Infectious diseases are a major public health problem throughout the world, and infections can be complicated by sepsis, an organ dysfunction caused by a dysregulated response to infection [1]. Septic shock is a subset of sepsis, in which circulatory collapse and metabolic dysfunction are associated with end-organ failure and high mortality. Worldwide, an estimated 48.9 million cases of sepsis occur each year, resulting in 11 million deaths [2]. Within the Global Burden of Disease, which records hierarchical levels of causes of health loss, the proportion of global deaths from sepsis (which is associated with 33 clinically significant bacterial pathogens) establishes it as the second leading level 3 cause of worldwide mortality [3,4].

The diagnosis and management of sepsis are based on criteria and algorithms published in international guidelines [5]. The definition of sepsis has evolved over time. In 1992, it was first defined, by consensus, to be a suspected infection associated with a systemic inflammatory response syndrome (SIRS) [6]; categories of sepsis, severe sepsis and septic shock were then delineated to define sepsis along a spectrum that corresponds to the host response to infection. A 2001 definition retained the unchanged 1992 definition, but expanded the list of signs and symptoms of sepsis to reflect clinical bedside experience and incorporate general, inflammatory (e.g., increased plasma C-reactive protein or increased plasma procalcitonin), hemodynamic, and tissue perfusion (e.g., hyperlactatemia) parameters [7]. Notably, although one objective of the 2001 meeting was to incorporate biomarkers into a new sepsis definition, it was held at the time that the development of a biomarker-based definition was premature. In 2016, a third set of definitions of sepsis was proposed, which defined it as organ dysfunction caused by a dysregulated host immune response (a host and immune response with pro- and anti-inflammatory pathway) to infection (sepsis = infection + systemic inflammation response + organ dysfunction) [1]. In this new definition, severe sepsis was no longer included as a category, as all sepsis was, by its very nature, deemed to be severe.

In the case of suspected sepsis, it is recommended to confirm or rule out bacterial infection and to continuously reassess, with a view to the initiation, modification, or discontinuation of antimicrobial therapy [5]. Unfortunately bacterial cultures, which are considered the gold standard for diagnosing bacterial infections, typically require 10 to 24 h to yield etiologic results [11], and an additional 8 to 24 h to yield antibiotic susceptibility results [12]. Although clinicians may already adjust antimicrobial treatment in about 20% of cases on the basis of the Gram results of a positive blood culture, and in an additional ~35% of cases on the basis of a matrix-assisted laser desorption-ionisation-time-of-flight (MALDI-TOF) identification obtained from a positive blood culture pellet [13], rapid antimicrobial susceptibility results are still warranted in nearly 50% of subjects with Gram negative bacteraemia. Even though faster technologies that provide antibiotic susceptibility results in 3–7 h have been developed to drive the tailoring of empiric antibiotic treatment, their turnaround time is insufficiently rapid to impact initial empirical therapy [13]. Furthermore, bacterial cultures can yield false negative results, potentially because of prior antibiotic therapy (high sensitivity of Streptococcus and Neisseria species) or inadequate sampling/an improper technique (air in anaerobic bottles may prevent the growth of most strict anaerobes).

It is proposed that antibiotic therapy should be initiated as soon as possible (within 1 h of recognition) in cases where there is a suspected septic shock (SSC strong recommendation, low quality of evidence) or a high likelihood of sepsis (SSC very low quality of evidence) [5]. Initiating broad-spectrum antimicrobial therapy in all patients with suspected infection solely on the basis of rapid clinical assessment (including history, clinical examination and tests that evaluate both infectious and non-infectious causes of acute illness) (SSC Best Practice Statement) [5], will result in unnecessary treatments in 60% to 70% of patients who ultimately do not have sepsis [14]. This mismatch occurs partly because the clinical presentation of viral and bacterial infections, as well as severe inflammatory processes, can be very similar. Such over-usage of antibiotics is problematic since: 1) antibiotics are associated with secondary toxic effects; and 2) contribute to the development of antimicrobial resistance (AMR), a significant concern in clinical practice. Overtreatment with antibiotics was particularly problematic during the COVID-19 pandemic [15,16], and it has more recently been estimated that, more generally, 4.95 million (3.62–6.57) worldwide deaths per year are associated with AMR [17].

Biomarkers of infection, such as procalcitonin (PCT), have been extensively studied to facilitate the rapid diagnosis of bacterial infection in patients with suspected sepsis. These biomarkers typically do not significantly enhance the sensitivity of infection diagnosis, due to the already high sensitivity achieved through scoring systems based on clinical signs and symptoms. Consequently, PCT is not recommended to aid decisions on when to start antimicrobials (SSC weak recommendation, very low quality of evidence) [5].

Ideally, biomarkers could play a role in improving specificity and restricting antimicrobial treatment to those patients with bacterial infection; a biomarker with a high negative predictive value (NPV) could also help rule out bacterial infection in patients with suspected sepsis and prevent unnecessary antibiotic treatments. The researchers therefore believe that biomarkers, could serve as a valuable additional tool when used in this setting, and could guide clinicians on the decision “to give or not to give antibiotics?”.

Once antimicrobial therapy has been initiated, it should be reassessed daily for continuation, modification, or discontinuation, based on the results of bacterial cultures and clinical evolution (SSC weak recommendation, very low quality of evidence) [5]. Ultimately, the goal is to tailor antimicrobial therapy, both to each patient and to a given infection. A biomarker-based approach of this kind is aligned with the suggestion to privilege shorter (over longer) durations of antimicrobials (SSC weak recommendation, very low quality of evidence). For this purpose, the combination of PCT with clinical evaluation is suggested, as this will aid decisions on when to discontinue antimicrobials (SSC weak recommendation, low quality of evidence) [5]. A similar approach that uses C-reactive protein CRP has already been used but is not recommended in this instance [5,10].

The diagnosis of the dysregulated immune response is based on the inflammatory response and organ dysfunction, according to the SSC 2021 guidelines [5]; and the diagnosis of systemic inflammation response is based on the systemic inflammatory response syndrome (SIRS) scoring system, rather than a biomarker value, such as CRP. The SIRS criteria for adults and children remain the same as in the (1992) original description [6], and incorporate three non-specific clinical parameters (temperature, heart rate, and respiratory rate), as well as laboratory testing for either circulating blood leukocytes or band forms.

As part of the clinical operationalization of the 2016 sepsis definition, the sequential organ failure assessment (SOFA) score [1], which assesses the function of six organ systems impacted by sepsis (lung, circulation/heart, brain, liver, kidney, coagulation), has been recommended for characterizing organ dysfunction. It is a relatively complex assessment that combines five clinical parameters and four laboratory results, and requires venous and arterial blood tests and approximately two hours for calculation [5]. In pediatric cases, the organ failure score is the pediatric logistic organ dysfunction (PELOD) score, which assesses the function of six organs [9].

The quick SOFA (qSOFA) score has been proposed as a sepsis screening tool, particularly in settings where all the parameters required to measure SOFA are unavailable. It is a simpler and quicker version that takes less than five minutes, and only measures three non-specific clinical parameters (blood pressure, mental status, and respiratory rate). Additional screening tools for organ dysfunction and illness severity include the national early warning score (NEWS) and the modified early warning score (MEWS): NEWS evaluates eight non-specific clinical parameters and three organs (lung, circulation/heart, and brain), and MEWS evaluates six non-specific clinical parameters and the same three organs. Neither of these scores requires blood sampling nor are they used in children. The 2021 SSC guidelines recommend discontinuing the use of qSOFA in favour of SIRS, NEWS, or MEWS, on the grounds of its poor sensitivity. (SSC Strong recommendation, moderate quality of evidence) [5].

Measuring serum lactate is suggested for adult patients suspected of having sepsis (SSC Weak recommendation, low quality of evidence) [1,5], and it can also be used as an additional assay to guide volume resuscitation in sepsis or septic shock cases (SSC Weak recommendation; low quality of evidence) [5]. However, it is neither sensitive nor specific enough to aid the early diagnosis of sepsis or dysregulated immune response and organ dysfunction. Lactate assay is often used in pediatrics but is not included in the 2020 pediatric guidelines [10].

In short, the diagnosis of sepsis is based on the combined recognition of suspected infection + systemic inflammation response + organ dysfunction. At present, these three components are based on rapid clinical assessment that only uses non-specific tests and clinical scores (tools), not biomarkers (Table 1) [5].

With the evolving definition of sepsis and the absence of a reference test, clinicians are often confused about how to clinically diagnose sepsis. For example, in Switzerland, more than half of clinicians (53.8%) pragmatically use a mixture of the various sepsis-associated scores (e.g., SIRS, MEWS, NEWS, qSOFA, SOFA) [18]. Moreover, when sepsis is suspected, a total of 89.7% measure circulating blood leucocytes; 92.3% use CRP; 84.6% measure PCT; and 100% measure lactate. This variability suggests that clinicians might benefit from using accurate biomarkers rather than clinical scores in their daily practice.

In 2009, the International Sepsis Forum Colloquium on Biomarkers of Sepsis proposed to develop a systematic framework for the identification and validation of biomarkers of sepsis, and to promote collaboration between investigators, the biomarkers industry, and regulatory agencies [19]. Unfortunately, the numerous studies since carried out have remained rather heterogeneous and have not made it possible to identify and accurately compare one or more sepsis biomarkers with the capacity for diagnostic (including screening and monitoring), guiding therapeutic (theragnostic) or risk stratification/measuring surrogate endpoint (prognostic) purposes.

Although most biomarker studies seek to show their utility in diagnosing suspected infection and sepsis, the use of biomarkers for rapid clinical assessment would ideally effectively and quickly (within 1 h) rule out the diagnosis of suspected infection and sepsis. The ideal diagnostic and theragnostic biomarker for rapid diagnosis of infection and sepsis should have a high negative predictive value (NPV) that minimizes false negatives for infection and sepsis (and therefore helps clinicians feel confident about not giving antibiotics after a negative test result) and a moderate positive predictive value (PPV) that minimizes false positives, and therefore helps confirm the clinical suspicion of sepsis.

Ideally, diagnostic and theragnostic biomarkers should also be able to detect sepsis early, even before clinical suspicion arises. This early detection is only possible when there is routine screening (repetitive measurements) and would be most applicable in the pre-symptomatic diagnosis of hospitalized patients with a high risk of nosocomial sepsis. While screening is recommended for acutely ill and high-risk patients (SSC Strong recommendation, moderate quality of evidence) [5], the tools that could be used are not clearly specified, and could include variables analyzed by manual methods or the automated use of the electronic health record (EHR) analysis (with or without artificial intelligence). Variables that need to be considered include existing scores, vital signs, signs of infections, non-specific tests, and others (Table 1).

A diagnostic and theragnostic biomarker should be able to monitor the progression of sepsis and assist decisions related to antimicrobial continuation, modification, or discontinuation, which should be made in conjunction with clinical evaluation (SSC Weak recommendation, very low quality of evidence).

Finally, one or more prognostic biomarkers could also make a useful contribution to risk stratification and surrogate endpoints undertaken for triage and resource allocation purposes. In summary, when one or many more diagnostic, theragnostic and prognostic biomarkers are used in combination with clinical scores (to improve pre-test probability) or a machine learning system, this should contribute to the diagnosis, screening, monitoring, risk stratification and reporting of surrogate endpoints for sepsis or its components (i.e., infection, systemic inflammation response and organ dysfunction) (Table 1).