Biomarkers for sepsis: more than just fever and leukocytosis—a narrative review

1A biomarker describes a measurable indicator of biological status in normal and pathogenic processes. It may be helpful as a theranostic for identifying suitable patients for therapeutic intervention and titrating the degree and/or duration of intervention. A biomarker should be accurate and reproducible. In the ideal scenario, the biomarker (or combination of biomarkers) should offer both high specificity and sensitivity for diagnosing a condition, but either alone may be adequate as a 'rule-in' or 'rule-out' test.

2Sepsis represents a dysregulated immune response to infection that leads to organ dysfunction. Host response biomarkers play a critical role in diagnosis, early recognition of organ dysfunction, risk stratification, prognostication, and patient management, including antibiotic stewardship. Biomarkers may also be helpful for trial enrichment to identify suitable patients and/or risk categorization for an intervention. A wide range of biomarkers, measured by a host of different technologies, are being investigated to discriminate a systemic inflammatory response syndrome (SIRS) rapidly, which is an excessive defensive body's response to a harmful stressor (for example, infection, trauma, surgery, acute inflammation, ischemia or reperfusion, or cancer) or early identification of infection-triggered organ dysfunction (sepsis). Also, the quick sepsis related organ failure assessment (qSOFA) is intended to raise suspicion of sepsis and encourage additional action; although, qSOFA is not a substitute for SIRS. These biomarkers include measurement of acute-phase proteins, cytokines, chemokines, damage-associated molecular patterns (DAMPs), endothelial cell markers, leukocyte surface markers, non-coding RNAs, miRNA, and soluble receptors, as well as metabolites and alterations in gene expression (transcriptomics). Biomarkers may help stratify septic patients into biological phenotypes, for example, hyperinflammatory versus immunosuppressive. Biomarkers can also be used to identify gut permeability, blood–brain barrier (BBB) permeability, probability of hospital readmission, and longer-term outcomes.

3The causative pathogen replicates and releases its constituents such as endo- and exotoxins, and DNA. These constituents are designated pathogen-associated molecular patterns (PAMPs). PAMPs are recognized by both pattern-recognition receptors (PRRs) and non-PRRs, which are essential components of the immune system. PRRs include several families, including Toll-like receptors (TLRs), nucleotide-binding oligomerization domain-like receptors (NOD)-like receptors (NLRs), a retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs), C-type lectin receptors (CLRs), and intracellular DNA-sensing molecules. Non-PRRs include receptors for advanced glycation end products (RAGE), triggering receptors expressed on myeloid cells (TREM), and G-protein-coupled receptors (GPCRs). The sensing of PAMPs by immune cell receptors triggers a cascade of signaling pathways that activate multiple transcription factors to promote the production and release of pro- and anti-inflammatory mediators such as acute-phase proteins, cytokines, chemokines, as well as antimicrobial peptides, which are needed to eliminate the invading pathogen.

4The host immune response and pathogen virulence factors will both trigger cell injury and/or induce cell stress. These results in the release of endogenous molecules (DAMPs), exacerbating the inflammatory response. DAMPs are recognized by the same immune receptors that recognize PAMPs. Many DAMPs have been identified, with some currently used as inflammatory biomarkers. Examples include proteins and cellular molecules related to nucleic acids, such as heat shock proteins (HSPs), the high mobility group box 1 (HMGB-1), and members of the S100 family. The immune response may induce vascular endothelial damage disrupting tight junctions (T.J.), increasing gut permeability, and potentially facilitating translocation of pathogens and/or their PAMPs from the gut to the bloodstream and lymphatics, thereby amplifying the systemic inflammatory response. In addition, an increase of BBB permeability allows circulating immune cells to enter the brain, triggering or exacerbating glial cell activation. These events could trigger an intense and excessive host response activating coagulation and fibrinolytic systems, activating or suppressing hormonal, bioenergetic, and metabolic pathways, and inducing macro- and microcirculatory changes with a net result of multiple organ dysfunction. In the past few decades, researchers have studied each inflammatory response stage during SIRS, sepsis, and septic shock, metabolites associated with inflammatory cascades, and cellular components that could be used as biomarkers. These biomarkers could help identify endothelial damage, intestinal permeability, organ failure, BBB breakdown and predict rehospitalization, short- and long-term mortality, and cognitive consequences in survivors.

5For this narrative review, we addressed the question, "Do biomarkers in patients with sepsis or septic shock predict mortality, MODS, or organ dysfunction?" Studies were identified by searching PubMed/MEDLINE (National Library of Medicine) databases for peer-reviewed journal articles published until October 2021. The abovementioned databases were searched with the following combinations of keywords: ("inflammatory response syndrome" OR "SIRS" OR "sepsis" OR "septic shock" OR "sepsis-associated encephalopathy" OR "SAE") AND ("markers" OR "biomarkers" OR "biological markers" OR "biological measures" OR "molecular predictor"). We omitted review articles, in vitro studies, and animal studies.

6The humoral innate immune response consists of multiple components, including fluid phase pattern recognition molecules (PRMs) and the complement system. PRMs include C-reactive protein (CRP), serum amyloid P component (SAP), and pentraxin 3 (PTX-3). The rise in CRP level is primarily induced by interleukin (IL)-6 and IL-1β acting on the gene responsible for CRP transcription during the acute phase of an inflammatory process. CRP is a pentameric acute-phase reactant protein whose conformation facilitates the ability to trigger complement activation and activate platelets, monocytes, and endothelial cells. Furthermore, CRP is one of the most widely used and investigated biomarkers. A prospective multicenter cohort study followed 483 adult patients who survived hospitalization for sepsis for up to one year. IL-6, high-sensitivity C reactive protein (hs-CRP), soluble programmed death-ligand 1 (sPD-L1), E-selectin, and intercellular adhesion molecule 1 (ICAM-1) were evaluated at five-time points during and after hospitalization. A comparison was made between a phenotype with hyperinflammation (high levels of IL-6 and hs-CRP) and a phenotype of immunosuppression (high sPD-L1 levels). Compared with a normal phenotype, both hyperinflammation and immunosuppression phenotypes had higher 6-month hospital readmission rates and 1-year mortality rates, both all-cause and attributable to cardiovascular or cancer.

7Pentraxin (PTX-3) is secreted by macrophages, dendritic cells, macrophages, fibroblasts, mesangial cells, and glial cells under pathogen or inflammatory stimuli. Plasma PTX-3 was assessed on days 1, 2, and 7 in 958 patients with sepsis or septic shock included in the Albumin Italian Outcome Sepsis (ALBIOS) study. The researchers assessed a possible association between PTX-3 levels and clinical severity, organ dysfunction, treatment, and mortality within 90 days. PTX-3 levels were elevated at the onset of sepsis and increased with illness severity and the number of organ dysfunctions. PTX-3 levels decreased between days 1 to 7, but this was less prominent in patients with septic shock. In a prospective observational analysis, PTX-3, IL-6, procalcitonin (PCT), and lactate combined showed excellent performance in predicting 28-day all-cause mortality among patients diagnosed with sepsis or septic shock and superior to the Sequential Organ Failure Assessment (SOFA) score.

8In a prospective pilot study of markers of complement activation in sepsis, higher C4d (3.5-fold), factor Bb (6.1-fold), C3 (0.8-fold), C3a (11.6-fold), and C5a (1.8-fold) levels were seen compared with healthy volunteer controls. In another study of 49 sepsis patients, 34 developed disseminated intravascular coagulation (DIC), and eight died. Patients with DIC had lower C3 levels and higher SC5b-9 levels. On stratifying by SC5b-9 quartile (ng/mL: low: < 260, moderate: 260–342, high: 343–501, highest: > 501), coagulation parameters were most deranged in the highest quartile with prolonged thrombocytopenia and higher mortality (33%).

9The activation of PRRs culminates in the stimulation of transcription factors resulting in the expression and secretion of proinflammatory cytokines, including tumor necrosis factor-alpha (TNF-α), IL-1-β, IL-6, and interferons (IFNs). These inflammatory mediators are required for host defense against pathogens and activation of the adaptive immune response. A retrospective study evaluated a broad panel of cytokines and found IL-1β, IL-6, IL-8, MCP-1, IL-10, and plasminogen activator inhibitor 1 (PAI-1) levels were increased in the acute phase of sepsis in both critically and non-critically ill patients. In addition, levels of IL-10 (days 1, 2, and 4), IL-6 and PAI-1 (days 2 and 4), and IL-8 (day 4) increased in critically ill patients compared to non-critically ill. In summary, hs-CRP, IL-6, and PAI-1 circulatory levels may have utility in stratifying a hyperinflammatory patient phenotype.

10DAMPs are endogenous danger molecules released from damaged or stressed cells. These molecules activate the innate immune system through interaction with PRRs. DAMPs contribute to the host defense but can also promote pathological inflammatory responses. Calprotectin, a protein found in the cytosol of neutrophils and macrophages, is released under cell stress or damage. In a mixed population study, plasma calprotectin levels were higher in sepsis than in trauma patients and other medical conditions. Calprotectin levels were higher in patients who did not survive for 30 days. Plasma PCT did not differ between the groups or as a prognosticator of the outcome. Receiver operating characteristic (ROC) analysis, used as a sepsis biomarker, had a higher area under the curve (AUC) value for calprotectin (AUC: 0.79) compared to PCT (AUC: 0.49).

11A prospective study evaluated IL-6, HMGB-1, and neutrophil gelatinase-associated lipocalin (NGAL) in 14 septic patients and 16 patients without sepsis admitted to the ICU. In patients with sepsis, IL-6 decreased levels were associated with ICU survival; NGAL levels rose in non-survivors, while HMGB-1 levels were unchanged in both survivors and non-survivors regardless of complications.

12The first step in endothelial and BBB injury is the breakdown and destruction of proteins followed by release into the bloodstream. These proteins or peptides can be evaluated as a marker of endothelial cells and BBB integrity. Plasma levels of occludin (OCLN), claudin-5 (CLDN-5), zonula-occludens 1 (ZO-1), PCT, and lactate were assessed in 51 septic patients. OCLN and ZO-1 were elevated with disease severity and positively correlated with the Acute Physiology and Chronic Health Evaluation II (APACHE-II) and SOFA scores and lactate levels. The predictive value for in-hospital mortality of ZO-1 was comparable to that of lactate levels, APACHE-II, and SOFA scores but superior to OCLN and PCT. In a case series of brain autopsies from adults who died from sepsis, 38% had no OCLN expression in the endothelium of cerebral microvessels. BBB damage was associated with higher maximum SOFA scores and PCT levels > 10 μg/L. BBB damage in the cerebellum was more common with CRP values > 100 mg/L. Soluble fms-like tyrosine kinase 1 (sFlt-1), soluble E-selectin (sE-selectin), soluble intercellular adhesion molecule 1 (sICAM-1), soluble vascular cell adhesion molecule 1 (sVCAM-1), and PAI-1 were evaluated in another studies. All these evaluated endothelial biomarkers were associated with sepsis severity. sFlt-1 had the strongest association with the SOFA score, while sFlt-1 and PAI-1 had the highest area under the operating receiver characteristic curve for mortality.

13Syndecan-1 is a structural component of the endothelium. Antithrombin, PAI-1, syndecan-1, VCAM-1, E-selectin, IL-1β, IL-6, IL-8, HMGB-1, and histone-H3 were increased in septic patients compared with healthy controls. Non-survivors had a higher syndecan-1 level compared with survivors. On day one, an association was seen between syndecan-1 levels and APACHE-II, SOFA, DIC scores, hemostatic markers, IL-1β, IL-8, and PAI-1. Day 1 syndecan-1 levels were also significantly higher in patients with DIC and had reliable discriminative power to predict both DIC development and subsequent mortality.

14The serum biomarker, calcium-binding protein B (S100B), reflects BBB disruption, glial cell injury, and activation. S100B is used to evaluate brain injury severity and predict outcomes from stroke, traumatic brain injury, encephalopathy, and delirium. A prospective cohort study demonstrated that day three values for predicting 180-day mortality were superior to day one (0.731 vs. 0.611). Patients with sepsis-associated encephalopathy also had elevated levels. In another observational study of 22 patients with septic shock, delirium was present in ten. The odds ratio for the risk of developing delirium with an S100B > 0.15 μg/L was 18.0. Patients with delirium had higher plasma levels of IL-6. S100B and IL-6 levels were positively correlated. S100B, PAI-1, angiopoietin (Ang)-2, ZO-1, and OCLN are the main biomarkers currently used to evaluate the vascular injury and BBB permeability.

15Critically ill patients show an increase in gut permeability, which may trigger a systemic inflammatory response syndrome and multiple organ dysfunction syndromes (MODS). Plasma zonulin levels were higher in sepsis patients compared to a post-surgical control group or healthy volunteers. In another study, serum levels of intestinal fatty acid-binding protein (I-FABP) were higher in patients with sepsis and higher still in those with septic shock. Serum D-lactic acid levels were also elevated with sepsis but did not differentiate severity. Neither I-FABP nor D-lactic acid could prognosticate.

16A non-coding RNA (ncRNA) is an RNA molecule transcribed from DNA but not translated into proteins. A microRNA is a small non-coding RNA molecule that functions in RNA silencing and post-transcriptional gene expression regulation. ncRNAs and mRNAs are being studied as sepsis biomarkers. For example, long non-coding metastasis-associated lung adenocarcinoma transcript 1 (lnc-MALAT1) and micro RNA (miR)-125a were increased in sepsis patients compared with healthy controls and positively correlated with APACHE-II score, SOFA score, serum creatinine, CRP, TNF-α, IL-1β, IL-6, and IL-8. The lnc-MALAT1/miR-125a axis was also a predictor of increased 28-day mortality risk. In another study lnc-MALAT1 expression was increased in acute respiratory distress syndrome (ARDS) patients compared to non-ARDS patients (AUROC: 0.674). Nonsurvivors compared to survivors (AUROC: 0.651) and positively correlated with APACHE-II and SOFA scores, CRP, PCT, TNF-α, IL-1β, IL-6, and IL-17. Long non-coding RNA maternally expressed gene 3 (lnc-MEG3), and the lnc-MEG3/miR-21 axis were increased, while miR-21 expression was decreased in sepsis patients compared with healthy controls. lnc-MEG3 (AUROC: 0.887) and the lnc-MEG3/miR-21 ratio (AUROC: 0.934) had high values for predicting elevated sepsis risk, while miR-21 (AUROC: 0.801) gave excellent predictive value for a reduced sepsis risk. A further study showed miR-125a and miR-125b expressions were elevated in sepsis patients compared with healthy controls and were predictive of sepsis risk—miR-125a (AUROC: 0.749) and miR-125b (AUROC: 0.839). No correlation was seen between miR-125a and CRP, TNF-α, IL-6, IL-17, and IL-23 in however, miR-125b was positively associated with these cytokines. miR-125a failed to predict 28-day mortality risk (AUROC: 0.588) in sepsis patients, whereas miR-125b was superior (AUROC: 0.699).

17Cell surface receptors are receptors incorporated into the plasma membrane of cells and act on cell signaling by receiving or binding to extracellular molecules. After detecting such molecules, the production of metabolites occurs. In one study, the cluster of differentiation (CD)-13, CD14, CD25, CD64, and human leukocyte antigen (HLA-DR) showed acceptable sensitivity and specificity for mortality prediction (CD13 AUROC:0.824; CD64 0.843; and HLA-DR 0.804) while CD14 and CD25 did not predict mortality. nCD64 expression, in a further study, nCD64, PCT, CRP, and SOFA scores were higher in septic patients, with nCD64 having the highest AUC (0.879) for differentiating a positive microbial culture. This was superior to PCT (0.868), SOFA score (0.701), CRP (0.609), and white blood cell (WBC) count. In predicting 28-day mortality, the combination of nCD64 and SOFA score had an AUROC of 0.91 versus 0.882 for the combination of PCT and SOFA.

18A meta-analysis of 19 studies enrolling 3012 patients evaluated the value of PCT (AUROC 0.84) and presepsin (0.87 AUROC) for diagnosing sepsis. The pooled sensitivities and specificities were 0.80 and 0.75 for PCT and 0.84 and 0.73 for presepsin. In one study, levels of presepsin, PCT, CRP, and WBC were higher in sepsis patients than in a SIRS group with AUROC values of 0.954 (presepsin), 0.874 (PCT), 0.859 (CRP), and 0.723 (WBC). The cut-off of presepsin for discriminating between sepsis and SIRS was 407 pg/ml, with sensitivity and specificity values of 98.6% and 82.6%, respectively. In a study of septic children, TREM-1 levels were higher in septic shock patients.

19Adrenomedullin (ADM) is synthesized in different tissues, including the adrenal cortex, kidney, lung, blood vessels, and heart. It has biological properties, including vasodilating, inotropic, diuretic, natriuretic, and bronchodilating. In one study, mid-regional pro adrenomedullin (MR-proADM) was an independent predictor of five different organ failures (respiratory, coagulation, cardiovascular, neurological, and renal), compared to lactate which predicted three (coagulation, cardiovascular and neurological), PCT two (cardiovascular and renal) and CRP (none). MR-proADM most accurately identified patients with a high likelihood of further disease progression compared to other biomarkers and clinical scores. A total of 1089 individuals with either sepsis (142) or septic shock (977) were included in a randomized controlled study. The MR-proADM level within the first 24 h after sepsis diagnosis was associated with 7-day mortality (AUROC 0.72 and p < 0.001) and 90-day mortality (AUROC 0.71 and p < 0.001). Patients with declining PCT levels but persistently high MR-proADM levels on day-1 or day-4 had a substantially higher mortality risk of 19.1 (8.0–45.9) and 43.1 (10.1–184.0), respectively. Adult patients hospitalized to ICU had their bioactive-ADM levels measured in this retrospective observational study. This study comprised a total of 1867 patients, 632 septic patients, and 267 septic shock patients. The median bioactive-ADM was 74 pg/mL in sepsis patients, 107 pg/mL in septic shock, and 29 pg/mL in non-septic patients. The association of elevated bioactive-ADM and mortality in sepsis patients and the ICU population resulted in O.R.s of 1.23 and 1.22, respectively. In addition, the MR-proADM is potentially removal by continuous renal replacement therapy (CRRT).

20N-terminal (N.T.)-prohormone BNP (NT-proBNP) is a non-active prohormone produced by the heart and released in response to changes in intracardiac pressure. Higher levels of NT-proBNP at 24 h after sepsis onset were associated with lower short physical performance battery (SPPB) scores at 12 months and lower handgrip strength at 6-month and 12-month follow-up. NT-proBNP levels during the acute phase of sepsis may thus be a valuable indicator of a greater risk of long-term impairment in physical function and muscle strength in sepsis survivors. In another study, NT-proBNP levels on admission were higher in-hospital non-survivors (7908 pg/mL) compared with survivors (3479 pg/mL). AUROC curves of admission and 72-h NT-proBNP levels for hospital mortality were 0.631 and 0.648, respectively.

21PCT is produced in the thyroid's C cells and converted to calcitonin, with no PCT released into the bloodstream. During inflammatory processes, PCT is produced directly by stimulating bacterial components or induced by various inflammatory mediators such as IL-6 and TNF-α. PCT and presepsin had similar performance in predicting positive sepsis results with AUROC values of 0.75 and 0.73, respectively. Another investigation gave AUROC values of 0.87 for PCT and 0.78 for presepsin in predicting bacteremia. Plasma levels of presepsin and PCT were progressively higher in sepsis and septic shock than in non-septic patients. Presepsin was superior to PCT in diagnosing severe community-acquired pneumonia, while PCT was marginally superior in another study of septic patients admitted to intensive care. On the other hand, MR-proADM had a better predictive value for septic shock. This study concluded that PCT, MR-proADM, and presepsin are complementary biomarkers that could have utility in the management of septic patients. In an intention-to-treat study comparing PCT versus no PCT guidance, there was no significant difference in 28-day mortality (25.6% PCT versus 28.2% no PCT, p = 0.34). The use of procalcitonin did not impact investment decisions as measured by the frequency of therapeutic and diagnostic interventions..

22High levels of resistin collected on day 1 of ICU admission were associated with an increased likelihood of developing new organ failure, whereas high myeloperoxidase (MPO) levels on day one were associated with an increased risk of developing incident organ failure for clotting and kidney systems.

23Soluble trigger receptor expressed in the myeloid cell-1 (sTREM-1) is a TREM family member. This receptor offers excellent potential as a biomarker for infectious diseases as it can be measured in different biological fluids, including serum, pleural fluid, sputum, and urine. However, a meta-analysis of 2418 patients enrolled in 19 studies showed serum sTREM-1 had only moderate accuracy in diagnosing patients with suspected sepsis. Combining sTREM-1 with clinical variables offered more significant mortality discrimination compared to clinical variables alone. In a multicenter prospective cohort study, soluble tumor necrosis factor receptor type 1 (sTNFR1) levels > 8861 pg/ml predicted 30-day mortality.

24Patients with sepsis or septic shock displayed higher levels of the soluble form of the urokinase plasminogen activator receptor (suPAR), PCT, and lactate on days 1, 2, 4, and 7 of admission, with lactate and suPAR being the best risk stratifies for suspected infection. Levels of suPAR and PCT levels were higher in sepsis patients than in a SIRS group with AUROC values of 0.89 and 0.82, respectively. Serum sPD-L1 levels were increased in non-survivors compared with survivors with similar prognostic accuracy for 28-day mortality as APACHE-II and SOFA scores. See Tables 1 and 2 for further information, as well as Fig. 1.

25Despite significant advances in treating septic patients, this disease continues to be associated with high mortality rates and high long-term cognitive dysfunction. Extensive research in the area is being performed to validate biomarkers, facilitate sepsis diagnosis, and allow an early intervention that, although primarily supportive, can reduce the risk of death. Sepsis sometimes shows a hyperinflammatory response pattern and may be followed by an immunosuppressive phase, during which multiple organ dysfunction is present. A biomarker or a panel of biomarkers could be a new avenue to predict, identify, or provide new approaches to treat sepsis.