Introduction

Sepsis is a leading cause of death in critically ill patients despite the use of modern antibiotics and resuscitation therapies [1]. The septic response is an extremely complex chain of events involving inflammatory and anti-inflammatory processes, humoral and cellular reactions and circulatory abnormalities [2, 3]. The diagnosis of sepsis and evaluation of its severity is complicated by the highly variable and non-specific nature of the signs and symptoms of sepsis [4]. However, the early diagnosis and stratification of the severity of sepsis is very important, increasing the possibility of starting timely and specific treatment [5, 6].

Biomarkers can have an important place in this process because they can indicate the presence or absence or severity of sepsis [7, 8], and can differentiate bacterial from viral and fungal infection, and systemic sepsis from local infection. Other potential uses of biomarkers include roles in prognostication, guiding antibiotic therapy, evaluating the response to therapy and recovery from sepsis, differentiating Gram-positive from Gram-negative microorganisms as the cause of sepsis, predicting sepsis complications and the development of organ dysfunction (heart, kidneys, liver or multiple organ dysfunction). However, the exact role of biomarkers in the management of septic patients remains undefined [9]. C-reactive protein (CRP) has been used for many years [10, 11] but its specificity has been challenged [12]. Procalcitonin (PCT) has been proposed as a more specific [13] and better prognostic [14] marker than CRP, although its value has also been challenged [15]. It remains difficult to differentiate sepsis from other non-infectious causes of systemic inflammatory response syndrome [16], and there is a continuous search for better biomarkers of sepsis.

With this background in mind, we reviewed the literature on sepsis biomarkers that have been used in clinical or experimental studies to help better evaluate their utility.

Materials and methods

The entire Medline database was searched in February 2009 using the key words 'sepsis' and 'biomarker'. All studies, both clinical and experimental, which evaluated a biomarker were included. For each identified biomarker, the Medline database was searched again using the biomarker name and the key word 'biomarker'.

Results

A total of 3370 studies that assessed a biomarker in sepsis were retrieved; 178 different biomarkers were evaluated in the 3370 studies. The retrieved biomarkers and the major findings from key studies using these biomarkers are listed in Tables 1, 2, 3, 4, 5, 6, 7, 8 and 9. Of the 178 biomarkers, 18 had been evaluated in experimental studies only, 58 in both experimental and clinical studies, and 101 in clinical studies only. Thirty-four biomarkers were identified that have been assessed for use specifically in the diagnosis of sepsis (Table 10); of these just five reported sensitivity and specificity values greater than 90%.

Table 1 Cytokine/chemokine biomarkers identified in the literature search (with some selected references)
Full size table
Table 2 Cell marker biomarkers identified in the literature search (with some selected references)
Full size table
Table 3 Receptor biomarkers identified in the literature search (with some selected references)
Full size table
Table 4 Coagulation biomarkers identified in the literature search (with some selected references)
Full size table
Table 5 Biomarkers related to vascular endothelial damage identified in the literature search (with some selected references)
Full size table
Table 6 Biomarkers related to vaosdilation identified in the literature search (with some selected references)
Full size table
Table 7 Biomarkers of organ dysfunction identified in the literature search (with some selected references)
Full size table
Table 8 Acute phase protein biomarkers identified in the literature search (with some selected references)
Full size table
Table 9 Other biomarkers identified in the literature search (with some selected references)
Full size table
Table 10 Biomarkers that have been assessed for use in the diagnosis of sepsis
Full size table

Discussion

A multitude of biomarkers has been proposed in the field of sepsis, many more than in other disease processes; for example, a study of patients with myocardial infarction revealed 14 biomarkers suitable for diagnosis and determination of prognosis [17] and in patients with Alzheimer's disease, just 8 biomarkers were identified [18]. This large difference in the numbers of biomarkers for sepsis is likely to be related to the very complex pathophysiology of sepsis, which involves many mediators of inflammation [19], but also other pathophysiological mechanisms. Coagulation, complement, contact system activation, inflammation, and apoptosis are all involved in the sepsis process, and separate markers for each (part of each) system have been proposed (Tables 1 to 9). Additionally, the systemic nature of sepsis and the large numbers of cell types, tissues and organs involved expand the number of potential biomarker candidates, compared with disease processes that involve individual organs or are more localized.

It is interesting to note that most of the biomarkers we identified have been tested clinically and not experimentally. This is likely to be in part related to difficulties creating an experimental model that accurately reflects all aspects of human sepsis, problems with species differences, and problems in determining end-points in animal studies. Additionally, as the sepsis response varies with time, the exact time period during which any specific biomarker may be useful varies, and this is difficult to assess reliably in experimental models. Moreover, as there is no 'gold standard' for the diagnosis of sepsis, the effectiveness of a biomarker needs to be compared with current methods used to diagnose and monitor sepsis in everyday clinical practice, i.e., by the combination of clinical signs and available laboratory variables [20]; experimental models cannot be used for this purpose.

Our study revealed that there are many more potential biomarkers for sepsis than are currently used in clinical studies. Some of these markers may require considerable time, effort and costs to measure. Some are already routinely used for other purposes and easily obtained, such as coagulation tests or cholesterol concentrations. In many cases, the reliability and validity of the proposed biomarker have not been tested properly [8]. Of the many proposed markers for sepsis, acute phase proteins have perhaps been most widely assessed. PCT has been used particularly extensively in recent years. The specificity and sensitivity of PCT for the diagnosis of sepsis is relatively low (typically below 90%), regardless of the cut-off value [21, 22]. Raised PCT levels have also been reported in other conditions associated with inflammatory response, such as trauma [23], major surgery [24] and cardiac surgery [25]. Although CRP is often reported as inferior compared with PCT in terms of sepsis diagnosis, it is frequently used in clinical practice because of its greater availability. Elevated concentrations of serum CRP are correlated with an increased risk of organ failure and death [26], and the study of its time course may be helpful to evaluate the response to therapy in septic patients [11].

Another group of compounds that has been widely assessed as potential biomarkers are the cytokines. These are important mediators in the pathophysiology of sepsis, and most are produced fairly rapidly after sepsis onset. In a clinical study, levels of TNF and IL-10 were increased within the first 24 hours after admission of the patient [27]. However, blood cytokine concentrations are rather erratic and their time course is not clearly in concert with the course of sepsis [27, 28], making interpretation difficult.

The diagnosis of sepsis is a challenge. Clinical and standard laboratory tests are not very helpful because most critically ill patients develop some degree of inflammatory response, whether or not they have sepsis. Even microbiological assessment is unreliable because many culture samples do not yield microorganisms in these patients. However, biomarkers have also not been shown to be a great asset in the diagnosis of sepsis. Indeed, relatively few biomarkers have been evaluated as diagnostic markers (Table 10). Our search retrieved only 10 biomarkers that have been assessed for their ability to distinguish septic patients from non-septic patients with systemic immune response syndrome. However, none of these biomarkers has been tested for both sensitivity and specificity, and there is therefore no biomarker clearly identified as being able to differentiate sepsis syndrome from an inflammatory response due to other causes.

Early diagnosis of sepsis is also an important issue as early institution of appropriate therapy, including antibiotics, is associated with improved outcomes. We identified 16 factors that have been evaluated specifically for the early diagnosis of sepsis; five of these had reported sensitivity and specificity of more than 90%. IL-12 was measured in newborns at the time when sepsis was first suspected clinically and was higher in patients with sepsis than in those without [29]. Interferon-induced protein 10 (IP-10) was higher in neonates with sepsis and necrotizing enterocolitis than in neonates who had only necrotizing enterocolitis [30]. These two biomarkers have not been evaluated for this purpose in adults. Group II phospholipase 2 (PLA2-II) was reported to have high sensitivity and specificity for the diagnosis of bacteremia in critically ill adult patients within 24 hours after admission [31]. CD64 had high sensitivity and specificity for the early diagnosis of sepsis in adults, but could not reliably distinguish viral from bacterial infections, or local infection from systemic sepsis [32]. Neutrophil CD11b could distinguish septic pediatric patients from those with possible infection with good sensitivity and specificity [33]. The sensitivity and specificity of the other 11 biomarkers used to diagnose early sepsis were not reported or were less than 90%.

Biomarkers can be more useful to rule out sepsis than to rule it in. We identified three biomarkers with high negative predictive value to rule out sepsis: PCT (99% at a cut-off value of 0.2 ng/ml) [34]; activated partial thromboplastin time (aPTT) waveform (96%) [35]; and fibrin degradation products (100% for Gram-negative sepsis by ELISA assay) [36]. It is important to emphasize that culture-positive sepsis was generally used as the gold standard in all these studies, although cultures may remain negative in many patients with sepsis.

The majority of the biomarkers that we identified in our search were assessed for their ability to differentiate patients likely to survive from those likely to die. Indeed, any biomarker is expected to have some prognostic value and sepsis biomarkers are no exception; however, this is not an absolute rule because some sepsis biomarkers failed to have prognostic value [37–39]. Moreover, sensitivity and specificity were tested in only some of the proposed prognostic markers, and none had sufficient (more than 90%) sensitivity and specificity to predict which patients were at greater risk of dying due to sepsis. Other biomarkers were assessed for their ability to predict the development of multiple organ failure and to evaluate response to therapy. It is known that the extent of infection and the severity of organ failure has a significant impact on the prognosis of patients with sepsis. Additionally, the response to therapy varies among patients. Recently, the PIRO model has been proposed as a way of stratifying septic patients according to their Predisposing condition, the severity of Infection, the Response to therapy and the degree of Organ dysfunction [20]. In the future, sepsis biomarkers may contribute to this model of classification rather than just being used as prognostic markers.

No biomarker has, therefore, established itself sufficiently to be of great help to clinicians in everyday clinical practice. As each biomarker has limited sensitivity and specificity, it may be interesting to combine several biomarkers [40, 41]; however, this hypothesis requires further study. A clinical study showed that the combination of aPTT waveform with PCT increased the specificity of the aPTT waveform in the diagnosis of sepsis [35]. Studies using panels of sepsis biomarkers have also provided encouraging results [42–44]. The cost-effectiveness of all these methods must also be evaluated.

In this study, we tried to categorize the sepsis biomarkers according to their pathophysiological role in sepsis. A useful sepsis marker must not only help to identify or rule out sepsis, but it should also be able to be used to guide therapy. It has been shown that using PCT levels to guide therapy reduces antibiotic use and may be associated with improved outcomes [45, 46]. The use of novel therapies that modify the pathophysiological process of sepsis may also be guided by biomarkers [47, 48]. A study is underway to evaluate the value of protein C levels to guide the administration of activated protein C (clinicaltrials.gov identifier NCT00386425). In the future, sepsis biomarkers may help us administer these therapies to the right patient at the right time.

Conclusions

Our literature review indicates that there are many biomarkers that can be used in sepsis, but none has sufficient specificity or sensitivity to be routinely employed in clinical practice. PCT and CRP have been most widely used, but even these have limited abilities to distinguish sepsis from other inflammatory conditions or to predict outcome. In view of the complexity of the sepsis response, it is unlikely that a single ideal biomarker will ever be found. A combination of several sepsis biomarkers may be more effective, but this requires further evaluation.

Key messages

• More than 170 different biomarkers have been assessed for potential use in sepsis, more for prognosis than for diagnosis.

• None has sufficient specificity or sensitivity to be routinely employed in clinical practice.

• Combinations of several biomarkers may be more effective than single biomarkers, but this requires further evaluation.