Recent advances in biomarkers for detection and diagnosis of sepsis and organ dysfunction: a comprehensive review

Abstract

Sepsis is life-threatening inflammatory disease, and its pathogenesis and prognostic factors remain unclear. In addition, the symptoms and signs of sepsis patients lack specificity, which makes the diagnosis, treatment and prognosis evaluation of sepsis extremely difficult. Recently, with the emergence and development of detection technologies, various sepsis-related biomarkers have emerged. Biomarkers could be considered as indicators of either infection or dysregulated host response or response to treatment and/or aid clinicians to prognosticate patient risk. Therefore, searching for reliable biomarkers and evaluating their role in sepsis is envisaged to aid clinical decision-making. This article reviews the advances in research on sepsis biomarkers and their application in the early prediction of organ dysfunction to improve our understanding of current sepsis biomarkers and provide a reference for the application of biomarkers in the clinical diagnosis, treatment, and prognosis of sepsis. Besides, we propose that the combining multiple biomarkers is expected to be a more accurate and comprehensive strategy to evaluate the condition and prognosis of sepsis patients.

Introduction

Sepsis is an inflammatory disease resulting from a dysregulated host response to infection, leading to organ dysfunction and high mortality in patients [1, 2]. A total of 48.9 million sepsis patients were diagnosed in 2017, among which 11 million died, with an annual mortality rate of 19.7% [3]. Similarly, 50 million yearly sepsis cases were reported globally in a study published in 2020, with a mortality rate of over 1/5, with no evident decline [4]. Severe sepsis often manifests as multiple organ failure, which may involve heart, liver or kidney damage as well as cognitive impairment. Septic shock, that is, despite fluid resuscitation, vasopressor is still needed or lactate > 2 mmol/L after resuscitation, typically represents the terminal phase of this condition [5, 6]. Sepsis persists as a critical worldwide healthcare challenge and a leading cause of mortality, even with progress in deciphering its underlying mechanisms. Timely identification is vital for optimizing patient outcomes, as delayed treatment significantly impacts survival [7]. Despite decades of research, the pursuit of a universally reliable diagnostic method for sepsis continues to evade researchers. Sepsis diagnosis presents an intricate clinical challenge, which requires the integration of multifaceted laboratory analyses, including biochemical markers, hematologic indices, immunologic parameters, and pathogen identification methodologies [8]. Therefore, actively searching for highly specific and sensitive indicators is crucial for early identification, early intervention and improvement of prognosis of sepsis.

The septic process triggers the liberation of numerous biochemical mediators from altered tissues, which can serve as biomarkers. Biomarkers are indicators that can objectively measure and evaluate normal biological and pathogenic processes and pharmacological treatment responses [9]. Emerging research demonstrates the pivotal diagnostic value of sepsis-biomarkers in clinical management, attributed to their capacity to quantify infectious and discriminate between microbial etiologies (bacterial, viral, fungal) as well as systemic versus localized infections [10]. Beyond diagnostic utility, these biomarkers enable prognostic stratification through three key mechanisms: (1) guiding antimicrobial stewardship protocols, (2) monitoring therapeutic responsiveness and clinical recovery trajectories, and (3) forecasting sepsis-associated complications [11]. Furthermore, some biomarkers now enable precise characterization of pathophysiological process of sepsis, including vascular endothelial compromise, gastrointestinal barrier dysfunction, and multiorgan deterioration while simultaneously forecasting critical clinical trajectories, such as post-discharge readmission probabilities, short- and long-term mortality [12, 13]. Current research initiatives are actively exploring innovative diagnostic approaches utilizing diverse biomarker combinations analyzed through various detection techniques. These methodologies aim to target the early identification of infection-induced multiple organ failure at its nascent stages.

To date, more than 250 sepsis biomarkers have been identified. New biomarkers are continuously being discovered due to the complexity of sepsis and improvements in the sensitivity of various detection technologies. A clinically optimal biomarker meeting rigorous sepsis criteria must demonstrate exceptional sensitivity and specificity, exhibit reliable reproducibility, display dynamic correlation with disease progression, and deliver information about prognosis [14, 15]. Nevertheless, selecting a suitable biomarker to guide clinical sepsis diagnosis, treatment, and prognosis remains challenging. For instance, while C-reactive protein (CRP) has served as a clinical indicator for decades, its specificity has been challenged [16]. Procalcitonin (PCT) has been regarded as a characteristic biomarker after bacterial infection, but serum PCT levels also increase in non-infectious conditions (e.g., trauma, major surgery, pancreatitis, and kidney injury), so there are contradictory results in meta-analysis of PCT as a diagnostic marker of sepsis [17].

In conclusion, this paper reviews the research progress on sepsis biomarkers in recent years, aiming to strengthen the understanding of current sepsis biomarkers. However, the manuscript moves beyond merely cataloging traditional or single-category biomarkers for sepsis. Instead, it aims to establish a multidimensional, hierarchical, and clinically translatable evaluation framework. We systematically elaborate on the paradigm shift from single diagnostic biomarkers to multi-biomarker panels, and further to integrated predictive tools powered by multi-omics and AI technologies. This progression underscores the comprehensive value of biomarkers in early sepsis detection, risk stratification, organ-specific dysfunction diagnosis, and real-time therapy monitoring. Finally, we propose a forward-looking, biomarker-guided precision medicine framework for sepsis and organ dysfunction, encompassing strategies for dynamic real-time monitoring, therapy responsiveness evaluation, and the adjustment of personalized treatment regimens (Fig. 1).

Fig. 1.

Integrated biomarker algorithm for sepsis management

Materials and methods

To comprehensively review existing evidence in the field of sepsis biomarkers, we conducted a systematic literature search. The Search keywords included: "Sepsis", "Biomarkers", "Shock, Septic", "Early Diagnosis and prognosis", and “Organ dysfunction”. The search period covered January 2015 to May 2025, with databases, including PubMed, Web of Science Core Collection, Embase, and the Cochrane Central Register of Controlled Trials (CENTRAL).

Common biomarkers of sepsis

The aim of this review is to inform clinicians about several common host–response biomarkers of sepsis and provide reference for clinical decision-making (Fig. 2). These biomarkers include measurement of acute-phase proteins, inflammatory cytokines, chemokines, cell membrane receptors, soluble receptors, as well as metabolomics (Table 1).

Fig. 2.

Common host–response biomarkers of sepsis and its pathophysiological process

Table 1.

Biomarkers related to sepsis identified in the literature search

Biomarkers	Sensitivity	Specificity	AUC	FDA/EMA approval	Comment	References
CRP	70–90%	50–70%	0.70–0.85	Yes	Rapid detection, low cost and wide clinical application; It is mainly used for inflammatory dynamic monitoring and efficacy evaluation, rather than diagnosis	[18, 19]
PCT	75–85%	70–85%	0.75–0.90	Yes	Ideal early infection marker; Identify bacterial infections and guide antibiotic withdrawal	[20, 21]
HBP	80–90%	75–85%	0.80–0.95	No (In the clinical transformation)	An excellent index for predicting septic shock and organ failure; Reflect vascular endothelial injury;	[23]
SAP	65–75%	60–70%	0.75	No (In the research phase)	Immunoregulatory function markers; Have an excellent ability to predict sepsis-related mortality; Associated with apoptosis cell clearance and inflammation resolution	[26]
IL-6	80–90%	65–75%	0.75–0.88	Yes	The core driving factor of cytokine storm; A sensitive indicator reflecting the intensity of inflammatory response and prognosis;	[27, 28]
HMGB1	70–80%	65–75%	0.70–0.82	No (In the research phase)	Late inflammatory mediators, representing persistent tissue damage and immune activation; Associated with poor long-term prognosis	[30, 31]
sTREM-1	85–95%	75–85%	0.80–0.90	No (In the clinical transformation)	High specificity in bacterial/fungal infection; It is helpful to distinguish infectious and non-infectious inflammation	[37–39]
suPAR	75–85%	65–75%	0.70–0.85	No (In the clinical transformation)	Strong prognostic markers; Reflect chronic immune disorder and poor immune reserve	[43]
HLA–DR	–	–	–	No (As a research tool)	Reflect the immune function of septic patients; A correlation between decreased HLA–DR expression and poor sepsis prognosis	[47, 48]
PD-1	–	–	–	No (As a research tool)	Key markers of t cell depletion; High expression is related to immunosuppression and poor prognosis	[49]

Acute phase proteins

CRP and PCT are the most employed acute-phase proteins in clinical sepsis evaluation. CRP production can be triggered by interleukin-6 (IL-6) in the liver tissue due to tissue injury, inflammation, or infectious stimulation, which results in a significant increase in CRP levels at 4–6 h post-stimulation, doubles at 8 h, and reaches a peak level at 36–50 h [18]. Since CRP levels can also increase in various other diseases, such as myocardial infarction, chronic obstructive pulmonary disease, and acute pancreatitis, it lacks specificity for sepsis diagnosis. Despite its inherently low specificity in diagnosing sepsis in adults, it is commonly employed for early onset sepsis screening (occurring during the first 24 h) because of its high sensitivity [19]. PCT has emerged as a novel diagnostic tool, demonstrating superior efficacy in identifying systemic bacterial infections at early stages. A multicenter randomized study found that changes in PCT levels can reflect antibacterial treatment efficacy, and monitoring PCT levels can shorten the treatment cycle of antibiotics and reduce adverse reactions without affecting the treatment effect of patients [20]. However, recent research has shown that PCT levels are independent of sepsis severity or mortality, necessitating further study of the implications of serum PCT levels in the differential diagnosis of SIRS and sepsis [21]. While PCT testing provides valuable diagnostic insights, clinicians should avoid relying exclusively on PCT levels when formulating antimicrobial treatment strategies. Interpretation of PCT assay results necessitates comprehensive evaluation of the patient’s clinical presentation. This multidimensional analysis ensures therapeutic decisions align with the individual’s pathophysiological context rather than isolated biomarker values [22].

In addition to CRP and PCT, some new acute-phase proteins have been shown to have certain advantages in early sepsis diagnosis and prognosis evaluation. Secreted by neutrophils during initial inflammatory responses, heparin-binding protein (HBP) induces vascular hyper-permeability, positioning it as a prognostic indicator for septic deterioration. A multi-center observational investigation quantified HBP levels within 60 min of emergency department in 371 individuals with signs of infection. Analytical findings revealed that the concentration of HBP in septic patients increased significantly, which was positively correlated with PCT, CRP, neutrophil/monocyte ratio and biomarkers of organ hypoperfusion (creatinine, bilirubin, lactate). At the 19.8 ng/mL threshold, HBP exhibited diagnostic performance metrics of 66.3% sensitivity, 44.9% specificity, 49.3% positive predictive value, and 62.2% negative predictive value. Therefore, HBP can be used as a tool for the early diagnosis of sepsis and for the risk of early death [23]. Despite its strong association with sepsis severity and its potential for early diagnosis, its utility is limited by the following factors [24, 25]: (1) limited specificity for sepsis. HBP is primarily a marker of neutrophil activation, which is a hallmark of sepsis but is not exclusive to it. Elevated HBP levels are also observed in other sterile inflammatory conditions characterized by significant neutrophil involvement. (2) Influence of renal function. HBP is cleared from the circulation primarily by the kidneys. Consequently, patients with impaired renal function, a common complication in sepsis, exhibit reduced clearance of HBP. This leads to artificially elevated plasma levels that may not accurately reflect the degree of neutrophil activation or sepsis severity, potentially confounding clinical interpretation. 3. Cost-effectiveness and availability. HBP testing is not yet widely available in routine clinical practice. The cost of implementing a new biomarker assay must be justified by a demonstrable improvement in patient outcomes or a significant advantage over existing, cheaper biomarkers (such as PCT or lactate). Similarly, the serum amyloid P (SAP) is an acute-phase protein mainly synthesized by the liver and plays a vital role in sepsis development. Kelly et al. [26] reported that the changes in the SAP curve between dead and surviving patients were different in septic patients, reporting SAP for the first time for predicting sepsis-related deaths. Therefore, it is speculated that SAP has an excellent ability to predict sepsis-related mortality.

Inflammatory cytokines

The massive release of cytokines in sepsis triggers an uncontrolled pro-inflammatory reaction that can lead to tissue damage and organ dysfunction. These inflammatory cytokines, including TNF-α, IL-1β, IL-6, and interferons, are required for host defense against pathogens and the activation of the adaptive immune response. When the body is infected, injured or burned, immune cells secrete IL-6. Functioning as a pivotal initiator in acute inflammation, IL-6 orchestrates dual immunological processes: facilitating clonal expansion and functional maturation of adaptive immune cells (T/B lymphocytes) while concurrently mediating hepatic production and systemic release of CRP through JAK–STAT signaling pathways. In addition, the reaction time of IL-6 to infection is shorter than that of CRP and PCT and is frequently used as a biomarker for early sepsis detection in China [27]. A recent study showed that IL-6 was an independent predictor of sepsis diagnosis, with a sensitivity and specificity of 68% and 83%, respectively, and an area under the receiver operating characteristic curve of 0.764. However, no correlation was observed between IL-6 and the 28-day mortality of sepsis patients [28], which indicated that IL-6 helped diagnose sepsis patients but could not evaluate their prognosis.

High-mobility group protein B1 (HMGB1) is reportedly involved in the pathogenesis of sepsis as a late and downstream lethal inflammatory mediator [29]. Studies have shown that the serum levels of HMGBl in the sepsis and septic shock groups were higher than those in the healthy control group, with significantly higher levels in the septic shock group, indicating that the increased expression of HMGBl is related to sepsis progression [30]. Another study have shown that the level of serum HMGB1 has a certain guiding effect on the prognosis and mortality of septic patients [31]. Mechanistically, extracellular HMGB1 can phosphorylate mitogen-activated protein kinase after binding to the receptor of advanced glycation end products, causing the nuclear transfer of activated protein 1 and nuclear factor κB, initiating inflammatory cascade reactions, worsening inflammatory reactions, aggravating the patient’s condition, and affecting prognosis [32]. Therefore, HMGB1 can be used as a potential biomarker for diagnosing and prognostic evaluation of patients with sepsis. However, the clinical application of HMGB1 as a sepsis biomarker remains limited by several factors [29, 30, 33, 34]: first, delayed release kinetics. It lacks diagnostic value for early detection. Second, low specificity. Third, the absence of internationally standardized testing methods for HMGB1. Finally, the kidney is the main organ for removing HMGB1.Therefore, elevated HMGB1 levels may more accurately reflect the severity of organ dysfunction than the inflammatory activity of sepsis itself, thus confusing its value as a specific marker of sepsis.

Soluble receptors and cell membrane receptors

Soluble receptors participate in signal transduction by binding to extracellular molecules, and cell membrane receptors also play a regulatory role in signal transduction. The molecular mechanism of sepsis involves complex signaling pathways, making sepsis a life-threatening syndrome. Soluble triggering receptor expressed on myeloid cell 1 (sTREM1) is an immunoglobulin expressed by neutrophils and monocytes. sTREM-1 is released in a soluble form during bacterial or fungal infection and can be detected in different body fluids [35]; hence, it can be used as a biomarker. A prospective cohort investigation by Asmaa et al. [36] assessed the clinical validity of sTREM1 for both sepsis detection and discrimination from non-infectious SIRS. Their findings revealed sTREM1 as a superior discriminative biomarker in ICU, with ROC analyses demonstrating exceptional performance. Comparative longitudinal analysis showed CRP achieved area under the curve (AUC) values of 0.87 (day 1) and 0.97 (day 7), while sTREM1 surpassed these metrics with near-perfect discrimination (AUC 1.00 on day 1; 0.93 on day 7), highlighting its time-dependent diagnostic superiority in sepsis identification. The sensitivity of sTREM1 was 100% and specificity 84% at a cutoff of 49 pg/mL. The synergistic integration of sTREM-1 with standard clinical parameters demonstrated superior mortality risk stratification compared to clinical variables alone [37]. Concurrently, multicenter validation studies established sTREM-1 > 8861 pg/mL as a critical threshold independently associated with 30-day mortality [38]. Some scholars, through meta-analysis, observed that sTREM-1 had moderate sensitivity (0.80) and specificity (0.75) in predicting the 28-day mortality of sepsis patients [39]. The critical range of sTREM-1 detection value was large (30–60,000 pg/mL), which might be due to differences in detection methods among studies and the heterogeneity of the subjects. First, the infection characteristics (gastrointestinal infection, lung infection, urogenital infection, blood infection) and the heterogeneity in the selection of patients might bring bias to the results. Second, severity of sepsis differed between studies might contribute to the bias. Finally, the use of different ELISA kits might contribute to the heterogeneities [40, 41]. These factors may lead to high range of cutoff value in different studies. In conclusion, the only application of sTREM-1 could not finish the tasks of an ideal biomarker for sepsis, which could help the recognition of syndrome, accurate diagnosis and prognosis and improve antibiotic stewardship.

The urokinase-type plasminogen activator receptor (uPAR) is often expressed on the surface of fibroblasts and endothelial cells. Under inflammatory stimulation, it detaches from the cell surface to form soluble uPAR (suPAR), widely found in various bodily fluids. suPAR demonstrates significant diagnostic potential in infectious disease. Maria et al. [42] revealed that integrating suPAR measurements with qSOFA scoring enhances prognostic precision for unfavorable outcome, enabling more timely therapeutic interventions. Furthermore, Huang et al. [43] established suPAR’s clinical utility in sepsis management through multicenter validation studies. Their meta-analysis quantified the biomarker’s diagnostic accuracy with sensitivity and specificity values of 0.76 and 0.78, respectively, supported by an AUC of 0.83 in sepsis detection. Prognostically, suPAR exhibited 0.74 sensitivity and 0.70 specificity for mortality prediction, achieving an AUC performance metric of 0.78 across diverse patient cohorts. In addition, AUC for differentiating sepsis from SIRS was 0.81, and the sensitivity and specificity were 0.67 and 0.82, respectively. Given the comparable prognostic efficacy between suPAR and PCT, current evidence supports incorporating suPAR into standardized sepsis evaluation indicators.

CD14 is a receptor for lipopolysaccharide and LPS-binding protein complexes, which can be divided into membrane-bound CD14 (mCD14) and soluble CD14 (sCD14). The sCD14-ST is a fragment produced by plasma protease cleavage of sCD14 during systemic inflammation [44]. Studies have identified an optimal diagnostic threshold of 1025.00 pg/mL for sCD14-ST in sepsis (specificity 83.0% and sensitivity 85.0%). Plasma sCD14-ST concentrations not only exhibit significant diagnostic value for identifying sepsis but also correlate with disease severity. Furthermore, dynamically monitoring the plasma levels of sCD14-ST may enhance the assessment of therapeutic responses and prognostic outcomes in sepsis [45]. In conclusion, these results suggested that sCD14-ST is a good indicator for early sepsis diagnosis, severity, and prognosis.

Biomarkers of the immunosuppressive phase in sepsis

Immunosuppression is a compensatory anti-inflammatory response caused by early inflammation in sepsis that often increases the risk of death in patients with sepsis. However, there is no direct clinical characteristic marker of immunosuppression [46], so it may have significant clinical value in identifying and predicting the biomarkers of immunosuppression in patients with sepsis. Monocyte human leukocyte antigen–DR (mHLA–DR) is a cell surface receptor of the major histocompatibility complex II on the surface of monocytes, macrophages, and dendritic cells. A correlation has been reported between decreased mHLA–DR expression and poor sepsis prognosis [47]. Xu et al. [48] pointed out that the significant changes in mHLA–DR in septic patients within 1 month after admission, indicating that the dynamic changes in mHLA–DR in peripheral blood can reflect the immune function of septic patients. Programmed death receptor 1 (PD-1) is a cell surface receptor expressed by activated B/T lymphocytes that can form two ligands, PD-L1 and PD-L2. PD-L1 has received immense attention as a biomarker related to sepsis-related immunosuppression than PD-L2. In sepsis, the expression of PD-L1 in neutrophils increased significantly, which could inhibit the apoptosis of neutrophils by PI3K/Akt pathway. Neutrophils with delayed apoptosis gathered in the lung, which aggravated lung injury and affected the prognosis of sepsis [49]. Therefore, it can be seen that low PD-1 level is associated with better prognosis in sepsis.

The combined application of multiple biomarkers

Owing to the complicated pathophysiological mechanism of sepsis, more than a single biomarker is required to complete the diagnosis, treatment, and prognosis evaluation of sepsis. Considering the limitations of applying individual biomarkers in sepsis, combining multiple biomarkers is envisaged as a more accurate and comprehensive approach for evaluating the condition and prognosis of sepsis patients.

The combined detection of PCT and CRP has important application value for the early diagnosis of bacterial infectious diseases, and serves as an ideal early infection marker. Studies have shown that patients with CRP > 59.25 mg/L (sensitivity 74.4%, specificity 65.4%) or PCT > 2.44 ng/mL (sensitivity 77.1%, specificity 68.4%) can be diagnosed as G⁻ bacterial sepsis, and when values are less than the above stated, they can be diagnosed as G⁺ bacterial sepsis [50]. Li et al. [51] evaluated the predictive capacity of conventional sepsis biomarkers through individual versus combinatorial analyses, seeking to define a biomarker panel to predict sepsis in severe patients. Their multi-parametric assessment revealed enhanced diagnostic precision when integrating PCT, CRP, and SAA measurements, with this tripartite panel demonstrating superior sepsis prediction capabilities compared to singular biomarker evaluation. A prospective study screened the best combination of biomarkers for the diagnosis of infection and sepsis in the emergency room. It was observed that the combined detection of mHLA–DR, PCT and IL-6 can be used for the diagnosis of sepsis patients, and the diagnostic efficiency is 0.89 under ROC curve. The diagnostic efficiency of mHLA–DR, hyaluronidase and creatinine in sepsis patients is 0.92 under ROC curve [52].

Pentraxin 3 (PTX-3) is a long-chain pentameric structural protein that can be secreted by monocytes, neutrophils, or vascular endothelial cells under the stimulation of inflammation or infection [53]. A prospective observational analysis had shown that when the critical values of PTX-3, PCT, IL-6, and lactic acid were 26.90 ng/mL, 0.47 ng/mL, 269.47 pg/mL, and 5.45 mmol/L, the combined detection of PTX3, PCT, IL-6 and lactic acid (the area under ROC curve = 0.778) was superior to sequential organ failure assessment score (the area under ROC curve = 0.712) in predicting the 28-day mortality of septic patients [54]. Lin et al. [55] explored the value of neutrophil/lymphocyte ratio (NLR) combined with red blood cell distribution width (RDW) in assessing the prognosis of emergency sepsis patients. The results showed that the areas under the ROC curve for NLR and RDW were 0.818 and 0.823, respectively, while the area under the ROC curve for the combined NLR and RDW was 0.891. NLR and RDW levels demonstrated significant prognostic relevance in critically ill sepsis patients. The combined use of NLR and RDW may serve as a predictive tool in acute sepsis cases, which deserves close attention from clinicians.

By integrating transcriptomic, proteomic, and metabolomic data, it is possible to elucidate the pathological mechanisms of sepsis at multiple levels and identify biomarkers with enhanced specificity and sensitivity. Transcriptomic analysis may reveal significantly upregulated inflammation-related genes (e.g., IL1β and TNF) in peripheral blood mononuclear cells of sepsis patients, while proteomics further validates the expression levels and post-translational modification states of the proteins encoded by these genes in serum. Metabolomics, through the analysis of metabolites (e.g., lactate and succinate) in plasma or urine, uncovers the metabolic disturbances associated with sepsis. Integration of clinical data with metabolomics would provide means to understand the septic patient's condition, stratify patients better, and predict the clinical outcome. Several very recent review articles have systematically reviewed the metabolic signatures of sepsis and septic shock, aiming to improve the diagnosis and prognosis of sepsis and septic shock through metabolomics [56, 57].

By integrating these data, researchers can build a network of multiple biomarkers for sepsis and identify potential diagnostic or prognostic biomarkers. For example, Huang et al. [58] constructed a multi-omics Spearman correlation network integrating core metabolites, proteins and renal function. Subsequently, metabolomics analysis was used to explore the dynamic changes of core metabolites in the serum of SA-AKI mice. Thirteen differential renal metabolites and 112 differential renal proteins were identified through a multi-omics study of SA-AKI mice. Ultimately, based on the identified core metabolites, they constructed a diagnostic model named IC3, using inosine, creatine, and 3-hydroxybutyric acid, to early identify SA-AKI (AUC = 0.90). Guo et al. [59] identified early metabolite biomarkers (such as sphingomyelin and lysophosphatidylcholine, AUC ≥ 0.85–0.94) from the plasma of trauma patients using ultrahigh-performance liquid chromatography–tandem mass spectrometry through untargeted metabolomics. This study identified early predictive and diagnostic biomarkers and explored metabolic pathways related to sepsis after trauma. Using NMR-based metabolomics, an investigation analyzing serum samples from 50 septic shock patients and 20 healthy controls revealed marked disturbances in amino acid, carbohydrate, and lipid metabolic pathways among critically ill individuals. Tracking these metabolic shifts during therapy could aid in developing tailored therapeutic strategies to enhance clinical outcomes [60]. Longitudinal analysis of metabolite fluctuations, particularly in ketone bodies, amino acids, choline, and NAG during treatment highlights their prospective utility in guiding real-time therapeutic monitoring and adjustments [61]. These studies suggest that multi-omics integration not only improves the efficiency of biomarker discovery in sepsis, but also provides a new perspective for understanding its complex pathogenic mechanisms.

Application of biomarkers in organ dysfunction

Organ dysfunction is a fatal complication of sepsis. A variety of well-established routine laboratory tests help physicians assess whether end-organ dysfunction has advanced the patient’s clinical status from sepsis to severe sepsis. However, highly specific and sensitive indicators remain lacking for the diagnosis of organ dysfunction. Sequential Organ Failure Assessment (SOFA) score is a core tool for the assessment of MODS severity in critically ill patients. It quantifies objective indicators of six organ systems to dynamically monitor the progression of organ failure, predict prognosis and guide clinical decisions. Therefore, we organizes the content of this section according to the components of the SOFA score system (respiratory system, coagulation system, liver, cardiovascular system, central nervous system, and kidneys), with a focus on the biomarkers used to indicate organ dysfunction in sepsis. This provides new insights for the application of biomarkers in the early diagnosis, treatment, and prognosis assessment of organ dysfunction (Fig. 3).

Fig. 3.

Application of biomarker combination in septic organ dysfunction

Sepsis-induced myocardial injury

Sepsis-induced myocardial injury (SIMI) is one of the most severe complications of sepsis and is a significant independent risk factor for death in patients with sepsis. However, because its early clinical manifestations are unclear, some patients with symptoms often miss the best treatment opportunity, affecting the clinical prognosis [62]. Traditional biomarkers such as myoglobin, creatine kinase isoenzyme, hypersensitive troponin T, brain natriuretic peptide, and N-terminal pro-brain natriuretic peptide have particular significance in the diagnosis and prediction of myocardial injury [63], but suffers limited sensitivity, specificity, and clinical value. Therefore, novel biomarkers are emerging.

Fatty acid-binding protein (FABP) is a chaperone protein that expresses heart-type FABP (H-FABP) in myocardial cells. Under normal physiological conditions, H-FABP does not exist in peripheral blood [64]. During ischemia or hypoxia of cardiomyocytes requiring fatty acid mobilization for energy supply, the levels of H-FABP in cardiomyocytes increase and are quickly released into the blood at the early stage of myocardial injury, making it a novel biomarker for myocardial injury [65]. A prospective observational study aimed to identify crucial indicators for the prompt and early assessment of SIMI. The investigation demonstrated that increased H-FABP levels upon hospital admission correlated with a greater risk of 7-day ICU mortality in patients with SIMI (P < 0.05). Quantifying H-FABP within the first 24 h of hospitalization emerged as a critical biomarker for both prompt detection and short-term outcome prediction in SIMI cases [66]. Chen et al. [64] conducted analytical validation of a multiplex biomarkers approach for stratifying SIMI and mortality risk. Their analysis conclusively identified h-FABP as an independent prognostic predictor of SIMI (P < 0.05) while proposing an optimized diagnostic framework combining three biomarkers to enhance predictive accuracy for both SIMI and mortality in sepsis. However, the clinical application of H-FABP as a specific diagnostic and prognostic biomarker for SIMI still has limitations. First, H-FABP is mainly eliminated by the kidney. Sepsis patients frequently develop acute kidney injury, which impairs the clearance of H-FABP, prolongs its plasma half-life, and leads to persistently elevated levels. Under such conditions, increased H-FABP may reflect the severity of renal dysfunction rather than real-time dynamic changes in myocardial injury, potentially misleading clinical interpretation [67]. Second, there is currently no internationally established optimal cutoff value of H-FABP for diagnosing myocardial injury in septic populations. Reported thresholds vary considerably across studies, which not only hampers cross-study comparability but also limits its utility as a standardized tool in clinical practice [68]. Finally, although H-FABP is a marker of cardiomyocyte injury, it does not distinguish between underlying pathogenic mechanisms [66].

The future direction no longer revolves around replacing troponin, but rather integrating it with novel mechanism-specific biomarkers to construct a more powerful diagnostic and prognostic prediction model. One proposed combinatorial approach includes: hs-cTnI + sST2 + cf-mtDNA + miR-208a. hs-cTnI serves as the “gold standard” for confirming cardiomyocyte injury. sST2 reflects cardiac-specific inflammatory and fibrotic stress. cf-mtDNA indicates intracellular mitochondrial damage and bioenergetic crisis. miR-208a offers high tissue specificity and may help differentiate injury subtypes [69]. In summary, the specific diagnosis of SIMI is advancing toward a mechanism-guided, multidimensional biomarker panel strategy. This integrated approach opens new avenues for early intervention and personalized therapy.

Sepsis-induced lung injury

Acute respiratory distress syndrome (ARDS) is a common complication of organ dysfunction in sepsis patients, and its mortality is extremely high. ARDS-specific biomarkers can objectively evaluate the pathophysiological process and the responsiveness to therapeutic intervention. The core pathological features of sepsis-induced ARDS involve dysregulated inflammation, endothelial barrier disruption, and alveolar epithelial injury. Accordingly, novel biomarkers have been developed to target these three mechanisms.

Sepsis-related ARDS is often accompanied by intense systemic and lung-specific inflammatory responses. IL-6, a key pro-inflammatory cytokine in sepsis, drives the “cytokine storm.” Elevated IL-6 levels are not only associated with sepsis severity but also predict a higher risk of ARDS development and worse clinical outcomes [70]. The destruction of the alveolar epithelial barrier leads to the influx of proteinaceous edema into the alveolar space, which is a hallmark characteristic of ARDS. Surfactant protein D (SP-D) is a low abundance hydrophilic protein (0.6%), which secreted by type Ⅱ alveolar epithelial cells. Emerging research has identified SP-D as a promising diagnostic indicator for ARDS, given its specific association with alveolar epithelial damage observed in recent investigations. The increase of SP-D level in early ARDS can not only assist the diagnosis, but also indicate the poor prognosis of patients. A prospective cohort study investigating the association between plasma SP-D concentrations and pulmonary injury in pediatric ARDS cases revealed that heightened SP-D levels correlated with more severe disease manifestations and unfavorable clinical outcomes in children experiencing acute respiratory failure [71]. Another separate longitudinal analysis identified a statistically significant association between serum SP-D concentrations at hospital admission and ARDS severity (P = 0.04), though no meaningful link was observed between levels of SP-D and mortality risk (P = 0.89) [72]. Clara cell secretory protein (CC16), which is synthesized by Clara cells and non-ciliated bronchiolar cells, is the most abundant secretory protein in respiratory secretions. The decrease of CC16 serum level is a sensitive sign of lung epithelial cell injury and dysfunction. A retrospective observational study evaluated the diagnostic and prognostic values of CC16 in critical care patients with ARDS. ROC curve analysis evaluated the diagnostic performance of CC16 at ICU admission, revealing a sensitivity of 90.4%, specificity of 79.8%, positive predictive value of 74.2%, and negative predictive value of 92.8% at a threshold of 33.3 ng/mL. Serum CC16 concentrations were markedly higher in ARDS patients (54.44 ± 19.62 ng/mL) compared to non-ARDS individuals (24.13 ± 12.32 ng/mL; P = 0.001). Elevated CC16 levels correlated strongly with ARDS severity. Furthermore, CC16 concentrations were linked to prolonged ICU stays but showed no association with overall hospitalization duration. These findings position CC16 as a potential diagnostic and risk-stratification biomarker for ARDS, though its utility in prognosticating outcomes remains limited [73]. Sepsis contributes to pulmonary edema primarily through a cytokine storm that directly damages pulmonary vascular endothelial cells, resulting in increased vascular permeability. Angiopoietin-2 (Ang-2) acts as an antagonist of the Tie-2 receptor, which can destroy vascular stability and promote vascular leakage. In patients with sepsis-induced ARDS, Ang-2 levels are significantly elevated and strongly correlate with disease severity and mortality. It is a critical molecular bridge linking sepsis to pulmonary endothelial injury [74].

Owing to the clinical heterogeneity and complex pathophysiology of ARDS, no single biomarker can currently provide specific diagnosis of sepsis-induced ARDS. The most promising strategy in the future lies in combining endothelial, epithelial, and inflammatory biomarkers to distinguish sepsis-induced ARDS (typically a hyperinflammatory endotype) from ARDS of other origins, ultimately enabling early precision diagnosis and personalized treatment.

Sepsis-induced kidney injury

Kidneys are among the most frequently involved organs in sepsis, where 50% of septic patients suffer acute kidney injury (AKI). Sepsis-induced acute kidney injury is associated with a worse prognosis and accounts for 70% of mortalities. Early identification of SA-AKI is envisaged to help with early intervention and prevent further renal function damage, which is pivotal for improving patient survival rates and reducing mortality [75].

The diagnosis of SA-AKI has long relied on functional parameters, such as serum creatinine (Scr) and urine volume, which are often delayed and non-specific, frequently resulting in missed optimal treatment windows [76]. The emergence of novel biomarkers has brought transformative changes to the precision management of SA-AKI. In terms of specific diagnosis, these biomarkers enable the differentiation of SA-AKI from other types of AKI and facilitate the early identification of tubular injury. For instance, urinary neutrophil gelatinase-associated lipocalin (uNGAL) rises significantly within hours after renal tubular stress, serving as a sensitive early warning indicator of AKI in septic patients [77]. Tissue inhibitor of metalloproteinases-2 (TIMP-2) and insulin-like growth factor-binding protein 7 (IGFBP7) are both inducers of G1 cell cycle arrest. Under stressors, such as ischemia, inflammation, or toxins, renal tubular epithelial cells secrete these proteins, leading to cell cycle arrest in the G1 phase. The product of urinary TIMP-2 and IGFBP7 concentrations has been approved by the FDA for predicting the risk of moderate to severe AKI. Extensive studies have confirmed its high predictive value in critically ill patients and it remains largely unaffected by non-renal factors, such as chronic kidney disease or systemic inflammation [78, 79].

Beyond early diagnosis, these biomarkers are increasingly being utilized to guide specific therapeutic decisions. Elevated biomarker levels can help identify high-risk SA-AKI patient populations who are most likely to benefit from targeted interventions. For example, patients with high [TIMP-2]•[IGFBP7] values may receive more aggressive hemodynamic optimization, avoidance of nephrotoxic agents, or earlier consideration of renal replacement therapy [80]. In summary, biomarkers provide crucial tools for facilitating a paradigm shift in SA-AKI management from passive diagnosis to active early warning, and from generalized approaches to individualized precision medicine.

Sepsis-induced liver injury

Owing to its special anatomical structure and physiological function, the liver is one of the most vulnerable organs involved in sepsis [81]. The incidence of liver injury in sepsis patients has been reported to be as high as 30%, with a 60% mortality rate [82]. Traditional liver function indicators, such as alkaline aminotransferase (ALT), aspartate aminotransferase (AST) and total bilirubin (TBiL), can indicate hepatocyte injury or cholestasis, but lack specificity and early warning value, and it is difficult to distinguish liver injury caused by sepsis from other causes (such as viral and drug-induced hepatitis). It is crucial to identify novel biomarkers for early diagnosis of sepsis-induced liver injury.

Recent studies have shown that some novel biomarkers in the serum correlate well with liver injury, providing a new direction for diagnosing and prognostic evaluation of liver cell injury [83]. As the most abundant miRNA in the liver, miR-122 is specifically expressed in the cytoplasm of hepatocytes. When hepatocytes are injured, miR-122 is rapidly released into the bloodstream, making it an earlier and more specific indicator of hepatic damage. Wang et al. [84] directly demonstrated that miR-122 levels in septic patients correlated with the severity of liver injury and exhibited superior diagnostic performance compared to ALT and AST. Similarly, glutamate dehydrogenase (GLDH) is a mitochondrial enzyme that is most concentrated in hepatocytes surrounding the central vein of the hepatic lobule. It is highly sensitive to ischemic and hypoxic injury, serving as a more specific indicator of centrilobular necrosis than ALT. The reliability of GLDH as a liver-specific marker for acute injury has been validated by Alhaddad et al. [85].

Integrating the novel, mechanism-specific biomarkers enables the construction of a more comprehensive prognostic prediction model, such as M30/M65 + sCD163 + HMGB1 combination model [86–88]. The M30/M65 ratio is helpful in determining hepatocyte death (apoptosis or necrosis). sCD163 assesses the activation status of hepatic Kupffer cells (indicating the local inflammatory burden). HMGB1 evaluates the magnitude of systemic immune activation. A combination of elevated levels of these markers strongly indicates severe liver injury accompanied by a systemic inflammatory storm, portending an extremely poor prognosis and necessitating heightened vigilance for impending multiple organ failure. Although these biomarkers are still in the research stage, they are driving the management of septic liver injury from passive biochemical monitoring to a new era of active and pathophysiological-based precision therapy.

Sepsis-induced coagulopathy

Sepsis-induced coagulopathy (SIC) represents a central event in the pathophysiology of sepsis, characterized essentially by widespread microvascular thrombosis that leads to inadequate organ perfusion and multiple organ failure. Although conventional coagulation tests such as platelet count, prothrombin time (PT), and activated partial thromboplastin time (APTT) may indicate coagulopathy, they fail to accurately capture the dynamic progression of SIC or differentiate critical states, such as hypercoagulability and impaired fibrinolysis. It is imperative to seek biomarkers that can specifically reflect SIC.

Thrombin–antithrombin complex (TAT) directly reflects the production of thrombin in vivo, and it is a sensitive indicator of the activation of coagulation system. Soluble thrombomodulin (sTM), released by damaged endothelial cells, is a specific indicator of endothelial barrier injury and microvascular dysfunction in sepsis. Its elevated levels strongly predict organ failure and poor prognosis[89]. The state of the fibrinolytic system is critical for the classification of SIC. A sharp increase in plasminogen activator inhibitor-1 (PAI-1) signifies “shutdown” of fibrinolytic system, which leads to the failure to effectively remove microthrombus [90]. Although D-dimer is not specific, its abnormally high value combined with other indicators (such as low platelet and high TAT) can effectively confirm SIC.

In recent years, the strategy of "precise anticoagulation" based on biomarkers is the current research frontier. For example, extremely high levels of PAI-1 or sTM may indicate a state dominated by fibrinolytic shutdown and endothelial failure, which carries a high risk of bleeding [91]. Dynamic monitoring of trends in D-dimer and platelet count can be used to evaluate the response to anticoagulant therapy and guide treatment duration [92]. This biomarker-driven management strategy for SIC aims to move beyond the universal treatment based on clinical scores (such as SIC scores), and stratify patients according to their real-time coagulation phenotypes, so as to maximize the curative effect and minimize the risk of bleeding.

Sepsis-associated encephalopathy

Sepsis-associated encephalopathy (SAE) is a severe neurological syndrome characterized by widespread brain dysfunction resulting from sepsis, significantly linked to higher mortality rates and prolonged complications in intensive care settings. The diagnosis of SAE mainly depends on clinical neurological assessment, lacking specific objective indicators. It is a hot and difficult point to find biomarkers that can diagnose SAE early and specifically and guide treatment.

The precise pathogenesis of SAE remains unclear. Current research suggests that active neuroinflammation, glial cell overactivation, blood–brain barrier (BBB) dysfunction and cerebral microcirculation dysfunction contribute to the pathophysiology of SAE [93]. Neuroinflammation in sepsis strongly activates astrocytes. In addition, glial fibrillary acidic protein (GFAP) serves as a specific biomarker for astrocyte activation and injury. Elevated serum GFAP levels indicate astrogliopathy in SAE, which are associated with BBB disruption and neuroinflammation [94]. Neurofilament light chain (NfL) is a highly sensitive and specific biomarker reflecting axonal damage. In patients with SAE, sNfL levels are significantly increased and closely correlated with delirium severity, cerebral atrophy, and long-term cognitive impairment. It is considered one of the most promising prognostic biomarkers for SAE currently available [95].

In sepsis, the core mechanism of SAE is that inflammatory factors lead to the destruction of BBB [96]. Disruption of BBB is initiated by the degradation of structural proteins, which subsequently enter circulation and may serve as biomarkers for BBB integrity [97]. In a cohort of 51 septic patients, plasma concentrations of occludin (OCLN), claudin-5 (CLDN-5), zonula–occludens 1 (ZO-1), PCT and lactate were analyzed. Elevated OCLN and ZO-1 levels exhibited significant correlations with clinical severity (APACHE-II and SOFA scores) and lactate concentrations, suggesting their utility in reflecting disease progression. ZO-1 demonstrated prognostic equivalence to lactate concentrations, APACHE-II, and SOFA scores but superior to OCLN and PCT [98]. A systematic review and meta-analysis investigated the clinical utility of serum S100B as a biomarker for diagnosing and predicting outcomes in SAE. Elevated serum concentrations of S100B in septic patients demonstrated a moderate correlation with SAE and poor clinical outcomes, particularly mortality. These findings suggest that S100B holds potential as both a diagnostic tool and prognostic indicator for SAE severity and patient survival [99].

Current evidences strongly support the use of biomarker panels (e.g., NfL + GFAP + CSF inflammatory markers + S100B) to comprehensively evaluate distinct pathological processes in SAE, including axonal injury, glial activation, neuroinflammation, and BBB disruption. Integrating such panels with clinical scoring systems could facilitate early detection, pathological stratification, and targeted therapeutic interventions for SAE.

Conclusions

The pathogenesis of sepsis is a complex process. Despite the decline in patient mortality due to sepsis over the last decade, its incidence continues to increase. Moreover, the symptoms and signs of sepsis lack specificity, making diagnosis, treatment, and prognosis evaluation challenging. During sepsis, an initial excessive inflammatory response may transition into a subsequent immunosuppressive state, frequently accompanied by multi-organ dysfunction. Utilizing a combination of biomarkers or biomarker panels may offer a promising strategy to predict and identify the condition earlier, as well as provide new approaches to treat sepsis. In recent years, the continuous emergence of new sepsis-related biomarkers has advocated the urgent need for optimal biomarkers or biomarker combinations for sepsis while also providing new hope for evaluating the condition and prognosis of sepsis patients. However, the high sensitivity and specificity of these biomarkers for early identification and prognosis still need to be improved, and further studies are required to confirm these findings.

Recent advancements in artificial intelligence (AI), single-cell RNA sequencing (scRNA-seq), and machine learning models integrating multiple biomarkers are revolutionizing the diagnosis and treatment of sepsis. First, AI can leverage large-scale multi-omics data (e.g., genomics, transcriptomics, proteomics, and metabolomics) to identify potential biomarkers associated with sepsis. AI algorithm can extract critical features from complex biological data and predict which molecular or gene are closely linked to the pathological processes of sepsis [100, 101]. This approach not only enhances the efficiency of biomarker discovery but also elucidates the molecular mechanisms underlying sepsis subtypes, paving the way for personalized treatment strategies. Second, the application of scRNA-seq provides unprecedented resolution in sepsis research. While traditional RNA sequencing provides only average expression levels of cell populations, scRNA-seq can reveal the gene expression profiles at the single-cell level, thereby identifying specific cell subpopulations that play pivotal roles in sepsis [102, 103]. By integrating AI algorithms, researchers can further analyze the dynamic changes in these cell subpopulations and uncover novel biomarkers and therapeutic targets. Finally, machine learning models integrating multiple biomarkers offer powerful tools for precise diagnosis and prognosis prediction in sepsis. By integrating diverse biomarkers, machine learning models can construct multidimensional predictive models, and significantly improve diagnostic accuracy and prognostic reliability [104, 105]. In summary, the synergy of these technologies not only accelerate the discovery of novel biomarkers but also advance precision medicine in sepsis, ultimately improving patient outcomes and survival rates.

Author contributions

Conceptualization: Leiyang Chen; searched literature: Xiuxiu Zhang and Peizhi Shi; writing and editing: Leiyang Chen; and supervision: Leiyang Chen, Xiuxiu Zhang and Peizhi Shi.

Funding

This article received no funding.

Data availability

No data to be shared.

Declarations

Ethical approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.