Biomarkers in Sepsis
According to the Biomarkers Definition Working Group, a biomarker is any “…characteristic that is objectively measured and evaluated as an indicator of normal biological processes, pathogenic processes, or pharmacologic responses to a therapeutic intervention.”1 While this definition admits such basic signs as fever and leukocytosis, it most commonly refers to a blood test or histologic finding that can be used to suggest a particular diagnosis, estimate disease severity, or inform the decision to prescribe specific treatments. In cardiology, for instance, an elevated serum troponin level indicates myocardial injury, while in cancer care, a breast tumor biopsy that is positive for the HER2 receptor indicates disease that is more likely to respond to treatment with trastuzumab.
The role of biomarkers in sepsis is not as well established as in cardiology and oncology, but has garnered increasing attention in recent years. In 2001, the revised consensus conference definition of sepsis was expanded to acknowledge the utility of certain biomarkers, including C-reactive protein (CRP) and procalcitonin (PCT).2 More recently, biomarkers have been used to stratify septic patients according to disease severity and to direct the timing and duration of fluid administration, antibiotic therapy, and other treatments. There are now more than 170 proposed biomarkers for sepsis, although only a few have been studied adequately in prospective clinical trials.3
Biomarkers may prove to play an important role in the diagnosis of sepsis, a syndrome that is traditionally defined by a set of highly sensitive but nonspecific clinical parameters.4 Current clinical criteria have shown a limited capacity to unambiguously identify patients with sepsis or to provide risk stratification. Ultimately, the goal of using biomarkers in sepsis is to bring quantification and exactitude to a diagnostic task that remains in many ways subjective. Toward this end, more than 2 dozen clinical trials investigating the role of biomarkers in sepsis are currently under way, with more still examining their use in guiding therapy.5
SPECIFIC BIOMARKERS IN SEPSIS
Lactate is widely used as a biomarker to identify septic patients in need of fluid resuscitation.6 The mechanism by which serum lactate levels increase in sepsis is multifaceted; it includes stimulation of glycolysis, increases in cytokine and catecholamine activity, and up-regulation of lactate production by bacterial endotoxin.7–9 In septic shock, inadequate tissue oxygen delivery results in anaerobic cellular metabolism, in which glycolysis terminates in the conversion of pyruvate to lactic acid, rather than its entry into the tricarboxylic acid cycle. Hyperlactatemia can also be seen in sepsis despite adequate tissue oxygenation because of thiamine deficiency, the presence of bacterial endotoxin, and diminished lactate clearance secondary to liver failure.8–10 On its own, hyperlactatemia is a nonspecific finding, as it can occur in other shock states, including hemorrhagic and cardiogenic shock.
Arterial blood lactate levels reflect the weighted sum of lactate production from all tissue sources. Venous lactate samples are easier to collect and have been shown to correlate well with arterial samples drawn simultaneously, being on average higher by about 0.18 mmol/L.11 In patients with infection, there is a linear relationship between serum lactate level and mortality.12 The current Surviving Sepsis guidelines recommend fluid resuscitation in patients with blood lactate levels ≥4 mmol/L, a value beyond which mortality risk has been shown to increase precipitously.12,13 More recent studies have suggested that lactate levels are prognostic for 28-day mortality even within the range considered normal.14
A recent systematic review of the role of lactate in predicting outcome examined 28 studies, concluding that although elevated lactate levels were associated with sequential organ failure and 28-day mortality, the overlap in lactate values between survivors and nonsurvivors meant that a specific cutoff with adequate performance characteristics could not be defined.15 Of greater prognostic value in the emergency management of sepsis is the rate of lactate clearance, usually defined as
In one retrospective study of prospectively collected data, mortality was 19% among patients with lactate clearance of at least 10% after 6 hours, compared to 60% in those whose lactate remained elevated.16 A prospective study of lactate clearance found that a decrease of ≥10% following 6 hours of resuscitation significantly predicted survival, with every 10% reduction corresponding to an 11% reduction in mortality.17 As such, serial lactate levels are often used as part of a “quantitative” resuscitation strategy, in which intravenous fluid boluses are given serially until lactate levels normalize. One study has shown this strategy to be noninferior to a resuscitation strategy that targets normalization of central venous oxygen saturation (ScvO2).18 Patients in the lactate clearance arm of this trial still had central venous lines placed and received most elements of early goal-directed therapy, including a target central venous pressure of ≥8 mm Hg.19 Lactate clearance may correlate better with survival in septic shock than with survival in other shock states.20
As a physiologically stressful state, sepsis is associated with activation of the hypothalamic–pituitary–adrenal axis, leading to an increase in cortisol levels.21 In a prospective study designed specifically to assess the predictive value of cortisol levels in sepsis, three prognostic groups were defined based on baseline cortisol levels and on patient response to a short adrenocorticotropic hormone (ACTH) stimulation test (250 μg).22 Patients with a favorable prognosis (26% mortality) had low baseline cortisol levels, which responded appropriately (increase of >9 μg/dL) to ACTH stimulation, while those with the worst prognosis (82% mortality) had high baseline cortisol levels that did not respond to ACTH. Patients in the intermediate group (67% mortality) had either low baseline levels without adequate ACTH response or high baseline levels with adequate ACTH response.
A subsequent study demonstrated that in patients with septic shock, ACTH nonresponders benefited from treatment with hydrocortisone and fludrocortisone (mortality rate 63% vs. 53%, p = 0.02)).23; however, the subsequent CORTICUS trial, which included patients who were less sick overall, failed to reproduce this result.24 Current Surviving Sepsis guidelines recommend against the routine use of ACTH stimulation to identify patients likely to respond to glucocorticoids, but do recommend administering steroids to patients who remain hypotensive despite adequate fluid resuscitation and vasopressor support.13
C-reactive protein is a pentameric protein secreted mostly by the liver, that activates the complement cascade and stimulates cell-mediated immunity.25 As an acute-phase reactant, CRP is nonspecifically increased in the setting of acute or chronic inflammation. Thus in addition to sepsis, other causes of elevated CRP levels include trauma, burns, surgery, chronic immune-mediated inflammatory diseases, and cancer.25
In diagnosing sepsis, a number of small, older studies have shown CRP levels to be sensitive (71% to 100%) but to lack specificity (40% to 85%).25 CRP levels begin to rise within 4 to 6 hours of an inflammatory stimulus and, in the case of sepsis, track the effectiveness of antimicrobial therapy.25,26 Early, adequate antibiotics produce a sharp decrease in CRP levels, which predicts a positive outcome; an increase, or even a slow decrease, should prompt consideration of broader spectrum coverage and a search for an uncontrolled source of infection.26–28
CRP levels have been shown to correlate with severity of sepsis, differing significantly in the settings of systemic inflammatory response syndrome (SIRS), sepsis, severe sepsis, and septic shock.25,29 In a prospective study of patients admitted to the ICU, increasing CRP levels correlated directly with length of stay, number of failing organ systems, incidence of infection, and mortality. Compared to patients with a CRP level <1 mg/dL on admission, those with levels >10 mg/dL had higher incidences of respiratory failure (65% vs. 28.8%), renal failure (16.6% vs. 3.6%), coagulopathy (6.4% vs. 0.9%), and death (36% vs. 21%).30 CRP levels have greater diagnostic performance than traditional signs of infection, such as fever and leukocytosis, and may be of particular value in elderly patients.27,28,31
Calcitonin, a hormone central to skeletal homeostasis that is expressed in the thyroid gland, is not known to play a significant role in infection and inflammation.32 By contrast, its larger precursor PCT has been shown to be ubiquitously expressed throughout the body in response to bacterial infection.33,34 Under such conditions, production of this “hormokine” can increase up to several thousandfold, a finding that has led to extensive investigation of its potential use as a sepsis biomarker.32
In experimental models, PCT levels rise within 3 hours of exposure to endotoxin, peak at around 24 hours, and persist in the circulation for up to 1 week.35 Importantly, PCT levels stabilize and then decrease in response to adequate antimicrobial therapy, while a failure to normalize reflects inadequate coverage and portends a worse outcome.36,37 In the setting of renal failure, PCT levels may be elevated in the absence of sepsis, but decrease with initiation of hemodialysis.38 Unlike other nonspecific markers of inflammation, such as white blood cells, CRP, and erythrocyte sedimentation rate, PCT levels are not typically affected in the settings of chronic inflammation or glucocorticoid use,34 but may be elevated in shock states, whether or not these are related to infection.39 Certain drugs can interfere with the measurement of PCT levels, most notably poly- and monoclonal antibody preparations.34 PCT levels typically are measured from serum samples, by means of an automated immunohistochemistry method.40 Healthy individuals typically have undetectable plasma PCT values (<0.05 μg/L). Infection is suggested by a level >0.25 μg/L to >0.5 μg/L, depending on the patient population and assay used. In severe sepsis and septic shock, levels in excess of 10 μg/L can be seen.
Some of the first clinical studies on PCT focused on its use in differentiating sepsis from noninfectious SIRS. A meta-analysis of these early works41 pooled 18 studies and showed that in critically ill adult patients, PCT had poor diagnostic performance. The value of both sensitivity and specificity was 71%, with an area under the summary receiver operating characteristic curve of 0.78. Likelihood ratios (LR + = 3.03, LR − = 0.43) were insufficient to confidently rule in or out a diagnosis of sepsis, based on a moderate pretest probability.
Despite these early findings, newer studies have correlated PCT levels with severity of illness in sepsis.34 One prospective study of 255 patients admitted to the ICU found that PCT levels were significantly correlated with the clinical subtypes of sepsis (median PCT = 1.5 μg/L), severe sepsis (median PCT = 4.5 μg/L), and septic shock (median PCT = 13.1 μg/L).42 A multicenter study of 1,156 immunocompetent hospital inpatients with sepsis, including patients admitted to the emergency department (ED), showed that PCT levels correlated with mortality both outside the ICU (8% vs. 20% for PCT < vs. > 0.12 μg /L) and in the ICU (26% vs. 45% for PCT < vs. > 0.85 μg /L).43 PCT has also been shown to be useful in identifying sepsis among immunocompromised ICU patients (sensitivity 100% and specificity 63% for PCT >0.5 μg/L).44
The finding that PCT directly reflects the effectiveness of antimicrobial therapy has led to an exploration of its role within a broader movement toward antibiotic stewardship in the ICU.45–47 Large studies, such as the PRORATA trial,48 provided evidence that basing the timing and duration of antibiotic administration on daily PCT levels could reduce the overall duration of therapy. In this study, 630 medical and surgical ICU patients with suspected bacterial infections were randomized to have antibiotics started and stopped either according to PCT levels (see Fig. 36.1) or at the discretion of the supervising clinician using local and international guidelines. Noninferiority analysis revealed no difference in 28-day and 60-day mortality between groups. Patients receiving PCT-guided antibiotic therapy received antibiotics for a total of 10.3 days (SD 7.7) and those in the control group for 13.3 days (SD 7.6) (23% relative reduction in days of antibiotic exposure). Analysis of secondary endpoints showed no difference between groups in terms of relapse, superinfection, or emergence of multidrug-resistant bacteria. A systematic review of 14 randomized clinical trials (RCTs) examining PCT algorithms for prescribing antibiotics came to similar conclusions, showing no significant differences in mortality between PCT-guided therapy and standard therapy, but a significant decrease in antibiotic exposure.
FIGURE 36.1 From: Bouadma L, Luyt CE, Tubach F, et al. Use of procalcitonin to reduce patients’ exposure to antibiotics in intensive care units (PRORATA trial): a multicentre randomised controlled trial. Lancet. 2010;375:463–474.
The current Surviving Sepsis Campaign guidelines have been revised to reflect newer evidence regarding the use of PCT in diagnosing and managing sepsis. These guidelines include a weak recommendation for using low PCT levels in deciding to discontinue empiric antibiotics in patients without evidence of infection. The authors do not, however, endorse the use of PCT in distinguishing sepsis from noninfectious SIRS, citing an inadequate evidence base for its use in this regard.
Some degree of reversible myocardial dysfunction is known to affect up to half of all patients with sepsis and septic shock, even in the absence of preexisting cardiac disease.49–51 Cardiac biomarkers including troponon I (TnI), which is known to correlate with myocardial injury, and the brain natriuretic peptides (BNP and NT-proBNP), which are known to correlate with ventricular stretch, have therefore been studied as biomarkers in sepsis.52
In a subgroup of 598 patients with severe sepsis from the PROWESS study, 75% had positive TnI levels at the time of enrollment into the trial.53 The multivariate logistic regression model derived showed that a positive TnI at baseline was an independent predictor of 28-day mortality (32.2% vs. 13.6%). Another prospective study found that in patients with septic shock, a positive TnI correlated with reduced left ventricular ejection fraction (46% vs. 62%), greater need for inotropic or vasopressor support (94% vs. 53%), and increased mortality (56% vs. 24%).54
Even in the absence of left ventricular dysfunction, BNP and NT-proBNP can be significantly increased in sepsis and, in one small study, were found to be comparable to levels seen in acute congestive heart failure.55 A recent meta-analysis of 12 prospective cohort studies examined the prognostic value of BNP and NT-proBNP in patients with sepsis and found that elevated levels of natriuretic peptides were a powerful predictor of all-cause mortality (OR 8.65).56
A number of additional biomarkers have been studied in sepsis, but few of these are readily available as commercial assays. Numerous cytokines, including IL-6, IL-8, IL-10, and MCP-1, have been found to predict survival in sepsis, but offer little diagnostic advantage over PCT levels.57,58 Soluble receptors, including the receptor of advanced glycation end products (sRAGE) and the triggering receptor expressed on myeloid cells 1 (sTREM-1), show promise as prognostic biomarkers in sepsis, but require further study.57,59 Proadrenomedullin, which, like PCT, is derived from the calcitonin gene family, has shown promising results in studies of pneumonia, demonstrating better diagnostic performance than both CRP and PCT.6,60,61
Sepsis is a physiologically complex state that is unlikely to be adequately identified and stratified by a single test. A reasoned approach, therefore, involves using clinical findings to develop a pretest probability of sepsis and then deploying the biomarker tests best suited to answer the question at hand.
In the ED, traditional markers such as hyper- or hypothermia and leukocytosis or leukopenia remain useful in distinguishing infectious from noninfectious processes, but may lack specificity in older or immunocompromised patients. In these cases, PCT levels >0.5 μg/L are suggestive of bacterial infection, with higher values being more specific and suggesting more severe disease. Since timely initiation of appropriate antibiotics is crucial, treatment decisions may have to be made in the absence of complete biomarker data. Much like the results of microbial culture samples, baseline values of PCT or CRP may play a key role in decision making over the first few days of hospital admission, and serial levels may prove useful in reassessing ED patients within the first 24 hours.
Serial lactate measurements are useful in guiding the initial resuscitation of the septic patient and in identifying those with persistently elevated lactate levels that are at risk of worse outcomes and may benefit from ICU admission. While studies on lactate clearance have used a 10% cutoff to define patients with improved survival, most survivors had clearance of approximately 40% or greater, with mortality lowest among those whose levels normalized.
In the ICU, the emergence of sepsis may be less obvious, as SIRS criteria are very frequently met. In septic patients who fail to improve or deteriorate anew, markers such as CRP and PCT may be useful in prompting either a change in antimicrobial coverage or a search for persistent or new sources of infection. The specificity of PCT in this setting, however, may be lower. ICU patients with sepsis from nosocomial infections have significantly lower PCT values than those with community-acquired infections (2.9 μg/L vs. 6.6 μg/L, respectively).37
In addition to the tests described above, newer technologies are being explored for their potential in diagnosing sepsis. Examples include the computational analysis of heart rate variability,62 PCR amplification of bacterial DNA,63 and high-throughput genomic technologies.64–66 Ultimately, the most valuable use of biomarkers in sepsis will come not from their ability to predict severity of illness or mortality, but from matching individual patients with the specific therapies to which they are most likely to respond.
OR, odds ratio; CI, confidence interval.