A Piece of the Puzzle: The Role of Molecular Testing in Antimicrobial Stewardship

INTRODUCTION

Antimicrobial resistance is a major global public health threat [1], and growing evidence suggests that antimicrobial-resistant infections are becoming increasingly common in children [2]. Early selection of appropriate antimicrobial therapy can improve clinical outcomes and decrease exposure to unnecessary antimicrobials [3]. For this reason, the 2016 Infectious Diseases Society of America guidelines for implementing an antimicrobial stewardship program (ASP) recommend the use of rapid diagnostic testing to supplement conventional culture (weak recommendation, moderate-quality evidence) [4]. Molecular testing can more rapidly identify pathogens and provide antibiotic susceptibility compared to culture. In this article we will provide an overview of the current molecular testing landscape, review the pros and cons of different molecular techniques, and discuss the role of ASPs in implementing new diagnostic techniques in clinical care.

There are several different types of molecular tests for identification of infectious diseases. Molecular tests are categorized by what specimens can be used for testing and whether they test for a broad or limited number of pathogens. Some molecular tests are performed directly from specimens such as whole blood before bacterial growth is detected and others are only performed after the blood culture becomes positive [5]. For the purposes of this article, we will focus on identification from blood specimens, though many techniques may also be used on other sterile or nonsterile specimens. Molecular tests for identifying bacterial and fungal organisms include mass spectrometry, nucleic acid amplification testing (NAAT), T2 magnetic resonance (T2MR), and sequencing. In this article, several specific commercial assays will be discussed, which does not constitute an endorsement of any particular company or assay. Each of these techniques has advantages and disadvantages with respect to speed, cost, sensitivity, and specificity (Table 1). An overview of their integration into diagnostic microbiology workflows is presented in Figure 1, with techniques classified by the current extent of their adoption.

Table 1.

Available Molecular Diagnostic Methods Using Blood Specimens

Method	Output	Requires Growth in Bottle?	Requires Growth on Plates?	Breadth of Coverage	Advantages	Disadvantages
MALDI-TOF MS	Species identification	Yes	Yes	Very high	Faster and more reliable than biochemical identification	Requires growth on plates; no drug susceptibility information
Targeted RT-PCR	Present/absent	Yes (for blood)a	No	Low—specific targets only	High sensitivity; low cost	Limited to single or a few targets
Rapid multiplex PCR	Selected species and resistance mechanisms	Yes (for blood)a	No	Moderate	Skips growth on plates; valuable resistance information Can detect polymicrobial infections	Relatively limited panels for both identification and drug susceptibility
T2MR	Identification of 5 Candida or 5 bacterial species	No	No	Moderate (within Candida); limited (bacteria)	Skips culture step; very fast	Requires separate order and tubes; currently limited coverage; no antimicrobial susceptibility information
Targeted next-generation sequencing	Identification of bacterial, fungal, or parasitic organisms	No	No	High	Increased sensitivity compared to untargeted sequencing	Less broad than untargeted metagenomic sequencing; may be less useful in polymicrobial infections; does not provide susceptibility or epidemiologic data
Untargeted metagenomic sequencing	Species and genotype information	No	No	Very high	Broadest testing; can identify rare pathogens; can provide susceptibility data; can be used for polymicrobial infections; epidemiologic analysis	Expensive; potential for reduced sensitivity; potential for false positives

Abbreviations: MALDI-TOF MS, matrix-assisted laser desorption ionization-time of flight mass spectrometry; T2MR, T2 magnetic resonance.

^aSome fungal multiplex PCRs (such as Aspergillus) do not require growth in culture bottles.

Figure 1.

A schematic overview of microbiologic analysis of blood samples, including conventional and molecular techniques. Abbreviations: MALDI-TOF MS, matrix-assisted laser desorption ionization-time of flight mass spectrometry; NAAT, nucleic acid amplification testing; T2MR, T2 magnetic resonance.

MATRIX-ASSISTED LASER DESORPTION IONIZATION-TIME OF FLIGHT MASS SPECTROMETRY

While no longer an emerging technology, matrix-assisted laser desorption ionization-time of flight mass spectrometry (MALDI-TOF MS) has significantly reduced time to identification of bacterial and fungal pathogens compared to conventional methods [6]. MALDI-TOF MS analysis produces a spectrometric signature that is compared to a comprehensive and growing library of organisms; identification failures are rare [7]. MALDI-TOF MS plus antimicrobial stewardship (AS) may significantly reduce the time to optimal therapy [8]. This method does require growth on culture plates prior to analysis and is thus significantly slower than NAAT (see Figure 1). Consider a patient with Gram-positive cocci growing in blood cultures and receiving empiric treatment with vancomycin and ceftriaxone. If Enterococcus faecalis is identified, local resistance patterns may permit targeting the regimen to ampicillin; Enterococcus faecium may prompt concern for vancomycin resistance and broadening to more effective therapy such as linezolid [9]. Such clinical decision-making must be tailored to local antimicrobial resistance patterns. When combined with AS, MALDI-TOF MS may generate significant cost savings and improve patient outcomes compared to conventional laboratory methods alone [10]. In the Enterococcus bacteremia example, clinical impact depends on expert interpretation of the results, which can take the form of real-time expert consultation or development of management protocols.

NUCLEIC ACID AMPLIFICATION TESTING

NAAT identifies pathogens by amplification and detection of genetic material. Commonly used platforms applied for rapid diagnosis in the United States include Verigene (Luminex, Austin, TX), ePlex BCID (GenMark Diagnostics, Carlsbad, CA), and Biofire (bioMerieux, Marcy-l’Etoile, France); these assays use microarray [11] or rapid multiplex PCR [12] for identification. Compared to MALDI-TOF MS, NAAT identifies a more limited set of organisms. MALDI-TOF MS asks, “What is this organism?” while NAAT asks, “Are any of the organisms on this list present?” Assays can be narrowly targeted, such as mecA PCR that rapidly identifies methicillin resistance, or multiplex, such as the panels listed above. NAAT offers 2 major advantages over MALDI-TOF MS: (1) identification directly from blood culture bottle at the time the culture is positive, without the need to subculture the organism and (2) identification of antimicrobial resistance determinants (Figure 1). Identification directly from blood culture bottles results in significantly reduced time to identification compared to MALDI-TOF MS and especially biochemical methods, with results typically available within 3 hours or fewer from growth. By identifying genes associated with antimicrobial resistance, such as mecA conferring methicillin resistance in staphylococci or carbapenemases in Gram-negative bacilli, antimicrobial therapy can be optimized much more rapidly [13]. Clinicians must understand that accuracy of genotypic prediction of antimicrobial resistance varies widely by organism and antimicrobial class. For example, detection of vancomycin resistance via vanA/B in enterococci has a high negative predictive value approaching 100% [14, 15]; similar performance is seen with mecA identification to detect methicillin resistance in staphylococci [16, 17]. In Gram-negative bacilli on the other hand, due to the genetic diversity of beta-lactamases and non-beta-lactamase mechanisms of resistance, failure to detect a known resistance mechanism does not necessarily predict antimicrobial susceptibility. For example, carbapenemases are responsible for only about 50% of carbapenem resistance, so carbapenem susceptibility should not be assumed if one of these determinants is not detected [18]. Nonetheless, rapid genus- or species-level identification may also permit antibiotic optimization while full susceptibility results are pending, such as identification of Pseudomonas aeruginosa in a patient not already receiving antipseudomonal coverage.

As with MALDI-TOF MS, the clinical impact of NAAT for pathogen identification depends on expert input. Clinicians without expertise in microbiology or infectious diseases, accustomed to responding to Gram-stain results and susceptibility reports, require guidance. In one randomized clinical trial, time to de-escalation to the most appropriate antibiotic therapy was significantly shorter in the group receiving AS guidance to interpret NAAT results, while there was no difference between the group receiving no NAAT results and the group receiving NAAT results but no ASP guidance [13].

NEXT-GENERATION SEQUENCING

Next-generation sequencing (NGS) allows for identification of a broad range of pathogens based on unique RNA or DNA sequences [19]. These tools, used for decades in research environments, are more recently being applied for clinical care and epidemiologic investigations of healthcare-associated infections. These tests can be run in commercial or institutional labs under CLIA regulation. Laboratories are required to validate the accuracy, sensitivity, specificity, and precision of these tests as laboratory-developed tests. NGS allows for high-throughput genomic analysis using repeated cycles of DNA synthesis in parallel in order to sequence millions to billions of nucleic acid fragments [20]. This approach is most useful in situations where biochemical reactions or culture are unsuccessful at identifying a specific pathogen.

NGS can be used for either targeted or untargeted “shotgun” metagenomic sequencing [19, 21]. Targeted sequencing utilizes primers that target amplification of conserved regions of bacteria, fungi, or parasites, such as 16S ribosomal RNA in bacteria. Untargeted metagenomic sequencing or shotgun sequencing is performed after extracting and fragmenting all of the DNA and RNA in a sample [21]. The fragments are then amplified and sequenced. One advantage of untargeted metagenomic sequencing is that it can detect multiple pathogens in a polymicrobial infection. However, if there is no enrichment step performed, the sensitivity of untargeted sequencing may be compromised by interference of human DNA. Some untargeted sequencing methods employ the use of a targeted probe enrichment step to increase the sensitivity for specific pathogens.

Microbial cell-free DNA (cfDNA) NGS has recently become more commercially available for the diagnosis of infectious diseases; one available platform is the Karius test (Karius Inc., Redwood City, CA) [22]. This technology uses an untargeted metagenomic sequencing strategy to identify microorganisms directly from blood. The sensitivity of this technology in one study of septic patients was determined to be 92.9% [22]. In patients who were pretreated with antibiotics in this study, cfDNA sequencing identified pathogens in 47.9% of cases, while traditional blood culture identified pathogens in only 19.6% of cases. In the same study, 22.8% of asymptomatic control subjects had cfDNA detected from a single organism at low concentrations [22]. This high false-positive rate demands caution in interpreting results. In our experience, this technology is most useful and cost-effective in immunocompromised patients with focal signs of infection and negative cultures. In a recent study of adult patients with leukemia and fever and neutropenia, the Karius test identified a possible pathogen in 41 of 55 patients (75%), compared to 10 of 55 (18%) by blood culture. The clinical significance of some of those identifications is uncertain. The Karius test identified multiple pathogens in 12 cases of intra-abdominal infections including enterocolitis and typhlitis. There were 4 identified fungal infections, including Aspergillus, Rhizopus, and Pneumocystis, which suggests significant potential benefit in this population [23]. Additionally, when there is a broad differential diagnosis and potential for polymicrobial infection, this technology may be beneficial.

The turnaround time for these tests is typically longer than for NAAT. Performance of these metagenomic methods may vary and infectious diseases physicians should be alert to the limitations of these tests. These tests are significantly more expensive than traditional culture or PCR and may not be reimbursed by insurance companies. Since these are new technologies, thresholds and standards are still being developed, and detection of contaminants and colonizers will likely present a significant obstacle to widespread adoption [19].

T2 MAGNETIC RESONANCE

T2MR technology detects pathogen directly from whole blood, using PCR amplification of pathogen DNA followed by hybridization of specific probes to the amplified DNA; the resulting nanoparticles are detected via magnetic resonance [24]. The T2Candida Panel (T2 Biosystems, Lexington, MA) applies this approach to detection of Candida spp., specifically C albicans, C parapsilosis, C tropicalis, C krusei, and C glabrata. Because this assay requires no culture step, mean time to detection is approximately 4 hours, compared to 42 hours for blood culture [25, 26]. Compared to blood culture as the gold standard, estimates for sensitivity and specificity are 91% and 94% [27]. Descriptions of real-world use remain limited, but one study found shorter time to both antifungal addition and discontinuation, without significant differences in clinical outcomes [28]. A recently FDA-approved additional application includes bacterial identification, with a panel including E faecium, Staphylococcus aureus, Klebsiella pneumoniae, P aeruginosa, and Escherichia coli. As with Candida, in clinical trials, time to identification was significantly reduced to a few hours [29]. However, only 48% of participants with positive blood cultures had a positive T2MR assay.

T2MR technology requires a new instrument and dedicated samples sent in addition to blood cultures. The required volume is 4 mL, which is likely to hamper implementation in infants and young children. Current FDA approvals are limited to the panels described above, including 5 Candida species and 5 bacterial species. Hence, these assays are unable to definitively rule out invasive infections, though they can rapidly rule them in. At this time, there is no antimicrobial susceptibility data provided by T2MR.

INTEGRATION OF MOLECULAR DIAGNOSTICS AND Antimicrobial Stewardship

We are living in an era of extraordinary innovation and progress in diagnostic microbiology. As these new diagnostic technologies enter clinical practice, it is critical that clinical microbiologists and experts in Infectious Diseases and Antimicrobial Stewardship collaborate to evaluate them and decide whether and when to adopt them. Key factors to consider include cost, labor, accuracy, speed, and anticipated real-world impact on clinical care.

1. Cost: Compared to classical microbiologic methods, molecular assays are usually more expensive, especially when considering capital investment costs for new devices. When, however, molecular assays can replace labor-intensive culture-based methods, they may actually be cost-saving. Further, with regards to patient care costs, inpatient stay and antibiotic exposure may be reduced as targeted antimicrobial therapy is initiated more quickly, especially when molecular diagnostics are combined with AS [10].

2. Labor: While most molecular assays are designed to require minimal technologist time, there is added effort. Additional training is generally required, though the amount of training varies by molecular assay. Some assays, like Biofire, are relatively automatic, whereas others, such as NGS, currently require more sophisticated skill level. To reap the benefits of rapid assays, 24-hour availability may be required.

3. Speed: Some tests, such as MALDI-TOF and NAAT from blood culture bottles, provide the same information as conventional methods but faster. Others, such as NGS, may have a longer turnaround time and limited availability, but can be useful when there is diagnostic uncertainty despite use of traditional microbiology methods.

4. Accuracy: Assays that replace or accelerate results should have equal or better sensitivity and specificity compared to conventional methods. Special care should be given to the likelihood of an inaccurate result that would adversely affect patient care.

5. Impact on patient care: Information alone will not have a significant impact, especially when provided by new tests with which clinicians are unfamiliar. Education and real-time expert interpretation are required.

Like new procedures, devices, and medications, safety and efficacy of new diagnostic microbiological methods must be proven prior to widespread adoption. Education about the positive and negative predictive value of these tests, indications for use, and interpretation of results will be necessary to ensure that they are used cost-effectively and appropriately for patient care. If providers receiving the results do not understand them, the advantage of greater speed or accuracy will be lost. Consider a patient admitted with fever and a central line, receiving empiric vancomycin plus cefepime with blood cultures pending. At 4 PM, the team receives notification that the blood culture is growing Gram-positive cocci in clusters; appropriately, no change is made. At 7 PM, the cross-covering team is notified that the isolate has been identified as S aureus with no methicillin resistance detected. Unsure of the reliability of the result, no change is made until full susceptibilities are available the next afternoon, exposing the patient to a full extra day of 2 broad-spectrum antibiotics. For rapid diagnostic methods, successful implementation will likely require just-in-time interpretation of results and recommended courses of action. ASPs, with their expertise in infectious diseases and healthcare implementation, are ideally positioned to fill this role through education, live support, and management protocols. Before implementing reporting of rapid diagnostic results, infectious diseases clinicians and microbiologists should educate providers on appropriate interpretation. Using local antimicrobial resistance patterns, ASPs should design protocols for clinicians to act on these results immediately; direction to these protocols can be appended to the microbiology report. An example of such a protocol is provided in Table 2. Notification of rapid molecular diagnostic results should ideally go to the responsible ordering provider rather than the patient’s nurse, as is often the case with positive blood culture results. At one author’s institution, overnight blood culture NAAT results are reported by page to the covering pharmacist, who has sufficient expertise to advise the responsible ordering provider according to protocols. Ideally, expert consultation, in the form of ASP or Infectious Diseases clinicians, should also be available around the clock if specific questions arise.

Table 2.

An Example of an Antibiotic Management Protocol for Use With the Verigene Gram-Positive Blood Culture (BC-GP) Rapid Molecular Assay

BC-GP Rapid Result		Susceptibility Result	First-Line Therapy	Notes
Streptococci	Streptococcus pyogenes (Group A Strep)	NA	Penicillin G +/− clindamycin
	Streptococcus agalactiae (Group B Strep) or S anginosus	NA	Penicillin G
	Streptococcus spp. (includes viridans group and S pneumoniae)	NA	Ceftriaxone +/− vancomycin	Use vancomycin for severe disease or concern for meningitis
Enterococci	Enterococcus faecalis	NO vancomycin resistance	Ampicillin
		Vancomycin-resistant	Linezolid or high-dose daptomycin
	Enterococcus faecium	NO vancomycin resistance	Vancomycin	Linezolid preferred over daptomycin in children
		Vancomycin-resistant	Linezolid or high-dose daptomycin
Staphylococci	Staphylococcus aureus	NO methicillin resistance	Nafcillin or cefazolin
		Methicillin-resistant	Vancomycin
	Staphylococcus lugdunensis	NA	Vancomycin
	Staphylococcus epidermidis OR Staphylococcus sp. Staphylococcus aureus not detected (ie, coagulase-negative staph)	NA	Vancomycin OR discontinue antibiotics if likely contaminant
Indeterminate	Rapid testing unable to identify the organism. No recommendation can be given.

This Table is adapted from guidance developed at one author’s institution. Individual hospitals should develop their own management guidance based on local susceptibility patterns.

While MALDI-TOF MS and NAAT are often implemented into standard clinical microbiology workflow, NGS and T2MR testing must be ordered separately. Consequently, clinicians will need expert guidance on appropriate scenarios for use of these tests. In hospitals that adopt T2MR, ASPs should develop standard ordering criteria and results interpretation support. Decisions to order NGS are more measured and case-by-case; at this time, Infectious Diseases consultation or approval is likely more appropriate. As new techniques increase the sensitivity to detect and identify bacterial and fungal species, new questions will arise about specificity and risk of contamination. In the study noted above, NGS for cfDNA identified possible pathogens in over 20% of asymptomatic individuals, an order of magnitude higher than the incidence of contaminated blood cultures. Unexpected results will likely create dilemmas requiring careful interpretation, and clinicians will need to temper enthusiasm for newfound diagnostic capacity with reasonable skepticism. Cost of these technologies should also be taken into consideration, as they may have low yield in some settings. Infectious diseases and clinical microbiological experts will increasingly need to stay abreast of the benefits and limitations of these technologies. It is also important for infectious diseases physicians and clinical microbiologic experts to consider when restrictions on ordering these novel diagnostic tools should be put in place to avoid overuse.

Notes

Acknowledgement. The authors would like to thank Shangxin Yang for assistance with the preparation of this manuscript.

Potential conflicts of interest. Both authors: No potential conflicts of interest. Both authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.