[18] and CRP levels are widely used as a relatively nonspecific marker of inflammation. Many studies have now been published that demonstrated increased CRP levels in patients with sepsis. [1921] As an acute phase protein, however, CRP levels are increased to some extent in most conditions associated with inflammation, including rheumatoid arthritis, Crohn's disease, acute myocardial infarction and pancreatitis, to list but a few. [22] As such, CRP is a very nonspecific marker of sepsis. Nevertheless, it is a cheap and widely available test and when levels are raised in a patient with signs suggestive of sepsis, it provides useful supporting evidence. In 112 ICU patients, Pvoa et al.reported that a serum CRP concentration of >8.7 mg/dl had a sensitivity of 93.4% and a specificity of 86.1% for infection; [19] the combination of CRP >8.7 mg/dl and temperature >38.2C increased the specificity for infection diagnosis to 100%. Because baseline CRP levels are often raised as a result of comorbid chronic inflammatory conditions, changes in concentrations over time are more useful than single values. [2325] Pvoa et al. [23] showed that CRP concentrations increased over time in infected patients, but remained unchanged in non-infected patients. A maximum daily CRP variation of greater than 4.1 mg/dl predicted nosocomial infection with a sensitivity of 92.1% and a specificity of 71.4%; moreover, when combined with a CRP concentration greater than 8.7 mg/dl, the sensitivity and specificity increased to 92.1 and 82.1%, respectively. Similarly, as a prognostic marker, a decrease in CRP level after 48 h was associated with a mortality rate of 15.4%, while an increased CRP level was associated with a mortality rate of 60.9% in patients with CRP concentrations >10 mg/dl on ICU admission (p < 0.05). [24]
Because CRP levels decrease as sepsis resolves, it has been suggested that levels could be used to guide antimicrobial therapy. In 50 adult ICU patients with sepsis, Schmit & Vincent [26] reported that an increase in CRP of at least 2.2 mg/dl in the first 48 h was associated with ineffective initial antibiotic therapy with a sensitivity of 77% and a specificity of 67%. Similarly, in patients with ventilator-acquired pneumonia, serum CRP levels were significantly lower in patients with appropriate antibiotic treatment than in those with inappropriate empirical treatment at 96 h (10.3 10 mg/dl vs 19.2 14 mg/dl; p < 0.05), [27] and in patients with community-acquired pneumonia, a less than 60% decrease in CRP levels by 3 days after admission (odds ratio [OR]: 6.98; 95% CI: 1.5631.33) was associated with an increased risk of having received inappropriate empiric antibiotic treatment. [28] However, use of CRP levels to guide therapy has not been tested prospectively in adult patients with sepsis. Procalcitonin PCT was described more recently than CRP and is not routinely measured in all hospital laboratories. PCT levels have been shown to be raised in patients with sepsis [29,30] and may be particularly useful in distinguishing bacterial from other forms of infection. [31] In patients with community-acquired pneumonia, PCT values >0.25 g/l had a sensitivity of 96% and specificity of 40% for predicting bacteremia. [32] Similarly, in patients with febrile urinary tract infection presenting to the emergency department, PCT values >0.25 g/l had a sensitivity of 95% and specificity of 50% for a diagnosis of bacteremia. [33] In patients admitted to the emergency department with symptoms of systemic infection, a PCT threshold of 0.15 ng/ml had a sensitivity of 75%, specificity of 79% and negative predictive value of 98% for a diagnosis of bloodstream infection. [34] Several studies have suggested that PCT is a more reliable marker of sepsis than CRP. [3538]
Higher PCT levels have been associated with increased mortality rates. [39,40] In ICU patients with sepsis, Giamarellos-Bourboulis et al. [39] reported mortality rates of 25.6% in those with PCT less than or equal to 0.85 ng/ml, but 45.3% in those with PCT greater than 0.85 ng/ml (OR for death: 2.404; 95% CI: 1.3854.171; p = 0.002). As with CRP, PCT levels are raised in other inflammatory conditions, including pancreatitis, acute myocardial infarction and postcardiac surgery. Trends in concentrations over time are again of more value than single measurements. Karlsson et al. [41] reported that although PCT concentrations did not differ between hospital survivors and nonsurvivors, mortality rates were lower in patients whose PCT concentration decreased by more than 50% in 72 h than in those in whom levels decreased by less than 50% (12.2 vs 29.8%; p = 0.007). Nevertheless, the precise role of PCT remains unclear with meta- analyses reporting conflicting findings. Indeed, one review concluded that 'Procalcitonin should be included in diagnostic guidelines for sepsis and in clinical practice in intensive care units', [36] whereas another published a year later noted that 'The findings do not lend support to the widespread use of the procalcitonin test in critical care settings', [42] although differences in the search criteria and studies included in these meta-analyses may explain these different conclusions. The use of PCT levels to guide antibiotic therapy has been tested in several clinical trials in different groups of infected patients. However, despite reported beneficial effects of this approach on antibiotic utilization in many of these studies, [4347] with no adverse effects on patient outcomes, not all studies support these findings. [48] Several meta-analyses of studies using this approach have also been conducted. Schuetz et al. [49] recently reported, in a meta- analysis of 14 trials in patients with acute respiratory infection, that use of PCT to guide initiation and duration of antibiotic treatment was effective in reducing antibiotic exposure in these patients without increasing rates of treatment failure or mortality. Another meta-analysis suggested that although PCT-guided antibiotic therapy was associated with a reduction in antibiotic usage and reduced costs, an associated 7% increase in hospital mortality could not be excluded. [50]
Cytokine Levels Various cytokines, key mediators of the sepsis response, have been assessed for their role as biomarkers of sepsis. Serum levels of IL-6, TNF and IL-8, to mention just a few, have been shown to be increased in patients with sepsis. [29,51,52] Raised serum levels have also been shown to be associated with development of organ function and mortality. [51,53] These cytokines are all involved in the inflammatory response to infection, but levels are also increased in patients with other inflammatory processes, such as arthritis and myocardial infarction. Several recent studies have used multiplex technology to measure multiple cytokine levels simultaneously. In a multicenter cohort study of 60 patients with severe sepsis, Bozza et al. [54] evaluated the prognostic value of 17 cytokines (IL-1, -2, -4, -5, -6, -7, -8, -10, -12, - 13, -17, IFN-, granulocyte colony-stimulating factor [G-CSF], granulocytemacrophage colony-stimulating factor, MCP-1, MIP-1 and TNF-). Levels of cytokines were correlated with disease severity and development of organ dysfunction. IL-1, -4, -6, -8, MCP-1 and G-CSF had good accuracy for predicting early mortality, whereas IL-8 and MCP-1 had the best predictive value for 28-day mortality. In multivariate analysis, only MCP-1 was independently associated with prognosis (OR for hospital mortality: 1.41; 95% CI: 1.021.93; p = 0.036). In 30 patients with sepsis, Mera et al. [55] also measured 17 cytokines and reported that the initial levels of IL-8 were the most predictive for fatal outcome. Levels of IL-1, -6, -8, -12, IFN-, G-CSF and TNF- remained high in nonsurvivors. In a study of 126 emergency department patients with SIRS or sepsis, in which levels of 22 cytokines were analyzed, only soluble IL-2 receptor (sIL-2R) levels were independently associated with a diagnosis of severe sepsis. IL-1, sIL- 2R and IL-8 levels were independently associated with a diagnosis of septic shock. [56] However, using principal component analyses and hierarchical clustering analyses, the authors demonstrated that there were no typical cytokine profiles associated with a diagnosis of severe sepsis or septic shock. Soluble Triggering Receptor Expressed on Myeloid Cells-1 Soluble triggering receptor expressed on myeloid cells-1 (sTREM-1) is a member of the immunoglobulin superfamily, the expression of which is upregulated in the presence of bacteria or fungi. [57,58] Serum levels of sTREM-1 are increased in patients with sepsis, [5962] and several studies have suggested that they are more sensitive and specific for infection than other markers, including CRP and PCT. [59,60] However, in emergency department patients, the specificity and sensitivity for diagnosis of sepsis were only 59 and 60%, respectively. [63] In addition, a decrease in sTREM-1 levels over time was associated with a favorable outcome, [58] suggesting a potential place in following response to therapy. Levels of sTREM-1 have also been studied in other body fluids. Raised sTREM-1 levels have been reported in the bronchoalveolar lavage fluid of patients with pneumonia, [64,65] and urinary levels have been associated with early diagnosis of sepsis, disease severity and prognosis. [66] In logistic regression analysis, urinary sTREM-1 levels were a risk factor for development of acute kidney injury in patients with sepsis (OR: 1.02; Wald coefficient: 5.246; p = 0.022). [66] In a recent study in emergency department patients with sepsis who underwent early goal-directed resuscitation, sTREM-1 concentrations were significantly higher at admission and pre-early goal-directed resuscitation in nonsurvivors than in survivors (514.1 pg/ml [interquartile range: 412.71749.5 pg/ml] vs 182.4 pg/ml [interquartile range: 54.3327.0 pg/ml]; p = 0.001); PCT and CRP levels were the same in the two groups. [67] In multivariate regression analysis, sTREM-1 values at admission were independently associated with poor prognosis (OR for 28-day mortality: 2.78; 95% CI: 1.1166.945; p = 0.028). CD64 CD64 is a membrane glycoprotein, which binds the Fc part of IgG. The expression of CD64 on the neutrophil membrane is negligible in health, but upregulated in response to proinflammatory cytokines as, for example, in sepsis and other inflammatory conditions, such as arthritis. Neutrophil CD64 expression has been reported to have variable sensitivity and specificity for sepsis diagnosis and prognosis. [63,6871] In a study of 132 emergency department patients with fever, 87% of whom had bacterial infection, the CD4 index was higher in patients with infection than in those without (3.7 3.2 vs 2.5 2.3; p = 0.03); the area under the receiver operating characteristic curve (AUROC) for detection of bacterial infection was 0.66 (95% CI: 0.520.8) and for prediction of survival it was 0.71 (95% CI: 0.570.85). [69] In 163 infants with suspected sepsis, the sensitivity and specificity of the CD4 index using a cutoff of 2.30 were 70 and 62%, respectively; when combined with an absolute neutrophil count of <7500 or >14,500 cells/mm 3 , the sensitivity increased to 95% with a negative predictive value of 93%. [70] In 293 ICU patients with at least two SIRS criteria, using a cutoff of 2.2, the CD64 index predicted bacterial infection with a sensitivity and specificity of 63 and 89%, respectively, and positive and negative predictive values of 85 and 70%, respectively. [71] In 631 emergency department patients with suspected sepsis, the sensitivity of CD4 for diagnosing sepsis was 66% and the specificity 65% with positive and negative predictive values of 79 and 49%, respectively. Clot Waveform Analysis In patients with sepsis, the optical transmission waveform obtained during measurement of the activated partial thromboplastin time is altered, giving a biphasic pattern. The presence of this abnormal waveform has been used to support a diagnosis of sepsis. [72] In 331 ICU patients, 59.3% of patients with a biphasic waveform on admission and 45.3% with a biphasic waveform during the total ICU stay developed sepsis. The sensitivity of the waveform analysis for detection of sepsis was poor, varying between 22 and 55% at admission and between 48 and 74% during the entire ICU stay, depending on the threshold value of the slope used. The specificity for sepsis, however, was good, varying between 92 and 98% at admission and between 81 and 94% during the total ICU stay. [73] In 187 patients with SIRS, presence of a biphasic waveform (90% sensitivity, 92% negative predictive value) was better than PCT (71% sensitivity, 67% negative predictive value) or CRP (88% sensitivity, 89% negative predictive value) for diagnosis of severe sepsis and septic shock. [74] The biphasic waveform on day 3 also had better specificity (91%) and negative predictive value (98%) for prediction of 28-day sepsis-related mortality (CRP specificity 57%, negative predictive value 95%; PCT specificity 83%, negative predictive value 95%). [74] Combination of an abnormal waveform plus a PCT level >1 ng/ml on admission to the ICU was associated with a sensitivity of 79% and specificity of 96%, with a negative predictive value of 96%. [75]
Growth Arrest-specific Protein 6 Gas6 is an intracellular vitamin-K dependent protein released by leukocytes and endothelial cells in response to injury. Gas6 has roles in cell survival (antiapoptotic), proliferation, migration and adhesion. [76,77] In a matched casecontrol study, plasma Gas6 levels were higher in patients with severe sepsis (median: 110 [interquartile range: 75139] ng/ml) than in those with organ failure unrelated to infection (median: 85 [56101] ng/ml; p = 0.026) and in healthy control subjects (median: 54 [4968] ng/ml; p = 0.0001). [77] In the patients with severe sepsis, Gas6 levels correlated disease severity as assessed using the APACHE II score (Spearman correlation coefficient: 0.45; p = 0.012) and the Organ Dysfunction and Infection (ODIN) score (nonparametric test for trend, p = 0.030). [77] In 232 patients with fever and suspected infection, Gas6 levels were significantly higher in patients with sepsis than in controls (healthy blood donors; p < 0.001); levels were also significantly higher in patients with severe sepsis than in those with sepsis (p < 0.001). [78] In 45 ICU patients with septic shock, Gas6 levels correlated with the Sequential Organ Failure Assessment (SOFA) score (Spearman rank-order coefficient: 0.37; p = 0.01) and with hepatic function as assessed by aspartate transaminase concentrations (Spearman rank-order coefficient: 0.42; p = 0.006) and prothrombin time (Spearman rank-order coefficient: 0.45; p = 0.02). Gas6 concentrations were also higher in patients who required renal support than in those who did not (median: 76.5 [52164] pg/ml vs 10.5 [1.5 80.5] pg/ml; p = 0.04). Gas6 levels were also strongly correlated with plasma PCT concentrations (Spearman rank-order coefficient: 0.44; p = 0.005). There were no differences in Gas6 levels among survivors and nonsurvivors. [76]
Angiopoietin Angiopoietin (Ang)-1 and -2 are endothelial-derived vascular growth factors that have contrasting roles in endothelial activation during sepsis: Ang-1 stabilizes the endothelium, whereas Ang-2 promotes loss of barrier integrity and vascular leak. [79] Increased levels of Ang-2 have been reported in patients with severe sepsis compared with those with no evidence of inflammatory response (p < 0.05). [80] In patients with sepsis, Ang-2 levels correlated with disease severity as measured by the APACHE II (r = 0.395; p = 0.007) or SOFA (r = 0.402; p = 0.006) scores and were higher in nonsurvivors than in survivors (24.9 ng/ml [21.538.0] ng/ml vs 13.5 [8.121.6] ng/ml; p = 0.02). [81] In a study of 70 patients with severe sepsis, survivors had higher peak Ang-1 levels (median: 13 vs 10 ng/ml; p = 0.019) and lower nadir Ang-2 levels (median: 2.8 vs 6.2 ng/ml; p = 0.013) than did nonsurvivors. [79] Lower Ang-1 levels (<5.5 ng/ml) on admission were independently associated with an increased risk of death at 28-days compared with higher Ang-1 levels (OR: 0.282; 95% CI: 0.0860.93; p = 0.03). [79] Raised levels of Ang-2 have also been reported in patients with trauma and after cardiopulmonary bypass surgery. Soluble Urokinase-type Plasminogen Receptor Urokinase-type plasminogen receptor (uPAR) is expressed on various cell types, including neutrophils, lymphocytes, monocytes, macrophages and vascular endothelial cells. uPAR is involved in multiple immunological functions, including migration, adhesion, angiogenesis, fibrinolysis and cell proliferation. [82,83] During inflammatory disease processes, such as sepsis, uPAR is cleaved from the cell surface by proteases, releasing a soluble form of the receptor, soluble uPAR (suPAR). Elevated levels of suPAR have been reported in patients with sepsis compared with patients without (on admission, median: 11.05 [1.8720] ng/ml vs 7.62 [020] ng/ml; p < 0.001). [82] However, in this prospective study of 273 critically ill patients, 197 of whom had sepsis, the AUROC for prediction of sepsis was only 0.62, compared with 0.86 for CRP and 0.78 for PCT. In 151 patients with SIRS and suspected community-acquired sepsis, 96 of whom had bacterial infections, the AUROC for diagnosis of bacterial sepsis was 0.50 for suPAR, 0.72 for PCT and 0.81 for CRP, again suggesting that suPAR had lower diagnostic capability than did these other biomarkers. [84] In an ongoing study in critically ill patients, a cutoff suPAR value of 5.5 ng/ml had a sensitivity of 75% and a specificity of 72% for diagnosing sepsis (AUROC: 0.75; 95% CI: 0.660.83). [85] Several studies have suggested that suPAR may be of more value as a prognostic marker than for diagnosis. [82,83,86,87] In 543 acutely ill medical patients, baseline suPAR levels were significantly lower in 30-day (4.97 ng/ml vs 7.31 ng/ml; p = 0.0002) and 90-day (4.87 ng/ml vs 7.29 ng/ml; p < 0.0001) survivors than in nonsurvivors. [87] Increased suPAR levels remained significantly associated with 30-day mortality (hazard ratio [HR]: 1.10; 95% CI: 1.011.20) and 90-day (HR: 1.11; 95% CI: 1.04 1.19) after adjusting for age, sex, CRP levels and the Charlson Score. In a recent meta-analysis of six studies that had assessed the prognostic role of suPAR, suPAR had better prognostic ability and was superior in predicting mortality than other commonly used markers, including CRP, PCT and sTREM-1, although it performed less well than the SAPS II score. [86] In another recent study, Giamarellos-Bourboulis et al. developed and validated a mortality risk prediction rule with four levels of risk, using combinations of APACHE II values (<17 or >17) and serum suPAR concentrations (<12 ng/ml or >12 ng/ml). The score had a negative predictive value of 94.5% in patients with sepsis. [88]
HLA The above mentioned biomarkers have largely been indicative of the proinflammatory phase of sepsis, but sepsis is also associated with an anti-inflammatory state, which, if prolonged, can lead to immunodepression or so-called immune 'paralysis'. Monocyte HLA-DR is part of the major histocompatibility class II antigen complex. The main role of HLA-DR in infection is in antigen presentation to T-helper cells resulting in release of proinflammatory cytokines, but in patients with sepsis, this pathway can be impaired leading to immunodepression. [89] Several studies have suggested that HLA-DR expression may be indicative of prognosis. In an early study in surgical patients with sepsis, Volk et al. reported a survival rate of 81% in patients with normal monocyte HLA-DR expression, but only 19% in patients with a persistently (at least 5 days) reduced HLA-DR expression of <30%. [90] More recently, in patients with septic shock, Monneret et al. also demonstrated significant differences in HLA-DR expression in survivors and nonsurvivors; the percentage of monocytes expressing HLA-DR was reduced in all patients during the first 48 h of septic shock, but on days 34 the percentage increased in survivors but not in nonsurvivors (43 vs 18%; p < 0.001). [91] In multivariate logistic regression analysis, low monocyte HLA-DR expression (<30%) at days 34 was significantly associated with mortality after adjustment for other confounders with an adjusted OR of 6.48 (95% CI: 1.6225.93). In a recent retrospective study, Trimmel et al. were, however, unable to identify a specific cutoff value for HLA-DR expression that was correlated with mortality, raising questions about the value of absolute values for risk prediction. [92] Supporting this, in patients with severe sepsis, Wu et al. reported that there were no differences between survivors and nonsurvivors using a 30% cutoff for HLA-DR expression; however, a change in monocyte HLA-DR expression of 4.8% over the first 3 days of ICU admission did allow discrimination between survivors and nonsurvivors with a sensitivity of 89.0% and a specificity of 93.7%. [93]