Early Diagnosis of Sepsis Using Serum Biomarkers

Terence Chan; Frank Gu


Expert Rev Mol Diagn. 2011;11(5):487-496. 

In This Article


To be considered clinically relevant, a biomarker must have a high DA;[32] that is, it must have high sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) (Table 1). Many studies also calculate the likelihood ratio to determine if test results are truly indicative of disease or due to random chance.[32] Indicators such as the positive and negative likelihood ratios, which are calculated based on sensitivity and specificity, are also useful for assessing the strength of a diagnostic test (Table 1). These values are commonly graphed for multiple cutoff values (CVs) as a receiver operating characteristic curve. The area under the curve value is used to determine the best diagnostically relevant CV. At a CV where area under the curve is equal to 1, an ideal biomarker would possess 100% sensitivity, specificity, PPV and NPV.

As sepsis can be caused by bacterial, fungal and viral infections, the biomarkers reviewed are categorized based on their most common diagnostic usage.

Biomarkers for Bacterial Infections

C-reactive Protein C-reactive protein (CRP) is a general acute-phase reactant protein that rises in concentration up to 1000-fold in the blood in response to inflammation and infection.[15,20,28,31–36] Sensitivity values have been reported from 30 to 97.2%, specificity values of 75–100%, PPV values of 31–100% and NPV values of 81–97% (Table 2).[5,20,27,28,34,37–40] The large disparity in DA values is mainly due to the wide range of CVs used in different studies, ranging from 0.2 to 110 mg/l. The rapid improvement of assays used in studies to detect CRP have also dramatically affected the determined CVs. A 2001 study on neonates reported a CV of 40 mg/l when using the NycoCard CRP test (Axis-Shield PoC AS, Dundee, Scotland),[40] whereas a 2010 study on neonates using the automated immunoassay IMx system (Abbott Laboratories, IL, USA) reported CVs of 0.2–0.6 mg/l.[39]

The use of different assays is another cause of variation in results. Each assay has a unique functional detection limit. Among the 15 different studies reviewed that measured CRP and mentioned the equipment used, ten different assay systems from various manufacturers were used. The variation in equipment also reflects the various different study designs between studies. While some studies focus on neonates,[5,14,27,28,38,40–42] others focus on children,[27,33,37,38,40,43,44] adults[16,33,34,42,45–48] and the elderly.[21,33,39,42] Serum CRP levels also vary with age in healthy individuals.[42] The authors of the study suggested that a CV of 10 mg/l be used for elderly patients over 65 years, while a CV of 5 mg/l should be used for adults and infants between 3 and 7 days old. CRP levels are also naturally higher in older children compared with younger children.[43]

Studies have shown CRP to be highly sensitive[5,16,21,38,40] or highly specific for infection.[27,28] However, it is unclear if CRP can be used to distinguish Gram-positive from -negative bacterial infections.[16] CRP has also been studied as a diagnostic marker for other illnesses.[20,39,47,49,50] A recent study on community-acquired pneumonia showed CRP as a prognostic marker in determining the severity of community-acquired pneumonia and the need for hospitalization when using a CV of 110 mg/l.[39] CRP is also reported to distinguish bacterial from fungal infections, where levels above 100 mg/l are indicative of bacterial infection and elevated levels below 100 mg/l are indicative of fungal infection.[21,45] Several studies have examined the diagnostic relevancy of CRP in viral infections,[47,51,52] and current data suggest that CRP is an unreliable marker for viral infections on its own.[47,52]

Procalcitonin Procalcitonin (PCT), a precursor to the hormone calcitonin, is another candidate biomarker for bacterial infection. It is reported to be present at very low concentrations of 0.033 ng/ml in the serum of healthy individuals,[17] and is known to increase by up to 1000-fold under inflammatory conditions.[17,53–55] It is reported to rise within 2–4 h of infection and peak at 6–8 h.[4,14,54] Persistent elevated levels are indicative of the continual presence of infection or sepsis.[17] The rapid upregulation and sustainment of PCT levels in the serum during infection makes it an ideal biomarker. It is also a stable molecule, an important biomarker quality, remaining stable during blood preparation methods and freezing procedures[38,56,57] and after long-term storage.[56]

Procalcitonin has been shown to have high DA,[4,21,34,37,38,40,43,45,53,54] reporting sensitivity values of 74.8–100%, specificity values of 70–100%, PPV values of 55–100% and NPV values of 56.3–100% (Table 2).[4,9,21,34,37,38,40,53] Similar to the studies on CRP, the variation in DA values can also be explained by variations in CVs, equipment used and study designs. Although many studies have used a CV of 0.5 ng/ml, a range of CVs have been reported (Table 2). A 2001 study on critically ill children with sepsis reported a CV of 8.05 ng/ml, using an immunoluminometric assay (BRAHMS, MD, USA) available at the time.[40] By comparison, a 2009 study on critically ill children with sepsis determined a CV of 0.28 ng/ml, using the newer immunoluminometric LIAISON BRAHMS PCT assay (DiaSorin, MN, USA).[27]

As a biomarker for bacterial infection, most studies find PCT to be a useful and accurate biomarker[14,17,58–61] and more useful than other common inflammatory markers.[21,22,34,37,40,43,53] Several studies have reported significantly elevated levels of PCT in patients with sepsis compared with those without sepsis.[9,17,20,21,23,33,40,46,53,54,62] Studies have also reported PCT being used as a prognostic marker,[21,22,25,33,54,63,64] indicating that levels of PCT are a good indicator of response to treatment,[22,25,54,63] severity of sepsis[21,54] and mortality from sepsis.[33,54,64] Similar to the case of CRP, it is unclear if PCT can be used to distinguish between Gram-positive and -negative bacterial infections,[17,22,34] which are treated with different strategies.

Procalcitonin has also been used to distinguish fungal and viral infections from bacterial infections. During viral infections, PCT levels are reported to remain at low levels, often at concentrations found in healthy individuals.[17,54,63] A study on 122 children with viral infection reported that the maximum PCT level observed was 0.7 ng/ml.[65] PCT can also be used to differentiate bacterial from viral meningitis.[54,61] In comparison, fungal infections tend to cause mild elevations in PCT concentration compared with levels seen in bacterial infections.[17,21,22,45,53,54] Studies on invasive aspergillosis and invasive candidasis have reported significantly higher levels of PCT in patients with bacterial sepsis than those with fungal infection.[21,45] PCT also shows potential as a diagnostic marker for pneumonia,[25,54] abdominal infections,[25,54,59] urinary tract infections,[49,54,59] lower respiratory tract infections,[54,56,63] myocardial infarction[66] and as a biomarker to guide antibiotic therapy in patients with community-acquired pneumonia.[22,25,32,56,67,68]

Serum Amyloid A Serum amyloid A (SAA) is an apolipoprotein reported to have potential for diagnosing sepsis.[4,28] SAA is expressed at levels up to 1000-times higher after 8–24 h from the onset of sepsis.[4,69] Compared with CRP levels, SAA levels are reported to rise faster and higher after the onset of sepsis and remain at higher relative elevations.[28] Similar to PCT and CRP, studies using SAA have reported various DA values and CVs (Table 2), likely a result of the different assays used, as well as the development of assays with lower detection limits. Serum SAA levels of less than 15 mg/l for the elderly over 65 years and less than 10 mg/l for adults and newborns aged 3–7 days are indicative of healthy states.[20]

Although SAA is mainly studied as a biomarker for bacterial infection, a recent study by Kajiya et al. on patients with viral infections reported elevated SAA levels above the healthy cutoff of 10 mg/l among the infected patients.[47] However, SAA may not be clinically useful because it may be too sensitive, as it has been reported to increase during minor viral infections and in patients not presenting symptoms.[47]

Biomarkers for Fungal Infections

Mannan (M) and antimannan (AM) antibodies are used exclusively to diagnose invasive fungal infections, due to the presence of M in the cell walls of the invasive fungal organisms.[30] M and AM levels are elevated in blood during invasive fungal infections,[45] such as candidiasis and aspergillosis, making them potentially useful biomarkers for diagnosing fungal causes of sepsis. Reported DA values and CVs for M and AM vary between studies, and are either measured as separate biomarkers[29,45] or in combination (Table 2).[29,70] Arendrup et al. reported that combining M and AM tests provides the best DA values.[29] Similar to the biomarkers used for bacterial infections, DA value variations are likely due to the different assays and study methodologies. Assays with lower detection limits are also required, as all three studies reviewed reported CVs lower than the manufacturer's suggested diagnostic CVs.

The main disadvantage of using M and AM tests alone is the high rate of false-positives and -negatives that M assays produce,[30] requiring that these tests be used in conjunction with other diagnostic tests. β-D-glucan tests have been used in combination with M and AM tests since β-D-glucan tests are highly sensitive and specific for invasive mycosis, and are not species specific.[30,70] Using M and AM as biomarkers also requires frequent serial measurements, as M is cleared relatively quickly from the blood.[30,71]

Biomarkers for Viral Infections

IFN-γ-inducible protein 10 (IP-10), a proinflammatory chemokine, is a promising biomarker for diagnosing viral infections due to its role in the host response to viral infections.[72] A study by Ng et al. on bacterial infections in very low birth weight infants concluded that IP-10 could be a good diagnostic marker during initial measurements, reporting significant elevated IP-10 levels in infected patients at both initial and 24 h measurements, with no overlap of ranges.[72] IP-10 levels may also correlate with the severity of infection.[72]

Differences in DA values and CV exist between IP-10 studies (Table 2),[52,72,73] a result of the complex nature of viral infections. The large variety of viral illnesses means studies must focus on a specific virus infection, requiring very specific study populations and assays. However, current studies report IP-10 to be a potentially useful biomarker.[52,72,73] It is known from previous studies that IP-10 is released in response to viruses such as rhinovirus, respiratory syncytial virus, hepatitis B and C viruses, and H5N1 influenza.[73,74] In addition, IP-10 in the serum has been shown to be useful as a prognostic marker in guiding treatment in hepatitis C patients.[75]

Other Biomarkers Requiring Further Research

IL-10 is a potentially useful biomarker of sepsis. Zeitoun et al. reported IL-10 to be a promising diagnostic marker of both early- and late-onset sepsis among neonates, where a CV of >17.3 pg/ml resulted in respective DA values of 92, 84, 80 and 89%.[15] A study on very low birth weight infants found IL-10 to be one of only three markers with diagnostically relevant sensitivity and specificity values (≥80%) when using a CV of 7.6 pg/ml for late-onset sepsis.[72] In general, circulating cytokines, such as interleukins, have a short half-life, which can result in false-negatives, ultimately limiting their DA.[14,28]

A potentially useful biomarker specific to bacterial infections is lipopolysaccharide binding protein. Serum lipopolysaccharide binding protein levels are known to increase during bacterial and fungal infections, but not during viral infections.[76] A study by Pavcnik-Arnol et al. on critically ill neonates and children with sepsis reported lipopolysaccharide binding protein to be a better biomarker for sepsis than PCT, CRP and soluble CD14.[76] Soluble triggering receptor expressed on myeloid cells-1 is another potential biomarker, as it is not upregulated during noninfectious inflammatory diseases,[25] a property not found in other biomarkers such as PCT. Gibot et al. reported significantly elevated plasma soluble triggering receptor expressed on myeloid cells-1 levels in sepsis patients compared with systemic inflammatory response syndrome patients and found it to be a diagnostically accurate biomarker.[46]

Toll-like receptor 2 (TLR-2) and neutrophil CD64 receptor (nCD64) are two potentially useful biomarkers only present on cell surfaces. During infection, both experience upregulated expression on the surface of monocytes[47] and neutrophils,[27,77] respectively. Kajiya et al. found significantly elevated TLR-2 expression during viral and bacterial infections.[47] The study used patients with influenza A and B, respiratory syncytial virus, mumps, varicella-zoster virus and cytomegalovirus, indicating TLR-2 could be a potential biomarker of viral infections. nCD64 has been widely used as a sensitive biomarker of bacterial infection in settings that allow for laboratory testing.[78] nCD64 expression is not affected by underlying or inflammatory diseases and is not differentially expressed between systemic and local bacterial infections.[77] Although different studies have used differing units to calculate the CVs, the DA values reported for nCD64 are generally diagnostically relevant; for example, reported sensitivity values range from 71 to 97%, 71 to 100% for specificity, 53 to 100% for PPV, and 75.9 to 98% for NPV.[9,15,27,77,79,80]


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: