Multinational evaluation of the BioFire® FilmArray® Pneumonia plus Panel as compared to standard of care testing

This study compared standard of care testing (SOC) to BioFire® FilmArray® Pneumonia plus Panel (PNplus). PNplus detects 15 bacteria with semiquantitative log bin values, 7 antibiotic resistance markers, three atypical bacteria (AB), and eight viral classes directly from bronchoalveolar lavage-like specimens (BLS) and sputum-like specimens (SLS). Fifty-two laboratories from 13 European countries and Israel tested 1234 BLS and 1242 SLS with PNplus and SOC. Detection rates and number of pathogens/samples were compared for PNplus pathogens. PNplus bin values and SOC quantities were compared. Three thousand two hundred sixty-two bacteria in PNplus were detected by PNplus and/or SOC. SOC detected 57.1% compared to 95.8% for PNplus (p ≤ 0.0001). PNplus semiquantitative bin values were less than SOC, equal to SOC, or greater than SOC in 5.1%, 25.4%, and 69.6% of results, respectively. PNplus bin values were on average ≥ 1 log than SOC values (58.5% 1–2 logs; 11.0% 3–4 logs). PNplus identified 98.2% of MRSA and SOC 55.6%. SOC detected 73/103 AB (70.9%) and 134/631 viruses (21.2%). PNplus detected 93/103 AB (90.3%) and 618/631 viruses (97.9%) (p ≤ 0.0001). PNplus and SOC mean number of pathogens/samples were 1.99 and 1.44, respectively. All gram-negative resistance markers were detected. PNplus and SOC results were fully or partially concordant for 49.1% and 26.4% of specimens, respectively. PNplus was highly sensitive and detected more potential pneumonia pathogens than SOC. Semiquantification may assist in understanding pathogen significance. As PNplus generates results in approximately 1 h, PNplus has potential to direct antimicrobial therapy in near real time and improve antimicrobial stewardship and patient outcomes. Supplementary Information The online version contains supplementary material available at 10.1007/s10096-021-04195-5.


Introduction
Determining the etiology of community-acquired pneumonia (CAP), hospital-associated pneumonia (HAP), and ventilatorassociated pneumonia (VAP) can be complicated [1]. Based on traditional test methods such as Gram stain and culture, diagnostic yield can be low [2][3][4]. Additionally, poor specimen quality may yield inconclusive or difficult to interpret results, sampling may require an invasive procedure, and/or patients may be on empiric therapy prior to specimen collection, reducing diagnostic yield [5]. Although international guidelines from the Infectious Diseases Society of America (IDSA)/American Thoracic Society (ATS) and the European Respiratory Society (ERS) recommend diagnostic testing for moderate to severe CAP [6,7], actual compliance with guidelines can be low (16.7% and 23.9%, respectively) and differs by geography and disease severity [8]. Aside from influenza A/B, often no viral testing is performed, despite evidence that other viruses are a significant cause of CAP in all age patients [9][10][11]. Consequently, broad use of empiric antibiotic treatment for undiagnosed viral infections has been associated with increased antibiotic resistance globally [12][13][14].
In HAP and VAP, empiric therapy often includes broadspectrum antibiotics for both gram-positive and gramnegative bacteria due to the risk of infection with multidrugresistant pathogens [15,16]. Identification of specific pathogen(s) is of increasing importance to allow prompt initiation of

Materials and methods
Clinical sites, specimens, and SOC testing Forty-eight academic medical center laboratories and four independent medical laboratories from 13 countries (Austria, Belgium, Denmark, Israel, Italy, France, Germany, Netherlands, Portugal, Spain, Sweden, Switzerland, and the UK) tested 2476 unique specimens (1234 BLS; 1242 SLS) from adult and pediatric patients suspected of pneumonia (Supplemental Table 7). Specimen selection was at discretion of study site. SOC was performed per institutional policies and healthcare provider prescription. SOC varied by site and all included bacterial culture and phenotypic susceptibility testing when indicated. Additional test methods were at the discretion of the laboratory and may have included additional cultures (fungal, viral, mycobacterial cultures) as needed, urinary antigen testing (Streptococcus pneumoniae, L. pneumophila), immunofluorescence tests, and nucleic acid amplification tests (NAATs) for selected bacteria, viruses, and antibiotic resistance markers. Methicillinsusceptible Staphylococcus aureus (MSSA) was differentiated from methicillin-resistant S. aureus (MRSA) using phenotypic methods, mecA NAAT, and/or PBP2a latex agglutination at the discretion of the laboratory. The investigators were instructed to perform the BioFire PNplus Panel in accordance with the manufacturer's instruction for use. Specimens were split and BioFire PNplus Panel testing was performed using either fresh specimens or from a frozen aliquot. Data was deidentified, no protected health information was provided, and participation was in accordance with local institutional ethical guidelines.
BioFire PNplus and SOC results were compared for the following:

Detection of BioFire PNplus Panel pathogens by BioFire
PNplus Panel and SOC: Results were evaluated for all specimens and by specimen types. SOC detected or not detected was determined in consideration of all test results reported. SOC C. pneumoniae, L. pneumophila, M. pneumoniae, and viral results were counted as not detected when a negative or no result was provided as patterns of testing varied extensively by institution. All SOC results were considered true positive. Results of BioFire PNplus Panel were considered true positive or true negative based on performance data established in US FDA clinical studies [20]. Mean
Including   [26]. Similarly, a VAP study using a research use only version of BioFire PNplus Panel demonstrated for bacteria an 89.0% PPA and 95.9% NPA with SOC [27]. BioFire PNplus Panel reflects high performance compared to SOC, yielding additional clinically actionable results that may be missed by SOC.
Prevalence of atypical pathogens in this specimen set was low (4.17%), with L. pneumophila the most frequently detected. Low SOC percent detections of M. pneumoniae (48.78%) and C. pneumoniae (57.14%) compared to BioFire PNplus Panel (90.24% and 100%, respectively) were mainly due to lack of testing and therefore a missed opportunity in CAP to either limit treatment to a macrolide or fluoroquinolone or stop treatment if tested negative. This missed opportunity for applying antimicrobial stewardship principles and streamlining therapy is of particular importance due to adverse effects of fluoroquinolones [28].  [30]. Virus detection in absence of a bacterial pathogen and in conjunction with clinical presentation, chest radiograph, and other diagnostic tests, such as a low procalcitonin, could support antimicrobial stewardship and discontinuation of antibiotics in the setting of CAP [31].
BioFire PNplus Panel identified more codetections (41.85%) compared to SOC (19.21%), which was mainly influenced by lack of SOC viral testing. Codetections were commonly identified by Webber [24]. All three studies found the majority of codetections contained 2 pathogens, but could rarely contain 5 to 6+ pathogens, similar to our results. In this study, BioFire PNplus Panel on average identified more potential pathogens (1.99) per specimen than SOC (1.44). Detection of multiple pathogens raises interpretation questions that need to be viewed in light of clinical parameters, pathogens detected, and abundances.
Use of BioFire PNplus Panel has led to concerns that identification of more bacteria than SOC, which may be colonizers, could lead to antibiotic overtreatment. Specimen types should be considered. SLS are prone to more oropharyngeal contamination compared to BLS. H. influenzae, S. pneumoniae, and M. catarrhalis can be normal flora, and hospitalized or ventilated patients may be colonized with gram-negative bacilli and S. aureus. Although laboratory and reproducible within ± 0.5 log 10 copies/mL and correlated with another quantitative molecular method [20]. Despite low concordance with quantitative culture, particularly when values were < 10 6 (3.1-38.9%), concordance improved to 90.9-100% when quantitative culture values were > 10 6 . There were few instances when BioFire PN Panel did not detect a bacterium or reported values lower than quantitative culture, which is similar to what we report in this study. Lee et al. also demonstrated an overestimation of quantification by BioFire PN Panel [24]. Buchan et al. demonstrated that PN values were frequently higher than culture values, resulting in semiquantitative agreement (within the same log 10 value) of 43.6% [32]. Gastli  Major limitations of this study were a lack of clinical information and comprehensive gram-negative phenotypic antibiotic susceptibility data, such as that described by Murphy et al. [20], needed to better understand the relevance of resistance marker results and impact on patient care. Discordant analyses were not performed due to the large number of participating sites and the number of specimens tested. Discordant analyses would have required some type of confirmatory testing for 150 culture isolates (BioFire PNplus Panel negative) and 1898 individual PCRs or direct sample sequencing for the confirmation of pathogens only detected by BioFire PNplus Panel (SOC negative). However, this study therefore does highlight the need for microbiologists and clinicians to address issues relating to discordant results and test interpretation. Additionally, detailed information regarding testing for viral or atypical bacterial pathogens for each sample was not available so no direct performance comparison could be made. Gram stains were not systematically reported and it is not known if a quality score was a testing requirement. However, considering the high sensitivity and specificity of BioFire PNplus Panel as demonstrated in the US FDA clinical studies [20], in combination with detection of key gram-  [34]. This data provides an early indication that proper use and interpretation of BioFire PNplus Panel could lead to targeted therapy not an increase in inappropriate antimicrobial usage. Prospective interventional studies in progress will provide data on interpretation of bin values, detection of resistance genes, and clinical impact of a rapid diagnosis. Finally, the BioFire PNplus bin comparison to SOC reporting was difficult to standardize as culture reporting can vary from technologist to technologist and laboratory to laboratory, which could lead to interpretive error. However, despite using 3 different interpretative criteria, results did not significantly differ. Strengths of the study include the large specimen size, the even distribution of specimen types, and geographical diversity of testing sites with differences in both ordering practices and results reporting.
The clinical laboratory plays a vital role in diagnosis of CAP, HAP, and VAP but faces numerous challenges due to testing complexity [1. 5, 8-10]. Often poor-quality sputum specimens are submitted and without quality rejection screening by Gram stain, culture results can be misleading or negative. HAP and VAP patients pose a different dilemma since these patients quickly become colonized with S. aureus and various gram-negative bacilli which may lead to pneumonia with multidrug-resistant strains [15,16]. Specimens can contain a diversity of pathogens including bacteria, viruses, and fungi [4,[6][7][8][9][10][34][35][36][37][38]. Time to traditional bacterial detection is 24-72 h and antibiotic susceptibility data takes an additional 24-48 h. Comprehensive viral diagnostics are often not performed aside from influenza A/B testing, may not be performed 24/7, or require referral to a reference laboratory delaying time to results. Urinary antigen tests for S. pneumoniae provide results in < 30 min but can be false negative and false positive, particularly in children [39]. Urinary antigen tests for L. pneumophila are restricted in serotype detection and can have sensitivities of < 50% [40,41]. Serology can be difficult to interpret and may require an acute and convalescent serum collected weeks apart. NAATs are the gold standard for the detection of the atypical bacteria but may not be routinely performed. Consequently, testing imitations lead to empiric HAP/VAP treatment with broad-spectrum antibiotics, especially in regions with high antimicrobial resistance rates. The switch to targeted therapy can take days, increasing risk of antimicrobial resistance and adverse events such as acute kidney injury and C. difficile disease [18]. Finally, if specimens are obtained after the start of antibiotic treatment, results may be altered or negative, without identifying the etiologic agent [5].
In conclusion, the BioFire PNplus Panel meets the challenges associated with routine test methods including poor pathogen recovery, lack of diagnostic comprehensiveness, and delayed time to result [42]. However, several factors need to be considered including the lack of a specimen quality marker and the inability to report the presence or absence of normal flora [43]. Although a Gram stain is not required prior to testing, good laboratory practice should still be followed to insure sample quality [43]. Pretreatment or dilution of samples would affect both the sensitivity of the assay and semiquantitative results and therefore is not recommended in the manufacturer's instructions for use. Interpretation challenges include understanding the increased detection rates, significance of the bin value, the differentiation between colonization and infection, and the presence of gram-negative resistance markers without direct linkage to a specific pathogen. However, approved for use with BLS and SLS, BioFire PNplus allows for easy specimen testing for CAP [6,7] and meets IDSA/ATS recommendations [13] for non-invasive diagnostic testing as a preferred method for VAP and ERS guidelines to test distal quantitative specimens [16]. Additionally, studies that used BioFire PNplus panels in COVID-19 patients demonstrated not only improved diagnosis of bacterial coinfections but enhanced options for appropriate therapy [35,44,45]. Verroken et al. demonstrated that BioFire PNplus Panel speeded up therapeutic changes in 46.9% of COVID-19 patients, five patients having antibiotics stopped and one third remained antibiotic free [35]. BioFire PNplus is rapid, simple to perform, and highly robust, with only 0.53% of the specimens in this study yielding invalid results. Detection of pathogens and antibiotic resistance markers can be used to inform immediate treatment decisions and improve patient outcomes.
Funding BioFire Pneumonia plus Panels were provided free of charge by bioMérieux, France. No other compensation was provided to the test sites.
Data availability The datasets generated during and/or analyzed during the current study are not publicly available due to restrictions by individual contributors but are available as composite data from the corresponding author on reasonable request.

Declarations
Competing interests Christine C. Ginocchio, Carolina Garcia-Mondragón, and Barbara Mauerhofer are employees of bioMérieux. Cory Rindlisbacher is an employee of BioFire Diagnostics. EME Evaluation Program Collaborative sites did not receive any other resources to perform this study other than the test panels. The study design, authors, data analysis, interpretation of data, and writing of report were performed by bioMérieux employees as per author contribution listed above. Data was reviewed and approved by the EME Evaluation Program Collaborative sites.
Ethics approval Ethical approval was in accordance with individual institution requirements.
Consent to participate Not applicable.

Consent for publication All authors and contributors have read and approved the manuscript.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.