Culture-independent identification of bloodstream infections from whole blood: prospective evaluation in specimens of known infection status

ABSTRACT

Sepsis caused by bloodstream infection (BSI) is a major healthcare burden and a leading cause of morbidity and mortality worldwide. Timely diagnosis is critical to optimize clinical outcome, as mortality rates rise every hour treatment is delayed. Blood culture remains the “gold standard” for diagnosis but is limited by its long turnaround time (1–7 days depending on the organism) and its potential to provide false-negative results due to interference by antimicrobial therapy or the presence of mixed (i.e., polymicrobial) infections. In this paper, we evaluated the performance of resistance and pathogen ID/BSI, a direct-from-specimen molecular assay. To reduce the false-positivity rate common with molecular methods, this assay isolates and detects genomic material only from viable microorganisms in the blood by incorporating a novel precursor step to selectively lyse host and non-viable microbial cells and remove cell-free genomic material prior to lysis and analysis of microbial cells. Here, we demonstrate that the assay is free of interference from host immune cells and common antimicrobial agents at elevated concentrations. We also demonstrate the accuracy of this technology in a prospective cohort pilot study of individuals with known sepsis/BSI status, including samples from both positive and negative individuals.

IMPORTANCE

Blood culture remains the “gold standard” for the diagnosis of sepsis/bloodstream infection (BSI) but has many limitations which may lead to a delay in appropriate and accurate treatment in patients. Molecular diagnostic methods have the potential for markedly improving the management of such patients through faster turnaround times and increased accuracy. But molecular diagnostic methods have not been widely adopted for the identification of BSIs. By incorporating a precursor step of selective lysis of host and non-viable microorganisms, our resistance and pathogen ID (RaPID)/BSI molecular assay addresses many limitations of blood culture and other molecular assay. The RaPID/BSI assay has an approximate turnaround time of 4 hours, thereby significantly reducing the time to appropriate and accurate diagnosis of causative microorganisms in such patients. The short turnaround time also allows for close to real-time tracking of pathogenic clearance of microorganisms from the blood of these patients or if a change of antimicrobial regimen is required. Thus, the RaPID/BSI molecular assay helps with optimization of antimicrobial stewardship; prompt and accurate diagnosis of sepsis/BSI could help target timely treatment and reduce mortality and morbidity in such patients.

INTRODUCTION

Sepsis caused by bloodstream infection (BSI) is a major healthcare burden and a leading cause of morbidity and mortality worldwide. In the US alone, sepsis accounts for over 850,000 emergency department visits and 250,000 deaths annually, and suspected sepsis is a documented reason for up to 6% of all patient hospitalizations (1–3). A key driver of sepsis-associated morbidity and mortality is delay in the application of appropriate antimicrobial treatment, as the prognosis for such patients deteriorates hourly (2, 4). For this reason, best practices for treating patients with suspected sepsis/BSIs include, among other interventions, timely collection of blood cultures for diagnostic purposes and empiric, broad-spectrum antimicrobial therapy (5–7). Once a causative pathogen is identified and characterized, antimicrobial treatment can be transitioned to a more targeted, narrower-spectrum pharmacologic agent.

Such an approach, however, has numerous limitations. Blood cultures are time-consuming and require skilled personnel and a well-equipped laboratory. Turnaround time for a positive test result typically ranges from 16 to 48 hours, although 5–7 days of incubation may be required before a negative test can be reported (8). Blood cultures are prone to contamination (either at the time of collection or in the laboratory), may not detect infections caused by microorganisms which are fastidious, anaerobic, or otherwise challenging to culture, and may be confounded by mixed infections (9). The culturing process may also be inhibited by medications and fail to detect BSIs in patients who recently received antimicrobial therapy, regardless of the spectrum of coverage (10–13). While advances in blood culture media with antimicrobial absorbing additives may have mitigated this risk to some degree, it remains a significant concern (12, 14). Indeed, the recent FABLED trial found that the sensitivity of diagnostic blood cultures was markedly reduced, by approximately 50%, if the blood specimens were collected after antibiotic administration (11).

For these reasons, there is significant enthusiasm for developing a complementary and/or alterative, orthogonal approach to detecting microbial pathogens in blood using a culture-independent process (15–19). Such an approach could identify BSIs more quickly and reliably while overcoming some of the challenges related to contamination, mixed infections, and antimicrobial interference.

Molecular approaches that target the genomic material of microorganisms represent a particularly promising method of detecting BSI directly from whole blood in a culture-independent manner but face a number of technical obstacles. For example, sepsis/BSIs can be caused by a very low concentration of microorganisms in the circulation, often less than 100 colony-forming units (CFUs)/mL (20–22). While blood culture overcomes this challenge by providing time and resources for microorganisms to proliferate in vitro, molecular approaches employ an enzymatic amplification step [such as polymerase chain reaction (PCR)] to markedly increase the quantity of genomic material derived from the infecting microorganism.

However, the process of enzymatically amplifying the microbial genomic material derived from blood-borne microorganisms has its own set of technical challenges. One concern is the potential for interference by the host’s genomic material, resulting in a reduction in overall amplification yield as well as increased generation of non-relevant amplicons due to mis-priming. In the blood of a typical septic patient, host genomic material, most importantly but not only from white blood cells, may be present at a level up to 10¹⁰-fold higher than the microbial genomic material (23, 24). Additional concerns include other endogenous components of whole blood (such as heme) and exogenous substances, such as anticoagulants added at the time of collection (such as EDTA or heparin) which can interfere, either wholly or partially (thus reducing yields), with enzymatic amplification processes. To minimize the effect of these potential interferents, these assays are typically designed for small sample volumes (<100 µL). However, it is the small sample volume aspect of these enzymatic amplification processes which may render them inadequate for this application without upstream sample preparation considerations. No matter how powerful an assay is, if the microorganism is not present in the test volume, there is nothing to be detected.

In addition, there is the question of clinical utility of molecular diagnostic results. Since it relies on microbial proliferation, blood culture identifies the presence of viable microorganisms. Molecular tests, on the other hand, identify the presence of microbial genomic material after induced cell lysis. DNA contaminants or DNA from an unrelated and/or a previously resolved infection may also be detected, making clinical interpretation difficult (25–27).

A novel approach to overcoming many of the challenges faced by molecular testing for sepsis/BSI has previously been described (28). A platform utilizing this approach, known as RaPID (resistance and pathogen ID; HelixBind, Boxborough, MA), relies on a precursor step to selectively lyse host and non-viable microbial cells using a proprietary selective lysis solution (SLS) (28, 29) and thereafter remove the contaminating genomic material, including any preexisting cell-free DNA (cfDNA). This step not only removes microbial DNA that is not derived from an active infection (thereby reducing the false-positive rate) but also allows for the sampling of larger blood volumes (reducing the false-negative rate and improving sensitivity) (28). Moreover, this enrichment is achieved without centrifugation, which is known to cause mechanical cell lysis and reduce assay sensitivity (30–32). Thereafter, the surviving viable microorganisms are lysed, and the released genomic material is enzymatically amplified and surveyed using a series of specific, unique γ-modified peptide nucleic acid (γPNA) probes capable of hybridizing to intact duplex DNA rather than only to single-strand DNA. Leveraging this platform, an assay (termed RaPID/BSI) targeting the 20 most common infection associated with sepsis/BSI was developed.

The assay described above was previously evaluated in a retrospective cohort of 61 patients, all suspected of sepsis/BSI where testing was performed on archived samples drawn at around the same time of the standard-of-care diagnostic culture (28). Clinical status was determined solely by the standard-of-care blood culture result. A potential weakness of that study design was the age of the archived samples (typically 2–3 days after blood draw) as well as the non-sterile procedures associated with the collection, processing, and handling of these samples before release for testing. Another potential weakness was the lack of an objective measure of true infection status (e.g., prior cultures, blood, or otherwise) beyond blood cultures.

The current prospective study is both larger and designed to further evaluate this assay in freshly drawn specimens from individuals whose true infection status can be more confidently characterized. In a first prospective cohort, healthy individuals are enrolled, and blood samples drawn for testing with RaPID/BSI. By targeting only healthy individuals, we assume that any positive finding by RaPID/BSI can be considered to be a false-positive finding. In a second prospective cohort, independent blood samples are drawn from subjects with proven sepsis/BSI (via standard-of-care diagnostic blood cultures). Any positive finding by RaPID/BSI should match the microorganism identified by the initial (i.e., trigger) positive culture or any follow-up standard-of-care diagnostic blood cultures. This study tests the ability of the RaPID/BSI assay to accurately detect and identify active infections residing in the bloodstream. Additionally, by the nature of this second cohort study, we can evaluate the performance of RaPID/BSI in patients with concurrent antimicrobial therapy in comparison to blood cultures (both the trigger blood culture and follow-on blood cultures), which remain the “gold standard” for diagnosis.

MATERIALS AND METHODS

Technical experiments

Testing the ability of RaPID/BSI to remove host DNA without disrupting viable pathogenic microbes

Microbial species of interest were acquired from American Type Culture Collection or CDC & FDA Antibiotic Resistance Isolate Bank. The following microbial species were acquired and tested: For Staphylococcus species, Staphylococcus aureus (ATCC 43300) and Staphylococcus epidermidis (ATCC 51625); for Enterococcus species, Enterococcus faecalis (ATCC 29212) and Enterococcus faecium (ATCC 700221); for Streptococcus species, Streptococcus agalactiae (ATCC 13813), Streptococcus pyogenes (ATCC 12344), and Streptococcus pneumoniae (ATCC 6303); Escherichia coli (ATCC BAA-2469), for non-E. coli Enterobacterales, Klebsiella oxytoca (ATCC 49131), Klebsiella pneumoniae (ATCC BAA-1705), and Serratia marcescens (CDC & FDA AR Bank #0091); Pseudomonas aeruginosa (ATCC 10145); Acinetobacter baumannii (ATCC 19606); for Candida species, Candida albicans (ATCC 90028) and Candida krusei (ATCC 14243). Each microorganism was cultured in the appropriate broth and cultures diluted in PBS to ~10⁴ CFUs/mL as determined by optical density measurements at 600 nm (OD₆₀₀). Thereafter, 0.1 mL of the diluted culture was added to 0.9 mL freshly drawn ETDA anticoagulated human whole blood and mixed by inversion. SLS was prepared as described (29) and added to the spiked blood specimens. After incubation with mild shaking for 8–10 min at room temperature, 0.1 mL of the mixture was plated onto the appropriate growth substrate and incubated at 35 ± 2°C until colonies were sufficiently large to be readily counted, typically 18–36 hours. In general, the study aimed to achieve ~100–150 colonies/plate for each of the microorganisms of interest. Experiments were performed in triplicate. Controls included spiked blood samples with PBS in place of SLS. Final colony counts/plate were compared between cases (samples treated with SLS) and controls (samples with PBS in place of SLS), and results expressed as percent survivability.

Assessing the efficacy of human DNA depletion in the presence of elevated amounts of host immune cells

Sepsis/BSI is often associated with a significant host immune response. High levels of white blood cells (leukocytes) in whole blood collected from such patients may interfere with the performance of the RaPID/BSI assay, either directly (due to the presence of large numbers of intact leukocytes) or indirectly (due to the release of large amounts of host genomic material during the SLS-mediated selective cell lysis process). To investigate this possibility, intact leukocytes were isolated from human whole blood using ammonium chloride incubation followed by differential centrifugation as reported, resuspended in PBS, and quantified using fluorescence cell cytometry (Cellometer K2 Fluorescent Viability Cell Counter, Nexcelom Bioscience, Lawrence, MA), according to the specifications of the manufacturer. Once isolated and quantified, leukocytes were spiked into freshly drawn human whole blood at predetermined loads to simulate blood drawn from a subject experiencing leukocytosis. The human whole blood used in these studies had an average leukocyte count of 6.7 × 10⁶ cells/mL (for reference, normal range in a healthy adult is 4.5–11.0 × 10⁶ cells/mL). Isolated lymphocytes were added to simulate levels of leukocytosis ranging from ~1.5 × 10⁷ to ~2.5 × 10⁷ cells/mL, the latter being a suitable endpoint as suggested by CLSI Guideline EP07.

After incubating the spiked whole-blood samples with SLS, human DNA was extracted using magnetizable, electro-reactive, µ-particles (MERPs) as described (28). MERPs bind cfDNA, and the MERP-DNA complexes are immobilized using a strong magnetic field. Transfer of the blood/SLS mixture into a fresh vial results in a blood/SLS mixture free of DNA. This is a standard component of the RaPID/BSI assay. To test the efficacy of this process and whether it is influenced by leukocyte concentration, the amount of DNA extracted after SLS treatment of whole blood with/without leukocyte supplementation was measured and compared with the amount of DNA extracted in a parallel series of experiments using cetyltrimethyl ammonium bromide (CTAB; MilliporeSigma, Burlington, MA) 10 mM for 10 min, a powerful cationic detergent expected to lyse 100% of cells in the sample, in place of SLS. All MERPs (those isolated after SLS or CTAB incubation) were collected, washed, and the DNA eluted. The amount of DNA removed from whole blood as measured by both absorbance (BioDrop, BioChrom Ltd., Holliston, MA) and fluorescence (GloMax, Promega, Madison, WI) was compared between the two techniques. Results are reported as the percent of DNA extracted using SLS/magnetic capture using CTAB as the accepted “gold standard.”

Testing whether RaPID/BSI is subject to interference from antimicrobial agents

To evaluate whether common antimicrobial agents interfere with the performance of the RaPID/BSI assay, early log-phase cultures of one of three microorganisms were spiked into freshly drawn EDTA anticoagulated human whole blood at ~10 CFU/mL, chosen because this concentration is near the lower limit of detection for the assay. The three microorganisms were selected to represent Gram-positive bacteria (S. aureus, ATCC 43300), Gram-negative bacteria (E. coli, ATCC BAA-2469), and fungi (C. albicans, ATCC 90028). Contrived blood specimens were divided into equal fractions and spiked with one of five antibiotic cocktails: ampicillin sodium 152 µM, vancomycin hydrochloride 69 µM (both from Fisher Scientific, Hampton, NH), gentamicin sulfate 21 µM, meropenem trihydrate 186 µM, or combined piperacillin 117 µg/mL/tazobactam sodium 19 µg/mL (all from ThermoScientific, Waltham, MA), each representing roughly threefold the clinical reference level recommended by CLSI guidelines EP07 and EP37. Aliquots of each were analyzed by RaPID/BSI as soon as feasible after preparing the samples, typically within 1 hour. Experiments were run three times for each of the microbial/antibiotic combinations. Results were expressed as the ability of the RaPID/BSI assay to correctly identify the microorganism in each of the microbial/antibiotic combinations.

Prospective evaluation of freshly drawn whole-blood samples

Study population and sample collection from healthy individuals

Human whole-blood samples were collected prospectively from healthy individuals in eastern Massachusetts by a third-party licensed phlebotomist between July 2020 and May 2023. Subjects were eligible for inclusion if they were older than 18 years of age, generally felt well, had not previously donated blood or otherwise provided blood specimens for the previous 10 weeks, did not exhibit any signs of fever or illness, and had provided verbal and written informed consent. After obtaining consent, a standard 10 mL K₂EDTA (“purple top”) specimen tube was filled using standard sterile techniques, with a target volume of 8–10 mL venous blood. Samples were de-identified, labeled with a unique study subject identification number, placed at 2–8°C, and provided for testing with RaPID/BSI within 24 hours as described below. The study was approved by Pearl Pathways, LLC in Indianapolis, INC (IRB No. 20-HELB-104).

Study population and sample collection from patients with proven sepsis/BSI

A prospective cohort study was conducted at Tufts Medical Center, Boston, MA between October 2019 and August 2021. Patients were eligible for inclusion if they were older than 18 years of age with proven (blood culture positive) BSI within 24 hours of blood draw. This initial result and associated draw served as a trigger. Eligible patients were approached, and written informed consent obtained. Thereafter, a fresh venous blood draw was performed within 6 hours of the reported positive trigger blood culture result using standard sterile technique and specifically avoiding in situ intravenous access lines or central lines. The study was approved by the Tufts Health Sciences Institutional Review Board (IRB No. 13353).

After obtaining informed consent, two K₂EDTA (“purple top”) specimen tubes were filled with a target volume of 10 mL each. Samples were de-identified by research staff, labeled with a unique study subject identification number, placed at 2–8°C, and provided for testing within 24 hours with RaPID/BSI as described below. Supervising clinicians were blinded to the results of the RaPID/BSI assay. Medical records were reviewed and relevant information abstracted by trained research staff, including baseline demographic characteristics, clinical features, antimicrobial therapy, and results of laboratory tests, including blood culture findings and, if positive, the genus/species of the microorganism identified. All data were de-identified and labeled with a unique study subject identification number only.

Sample processing for prospectively drawn samples

To detect the presence of microorganisms in whole blood, samples were analyzed using the RaPID/BSI assay at a central laboratory as previously described (28). Briefly, RaPID/BSI is a sample-in/results-out assay that utilizes five consecutive processes for the direct detection and subsequent identification of one or more of the 20 most clinically prevalent BSI-causing pathogens, including Gram-positive bacteria (Staphylococcus aureus, Staphylococcus epidermidis, Enterococcus faecalis, Enterococcus faecium, Streptococcus agalactiae, Streptococcus pyogenes, and Streptococcus pneumoniae), Gram-negative bacteria (Escherichia coli, Klebsiella aerogenes, Enterobacter cloacae, Klebsiella oxytoca, Klebsiella pneumoniae, Serratia marcescens, Pseudomonas aeruginosa, and Acinetobacter baumannii), and yeast species (Candida albicans, Candida glabrata, Candida krusei, Candida parapsilosis, and Candida tropicalis). The five steps include: (i) selective lysis of host/human cells using SLS followed by removal of host genomic material and cfDNA, (ii) microbial lysis, (iii) isolation and purification of microbial DNA, (iv) PCR-based enzymatic amplification of microbial DNA, and (v) species-level identification using duplex DNA invading γPNA (28).

Samples containing less than 2.5 mL of blood were not processed. All other samples were analyzed by RaPID/BSI within 6 hours of arrival. While staff at the central laboratory were not blinded to the location from which the blood samples were provided and thus could potentially infer the study population, they were blinded to any diagnostic information, including culture results for patients with proven sepsis/BSIs. Samples that yielded positive results underwent further processing, wherein remnant isolated and amplified microbial genomic material underwent an additional manual detection assay utilizing either a pan-bacterial duplex DNA invading γPNA or a pan-Candida duplex DNA invading γPNA. If either assay was positive, the remnant genomic material underwent a second PCR process and subsequent bi-directional DNA sequencing to confirm the identified test menu target.

RESULTS

Technical experiments

Incubation with SLS does not disrupt viable pathogenic microbes

To assess for microbial survival in the presence of SLS, freshly drawn human whole blood was spiked with common BSI-causing microorganisms and cultured on appropriate growth media. Colony counts for each species were then compared to controls in which SLS was replaced with PBS, and results expressed as percent survivability. Data demonstrate that incubation with SLS does not adversely affect the viability of microbial cells for the microorganism species tested (Table 1).

TABLE 1.

Effect of the selective lysis solution on viability of select bloodstream infection-causing microorganisms

Microorganism	Survivabilitya (%)
Staphylococcus spp.	99.8 ± 2.1
Enterococcus spp.	100.3 ± 6.2
Streptococcus spp.	98.6 ± 2.9
Non-E. coli Enterobacterales	103.1 ± 2.9
E. coli	97.8 ± 3.6
P. aeruginosa	98.1 ± 5.6
A. baumannii	101.8 ± 2.2
Candida spp.	99.8 ± 2.1

^a Data are mean ± SD from five to seven experiments performed in triplicate.

Performance of RaPID/BSI is not adversely affected by high concentrations of immune cells

To assess for DNA removal efficiency, the amount of DNA extracted from human whole blood following SLS-mediated selective cell lysis and magnetic capture (routine elements of the RaPID/BSI assay) was compared to incubation with a powerful detergent expected to lyse 100% of cells in the sample. This was tested in human whole blood with typical and artificially elevated levels of leukocytes. Irrespective of the leukocyte content, SLS/magnetic capture effectively removed more than 99% of all DNA from human whole blood, including both unbound cfDNA and that contained within leukocytes (Table 2).

TABLE 2.

Efficacy of RaPID/BSI in removing human DNA from whole blood

Leukocyte count (cells/mL)	Percent of normal leukocyte counta (%)	Source of sample	Human DNA removal yieldb (%)
~6.7 × 106	~100 (baseline)	Human whole blood	99.98 ± 0.01
~1.5 × 107	~220	Human whole blood spiked with isolated human leukocytes	99.97 ± 0.01
~2.5 × 107	~370	Human whole blood spiked with isolated human leukocytes	99.94 ± 0.02

^a Normal leukocyte count in a healthy adult is 4.5–11.0 × 10⁶ cells/mL (for baseline, the average leukocyte concentration of human whole blood was used in the experiments, namely 6.7 × 10⁶ cells/mL).

^b Data are expressed as % of human DNA isolated using SLS/magnetic capture versus CTAB, mean ± SD from three to nine experiments.

Performance of RaPID/BSI is not adversely affected by the presence of antimicrobial agents

One of the major limitations of blood culture is interference by antibiotics (13). To investigate the effect of antibiotics on the performance of the RaPID/BSI assay, freshly drawn human whole blood was spiked with one of three microorganisms selected to represent Gram-positive bacteria (S. aureus), Gram-negative bacteria (E. coli), and fungi (C. albicans) at concentrations near the lower limit of detection for the assay. Aliquots of contrived blood specimens were analyzed using the RaPID/BSI assay after the addition of one of five clinically relevant antibiotic cocktails at supraphysiological concentrations. In all cases, the assay was able to successfully identify the target microorganism with no readily discernable interference by antibiotics (Table 3). Although RaPID/BSI currently provides qualitative assay results, the assay yields a detection signal related to the microbial load detected (28). These signals were provided along with those from same-day testing of blood specimens spiked with each of the three microorganisms but without the addition of any antimicrobial agent. A comparison of these signals is presented, demonstrating no consistent change associated with the addition of the tested antimicrobials.

TABLE 3.

Effect of select antimicrobial agents on the performance of RaPID/BSI

Microorganism	Antimicrobial agent	Concentration	Identification of the target microorganism in the presence of the antimicrobial agent
			Success rate % (N)a	Normalized RaPID/BSI assay signal (mean ± SD)
				(−) Agent	(+) Agent
S. aureus	Ampicillin	152 µM	100% (3/3)	1.00 ± 0.18	0.98 ± 0.08
	Vancomycin	69 µM	100% (3/3)		1.06 ± 0.24
	Gentamicin	21 µM	100% (3/3)		1.02 ± 0.18
	Meropenem	186 µM	100% (3/3)		1.22 ± 0.15
	Piperacillin and tazobactam	117 and 19 µg/mL	100% (3/3)		0.94 ± 0.06
E. coli	Ampicillin	152 µM	100% (3/3)	1.00 ± 0.15	0.94 ± 0.14
	Vancomycin	69 µM	100% (3/3)		1.14 ± 0.04
	Gentamicin	21 µM	100% (3/3)		1.19 ± 0.15
	Meropenem	186 µM	100% (3/3)		1.20 ± 0.04
	Piperacillin and tazobactam	117 and 19 µg/mL	100% (3/3)		0.91 ± 0.13
C. albicans	Ampicillin	152 µM	100% (3/3)	1.00 ± 0.12	1.05 ± 0.16
	Vancomycin	69 µM	100% (3/3)		1.03 ± 0.24
	Gentamicin	21 µM	100% (3/3)		1.04 ± 0.14
	Meropenem	186 µM	100% (3/3)		0.93 ± 0.02
	Piperacillin and tazobactam	117 and 19 µg/mL	100% (3/3)		0.89 ± 0.10

^a Successful detection is defined as an obtained signal from the detection channel being above the established cutoff, determined as greater than the limit of blank (>3× SD the mean signal obtained in the same detection channel from a non-contrived sample).

Prospective evaluation of freshly drawn whole-blood samples

Evaluation of specimens from healthy individuals

To assess the false-positive rate of RaPID/BSI, we conducted a prospective cohort study using freshly drawn specimens collected external to a healthcare setting from volunteer adult individuals who exhibited no apparent signs of illness and generally felt well. Using this approach, we assumed that none of the enrolled individuals were clinically septic and/or harbored a bloodstream infection and that, therefore, any positive finding by RaPID/BSI could be confidently considered a false-positive result.

Results for this study arm are summarized in Table 4. Of the 156 specimens collected from 156 unique visits/study enrollment events and analyzed by RaPID/BSI, nine (5.8%) were positive. Of these, six were positive for the detection channel specific to S. epidermidis and three for the detection channel specific to multiple non-E. coli Enterobacterales species, including Klebsiella aerogenes, Enterobacter cloacae, Klebsiella oxytoca, Klebsiella pneumoniae, and/or Serratia marcescens. Across the twelve detection channels included in the RaPID/BSI assay, this yields a specificity of 99.5% (1,863/1,872) across the assay and 94.2% (147/156) on a per assay basis (Table 5) resulting in a clinical false-positive rate of 5.8%.

TABLE 4.

Results of RaPID/BSI analysis of specimens from healthy donors

Microorganism	Positive findings	Channel specificity
S. aureus	0	100% (156/156)
S. epidermidis	6	96.2% (150/156)
S. agalactiae/S. pyogenes	0	100% (156/156)
S. pneumoniae	0	100% (156/156)
E. faecalis	0	100% (156/156)
E. faecium	0	100% (156/156)
E. coli	0	100% (156/156)
Enterobacter spp./Klebsiella spp./S. marcescens	3	98.1% (153/156)
P. aeruginosa	0	100% (156/156)
A. baumannii	0	100% (156/156)
C. albicans/C. parapsilosis/C. tropicalis	0	100% (156/156)
C. glabrata/C. krusei	0	100% (156/156)

TABLE 5.

Summary of RaPID/BSI analysis of specimens from healthy donors

Description	Specificity across assay	Per test specificity
All RaPID/BSI channels	99.5% (1,863/1,872)	94.2% (147/156)
Without S. epidermidis channel	99.8% (1,713/1,716)	98.1% (153/156)

RaPID/BSI is susceptible to S. epidermidis contamination in the same manner as standard-of-care blood cultures. These contaminations occur most often when commensal bacteria on the skin enter the specimen collection tube at the time of venipuncture, with contamination rates varying widely across institutions from 0.6% to 6% (target is <3% per the American Society for Microbiology and the Clinical Laboratory Standards Institute) (33–35). If RaPID/BSI false-positive results associated with S. epidermidis were a result of the collection procedure, this would correspond to a specimen collection contamination rate of 3.8% (6/156). Excluding results from the S. epidermidis detection channel would result in a specificity for RaPID/BSI of 99.8% across the assay (1,713/1,716) and 98.1% (153/156) on a per assay basis (Table 5).

Evaluation of specimens from individual with proven BSIs

We conducted a second prospective cohort study using freshly drawn specimens from patients with proven (blood culture positive) sepsis/BSI. Specimens were drawn for testing soon after a positive blood culture was reported by the laboratory and written consent obtained. Time from initial trigger draw until the experimental draw ranged from 19 to 29 hours (mean ± SD, 24 ± 2.8 hours). A total of 12 individuals agreed to participate in the study. One individual subsequently withdrew consent (Subject 007), leaving 11 study subjects in the final analysis (Fig. 1).

Fig 1.

Recruitment of subjects with known BSIs/sepsis.

RaPID/BSI results are summarized in Table 6. For each of the eleven subjects, the trigger blood culture recovered a single organism, one of which (Subject 003) was deemed a probable contaminant due to the nature of the microorganism combined with only a single-positive bottle from the four collected. Seven of the pathogenic microorganisms identified by blood culture were represented on the RaPID/BSI panel. Five of these (71%) were recovered in the follow-up blood draw, and all were correctly identified by RaPID/BSI. Four of the pathogens identified by blood culture were not on the RaPID/BSI panel, including the probable contaminant (Subject 003). Of these “off-menu” pathogens, two were found by a RaPID/BSI channel designed to detect (but not identify the species) the 16S/18S rDNA from off-menu microorganisms. Since the assay effectively removes all cell-free genomic material residing in the blood prior to microbial lysis, any 16S or 18S rDNA detected must have originated from viable microbial cells. For all subjects in which the microorganism was identified by RaPID/BSI as well as the two in which remnant 16S rDNA was detected, bi-directional Sanger DNA sequencing was performed to confirm species identity. For the two off-menu pathogens detected, the species identified by DNA sequencing matched the trigger blood culture results. In sum, all seven positive results and pathogen identifications provided by RaPID/BSI matched blood culture results (Table 6). In no instance did RaPID/BSI detect a microorganism that was not identified by either the trigger or follow on blood culture.

TABLE 6.

RaPID/BSI results for subjects with confirmed BSIs

Subject no. (n = 11)	Trigger blood culture result	Hours from trigger draw to follow-up draw	Follow-up RaPID/BSI result	DNA sequencing of RaPID/BSI	Same day follow-up clinical blood culture result
001	S. aureus	27	S. aureus	S. aureus	No growth
002	E. faecalis	23	Negative	–	No growth
003	Viridans Streptococci a,b	26	Negative	–	No growth
004	S. aureus	25	S. aureus	S. aureus	S. aureus (one of two bottles)
005	S. pseudintermedius b	19	Pan-bacteria positive	S. pseudintermedius	N/Ac
006	S. warnerib	20	Negative	–	CoNS (two of two bottles)
008	Pasturella spp.b	19	Pan-bacteria positive	Pasturella spp.	No growth
009	E. faecalis	24	E. faecalis	E. faecalis	E. faecalis (one of four bottles)
010	E. faecium	23	Negative	–	No growth
011	S. aureus	<24	S. aureus	S. aureus	No growth
012	S. aureus	<24	S. aureus	S. aureus	S. aureus (three of four bottles)

^a Likely contaminant.

^b Not on the panel of the 20 most common BSI-causing microorganisms included in the RaPID/BSI assay (“off-panel”).

^c N/A, not applicable.

By virtue of this study design, all RaPID/BSI assays were performed on a second blood sample collected up to 30 hours after an initial draw that resulted in the positive trigger blood culture result (i.e., up to 24 hours for a positive blood culture result and up to 6 hours for the repeat blood draw). If the patient’s antibiotic treatment had effectively eradicated the BSI during that time period, there would be nothing to detect in this second blood sample. This effect was seen with Subjects 002 and 010, which were positive for E. faecalis (subject was concurrently treated with a combination of vancomycin, cefepime IV, and linezolid added immediately prior to the additional blood draws) and E. faecium (subject was treated with linezolid), respectively, in the trigger blood culture and negative by both RaPID/BSI and follow-up blood cultures.

Of the 11 enrolled patients, 10 had follow-up blood cultures performed from a separate sample obtained within 36 hours of the initial draw, all around the same time that the RaPID/BSI samples were collected. Of these, four (40%) returned a positive result showing persistence of a BSI. In all cases, the organism identified was consistent with the initial blood culture (Subjects 004, 006, 009, and 012 in Table 6). Of those four cases, RaPID/BSI accurately identified three (75%); the final case represented a microorganism that was off-menu (Subject 006).

DISCUSSION

Blood culture remains the “gold standard” for the diagnosis of sepsis/BSI but has many limitations. Molecular diagnostic methods have the potential for markedly improving the management of such patients through faster turnaround times and increased accuracy. In this paper, we evaluated the performance of the RaPID/BSI assay that incorporates a novel precursor step which selectively removes the host/human genomic material as well as that from non-viable microbial cells prior to the requisite microbial lysis step, thereby isolating genomic material only from viable microorganisms in human whole blood for downstream processing (24, 25). We demonstrated that the assay is free of interference by host immune cells and multiple common antimicrobial agents. We also demonstrated the accuracy of this technology with specimens prospectively drawn from 156 subjects presumed to be sepsis/BSI negative as well as 11 subjects/patients with proven (culture positive) sepsis/BSI.

Despite its promise, the utility of molecular technology for the diagnosis of sepsis/BSI has been limited to post-culture analysis and has yet to be widely adopted for the identification of BSIs without prior specimen enrichment by culture. Existing commercially available assays for the direct detection and identification of microorganisms from human whole blood all utilize a single (total) lysis reaction pre-PCR amplification without the selective lysis of host (i.e., human) cells and/or the removal of host and cell-free DNA. As such, these assays are subject to detection of DNA unrelated to an active/ongoing infection, sometimes referred to as “DNAemia.” While false positivity is a concern for any diagnostic test, it is of particular concern for indications with inherent low true-positivity rates. Sepsis is one such indication, where symptoms are often non-descript, and the large majority of those considered at risk are ultimately found to not harbor a BSI (typical blood culture positivity rates are ~9%) (36).

Only a single platform has been cleared by the FDA for identifying BSIs directly from blood, offering an assay targeting five Candida species (37) and another targeting five bacterial species (38). Both assays demonstrate high rates of molecular-positive/blood culture-negative findings. Even after adjustment for other clinical evidence of sepsis/BSI (and therefore potential culture false negativity), these tests demonstrated relatively low positive predictive values (PPVs): 28% (10/36) for the Candida assay and 55% (104/190) for the bacterial assay (37, 38). Similarly, a CE-marked direct-from-blood assay targeting a combined menu of 19 bacteria and 6 Candida species provided one molecular-positive/blood culture-negative finding for every molecular-positive/blood culture-positive finding across multiple studies (39–42). In some instances, it was likely that the discordant (molecular false positive) result may have identified a case of true sepsis/BSI based on clinical features (range, 4%–68%), but a significant number had no clear explanation (39–42).

In a recent study, Blauwkamp et al. utilized next-generation DNA sequencing to detect microbial cfDNA in patients with suspected sepsis/BSI. They identified 94% (59/63) of blood culture-positive infections but an additional 171 cfDNA-positive results from patients with negative blood cultures. Even after adjusting for cases in which the clinical presentation was suggestive of probable sepsis/BSI, the PPV of the test remained relatively low at 50% (112/224) (26). These data suggest that cfDNA from non-viable microbial cells may be a source of false-positive tests in these molecular assays.

The precursor SLS/magnetic capture step included in RaPID/BSI should mitigate the risk of false-positive tests by virtue of removing over 99% of cfDNA and leukocyte-based DNA in human whole blood. As demonstrated, this precursor step is effective regardless of the leukocyte load (Table 2) and does not adversely affect the viability of circulating pathogenic microorganisms, including Gram-negative bacteria, Gram-positive bacteria, and yeast common in sepsis/BSIs (Table 1). The intent of this precursor step is to both improve sensitivity through improved sampling and reduce the likelihood of false-positive results.

In the current study, excluding calls likely associated with skin contaminants, subjects assumed to be free of sepsis/BSI were incorrectly classified by RaPID/BSI in only 1.9% (3/156) of cases (Table 5). In assessing specimens from subjects with confirmed sepsis/BSI, all microorganisms identified by RaPID/BSI were consistent with those reported on the trigger blood culture (Table 6). We note that all three of the RaPID/BSI false-positive findings among the 167 blood specimens (156 for those assumed free of sepsis/BSI and 11 for those with confirmed sepsis/BSIs) were associated with a single detection channel for non-E. coli Enterobacterales. To that end, we hypothesize that this is likely due to a potential contamination of assay reagents, a known concern, predominantly associated with a variety of Gram-negative bacterial species including non-E. coli Enterobacterales, plaguing reagent manufacturing of molecular diagnostic processes and a particular challenge when assaying samples with low microbial biomass (43–45). Thus, while the precursor SLS/magnetic capture step is a powerful addition to improve on the accurate diagnosis of subjects with likely sepsis/BSIs, it is not immune to such contaminations during manufacturing processes. As such, the importance of proper manufacturing processes and quality controls cannot be overstated, which in the absence of, could lead to unnecessary and/or incorrect treatments, potentially compromising the health of the individual.

A known challenge associated with blood culture is interference by antibiotics (46–48). Here, we demonstrate that RaPID/BSI is unaffected by the presence of multiple common clinically relevant antimicrobial agents, even at supraphysiological doses (Table 3). Not only is this capability helpful for establishing the initial diagnosis (by reducing false-negative test results), but it suggests also that—since the turnaround time is only ~4 hours—RaPID/BSI could be used to track, in close to real time, how quickly the pathogenic microorganism is being cleared from the patient’s blood or whether the antimicrobial regimen should be changed. In this way, RaPID/BSI could help target therapy and optimize antimicrobial stewardship, minimizing the duration of exposure to unnecessary antimicrobials (i.e., those lacking the appropriate spectrum of coverage) and thereby reducing side-effects and the emergence of antimicrobial resistance. Future tests incorporating larger test menus inclusive of antimicrobial resistance genes should lead to even greater benefits in this regard. Given the current paradigm, relying solely on blood cultures for identifying bloodstream infections, the incorporation of a culture-free diagnostic approach, such as RaPID/BSI, might also mitigate the need for follow-up blood cultures (46–48).

To account for the difficulty in establishing true infection status in the hospital setting, this study utilized two patient populations. A limitation of this approach is that the combined population may not reflect the actual true-positive rate in the hospital settings. Another limitation is the small number of proven sepsis/BSI subjects enrolled. This was due in large part to the strict time constraints built into the study to secure repeat blood samples as soon as possible after sepsis/BSI was confirmed by a positive blood culture in the laboratory. To improve recruitment in future studies, investigators may consider requesting IRB approval to draw additional research samples at the time of initial blood culture and consenting patients thereafter if the blood culture is positive for the purpose of reviewing medical records. A third limitation of this study was the lack of a laboratory verification of infection status for the first cohort of subjects, where absence of a BSI was assumed based on enrollment methodology and not confirmed via blood culture. Classical microbiological results could have helped identify or rule out commensal contamination, transient bacteremia, or other potential sources of RaPID/BSI clinical false-positive results. A fourth limitation is the panel size of the existing RaPID/BSI assay, which includes the 20 most clinically prevalent BSI-causing pathogens, including Gram-positive bacteria, Gram-negative bacteria, and fungal species (24). Four of the 11 organisms identified by blood culture were “off menu,” i.e., are not included in the RaPID/BSI panel (Table 6), although two of the four were identified through a separate channel combined with DNA sequencing and one was a suspected contaminant (Viridans Group Streptococci). While the panel is comparatively broad, expanding the panel by developing and validating additional γPNA detection probes for a wider scope of microorganisms could effectively address this limitation.

In this paper, we have evaluated the utility of the molecular direct-detection RaPID/BSI assay for the detection and identification of pathogenic microorganisms in human whole blood collected from patients with proven (culture positive) sepsis/BSI. The methodological advantages of this platform were confirmed and highlighted, including lack of interference by antibiotics, assay reagents (such as SLS), or sources of DNA other than viable micro-organisms (such as cfDNA or DNA from host immune cells). Larger prospective cohort studies in both high- and average-risk populations are needed to confirm these findings and evaluate the utility of this test in reducing the morbidity and mortality associated with sepsis/BSI.