Rapid ID and AST of PBC has shown to greatly impact patient care and improve clinical outcomes [24]. To our knowledge, this is the first study in Asia evaluating the ID performance of Pheno and BCID2 directly from PBC bottles and the first study to evaluate the AST performance of Pheno and direct Vitek AST using a saponin-based extraction method.
The identification performance for Pheno, BCID and BCID2 observed in this data set was comparable to previously published performance data for identifications by other rapid methods such as short term culture with MALDI [4], BCID [25,26,27,28,29], Bruker Sepsityper [30], and GenMark ePlex [31]. The expanded panel of BCID2 provides additional target groups for clinically significant causative agents of bacteraemia; in particular, B. fragilis and C. auris. B. fragilis accounts for 45% of all anaerobic blood culture infections [32] and are attributed to a high mortality rate up to 25% [33]. Sepsis caused by multidrug-resistant C. auris is reported to be associated with high morbidity and mortality rates ranging from 30 to 72% [34], therefore timely identification is crucial for appropriate treatment initiation and infection control on the spread of multidrug-resistant C. auris in clinical settings. Furthermore, the addition of mecA/C & MREJ in BCID2 allows the differentiation between MRSA and MR-CNS in polymicrobial infections where these organisms are co-detected.
Despite BCID2 being an advanced version of BCID with expanded target groups and resistant genes, we observed a few notable differences in 2 samples when tested on the BCID compared with the BCID2. One sample of a S. hominis was mecA-positive when tested on the BCID platform but failed to detect the mecA/C gene on the BCID2. Additionally, a polymicrobial specimen containing Proteus species and Enterococcus species with the BCID, whereas BCID2 detected Proteus species only and failed to detect the presence of Enterococcus species, due to the replacement of the Enterococcus species target with species-specific probes for E. faecalis and E. faecium only.
With respect to AST, overall performance was comparable across the three AST methods evaluated in this study. A higher rate of VME was observed with direct Vitek (5.3% compared with 3.1 and 3.8% of Pheno and colony Vitek, respectively). This was largely attributed to more VMEs observed for cefepime and ceftazidime with direct Vitek. Some publications suggested that this high rate of VMEs might be due to the inoculum effect as the inoculum size of Vitek is smaller than with BMD [35]. The data observed here for cefepime also highlight the importance of considering use of a reference method, such as BMD, when evaluating a new AST assay, although this poses a practical challenge for clinical laboratories. Some published data sets evaluating direct Vitek have used Vitek from colony as a comparator method, in which case certain shortcomings of the assay are never identified such as would be the case for cefepime here if BMD was not performed [36].
It is important to note that the Advanced Expert System (AES) corrections by Vitek2 were ruled out in this study for true MIC comparisons between different methods, however in routine laboratory setting, AES was also applied for results interpretation. For instance, it is well published that MICs for certain β-lactam antibiotics with some Enterobacterales which produce ESBLs are found to be susceptible in vitro however clinical failures have been observed in vivo [37]. The additional testing of Vitek2 system on the ESBL and cefoxitin screens allowed the correction of false susceptibility on cephalosporins and oxacillin, respectively. With such supplementary interpretation systems, one mE and 15 VMEs were resolved by AES for Enterobacteriaceae and one VME was resolved in Staphylococcus lugdunensis in direct. Similarly, for colony Vitek, eight VMEs were resolved by AES. Despite the fact that some of the VMEs could be rectified by applying AES rules in Vitek, cefepime accounted for most of the MIC discrepancies in both direct and colony Vitek while in Pheno cefepime only accounted for a low level of MIC discrepancies. MIC discrepancies with cefepime are of concern as the evolving science of pharmacokinetics-pharmacodynamics has become increasingly important in recent years in determining MIC breakpoints and have even adopted a dose-dependent breakpoint [19]. This emphasises the importance of an accurate MIC value.
The invalid rate of all tested isolates was mainly attributed to mucoid P. aeruginosa strains for all three methods and occasionally occurred in Enterobacterales (n = 2) and Enterococcus (n = 1) for Pheno and coagulase-negative Staphylococcus species (n = 1) for direct AST. It is suggested that testing on mucoid strains is a known limitation for automated AST systems [38]. Due to this limitation, supplementary testing on mucoid strains by disk diffusion is recommended.
The performance of Pheno against gram positive organisms was excellent, with no major or very major errors observed and only 3 minor errors with linezolid against Enterococcus spp. A low level of mEs were observed in the testing of Enterococcus to linezolid for both direct (n = 2) and colony AST (n = 3) and one VME was noted in S. lugdunensis to cefoxitin for direct AST, however it could be resolved by applying AES corrections.
The organisms represented in this data set include most of the clinically important resistant genotypes and phenotypes in Hong Kong [39]. In addition, colistin is a crucial last resort drug choice for multidrug resistant microorganisms, however recent study has shown that the asymptomatic faecal carriage of mcr-1-harbouring Enterobacteriaceae was 2.08% [40]. The ability of BCID and BCID2 in rapid detection of antimicrobials resistance genes, in particular BCID2 provided additional clinical significant resistant genes targets such as mecA/C and MREJ, bla
VIM and bla
IMP
bla
CTX-M,
bla
NDM and bla
OXA-48 and mcr-1 which allow clinicians to predict the antimicrobial resistant patterns and to execute appropriate and prompt treatment/ precautions. The overall concordance rate between resistant genes detected by BCID, BCID2 and phenotypic expressions detected by Pheno, direct and colony AST on most of the resistant genotypes/phenotypes was good. There were some isolates for which carbapenem resistance was not due to the presence of enzymes or the presence of CPE gene was not expressed phenotypically. This demonstrates the importance of both genotypic and phenotypic testing. While genotypic information is helpful to make predictions, phenotypic results may not always correlate according to the anticipated predictions.
Among the isolates with carbapenemase genes detected (n = 5), all were correctly identified and found to possess other ESBL genes (bla
CTX-M
n = 9, bla
SHV
n = 1, bla
TEM
n = 1); some of these isolates were also resistant to quinolones and aminoglycosides. CREs that were correctly identified with the corresponding resistant genes were also found to have resistant carbapenem phenotypes for all 3 phenotypic detection methods. However, the resistance of antimicrobials cannot solely depend on genotypic detection. In Hong Kong, a local study in 2016 examined the clonality and mechanism of resistance of CRE isolates. It was shown that only 10% were genotypic carbapenemase-producing Enterobacterales (CPE) while porin loss combined with AmpC and/or CTX-M type ESBL was the major mechanism of resistance of the CREs [39].
For CRE isolates lacking the targeted carbapenemase genes in this study, Pheno correctly identified non susceptibility to ertapenem (5/5). Noteworthy, the low resistance detection rate could be accounted for the low number of tested samples, further studies for CREs lacking carbapenemase genes would be recommended.
Conversely, detection of carbapenemase genes may not always confer an antimicrobial resistance phenotype, for isolates with carbapenemase genes without phenotypic expressions; both BCID and BCID2 correctly detected the presence of carbapenemase genes. Pheno correctly identified ertapenem as susceptible while for meropenem, all cases (n = 2) were failed to provide results, due to inconclusive assay results. For both direct and colony AST, the susceptibility of ertapenem and meropenem was correctly identified. Since genotypic and phenotypic expression of resistant genes do not always co-exist and correspond, therefore the adoption of both BCID/BCID2 (genotypic) in combination with direct AST (phenotypic) methods can allow clinicians to predict the effectiveness of antimicrobials in an accurate and timely manner. Furthermore, other clinical significant antimicrobial resistant microorganisms such as linezolid non susceptible Enterococcus species and multi drug-resistant A. baumannii, which resistant genes are not available on BCID and BCID2 detection panel, Pheno, direct AST and colony AST demonstrated satisfactory performance with low level of minor errors (Pheno n = 2, direct AST n = 1, colony AST n = 2).
Antimicrobial treatment choices for sepsis are often empirical and based on the susceptibility profile of the most common causative agents for sepsis. However, the causative agents profile may vary slightly depending on the institution and local epidemiology. The most commonly found causative agents for sepsis in 2019 at our institution were E. coli (32.9%), K. pneumoniae (6.9%), S. aureus (5.8%), E. faecalis (4.3%), P. aeruginosa (3.2%) and Coagulase-negative Staphylococcus (2.9%). The Accelerate PhenoTest® BC kit used in this evaluation includes 17 ID organisms covering the majority of these causative agents and 7 g-positive and 19 g-negative antimicrobials. In comparison, BCID provides additional 10 ID targets (totalling 27), which include H. influenzae, N. meningitides, L. monocytogenes, C. krusei, C. parapsilosis, C. tropicalis and species-level identification for organisms that are limited to genus level detection in Pheno, such as: K. oxytoca, K. pneumoniae, S. agalactiae, S. pneumoniae, and S. pyogenes, but only 3 resistance genes. BCID2 expands on BCID with an additional 5 ID and 7 resistance gene targets. The lack of species-differentiation of Pheno is not a major concern for most Enterobacterales because the breakpoints in CLSI M100 for Enterobacterales are classified in the same group. Nevertheless, for some organisms, identification to species level is of clinical significance, in particular Streptococcus species. It is critical to differentiate clinically significant species such as S. pneumoniae and S. pyogenes from those that are typically considered as contaminants, like Viridans-group Streptococcus species and other beta-hemolytic Streptococcus species. The rate for these groups of organisms at our institution in 2019 was only 0.4 and 0.7%, respectively.
On the other hand, Vitek panel provides a wider range of antimicrobial panel choice than Pheno for both gram negative and gram-positive microorganisms. Both Pheno and Vitek have novel antimicrobials such as ceftazidime-avibactam and ceftolozane-tazobactam on their panels, however, the newer Vitek card with these antimicrobials was not evaluated in this study. Studies have demonstrated that the usage of ceftazidime-avibactam and ceftolozane-tazobactam had significant improvement in activity against Enterobacterales with EBSLs and multi drug-resistant P. aeruginosa. Therefore, the availability of these novel antimicrobials on diagnostics devices could provide additional drug choice for clinicians when treating patients with these more complicated infections [41, 42].
An integrated ID and AST system such as Pheno allows for easy workflow and reduced time to result, however such integrated system might lead to another issue when interpreting both results in combination. There was 1 case in which E. gallinarum was incorrectly identified as E. faecium but was able to provide AST result. This misidentification could lead to inappropriate prescription of vancomycin since all E. gallinarum exhibit low level vancomycin resistance due to the presence of vanC gene. For this case, Pheno did demonstrate intermediate resistance for vancomycin but nonetheless, was still an incorrect ID.
On average, the time to result for PBC using conventional colony-based method is around 48–72 h due to the requirement of overnight incubation for colony ID and AST. Both (1) Pheno and (2) BCID/BCID2 and direct AST have significantly reduced the time to result compared to conventional colony based ID and AST: the TAT for Pheno was around 7 h (90 mins for ID and 7 h for AST) with 2 min hands-on time; whereas for BCID/BCID2 with direct AST, the total TAT was around 9 to 20 h (1 h for ID and 9–19 h for AST), with hands-on time of 15 min. Less hands-on time eased our laboratory technicians’ labour and streamlined workflow. Furthermore, faster TAT enabled our physicians to make faster and accurate clinical decisions [43]. Though cost effectiveness analysis is not performed in this study, it is believed that the improvement in TAT for ID/ AST would translate into better clinical outcome such as optimal use of antimicrobials, fewer complications, shorter hospital stay, etc. These would in return result in a more cost effective management for PBC cases [5]. Additional studies to look into the cost effectiveness in shortening the TAT for ID/ AST would be important for driving the change in practices.
Although empirical antibiotics are often given prior to the availability of ID and AST results, several studies have shown that incorrect use of antibiotics even for a short duration could associate with increased risk of acute kidney injury [44]. Therefore, rapid and accurate ID and AST results could reduce unnecessary exposure to drug toxicity and antimicrobial resistance.
In conclusion, both 1) Pheno and 2) BCID/BCID2 with direct AST methods achieved satisfactory ID and AST results. The direct Vitek AST method achieved good and comparable antimicrobial susceptibility performance with conventional colony Vitek AST. Although several studies have been published to evaluate the performance of positive blood direct extraction for ID however, the procedure and equipment required were more complicated and none were published on evaluating direct AST with the same sample preparation [13,14,15,16,17,18, 45,46,47,48,49,50,51,52,53,54]. With the direct Vitek method used in this study, along with the accurate ID and resistance genes included on the BCID and BCID2 panels, this integrated method can provide fairly rapid and accurate results. To implement proper methods, clinical laboratories should consider and evaluate different methodologies in order to select the best method to fit their routine workflow. Nevertheless, there are limitations of our study in that for some species, very few isolates were represented in this data set e.g. Enterobacter species (n = 1), Serratia marcescens (n = 1) and S. lugdunensis (n = 2). Therefore, further studies would be recommended for optimal determination of system performance on both ID and AST [43]. This study was also limited by the small numbers of fresh clinical samples. However, it has already demonstrated the accuracy and advantages of the various methods.
Although all these rapid methods demonstrated promising ID and AST results, the practice of gram-stain and sub-culturing of PBC would still be useful and important as these rapid methods have limitations, such as detecting and/or performing AST on polymicrobial samples, performing AST on mucoid strains such as P. aeruginosa, a known limitation for automated systems when performing colony AST [38]. Furthermore, some commercial platforms such as Verigene or ePlex require a prior gram stain result to facilitate the selection of the panel.
In laboratories where non daytime working hours were staffed by staff with limited training or not proficient in microbiology, the ease of lean workflow and result interpretations of Pheno are advantageous in such a setting with fewer technician interventions and techniques required, whereas despite the significant reduction in hands-on time for BCID/BCID2 with direct AST compared to conventional methods, the operation of direct AST still required highly trained microbiology staff and intervention time. Pheno can easily be set up around the clock whenever a PBC is available. Both workflows evaluated in this study provide different solutions for laboratories looking for more direct ID and AST methods but must be considered individually for each institution’s needs and practices.