throughput sequencing methods where required (i.e., bacterial metataxonomics and/or metagenomics), can inform clinical decision-making, aid in antimicrobial stewardship, and improve patient outcomes in suspected PSI.
Data Availability
All sequence data generated in this study are available via NCBI BioProject PRJNA972883; genome assemblies are available via NCBI BioProject PRJNA970939.
Conflicts of Interest Statement
The authors declare no conflict of interest. The funders had no role in study design, in the collection, analyses, or interpretation of data, in manuscript writing, or in the decision to publish the results.
Funding
This work was funded by Wishlist Sunshine Coast Hospital Foundation (award 2021-04-CRG to PTB, TB, OO, JG, DSS, and EPP) and Advance Queensland (awards AQIRF0362018 to EPP and AQIRF095-2020-CV to OSO).
Acknowledgements
We thank Emma Seaniger and Louise McIntosh for assistance with sample collection and study coordination, Michael Caffery and the Pathology Queensland laboratory scientists for their assistance with culture, and the Australian BioCommons team (led by Ashley Dungan at University of Melbourne, and Melissa Burke at University of Queensland) for providing QIIME 2 training.
References
