Harnessing the host response for precision infectious disease diagnosis

SUMMARY

Detection of the presence of infection and its etiology must be accurate and timely to facilitate appropriate antimicrobial use. Diagnostic strategies that rely solely on pathogen detection often are insufficient due to poor test characteristics, inability to differentiate colonization from infection, or protracted delay to result. Understanding the human response across different pathogens on a clinical and molecular level can provide more accurate, timely, and useful answers, especially in critical illness and diagnostic uncertainty. Improvements in understanding the human immune response including genomics, protein analysis, gene expression, and cellular morphology have led to rapid innovation of new host response-based diagnostic tests. This review describes the limitations of pathogen-focused technology and the benefits of examining the breadth of immune response to diagnose infection. It then explores biomarkers that have been studied for this purpose and scrutinizes the performance of host-based multianalyte testing. Currently cleared diagnostics and those in late-stage development are described in depth, with a focus on the purpose of testing and its utility for clinicians. Finally, it concludes by examining opportunities for further host response-derived diagnostic innovation.

INTRODUCTION

Traditional microbiological diagnostics often fall into one of two broad categories: diagnostics that directly detect the pathogen of interest and those that measure pathogen-specific adaptive immune responses, such as serology. However, a third approach that has been expanding is measuring the response of the host to infection through a combination of nonspecific biomarkers whose patterns are indicative of certain types or classes of infection. At its core, this diagnostic strategy moves away from pathogen-focused testing to best answer many questions most pertinent to clinicians.

Developing a diagnostic test that measures the host response to detect the general presence or absence of infection is not a new concept. The microscopic visualization of leukocytes in both blood and purulent wounds has long supported their role as a biomarker of infection (1). With the introduction of germ theory in the nineteenth century, diagnostic methods have primarily been pathogen-focused, including direct detection through culture, antigen processing, and more recently detection of microbial nucleic acids. The development and improvements in the detection, quantification, and analysis of host-based protein, transcriptional RNA, and DNA have led to the ability to characterize how the human host responds at the molecular level to the presence of an invasive pathogen.

Utilizing host-based physiology in the setting of infection has led to rapid development of several novel and unique tests. Many of these tests include multianalyte assays that can fingerprint the immune response in real-time. These measurements of the human host response aim to augment diagnostic efforts throughout the course of illness. These novel diagnostics have the potential to answer whether a patient has an infection, identify most likely pathogen types, forecast prognosis following an infectious diagnosis, and then monitor response to therapy.

Limitations of pathogen-focused diagnostic methods

The most widespread and traditional microbiological testing involves pathogen-specific identification. Common culture-based methods are a cornerstone of modern infectious disease testing as they allow for the isolation of a pathogen and can also be tested against various antimicrobial therapies, including antibiotic susceptibility (2). This process often takes several days, has variable sensitivity, and may be unable to reliably detect fastidious, atypical, or low-quantity pathogens (3, 4). Although the most widespread method of microbiological detection, cultures can be resource- and personnel-intensive and require rigorous oversight to ensure accuracy and best practices (5). Traditional culture methods have variable test performance by pathogen, body site, and method in which cultures were obtained and are less sensitive after antimicrobials have been administered (6, 7).

Microbial antigen detection is also a common diagnostic method employed by many microbiology laboratories. Detecting a specific antigen can frequently be performed rapidly at the point-of-care using lateral flow assays (LFA) (8). Similarly, antigen detection remains the mainstay of diagnosis for certain infections including Cryptococcus spp., Histoplasma spp., and Legionella pneumophila (9–12). Despite the convenience of LFAs and other antigen-detection methods, they require a priori suspicion of the specific pathogen of interest, are prone to false-negatives across diverse strains or serogroups, and may have lower sensitivity in the setting of lower pathogen burden (12, 13).

Nucleic acid amplification tests (NAAT), especially polymerase chain reaction (PCR) technologies, have transformed microbiology by facilitating the rapid and precise detection of numerous pathogens (14, 15). These advances in molecular technology allow for rapid detection of numerous bacterial, viral, fungal, and protozoal targets on a single sample within a matter of hours or even minutes. The implementation of NAATs has allowed for the widespread uptake of PCR and isothermal amplification devices, which rapidly accelerated during the course of the COVID-19 pandemic (16, 17). Despite these transformative properties, NAAT technologies are limited by their differential performance across sample types, particularly when physically accessing the nidus of infection requires invasive procedures or surgery (18, 19). NAAT techniques are generally unable to distinguish between true infection and colonization or when there are residual non-viable and non-infectious nucleic acids after resolution of illness. This can be a particularly critical issue when dealing with non-sterile sites such as sputum, urine, skin swabs, and the gastrointestinal tract (20–22).

In addition to direct pathogen detection, detecting a specific pathogen by measuring the adaptive host responses to that pathogen is a common technique, especially quantification of serologic IgG and IgM titers. Although not the focus of this review, these microbiologic tests represent an important method of measuring the host response for diagnosing an infection by a specific pathogen. This is particularly useful for fastidious or atypical pathogens such Bartonella spp., Borrelia spp., Brucella spp., Rickettsia spp., and Treponema pallidum. (23–26) These tests are limited by delay in detectable antibody response, the inability to reliably differentiate prior versus active infection, and cross-reactivity between pathogens (27).

In contrast to serologic testing, cellular-mediated immunity testing is the mainstay for diagnosing latent Mycobacterium tuberculosis (TB) infection. Tuberculin skin testing (TST) uses a TB-based purified protein derivative (PPD) injected intradermally to measure prior exposure and immunological memory but requires technical expertise in administering and reading, as well as a second follow-up visit (28). Intended to fill the limitations of TST, interferon-gamma release assays (IGRAs) are a single blood-based test that measure the host T-cell response in the setting of prior Mycobacterium tuberculosis exposure (29). Both strategies are significantly limited by false-positive results following BCG vaccination and false-negative testing from immunosuppression or T-cell anergy (30, 31). Use of IGRA testing is also being pursued to test for other pathogens that have prolonged and clinically challenging latency states, such as cytomegalovirus (32, 33).

TRADITIONAL BIOMARKERS OF THE INFLAMMATORY HOST RESPONSE

To differentiate infection from contamination or colonization, clinicians rely on complex medical decision-making. Understanding the host is instrumental and relies on physical exam, clinical history, laboratory findings, and biomarkers of inflammation and infection. Widely used biomarkers of the immune response are described here.

Complete blood count

The most used biomarker of the host response is the white blood cell (WBC) count. It is routinely found on complete blood counts and obtainable in essentially any clinical laboratory worldwide. A measurement >11.0 or <2.0×10⁹/L WBC is associated with states of physiologic stress, including infection. Measurement of types of leukocytes, including neutrophils, lymphocytes, and eosinophils, can often raise clinical suspicion for bacterial, viral, or parasitic causes of infection. Leukocyte differentials can be leveraged to investigate an infectious etiology. For example, an increase in the concentration of premature neutrophils (or left shift) may be suggestive of a bacterial infection, while a ratio of lymphocytes to monocytes < 2 has been shown to be predictive of influenza in patients presenting with symptoms of a respiratory infection (34, 35).

Although abnormal leukocyte counts are often associated with infection, particularly bacterial infections, they are nonspecific for infection alone; elevated WBC counts can be seen in any type of physiologic stress. Non-infectious causes of leukocytosis are broad and commonly include trauma, autoimmune disease, hemorrhage, thromboembolism, infarction, and hematologic malignancy. For example, in one study, only 48% of patients with WBC elevations 12–25 × 10⁹/L had likely bacterial infection, and only 74% had infection with >25×10⁹/L WBC (36).

Similarly, elevations in platelet count are seen in many inflammatory states, and platelet count derangements in critical illness have been associated with more severe infection (37, 38). The nonspecific and insensitive characteristics of complete blood counts make these less useful as an overall host response-based test in clinical practice (39). More recent advancements describing the morphology of white blood cells, including monocyte distribution width and leukocyte deformability, have also been leveraged into novel host response-derived diagnostics and are discussed later in this review.

Erythrocyte sedimentation rate & C-reactive protein

The erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) have been used clinically to describe inflammation for almost a century since they were first described. However, both test modalities lack appropriate sensitivity or specificity to serve as standalone biomarkers to diagnose infection.

The ESR measures the rate (in mm/hour) that red blood cells descend in a vertical tube and was first described in detail for assessment of inflammatory arthropathies in the 1930s (40). The rate at which this erythrocyte sediment forms is used to indirectly measure a host of inflammatory markers in serum, including fibrinogen and various immunoglobulins, and may increase within 48 hours of stimulus and then decrease over many weeks (41). No large high-quality studies have identified ESR as a consistent biomarker to distinguish bacterial from viral causes of infection; however, several studies have identified an association between elevated ESR and severity or duration of known bacterial illness (42–45).

In contrast, CRP is a single-protein biomarker produced in the liver and frequently released in states of physiologic stress in response to interleukin-6. It was described as early as 1930 in patients with pneumococcal pneumonia and staphylococcal infections and is postulated to enhance the immune response through activation of the classical complement response by binding pathogenic polysaccharides (46, 47). CRP is often expressed in low levels at times of physiologic stress outside the context of infection, although particularly expressed in bacterial infections. Rise in CRP can be reliably detected 5 hours after stimulus, with peak values typically around 48 hours after resolution and is not typically affected by renal replacement therapy, immunosuppression, or corticosteroids (48, 49). Large meta-analyses on CRP accuracy to differentiate bacterial from non-bacterial causes have shown sensitivity ranging from 75% to 86% and specificity 67% to 70% (50, 51). Critically, most of these studies were performed in the pre-COVID era, and since CRP levels are commonly elevated in COVID, this likely limits CRP’s utility as a single marker to differentiate viral and bacterial etiologies in the modern era.

Compared to the ESR, CRP has a more targeted immune mechanism and is a more specific marker for bacterial infections, including cellulitis and osteomyelitis (52). As CRP measures a single protein-based immune response, it generally increases more rapidly with infection and resolves more quickly, compared to a slower increase and decrease of the ESR. CRP and ESR are also commonly used in clinical practice to evaluate for the presence of chronic infections such as osteomyelitis, but data supporting this practice are limited (53). Both are routinely elevated in many forms of inflammation from both infectious and non-infectious causes, including rheumatoid arthritis, osteomyelitis, infective endocarditis, pancreatitis, systemic lupus erythematosus, inflammatory bowel disease, community-acquired pneumonia, and SARS-CoV-2 infection (48, 54). In some situations (like SARS-CoV-2 infection), CRP values over time may show some utility for monitoring the progression of infection (55). Problematically, these inflammatory syndromes may all present with fever and leukocytosis. The fact that CRP cannot consistently distinguish infectious versus non-infectious causes of systemic inflammation hampers clinical utility as a biomarker in clinical practice.

Concerningly, patients in early sepsis might not yet have meaningful CRP elevations, and a normal value might encourage clinicians to inappropriately withhold antibiotics (56). Examining CRP velocity (CRPv), defined as the difference in the first two CRP measurements divided by the number of hours between tests, has been proposed as a meaningful differentiator between bacterial and viral causes of inflammation. This is based on the observation that bacterial infections show a faster rise than viral or non-infectious causes (57). There are currently no well-accepted thresholds for CRPv to define bacterial infection versus other syndromes of interest (58). Implementing clinical algorithms that involve multiple point-in-time estimates can be challenging and must appropriately manage confounding non-infectious scenarios like myocardial infarction (59). Perhaps due to these difficult questions about optimal guidance and interpretation, no large-scale clinical implementation trials have been performed on CRP or CRPv alone to guide antibiotic decision-making. These clinical dilemmas have led to the expanded search for meaningful laboratory measures to assess the host response to infection, including procalcitonin.

Procalcitonin

One extensively studied biomarker to establish or exclude the presence of bacterial infection is procalcitonin (PCT), the prohormone of the calcium-modulating hormone calcitonin. In healthy states, it is minimally found in serum, as it is produced and cleaved in C-cells of the thyroid (60). During bacterial infection, it is produced but not cleaved by many peripheral tissues (61). Procalcitonin can typically be detected 4 hours after stimulus, has a half-life of approximately 24 hours, and peaks 24 hours after removal of the stimulus (62, 63). Notably, chronic kidney disease may cause elevations in procalcitonin, while corticosteroids and neutropenia may result in decreased PCT levels (64–67). As procalcitonin release is increased by numerous pro-inflammatory cytokines but inhibited by interferon-γ, it is a useful marker for differentiating bacterial versus non-bacterial causes of illness and thus has been leveraged into clinical algorithms for antibiotic stewardship (68, 69).

A large 2017 Cochrane review of randomized control trials showed procalcitonin-based clinical algorithms for respiratory tract infections could augment clinical decision-making, reduce antibiotic exposure, and were associated with lower aggregated mortality (70). A meta-analysis of 523 patients with bacteremia selected from 13 randomized controlled trials evaluating PCT-guidance described lower antibiotic exposure and similar improvements in mortality (71). However, all of these trials were conducted prior to Infectious Diseases Society of America (IDSA) and American Thoracic Society (ATS) guidance that antibiotic treatment duration for 5 days is appropriate for most patients with pneumonia (72). Several subsequent high-quality randomized controlled trials have failed to show clinical benefits for procalcitonin (73, 74). A large multicenter randomized controlled trial in United States (ProACT Trial) showed that procalcitonin-based guidance for suspected lower respiratory tract infections did not impact antibiotic decision-making or mortality (75). Because of these variable results from numerous clinical implementation studies, procalcitonin has seen limited utility as a single biomarker to guide diagnosis of bacterial infection and subsequent antibacterial therapy management. Additionally, the clinical accuracy of procalcitonin, as well as CRP and ESR, in the setting of severe immunocompromised states has shown inconsistent and variable results (76–78) (See Table 1).

TABLE 1.

Properties of traditional biomarkers

	Erythrocyte sedimentation rate	C-reactive protein	Procalcitonin
Physiologic source	Serum proteins and erythrocytes	Hepatic synthesis	All tissues
Mean detection time after stressor	Variable	5h	3–4h
Peak time after stimulus	Variable	48h	24h
Half-life	Slow, variable	20h	24h
Effect of immunosuppression	Decreased	No effect	Variable
Renal clearance	No effect	No effect	Cleared by renal filtration (or replacement therapy)
Bacterial versus viral differentiation	Poor	Single value: Poor Velocitya: Unclear	Moderate

^a CRP velocity is different in first two CRP measurements divided by the number of hours between tests.

Additional single biomarkers

Numerous single biomarkers measuring the host response to infection have been studied, particularly to better define prognosis once an etiology is known. Several of these biomarkers are discussed below but have notable limitations or inadequate overall performance metrics.

Interleukin-6

Interleukin-6 (IL-6) is a proinflammatory cytokine synthesized in lymphocytes, fibroblasts, and endothelium that is elevated in the presence of infections, with elevations found to correlate with poor prognosis in setting of sepsis (79). IL-6 primarily works as a pro-inflammatory cytokine by inducing production of numerous subsequent pro-inflammatory markers including CRP, fibrinogen, and serum amyloid A (SAA) (80). This results in CD4 and CD8 T cell differentiation, B cell antibody production, and hematopoiesis regulation. The broad pro-inflammatory nature of IL-6 can be seen in both infectious and non-infectious states, such as rheumatoid arthritis, giant cell arteritis, and juvenile idiopathic arthritis. Within confirmed infection, the proinflammatory state induced by IL-6 can be profound and may correlate with both prognosis and response to treatment in intensive care settings. A meta-analysis of 14 studies showed lower IL-6 levels (SMD −0.69) in sepsis survivors following treatment compared to baseline (81).

Utility of measuring IL-6 has also been explored in infectious states apart from bacterial sepsis. In the context of SARS-CoV-2 infection, elevated IL-6 serum levels have been measured as early prognostic markers and accurately predicted the likelihood of progression to mechanical ventilation (82). This measurement of host–cytokine response was leveraged into emergency use authorization (EUA) for IL-6 clinical assays, which in turn allowed for study of IL-6 inhibitor therapy in the setting of severe COVID-19 pneumonia (83). A meta-analysis of 27 randomized controlled trials studying IL-6 inhibitors, most prominently tocilizumab, showed reduction in mortality among patients with severe COVID-19 pneumonia without a significant increase in secondary infections (84). Given the nonspecific association with inflammation from diverse pathogens and auto-immune disease, IL-6 measurement has no well-defined role in differentiating the etiology of an acute illness, although the data do suggest potential uses for purposes of prognosis and immunomodulator selection. Evaluation of the accuracy of a point-of-care IL-6 detection device (Symphony IL-6 by Bluejay Diagnostics) to predict mortality in sepsis and septic shock is forthcoming (85).

Pro-adrenomedullin

Pro-adrenomedullin is a novel biomarker that has been described in detection of early bacterial infection. Adrenomedullin (ADM) is a peptide related to calcitonin that is similarly expressed in low levels by many tissues during healthy states and has potent vascular endothelial dilatory properties. It is upregulated in states of physiologic stress and increases cardiac output as well as microvascular permeability but may also provide a protective effect in regulating microcirculation in the setting of sepsis (86). Upregulation during sepsis is associated with more severe disease progression, but unstable serum kinetics make its prohormone pro-adrenomedullin (pro-ADM) a more appropriate biomarker to measure in blood.

Review of pro-ADM as a biomarker suggests that it may hold promise as an additional clinical decision-making tool to help triage early sepsis, identify those at highest risk for disease progression, and particularly those most vulnerable to consequences of vascular endothelial dysfunction (87). In a meta-analysis examining seven studies evaluating the accuracy of pro-ADM as a biomarker for sepsis, pooled characteristics included sensitivity 0.84, specificity 0.86, and AUC of 0.91 (88). Notably, in this review, the reference group used heterogeneous definitions for sepsis without clinical adjudication.

Additionally, the relationship of pro-ADM and vascular dysregulation likely limits its utility as a broad biomarker for infectious diagnoses, particularly in the setting of cardiac dysfunction (89). No well-performing cut-off values of pro-ADM have consistently defined sepsis and septic shock, which highlights the challenges of meaningful clinical implementation (90). Large-scale clinical utility studies will need to be performed to support broad clinical use at this time, which is not currently supported (91).

Pancreatic stone protein

Pancreatic stone protein (PSP) is a different host immune protein that has also been identified as a candidate biomarker in the setting of infection (92). In review of 23 diverse studies of PSP as biomarker for sepsis, it showed variable performance and did not clearly outperform existing biomarkers such as PCT (93, 94). A different review of PSP performance to discriminate ICU mortality and sepsis severity found a low-threshold PSP had sensitivity 0.96, but with poor specificity of only 0.16 (95). Importantly, no large-scale clinical implementation studies have been performed to define exactly what role it would play and if PSP values could be implemented to change clinical management, such as antibiotic initiation or cessation. A niche role for PSP as a single analyte biomarker might be in the setting of secondary infections with patients encountering severe burns, but more study on implementation would be needed (96).

Other protein biomarkers

Additional notable protein-based biomarkers include soluble triggered receptor on myeloid cells-1 (sTREM-1), myxovirus resistance protein A (MxA), and TNF-related apoptosis-inducing ligand (TRAIL). sTREM-1 is a transmembrane receptor expressed on numerous human cells and mediates septic shock (97). A systematic review of nine studies found that sTREM-1 was a moderate predictor of mortality in sepsis, and therapies that inhibit sTREM-1 are being studied related to septic shock (98, 99). MxA is a protein with antiviral activity that is induced by interferon and upregulated in states of active viral infection or autoimmune connective tissue disease but generally unaffected by bacterial infection (100–102). Similarly, TRAIL is a protein typically elevated in the setting of viral infection that was initially well-described as a candidate for adjunctive anti-neoplastic therapies (103, 104).

Individually, these biomarkers have poor operator characteristics, but when combined into a pauci- or multi-analyte test may begin to show improved discriminative performance (105) (See Table 2).

TABLE 2.

Candidate protein biomarkers

	Description	Limitations
Interleukin-6 (IL-6)	Prototypical cytokine involved in many pro-inflammatory states, associated with comparatively poor prognosis in the setting of sepsis and SIRS. Numerous IL-6 inhibitor therapies approved for auto-inflammatory conditions.	Broadly pro-inflammatory and nonspecific for infectious states.
Pro-adrenomedullin (Pro-ADM)	Serum stable pro-peptide of adrenomedullin. Proposed to function similarly to procalcitonin.	Primarily related to vascular dysregulation; no well-accepted cutoffs.
Pancreatic stone protein (PSP)	Glycoprotein, associated with sepsis severity. Has been studied to differentiate infectious states in burn patients.	Limited data, has not been shown to outperform existing single biomarkers.
Myxovirus resistance protein A (MxA)	Protein elevated in the setting of acute viral infections and myositis. Utilized in FebriDx point-of-care device.	Limited data as a single biomarker to differentiate infection. Associated with connective tissue autoimmune diseases.
sTREM-1	Cell surface receptor expressed on monocytes, macrophages, and neutrophils. Expressed in states of bacterial, fungal, viral, and parasitic infection, particularly sepsis. Associated with septic shock and poor malarial prognosis.	Poor ability to differentiate a specific infectious etiology of sepsis.
TNF-related apoptosis-inducing ligand (TRAIL)	Protein ligand cytokine produced by most tissues, elevated in viral infection compared to bacterial infection. Also studied as the target for cancer immunotherapy. Utilized in the MeMed BV test.	Limited data as a single biomarker for measuring the host response to infection.

Analyte panels as an alternative to single-biomarker measurements

Understanding the variation in levels of different biomarkers in both healthy and diseased states may prove pivotal to better diagnose infections through measurement of the immune response. One approach is to combine multiple immunologic analytes in a single assay to better define the underlying process of the host. In theory, using multiple complementary analytes combined into a larger diagnostic test allows for fingerprinting of the host response, improving the narrow or static view provided by a single biomarker.

Large-scale identification and quantification of proteins (proteomics) such as cytokines allows for researchers to select a subset of proteins linked to a given condition, quantify and model their relationship to each other, and then make a probability assessment about the likelihood of a condition from that sample. Additionally, the broad expansion of quantifiable nucleic acid amplification technologies has enabled nucleic acid identification as a new strategy for understanding the host response to infection, particularly RNA-based gene expression (transcriptomics) and epigenomic changes to DNA (epigenomics). This allows for a more upstream look at the central dogma of molecular biology—moving from a protein-based host detection strategy toward the detection of RNA and/or DNA changes (106).

The nuance of this approach depends greatly on the chosen targets of the assay and the algorithms developed to interpret their measurements. Assays may target any part of the immune response, as well as any upstream signaling that may occur. The fields of proteomics, transcriptomics, epigenomics, and differential leukocyte morphology can all play key roles in the development of innovative host response-derived diagnostics.

CONSIDERATIONS DURING DEVELOPMENT OF HOST RESPONSE-DERIVED TECHNOLOGIES

Advancing novel development of technologies to harness the host response to diagnose infection involves several integrated approaches. Regardless of the strategy chosen, there are several important considerations that underpin development.

New technologies should be designed to provide the best, most unambiguous information to clinicians that augments medical decision-making. In particular, understanding the context of clinical decision-making in which a test is deployed is crucial. For example, if the pretest probability of a condition or infection is not well-understood, then offering an ordinal result (i.e., low, medium, or high result) may not provide much clinical utility to a Bayesian decision-maker. In contrast, if a test result accurately aligns with clinical decisions, like stopping antibacterials for a likely viral pathogen, then it could be of high clinical value.

Host response-based diagnostic development has a unique pathway for development (See Fig. 1) as efforts should start with identifying a clinical question and then engineering a biomarker or host-response elements that accurately differentiate between those clinical groups of interest. The most appropriate diagnostic strategy may vary greatly depending on the patient setting, site of suspected infection, and current disease severity. It must also be robust and accurate despite variable types and levels of immunosuppression. Cost, ease-of-use, and turnaround time also have implications for how a test is used and interpreted.

Fig 1.

Host response-based diagnostic development.

Considerations uniquely important to host response-derived testing include the following:

1. Focus on shared canonical responses to pathogens or pathogen groups, such as bacterial, viral, or fungal infections.

2. For diagnostics that identify pathogen class (rather than a specific organism), tests need to target groups of organisms that support specific clinical decisions on antimicrobial use (such as providing routine antibiotics for community-acquired pneumonia).

3. Target diagnostic niches where pathogen-specific testing (a) is not available, (b) involves invasive sampling, (c) takes a prolonged period of time to result, or d) offers suboptimal performance.

4. For tests of prognosis rather than etiologic diagnosis, identification of elements of the early immune response to infection that correlate with later disease severity.

5. For all test targets, identification of conserved immune response elements that remain robust across heterogeneous populations, including variable iatrogenic or natural immunosuppression.

Identifying disease states with existing diagnostic challenges

One fundamental consideration is the disease state or clinical syndrome for which a test is designed. Three common infectious clinical scenarios in which host response-based testing can be deployed include sepsis, respiratory tract infections, and urinary tract infections. To optimize the likelihood of developing a successful product, investigators must choose a specific level of care and intended disease state to refine test performance and provide tailored clinical utility of a diagnostic.

The diagnostic uncertainty between sepsis and sepsis-mimickers, alongside poor outcomes associated with delayed or withheld antibiotics, highlights the need for improved clinical diagnostic technologies. Given the immune dysregulation associated with sepsis, measuring the host response has potential to accurately differentiate between sepsis and sterile systemic inflammatory response syndrome (SIRS). Rapid antibiotic administration following presentation of sepsis is associated with decreased mortality, with some estimates at 4% increase in mortality for each hour delay (107). Because of the urgency and importance of rapid antibiotic administration, a diagnostic test for sepsis must have exceptional sensitivity for bacterial infection, while still maintaining good specificity.

Additionally, gauging the response to a given therapy can be challenging in sepsis and related syndromes. Similar to serum lactate level monitoring in states of shock, a biomarker of the host response could play a pivotal role in describing a patient’s condition to clinicians in real-time (108). Currently available host-based biomarkers, such as CRP and PCT, have not shown sufficient accuracy to fill this clinical need (109).

In contrast to sepsis, a diagnostic test for respiratory tract infections would be most beneficial if able to accurately discern between (1) bacterial infection (2) viral infection, and/or (3) non-infectious causes of respiratory distress (110). A prognostic model that stratifies the risk of decompensation in pneumonia (rather than diagnosing it) might also be beneficial. However, ideal prognostic biomarkers would offer additive value (or improved performance characteristics) compared to existing clinical predictive algorithms such as CURB-65 and pneumonia severity index, which are well-established clinical support tools to predict mortality in community-acquired pneumonia (111).

Urinary tract infections (UTI) are very common, especially among women presenting to care at urgent care and emergency departments (112). Accordingly, a test to assist with UTI diagnosis must emphasize the ability to differentiate bacterial simple colonization vs invasive bacterial infection, as non-bacterial UTIs are uncommon and antibiotic stewardship considerations are paramount (see Table 3).

TABLE 3.

Ideal characteristics of host response-based testing for common infections^a

Clinical scenario	Determine bacterial vs viral	Colonization vs disease	Need to discriminate non-infectious causes	Prognosis and response to therapy
Sepsis	↑↑	↑	↑↑↑	↑↑↑
Respiratory tract infection	↑↑↑	↑↑	↑↑↑	↑↑
Urinary tract infection	↑	↑↑↑	↑	↑

^a Crucial: ↑↑↑; Important: ↑↑; Less important: ↑.

HOST RESPONSE-BASED DIAGNOSTICS CLEARED OR IN ADVANCED DEVELOPMENT

Overcoming the limitations of a single biomarker requires combining multiple analytes or a novel look at immune cellular function. Since the US FDA clearance of procalcitonin in 2017, there have been several new approaches and tests developed for diagnosis of infection by measuring the host response. Current US FDA-cleared diagnostic assays are discussed below.

MeMed BV

The MeMed BV test, which is US FDA-cleared, distinguishes acute bacterial vs viral infection by measuring three inflammatory proteins: CRP, TRAIL, and IP-10 (IFN-gamma-inducible protein 10). Through a machine learning algorithm, these protein measurements provide a score for the likelihood of bacterial or viral infection between 0 and 100 (low score suggesting viral infection, and high score indicating bacterial infection). This score is then put into one of five discrete categories: high likelihood viral, moderate likelihood viral, equivocal, moderate likelihood bacterial, and high likelihood bacterial. Although published FDA validation studies included patients with a variety of infectious syndromes, over 78% of enrolled patients had upper or lower respiratory tract infections.

In data submitted to the US FDA for 510(k) device clearance, the MedMed BV test showed increasing likelihood of a bacterial infection across MeMed BV score categories (113). In a secondary analysis using only adults with well-adjudicated bacterial or viral infections (indeterminate cases excluded) between March 2017 and October 2018, the MeMed BV test showed excellent sensitivity 0.981 and specificity 0.884 (114). However, given that almost a quarter of patients with the targeted respiratory syndrome were excluded from these calculations due to indeterminate adjudications (n = 101/415), the degree to which these performance numbers would reflect real-world use is unclear. Among children with suspected infection, the MeMed BV (run on the ImmunoXpert assay device) distinguished between bacterial and viral infections with sensitivity 0.938 and specificity 0.898, with 11.7% of outcomes reported as equivocal, using forced clinical adjudication as the reference comparator (115). The inclusion of an “equivocal” result improves the analytical performance and thus clinical certainty when acting on results but comes at the cost of limiting the number of subjects in whom the test provides an actionable result.

As these results were primarily gathered prior to the COVID-19 pandemic, it is not yet clear how these diagnostic data can generalize to modern respiratory viral epidemiology, which now includes SARS-CoV-2 as a major respiratory pathogen. Some early data suggest differential expression of the underlying analytes in COVID-19 compared to other respiratory viral infections that may affect performance (116). Separate from the diagnostic BV assay, however, a subsequent multicenter study of 349 patients with confirmed SARS-CoV-2 infection did show that a novel MeMed COVID19 Severity score (based on the same underlying biomarkers) was associated with a higher likelihood of severe COVID-19 once the diagnosis had been established (104).

This test run time is approximately 15 minutes and has been cleared to run on MeMed immunoassay devices (including the point-of-care MeMed Key) or the family of Diasorin Liason instruments (117). Clinical utility of this test is slated to be evaluated through a randomized controlled trial funded by the US Biomedical Advanced Research Development Authority (BARDA) (118).

FebriDx

The FebriDx device is a rapid, disposable point-of-care device cleared to determine the presence or absence of bacterial respiratory tract infection in outpatient or emergency department settings (119). This test uses a lateral flow assay to qualitatively assess elevations in C-reactive protein (CRP ≥2 mg/dL) and Myxovirus A (MxA ≥40 ng/mL). Given nonspecific elevations in CRP from both bacterial and viral infections alongside MxA elevation with viral infection, this test reports isolated positive CRP as a bacterial infection and all other scenarios (including positive CRP and MxA, isolated positive MxA, and both negative) as a non-bacterial result.

FebriDx performance was validated through enrollment of patients from October 2019 to April 2021 and ultimately included 496 participants with an assigned diagnosis, 73 (14.7%) of whom were diagnosed with respiratory bacterial infection. Compared to the reference standard of clinical algorithms followed by review adjudicated by a panel of experts, FebriDx showed a sensitivity 0.932 and specificity 0.884 for bacterial infection. Although it does not have FDA clearance to define viral-associated host immune responses, this test’s use of MxA does have a reported sensitivity 0.703 and specificity 0.884 to detect viral infection, with the same reference standard (120). In a separate study of 244 patients presenting to the ED with suspected respiratory infection in the Netherlands from March 2019 to November 2020, the FebriDx test showed sensitivity 0.87 for bacterial infection and specificity 0.67 with a sensitivity 0.49 and specificity 0.94 for viral infection (121). Notably, the performance of FebriDx has also been validated in studies that included SARS-CoV-2 infection, increasing the potential utility of this test in the post-COVID era (122).

Overall, FebriDx provides a simple point-of-care test with rapid results and test characteristics that may support changes in clinical management of acute respiratory illness in the outpatient setting. There are as yet no extensive data describing the performance in inpatient or intensive care contexts or in syndromes other than respiratory tract infections, so whether there are additional clinical situations where such a test may be useful remains to be demonstrated.

SeptiCyte

The SeptiCyte tests (SeptiCyte LAB and SeptiCyte RAPID) have been developed and cleared for use in patients admitted to intensive care units (ICUs) within the first day of suspected sepsis. The SeptiCyte LAB test was originally cleared by the US FDA in 2017 for differentiation of suspected sepsis using four transcriptional targets of genes involved in early sepsis or bacterial clearance (CEACAM4, LAMP1, PLA2G7, and PLAC8). This test required manual RNA extraction and had a typical resulting time of 6.5 hours (123). The SeptiCyte RAPID was then cleared in 2021, which used automated extraction and analyzed PLA2G7 and PLAC8 with a runtime of 65 minutes (124).

Clinical performance of SeptiCyte LAB was assessed through two prospective studies enrolling patients admitted to ICUs that met two or more SIRS criteria. These studies showed an increasing likelihood of sepsis as score 0–10 increased and then grouped patients into Bands 1 through 4 for more discrete sepsis description. Scores were then compared to a reference standard of retrospective clinical adjudication, which determined if a participant had non-bacterial SIRS, sepsis, or an indeterminate diagnosis. Bands 1 through 4 were grouped in order of increasing likelihood of sepsis. Band 1 represented a low (<16%) likelihood for sepsis, intermediate likelihood of sepsis for bands 2–3, and sepsis was >82% likely for band 4. SeptiCyte RAPID was then tested retrospectively on the same cohort and a small number of additional prospective specimens and found to be substantially similar. The SeptiCyte RAPID device instead used a score range 0–15 but retained the Band 1–4 scoring system to denote increasing likelihood of sepsis. Notably, the SeptiCyte LAB test showed variability based on the participant’s race, with Black race/African ancestry patients observed to have higher scores, and only the highest classification (band 4) associated with a higher likelihood of sepsis in these patients.

These assays are currently the only FDA-cleared gene expression devices to better differentiate sepsis based on the host response. However, they also highlight several challenges with the development, validation, and implementation of host response-derived diagnostics. The consequences of withholding early antibiotics from a patient with sepsis are high, necessitating an extremely high NPV for bacterial sepsis in order to allow biomarker-driven non-antibiotic-containing approaches. Thus, whether the reported NPV of 0.91 for a “Band 1” result of Septicyte RAPID is sufficient to withhold antibiotics in patients with signs and symptoms of severe sepsis remains a topic for debate (125).

This decision-making is in contrast to the clinical decision tree following a “bacterial” or “viral” result of diagnostics listed above. Thus, it is less clear how exactly clinicians should alter the management of patients across the variety of possible sepsis risk scores, particularly regarding use of antimicrobials. For instance, it might be suggested that patients in the lowest-risk category for bacterial sepsis have antibiotics withheld when they would otherwise have been given based on clinical suspicion for sepsis (although careful studies of outcomes are needed if this approach is utilized). However, what changes in management (if any) are appropriate for patients in whom sepsis is already suspected (and thus who typically would already receive antibiotics) when they have higher probability score results? For such stratified risk-based scores (especially in severely ill patients), clinical utility studies examining optimal approaches to various test results will be necessary to optimally define the best (and safest) means of implementing these promising assays. Finally, the reported difference in SeptiScore based on race highlights the importance of choosing transcriptional or genetic targets that retain similar performance across heterogeneous populations.

The SeptiCyte transcriptomic system is also being studied for discrimination of bacterial and viral infections but is not yet FDA-cleared for this purpose. SeptiCyte VIRUS and SeptiCyte TRIAGE are independent gene expression signatures to evaluate the presence of viral infection or bacterial infection, respectively, and designed to improve diagnostic precision for febrile patients without a microbiologic diagnosis, especially when utilized in a combinatorial approach. These signatures demonstrated a negative predictive value 0.97 for bacterial infection given low pretest probability and a negative predictive value 0.86 given high pretest probability in emergency departments, superior to WBC or CRP alone (126).

IntelliSep

Although substantial work describing the host response to infection relates to protein or transcriptional biomarkers, the IntelliSep test and Cytovale system instead describe immune activation by measuring the deformability of leukocytes (127). Cleared by the US FDA in December 2022 for patients with suspicion for sepsis, the IntelliSep Test uses microfluidics-based testing to quantify the viscosity and elasticity of leukocytes (128). Grounded in the understanding that activated leukocytes from sepsis have different biophysical properties than those of healthy controls, the IntelliSep test gives a score (called the IntelliSep Index) that is then categorized into three ranked ordinal bands of sepsis likelihood (low, moderate, and high likelihood). This risk stratification aims to understand which patients might otherwise need rapid triage tailored for sepsis against patients who either meet SIRS criteria but have alternative etiologies, such as congestive heart failure or a gastrointestinal bleed.

The IntelliSep test was validated through a prospective multicenter study at four US Emergency Departments from May 2021 to January 2022, enrolling 599 adults with suspected infection presenting to care, later adjudicated on whether sepsis was present using Sepsis-3 criteria. Compared to this reference standard of retrospective forced adjudication, the IntelliSep results showed correlation with sepsis likelihood. The lowest risk stratification score cohort showed 11.1% likelihood of sepsis compared to the highest risk cohort, which showed 49.4% likelihood for sepsis.

Different from other technologies described, the IntelliSep device does not attempt to differentiate by pathogen type or predict the presence of bacterial infections. Instead, as a rapid ED-centered tool, such a test could identify subjects at greatest risk for decompensation or severe outcomes. One potential benefit of this result would be to shorten the prolonged wait times for ED by identifying patients at low risk for progression who may be safely discharged home or to guide early treatment when sepsis is suspected. Given the presence of sepsis in 11.1% in even the lowest result band, a band 1 (low likelihood) result alone is likely not sufficient to rule out sepsis or withhold antibiotics when otherwise indicated. Careful implementation strategies and clinical utilization studies will be needed to determine the optimal means of utilizing the results of these tests to help critically ill patients.

Monocyte distribution width

The monocyte distribution width (MDW) describes peripheral blood monocyte cell volume variability to predict the development and severity of sepsis (129). As monocytes engage infection in the inflammatory response of sepsis, they play key roles in phagocytosis of pathogens, antigen presentation, and cytokine production (130). These functional changes have corresponding increases in cell volume, which can be quickly measured and reported. Using the Beckman Coulter Hematology Analyzer routinely used for complete blood and differential counts, the MDW test provides a numeric value, with results > 20.0 being associated with a higher likelihood of sepsis.

In prospective studies in emergency departments, MDW has been identified as a possible screening tool for early triage of sepsis (131, 132). Higher MDW values corresponded to increasing probability of sepsis compared to non-sepsis presentations (such as sterile SIRS and non-severe infection). Moreover, MDW was found to augment diagnosis of sepsis, both including and independent of SIRS or qSOFA criteria (133). Overall, MDW has shown performance characteristics similar to or exceeding those of procalcitonin or CRP in aiding the diagnosis of early sepsis in emergency departments, as described in a meta-analysis of 18 head-to-head studies at several international sites (134, 135). Given the ease-of-use within existing laboratory systems, low-cost, and moderate performance characteristics, MDW may have a role augmenting sepsis triage. However, more implementation studies are needed to understand the well-defined role it may have in antimicrobial discontinuation or clinical management (See Table 4).

TABLE 4.

FDA-cleared multianalyte host response-derived diagnostics for infectious diseases

Name of test	Analytical target(s)	Intended population	Interpretation of results	Device type, result time
MeMed BV	Proteins: CRP, TRAIL, and IP-10	Patients with suspected bacterial or viral infection, symptoms < 7 days	Ranked ordinal scale for bacterial or viral infection, includes equivocal result (8%–12% of samples)	Immunoassay, 15 minutes
FebriDx	Proteins: CRP and MxA	Outpatient, suspected acute respiratory infection and symptoms < 7 days, ages 12–64	Bacterial or non-bacterial	Disposable lateral flow assay, 10 minutes
SeptiCyte	mRNA: host biomarkers PLA2G7 and PLAC8, (LAMP1, CEACAM4 also only in LAB version)	Critically ill adult patients on their first day in ICU with suspected sepsis.	Ranked ordinal scale of four bands, with increasing likelihood of bacterial sepsis (vs noninfectious SIRS)	Real-time PCR 6.5 hours for SeptiCyte LAB 65 minutes for SeptiCyte RAPID
IntelliSep	Leukocyte biophysical properties (aspect ratio and visco-elastic inertial response)	Adults with signs and symptoms of infection presenting to emergency departments	Ranked ordinal scale of three bands, with increasing likelihood of sepsis (low, moderate, and high likelihood)	Microfluidics device, 10 minutes or less
Monocyte Distribution Width	Monocyte distribution width	Early sepsis, patients presenting to ED	Numeric value interpreted as MDW >20.0 predicted the higher risk of sepsis	Beckman-Coulter Hematology Analyzer, 40 seconds per sample

NEW AND DEVELOPING APPROACHES

In addition to the several US FDA-cleared host response-based tests for infectious diseases, there are several products currently being enrolled in clinical trials or otherwise in late-stage product development as of late 2024. Key foundational work on gene expression signatures and efforts in a late stage of development are described here.

Key foundational gene expression signatures

Gene expression profiling has been proposed to be a powerful tool to discriminate bacterial versus viral infections. Some of the first published gene expression signatures include a 35-gene signature developed by UT-Southwestern researchers comparing differential gene expression in children with influenza A, Gram-negative, or Gram-positive infections (136). Additionally, a team of Duke University researchers used a 30-gene signature from peripheral blood mononuclear cells to predict if a subject had influenza A infection, bacterial infection, or no infection (137). Further published efforts using gene expression to differentiate bacterial versus viral respiratory infection include the use of nasal transcriptional signatures by researchers at Washington University at St. Louis (138). Subsequent gene expression signatures published by researchers at University of Rochester also aimed to differentiate bacterial versus viral infection in adults hospitalized with respiratory tract infections (139, 140). Signatures developed by a team at Stanford University particularly highlight the need for sepsis-related research and exploration (141). Outside of bacterial versus viral profiles, published gene expression profiles for infants with RSV from researchers at The Ohio State University may predict disease severity (142). Numerous gene expression signatures have also been published to predict the likelihood of Mycobacterium tuberculosis (MTB) infection, as well as to distinguish latent versus active MTB infection (143–145).

Biofire & Biomeme (host response test [HR-B/V])

Some of the earliest works to develop a gene expression signature to differentiate bacterial versus viral infection, performed by Duke University researchers, have subsequently been tested on multiple diagnostic platforms (130). Within a cohort of 623 participants at four emergency departments, BioFire FilmArray platform tested 45 mRNA targets and was able to differentiate bacterial or viral infection better than procalcitonin, when compared to a reference standard of clinical adjudication (146).

Using a refined algorithm on a subset of this mRNA gene signature, the Biomeme Host Response Test (HR-B/V) also used two separate models to generate probabilities to discriminate bacterial from non-bacterial and viral from non-viral febrile presentations. The test was initially intended to be run on the Franklin ISP (Integrated Sample Prep), a proprietary device designed to provide near point-of-care testing up to 27 signal multiplex RT-PCR on blood or other samples. The Biomeme HR-B/V test is designed to run on whole or capillary blood samples of patients presenting to care with concern for systemic infection and to differentiate between bacterial and viral causes (147).

Cepheid (Xpert MTB host response [MTB-HR])

Cepheid GeneXpert RT-PCR-based systems are widely distributed worldwide and indispensable to pathogen-based detection efforts of Mycobacterium tuberculosis (MTB), with subsequent detection assays for numerous pathogens including SARS-CoV-2, Ebola, and Mpox (148). Current efforts to identify active MTB infection are pathogen-based and limited by cost, ability to produce sputum, and poor test sensitivity, particularly in people living with HIV. Cepheid is now in late-stage development and testing of a blood-based host-response assay that uses three genes differentially expressed in the setting of active MTB infection (including children), intended to supplement existing culture- or PCR-positive testing strategies (149, 150).

Inflammatix, Inc (TriVerity acute infection and sepsis)

The TriVerity Acute Infection and Sepsis Test System is a device developed by Inflammatix, Inc that uses a 29-gene target (mRNA) assay to assist with diagnosis of sepsis by differentiating both bacterial or viral causes, as well as predicting the severity of illness (151). Although not currently FDA-cleared, the intended use of the TriVerity device is broad, in that it aims to clinically differentiate the presence of infection, if an infection is bacterial or viral, and the severity of sepsis infections—within a single integrated test (152). As of late 2024, this device was mid-enrollment for a 510(k) validation study with subsequent plans to apply to US FDA clearance (153).

EMERGING IMPLEMENTATIONS

Describing the host response to infection can also be leveraged to better diagnose the infection of difficult-to-isolate pathogens, more rapidly identify infection of highly transmissible pathogens, and differentiate between colonization and invasive infection with more common microbes. Select emerging or needed areas for innovation are described here.

Generalizing to a global context

Although a promising technology, gene expression profiles may lack consistent test characteristics in more genetically and geographically diverse populations. Results based on patients from medical centers in North America or Western Europe may not be generalizable to environments with a higher prevalence of tuberculosis, malaria, or HIV; different epidemiology of respiratory pathogens; and higher burden of intracellular bacterial and rickettsial disease (154).

The use of large retrospective genomic data sets (including Gene Expression Omnibus) may offer one approach to address this concern, as exampled by an eight-gene signature developed from 69 data sets with generally positive characteristics, including sensitivity 0.902 and specificity 0.859, distinguishing bacterial versus viral infections across diverse populations and disease states (155). Alternatively, existing gene expression signatures may be derived from traditional research cohorts but should then be validated in clinically and epidemiologically distinct environments, as has been done with existing signatures in South Asian populations of Sri Lanka (156).

Vector-borne disease detection

Lyme disease, transmitted by Ixodes ticks infected with the bacteria Borreliella burgdorferi, is the most common vector-borne disease in the United States (157). Given the fastidious nature of this bacteria and poorly performing diagnostic tests available, the acute diagnosis of Lyme disease is made through relevant epidemiology and an appropriate clinical syndrome. Convalescent serology is occasionally helpful in confirming prior diagnosis but is neither sensitive nor specific for an acute diagnosis and rarely useful in guiding antimicrobial therapy. Controlled murine models have shown differential gene expression in synovial fluid when infected with Borreliella burgdorferi (158). A 31-gene Lyme disease classifier has been published to augment the diagnosis of patients early in the disease course, before traditional serology becomes positive (159). More accurately describing the host response in the setting of Lyme disease and developing an assay to run on existing platforms would be beneficial to assist with the current dearth of rapid point-of-care diagnostics for this common presentation. Some future efforts to longitudinally follow outcomes and gene expression of individuals with suspected Lyme disease may help with these efforts (160).

Spotted fever group rickettsial diseases also may be difficult to diagnose with traditional detection techniques. Some research suggest that spotted fever group rickettsial illnesses have differential gene expression from other similarly presenting syndromes in Sub-Saharan Africa, such as malaria or blood stream infections, which could be leveraged into improved diagnostics (161). Additional early work has described differential serum protein and gene expression for scrub typhus among febrile syndromes in Sri Lanka, which may prove useful in enhancing diagnostic innovation between scrub typhus and dengue fever (162, 163). One prime example is the development of a 20-gene signature able to accurately predict progression and prognosis of dengue fever in a diverse cohort (164).

Additionally, despite malaria accounting for over 600,000 deaths annually, there are no widespread or cleared biomarkers to assist with detection or disease severity (165). Many biomarkers have been studied, including CRP, PCT, and IL-6, with soluble triggering receptor expressed in myeloid cells-1 (sTREM-1) showing the best performance for predicting prognosis for malaria (166). Other more experimental host-response studies have suggested that the genetic and epigenomic responses to malaria may explain some variability in severity, including cerebral malaria (167).

Differentiating Clostridioides difficile infection and colonization

Clostridioides difficile infection (CDI) is a leading cause of morbidity in healthcare settings, particularly during prolonged hospital stays and after antibiotic exposure (168). The diagnosis of CDI typically involves establishing the presence of C. difficile in stool through PCR testing and then correlating with clinical syndrome and/or toxin expression testing. PCR testing can be nonspecific and reflect simple colonization, but both clinical phenotyping and toxin detection assays can be insensitive, leaving clinicians in a diagnostic quandary. At least one prospective study of hospitalized patients has demonstrated differential blood and stool cytokine expression patterns between colonization and invasive disease (169). Refining these results into an effective test that uses host response-derived patterns to identify true infection may help improve this clinical dilemma.

Sepsis subtyping and differential management

Understanding and stratifying sepsis based on the clinical phenotype and immune endotype is also an emerging area of research and development for diagnostics, particularly when paired with tailored immunomodulating agents or anticoagulants (170). Endotyping, or the identification of characteristic transcriptional patterns within a clinical syndrome, may allow for more targeted clinical interventions (171). This novel approach includes an effort by Endpoint Health to develop diagnostics that report hypercoagulable sepsis endotypes and then deploy the use of companion therapeutics (172).

Similarly, SphingoTec has performed early-stage host response-based biomarker-guided trials to investigate the role of adrecizumab, an investigative agent targeted to regulate endothelial dysfunction (173). This primary study enrolled patients in septic shock with elevated adrenomedullin levels, which triggered the use of adrecizumab, and demonstrated a favorable safety and efficacy profile. However, larger trials are still needed to validate these findings.

Asep Inc. is also in early stages of product development using host response-derived gene expression signatures to rapidly detect sepsis subtypes at initial presentation, which could later prove effective in differential management by endotype (174).

Cepheid Inc. has also published preliminary data on a six-gene signature to run on GeneXpert devices to differentiate between bacterial and viral infections, as well as a partnership between Danaher Corporation and University of Oxford to develop sepsis-based gene signatures (175–177).

Similarly, while identifying patients with early sepsis is critical, identifying patients who are at high risk for sepsis based on factors of the host response following an insult (such as surgery or non-severe infection) may help with triage and expedited work-up. One such effort is being studied by the biotech company Presymptom Health Ltd, which is developing a gene expression host response-based testing in development to predict sepsis up to 72 hours before clinical recognition (178).

Pre-symptomatic disease detection

Early detection of respiratory viral infections is crucial to minimize spread and appropriately triage respiratory viral disease. Early peak transmission by pre-symptomatic and asymptomatic individuals is a hallmark element of both influenza and COVID-19 disease (179, 180). Therefore, detecting impending viral infection at the earliest possible time is of great potential impact. However, studies of patients during the timeframe between exposure and onset of clinical illness are difficult and expensive to perform, and therefore data from these early timepoints are sparse.

When a novel respiratory virus or highly contagious bacteria capable of pandemic spread emerges, the availability of a host response-derived assay may prove critical. Specific alterations to RNA/DNA sequences or antigens could make existing detection methods obsolete, inaccurate, or difficult to access for emerging pathogens, as was seen early in spread of SARS-CoV-2 (181). In contrast, a host response-derived test may accurately detect a broad range of respiratory viral pathogens while being agnostic to specific alterations of a single pathogen, critical to future pandemic preparedness efforts while awaiting the development of pathogen-specific NAAT or antigen tests (182, 183).

Substantial data exist stating that early detection of viral infection, even during the pre-symptomatic phase, is possible through host response-based approaches. Conserved genomic responses can be used to correctly classify individuals with a broad array of naturally acquired respiratory viral infections well before the time when they would present for clinical evaluation and even before detectable viral shedding begins for most subjects (184, 185). The fact that this holds true across as many as nine different respiratory viruses, each with variable incubation, clinical progression, and duration, speaks of the potential power of a host response-based approach to early detection.

The diagnosis of emergent pathogens with high case fatality may also benefit from host response-based test development. In the case of Ebola disease and other related hemorrhagic fevers, early infection may be characterized by biomarkers, early IgM serologic markers, and certain physiologic responses (186). Better defining the early host response to infection, particularly if it can be translated to point-of-care tests using existing local technologies such as GeneXpert platforms, may be beneficial in early identification, treatment, and containment of outbreaks.

Fungal disease

Nearly all currently cleared and late-stage development of host response-based testing is targeted toward management of bacterial sepsis and/or differentiating bacterial versus viral infection. However, fungal pathogens also play a significant role in overall sepsis epidemiology, with invasive Candida species being common (187). Early-stage research efforts are underway to better understand differential host transcriptional patterns in the setting of candidemia, with neutrophil activation and heme biosynthesis noted to play a unique role compared to other causes of sepsis (188). The mainstay of sepsis management includes antibacterial therapy but less often antifungal therapy, leading to delays in appropriate antimicrobial therapy when fungal infection is ultimately detected. Incorporating a result for fungal infection within bacterial and viral host response-based testing for sepsis could help bridge this gap in clinical diagnostics. Additionally, traditional diagnostic approaches for non-sepsis-associated invasive fungal infections, such as Aspergillus or Mucorales infection, are equally limited, and host response-based approaches to these challenging diagnoses are another growing area of interest and need.

Physiologic response sensors

While markedly different from classical biomarkers, understanding the physiologic host response may provide personalized triage and care delivery earlier in the course of infection. Novel approaches include using smartwatches and other biometric monitoring devices to identify early changes to heart rate, peripheral blood oxygenation, skin temperature, and sleep patterns, while identifying pre-symptomatic infection or high risk for disease progression (189).

In controlled influenza challenge studies, smartwatches that measure the heart rate and heart rate variability have shown promising results in predicting disease 24–48 hours prior to symptom onset (190). In both real-world and controlled viral infection, early and pre-symptomatic viral infection has also demonstrated many shared transcriptional patterns (191). While both techniques separately hold promise in the early diagnosis and control of viral disease, a combined approach should be studied as an even more effective public health strategy.

Exhaled volatile detection

Detection of on-breath volatile compounds for the early diagnosis of infectious diseases is an emerging technology that may complement more advanced host response diagnostics. Volatile organic compounds (VOCs) and metabolites may be reflective of the early host cell and microbial interaction, especially for respiratory tract infections (192). Detecting and fingerprinting thousands of different exhaled biomarkers in the setting of infection has been proposed as a complementary detection method (193, 194). This strategy is in early development by companies such as UK-based Owlstone Medical’s Breath Biopsy Device, which has also been studied for detection and categorization of non-infectious hepatic and pulmonary disease (195). Additional work with devices such as the Aeonose has shown initial proof-of-concept for using VOCs in detecting M. tuberculosis and SARS-CoV-2 infections (196, 197). Using this technology as a complementary approach for detection and early management of outbreaks could provide utility in congregate settings such as military barracks, schools, and healthcare facilities, but further study on development and implementation of such devices is needed.

Artificial intelligence tools for precision diagnosis

Understanding the relative importance of innumerable data points in the electronic medical record (EMR) can be a daunting task for clinicians. Leveraging artificial intelligence (AI) tools to integrate both biological data, such as traditional protein biomarkers, with physiologic data may provide real-time insight that augments clinical decision-making (198).

FDA authorization of Prenosis, Inc. ImmunoScore represented first-time authorization for use of AI in a sepsis diagnostic tool (199). The company reports this test uses 22 predetermined inputs from the electronic medical record and classifies into one of four categories of risk for sepsis. As with non-AI-based technologies for sepsis determination, use of low- to high-risk outputs needs to be partnered with clear paths of Bayesian clinical decision-making when pretest probability is well-understood. Accordingly, although use of AI may appear to augment precise diagnosis of infections, appropriate clinical implementation studies are needed to define their role in practice.

CONCLUDING REMARKS

Despite advances in molecular microbiology and other techniques, diagnosing many infections in the acute clinical setting remains a significant challenge, with severe and profound effects on treatment and other clinical management decisions. In particular, pathogen-directed testing may fail to identify fastidious pathogens, poorly discriminate colonization from invasive disease, and be prone to contamination or detection of clinically insignificant pathogens. As detailed above, there is an increasingly impressive body of data demonstrating that tests which harness the host response to infection can augment the diagnostic armamentarium clinicians use to quickly determine information regarding the etiology and prognosis.

Single protein-based biomarkers, including CRP and PCT, remain the most utilized biomarkers of the host response to diagnose infection. However, these and many other single-analyte assays have limitations in their accuracy and applicability. The expanding field of multi-analyte diagnostics, built on proteomics, gene expression, immune cell morphology, and others, is increasingly demonstrating the promise to improve the clinician’s ability to diagnose infection in the acute setting. Notably, appropriately applying these tests to immunocompromised patients remains difficult, as they were often excluded from clinical trials. Challenges remain in determining optimal biomarkers for some specific clinical syndromes and tasks, but many of these host response-based tests are either available or already far into the translational diagnostic development pathway. As the field continues to mature, the promise for an array of novel host response-derived biomarker-based approaches to diagnosing acute infectious diseases appears to be strong.

Biographies

E. Wilbur (Will) Woodhouse is a physician in Infectious Diseases at Duke University Medical Center and Durham Veterans Affairs Medical Center. After studying Public Policy at Duke University, he received his Doctor of Medicine and Master of Public Health from University of North Carolina at Chapel Hill. He then completed Internal Medicine Residency at Vanderbilt University Medical Center in Nashville, TN. He is completing his fellowship in Infectious Diseases at Duke University. His interests include innovation and equitable access to novel diagnostics for infectious diseases, including those that harness the host response, alongside novel strategies to engage and identify patients in highest need of those diagnostics. He is also engaged in community outreach efforts for diagnostics and retention in care for viral hepatitis.

Micah T. McClain is an Associate Professor of Medicine, Infectious Diseases physician and Immunologist at Duke University and the Durham Veterans Affairs Medical Center. He obtained his BS in Microbiology, PhD in Immunology and MD all from the University of Oklahoma, and completed Internal Medicine Residency and ID Fellowship at Duke University. His research focuses on analyzing the host response to infection as seen through the lens of proteomic, transcriptional, and epigenetic signals in peripheral blood. This research is then translated to develop novel host biomarker-based diagnostics for respiratory illness, rickettsial diseases, febrile illness in travelers and Department of Defense personnel, fungal infections, and diagnostic and severity prediction tools for infections and other inflammatory states (such as organ rejection) in immunocompromised adults.

Christopher W. Woods is the Executive Director of the Hubert-Yeargan Center for Global Health; Director of the Center for Infectious Disease Diagnostic Innovation (CIDDI); professor in the Departments of Medicine and Pathology at Duke University; and an adjunct professor in the Emerging Infections Program at the Duke-National University of Singapore Graduate Medical School. Clinically, he serves as Chief of Infectious Diseases Durham VA Health System. Dr. Woods is a co-founder of Predigen, Inc., and currently serving as Acting Chief Medical Officer for Biomeme, Inc. Dr. Woods has a particular interest in the diagnostic capacity in the developing world and the epidemiology of emerging and re-emerging infectious diseases. As an infectious diseases clinician and medical microbiologist, his approach to host genomic response has been called a paradigm shift in the field of infectious disease diagnostics.