Lourens, Nicolien A. MBChB, MMed*; Bösenberg, Liesel H. MBChB, MMed*; Tintinger, Gregory R. MBBCh, MMed, PhD*; Ker, James A. MBChB, MMed, MD*; Fickl, Heidi PhD†; Sharp, Catherine DipMedTech†; Van Zyl, Marieta MBChB, MMed*; Anderson, Ronald PhD†
Physicians frequently need to evaluate patients with fever, which may be caused by a wide range of conditions, including sepsis. Early empiric antimicrobial therapy is important because this may reduce mortality in patients with sepsis. Although alterations in clinical parameters such as temperature, pulse rate, respiratory rate, and blood pressure indicate systemic inflammation or the systemic inflammatory response syndrome,1 they do not allow differentiation of infectious from noninfectious etiologies.
Microbial cultures of various body fluids are considered the criterion standard for diagnosing infection. However, microorganisms are cultured in less than half of those patients in whom sepsis is diagnosed clinically, and prior exposure to antibiotics may inhibit the growth of pathogens. Therefore, a biological marker that reliably predicts the presence of sepsis would be a useful adjunctive diagnostic test.
Neutrophils, monocytes, and macrophages play a pivotal role during the innate immune response to invading microbial pathogens.2 The recently identified Triggering Receptor Expressed on Myeloid cells 1 (TREM-1) is a pattern recognition receptor of the immunoglobulin superfamily that is up-regulated on the surfaces of these inflammatory cells after exposure to extracellular bacteria and fungi.3,4 Signal transduction pathways activated by ligand binding toTREM-1 are associated with enhanced production of proinflammatory cytokines and amplification of the immune response. The expression of TREM-1 reportedly allows differentiation of infectious from noninfectious causes of pneumonia when the soluble form of the receptor, sTREM, is quantified in samples of bronchoalveolar lavage fluid (BALF).5
The presence of sTREM in BALF of mechanically ventilated patients may be useful for diagnosing ventilator-associated pneumonia (VAP).5 Soluble TREM concentrations in the plasma of critically ill patients are significantly higher in those with sepsis compared with patients without a proven source of infection.6 This finding may be used to target empiric antibiotic therapy to patients with an elevated plasma sTREM.
Expression of TREM-1 on peripheral blood neutrophils may not distinguish pneumonia from noninfectious causes of interstitial lung disease,7 suggesting that biological fluids from the site of infection, such as BALF, may provide more useful diagnostic information. Soluble TREM has been measured in BALF, plasma, and cerebrospinal fluid (CSF), but not to our knowledge, in peritoneal or pleural fluid in humans. Furthermore, pneumonia due to intracellular pathogens such as Mycobacterium tuberculosis is not associated with significant up-regulation of TREM-1 expression,8 which may limit its usefulness in patients with tuberculosis.
The current study was undertaken to evaluate the role of sTREM-1 measured in samples of CSF, peritoneal fluid, and pleural fluid from patients with suspected meningitis, peritonitis, or pleuritis by comparing sTREM concentrations with the results of standard laboratory investigations to detect the presence of microbial pathogens in these fluid samples. Our results suggest that sTREM-1 can be detected and quantified in each of the body fluids tested and may be useful to predict the presence of diverse microbial pathogens, including M. tuberculosis, in pleural and peritoneal fluid samples. Detection of sTREM in CSF strongly supports the presence of inflammation within the brain or meninges, although a consistent correlation between sTREM and microbial pathogens could not be demonstrated.
Fluid for sTREM measurements was taken during the routine diagnostic evaluation of patients in the Departments of Internal Medicine at Pretoria Academic and Kalafong Hospitals (Pretoria, South Africa) with suspected meningitis, peritonitis (ascites), or pleuritis (pleural effusion). The physician's decision to submit CSF, peritoneal fluid, or pleural fluid for laboratory evaluation was based on standard diagnostic protocols, and after informed consent from the patient, fluid samples were sent to the laboratory for diagnostic testing, which included biochemical analysis, concentration of adenosine deaminase, microscopy, Gram stain, culture, and sensitivity. The remaining fluid sample was centrifuged at 1100 rpm for 5minutes before 0.5 to 1 mL of supernatant was withdrawn, labeled, and frozen, at −70°C, before measurement of sTREM concentrationsusing a capture enzyme-linked immunosorbent assay procedure (Quantikine R&D Systems). Briefly, 100 μL of sample was transferred to the well of a microplate precoated with monoclonal murine immunoglobulin G directed against human TREM-1. After 2 hours of incubation at room temperature, the plates were washed, and bound sTREM was detected and quantified by the serial addition to the wells of a biotin-labeled second antibody, streptavidin-labeled horseradish peroxidase, and enzyme substrate. Samples were collected from January to December 2005 and July to October 2007.
The presence of microbial pathogens in CSF was determined by means of rapid bacterial antigen tests to detect bacterial polysaccharide antigens of pneumococci, meningococci, Haemophilus influenzae, and Group B streptococci, as well as Gram and Ziehl-Neelsen staining to detect the presence of bacteria and mycobacteria, respectively, followed by appropriate cultures. An India ink preparation was used to reveal the capsule of Cryptococcus species. For peritoneal and pleural fluid samples, Gram and Ziehl-Neelsen staining followed by cultures of the fluid was used to detect microbial pathogens.
The presence of bacteria in peritoneal or pleural fluid detected by Gram stain alone in the absence of growth of a microbial pathogen was not by itself considered sufficient evidence to support a diagnosis of peritonitis or pleuritis because this may detect the presence of nonviable organisms. However, the presence of acid-fast bacilli with Ziehl-Neelsen staining of any fluid sample was regarded as diagnostic of tuberculosis infection. In addition, the total white blood cell (WBC) count and serum C-reactive protein (CRP) concentrations were recorded for each patient. The number of polymorphonuclear leukocytes, lymphocytes, and erythrocytes per microliter (μL) of CSF or peritoneal fluid for those samples included in a subsequent analysis of the results were retrieved from a computerized database.
This study was approved by the ethics committee of Pretoria Academic Hospital.
Results are expressed as the mean ± SEM, together with the median and range (10th:90th percentile). Levels of statistical significance were calculated using the Mann-Whitney U test for comparison of nonparametric data, and a P < 0.05 was considered significant.
A total of 265 patients (132 men and 133 women with a mean age of 39 ± 1.1 years) was included in this study, with 101, 53, and 111 samples of CSF, peritoneal fluid, and pleural fluid, respectively, analyzed during the study period. The most common clinical diagnoses at presentation (time of sample collection) were confusion, fever and headache, delirium, sepsis, cirrhosis, portal hypertension, alcoholic liver disease, hepatic failure, congestive cardiac failure, renal failure, spontaneous bacterial peritonitis, pneumonia, parapneumonic effusions, empyema, and pulmonary tuberculosis. The number of samples for each type of fluid in which a microbial pathogen was present or absent, and the corresponding sTREM concentrations, WBC counts and serum CRP concentrations are shown in Table 1, and the sTREM concentrations are shown in Figure 1.
Microbial pathogens were identified in 29 CSF samples with a mean sTREM concentration of 134 ± 47 pg/mL. The pathogens present in this group were M. tuberculosis (8), Cryptococcus neoformans (9), Streptococcus pneumoniae (6), enterococci (2), Staphylococcus aureus (1), Listeria monocytogenes (1), Escherichia coli (1), and Salmonella typhi (1). Soluble TREM was not detected in 13 of the 29 samples from which pathogens were identified. Microbial pathogens were absent in 72 CSF samples, and sTREM was undetectable in 63 of these. The mean sTREM concentration of those CSF samples without an identified microbial pathogen was 48 ± 28 pg/mL compared with 134 ± 47 pg/mL for those with an identifiable microorganism (P < 0.05). The cell counts of CSF samples with sTREM concentrations greater than 50 pg/mL in the absence of microbial pathogens, together with the corresponding WBC counts and serum CRP concentrations, are shown in Table 2. Markedly abnormal cell counts were present in the CSF of 3 of 4 samples, suggestive of an inflammatory process or hemorrhage within the brain or meninges. At a cutoff level for sTREM of 50 pg/mL, the sensitivity was0.41 (95% confidence interval [CI], 0.24-0.61), and the specificity was 0.93 (95% CI, 0.85-0.98) with a positive predictive value (PPV) of 0.7 (95% CI, 0.44-0.99), a negative predictive value (NPV) of 0.8 (95% CI, 0.7-0.88), and a positive likelihood ratio of 5.96.
Microbial pathogens were isolated from 19 samples of peritoneal fluid with a corresponding sTREM concentration of 890 ± 332 pg/mL. The pathogens identified were gram-negative enteric bacilli (5), streptococci (3), Candida species (2), S. aureus (2), Staphylococcus epidermidis (1), M. tuberculosis (3), Pseudomonas aeruginosa (2), and Salmonella species (1). Soluble TREM was not detectable in 5 of 19 peritoneal fluid samples, despite the presence of a pathogen. Of these, M. tuberculosis was present in 1 sample with streptococci (2), P. aeruginosa (1), and E. coli (1) cultured from the remaining samples. No microbial pathogens were found in 34 samples of peritoneal fluid with a mean sTREM concentration of 123 ± 32 pg/mL compared with 890 ± 332 pg/mL for those samples with microbial pathogens (P > 0.05). The measured sTREM was greater than 60 pg/mL in 11 of 34 samples in the absence of a pathogen, and the cell counts of these samples, together with the corresponding WBC counts and serum CRP concentrations, are shown in Table 2. Markedly abnormal numbers of cells were present in 7 of these patients suggestive of an inflammatory process, and 4 samples were macroscopically blood stained indicative of hemorrhagic peritonitis. Cell counts were not available for 2 samples.
Using a cutoff value for sTREM of 100 pg/mL, the sensitivity was 0.65 (95% CI, 0.4-0.9), and the specificity was 0.73 (95% CI, 0.55-0.87) with a PPV of 0.60 (95% CI, 0.36-0.8), an NPV of 0.77 (95% CI, 0.59-0.90), and a positive likelihood ratio of 2.4.
Microbial pathogens were isolated from 49 samples of pleural fluid with a corresponding sTREM of 915 ± 159 pg/mL. This differed significantly from the sTREM concentration of 302 ± 48 pg/mL for pleural fluid without anidentifiable pathogen (P < 0.001) (Table 1). At a cutoff level for sTREM of 300 pg/mL, the sensitivity and specificity were 0.61 (95% CI, 0.46-0.75) and 0.8 (95% CI, 0.67-0.89), respectively, with a PPV of 0.73 (95% CI, 0.57-0.86), an NPV of 0.69 (95% CI, 0.56-0.8), and a positive likelihood ratio of 3.0.
A comparison of the sTREM and adenosine deaminase (ADA) concentrations for each class of pathogen is shown inTable 3. The mean sTREM concentrations associated with bacterial infections (263 ± 102 and 970 ± 185 pg/mL), were significantly greater than those measured in the presence of M. tuberculosis (61 ± 33 and 776 ± 320 pg/mL) for CSF, and pleural fluid, respectively (P < 0.05). Tuberculosis was positively identified from 3 peritoneal fluid samples with sTREM concentrations of 0, 146 and 4000 pg/mL. A similar trend was observed when comparing bacterial and fungal pathogens, although the number of fungal isolates was relatively small. The measured concentrations of ADA in CSF and peritoneal fluid for each classofpathogen did not differ significantly. However, the pleural fluid ADA concentration was significantly higher with tuberculous pleuritis (117 ± 17IU/L) compared withbacterial infections of the pleural cavity (61 ± 9 IU/L) (P < 0.05).
The total WBC counts and serum CRP concentrations did not differ significantly between those groups with identifiable pathogens and those without (Table 1).
The mortality from severe sepsis and septic shock remains high. Early recognition of infection before the development of systemic sepsis may facilitate rapid institution of appropriate antimicrobial therapy and drainage of septic foci, both of which are essential to improving sepsis survival.9 Markers of inflammation such as serum CRP, procalcitonin, and WBC count have been used to indicate the presence of a systemic inflammatory response, but do not routinely allow differentiation of infective from noninfective causes of inflammation.10 Although early antibiotic therapy reduces sepsis-related mortality, the injudicious use of antimicrobial agents may promote the development of resistant microorganisms.9 Therefore, it is of paramount importance to distinguish sepsis from other diseases causing systemic inflammation to minimize the inappropriate use of antimicrobial agents. Microbial culture of normally sterile body fluids is routinely performed as the criterion standard for diagnosing infection, but culture results may be delayed for a number of days.
The sTREM is a potential new marker of infection,11 which can be measured rapidly in the laboratory and may facilitate the early diagnosis of sepsis. Measurement of sTREM in body fluids such as BALF has been used to differentiate VAP from other causes of lung infiltrates,5 and in the presence of low concentrations of sTREM in BALF, clinicians may consider withholding antimicrobial therapy. Bronchial fluid samples obtained by means of nondirected lavage may also predict the presence of VAP in mechanically ventilated patients.12 Plasma concentrations of sTREM greater than 60 ng/mL in patients admitted to a medical intensive care unit predicted sepsis with a positive likelihood ratio of 8.6.6
Although sTREM has been measured in plasma6 and BALF5 in humans, only 1 study has, to our knowledge, been performed on other body fluids such as pleural fluid,13 and the spectrum of pathogens leading to up-regulation of sTREM has not been fully elucidated.
The current study has evaluated the role of sTREM during the laboratory evaluation of patients with suspected meningitis, peritonitis, or pleuritis using samples of CSF, peritoneal fluid, and pleural fluid, respectively. Soluble TREM was readily detectable and could be quantified in all types of fluid. The concentrations of sTREM measured in CSF from which a potential pathogen was identified differed significantly from those samples without an identifiable pathogen. Although sTREM concentrations were elevated in most CSF samples in the presence of a microbial pathogen, this was not true for all samples and may imply lack of sensitivity of the assay or loss of soluble receptors during transport and processing of the fluid. The sTREM was detected in the CSF of patients with cryptococcal and tuberculous meningitis, and those with bacterial infections. However, because of an apparent lack of sensitivity of sTREM, the diagnosis of meningitis must rely on standard microbiological tests.
A small proportion of CSF samples (8/72) had detectable sTREM in the absence of microbial pathogens. Although an infective agent could not be identified in these patients, other forms of inflammation could conceivably lead to up-regulation of sTREM expression in this setting, as suggested by the abnormal cell counts detected in all but one of these patients with corresponding sTREM concentrations greater than 50 pg/mL. Soluble TREM expression is reportedly up-regulated in the presence of extracellular pathogens2 when these are recognized by pathogen recognition receptors on the surfaces of inflammatory cells. However, enhanced sTREM expression has been recently documented in the presence of Marburg and Ebola virus infections.12
This suggests that TREM expression may not depend exclusively on the presence of bacterial and fungal infections, but also occurs when cells of the innate immune system respond to other pathogens. Therefore, detection of sTREM in CSF should alert clinicians to the possibility of an inflammatory process within the brain or meninges.
Soluble TREM measured in aspirates of peritoneal fluid revealed a trend toward significantly higher concentrations in patients with an identifiable pathogen compared with those in whom no pathogen was isolated. Most of the peritoneal fluid samples with positive microbial cultures were associated with significantly elevated sTREM concentrations. However, sTREM was not detected in the peritoneal fluid of 4 patients with positive bacterial cultures and 1 patient from whom M. tuberculosis was isolated. This may be attributed to lack of sensitivity of sTREM in detecting infection of the fluid or loss of sTREM during processing of the samples.
Peritoneal fluid sTREM concentrations greater than 60 pg/mL were found in 11 patients in the absence of microbial pathogens. Interestingly, 4 of these had macroscopic evidence of hemorrhagic ascites that may result from trauma during the aspiration procedure, as well as peritoneal inflammation from tuberculosis or malignancy. Furthermore, abnormal cell counts in keeping with an inflammatory response were found in 3 of the remaining 5 patients. Thus, elevated sTREM concentrations in peritoneal fluid in the absence of a positive bacterial culture should be investigated further to exclude tuberculosis or malignancy, provided traumatic contamination of the fluid has been excluded. Although a threshold level for predicting infection remains to be determined, sTREM concentrations in peritoneal fluid of greater than 300 pg/mL may indicate bacterial peritonitis and should prompt clinicians to initiate empiric antimicrobial therapy. Furthermore, antimicrobial therapy could be delayed with sTREM concentrations below 100 pg/mL, thereby reducing both the costs of therapy and the risk for development of antibiotic resistance.
The concentrations of sTREM measured in pleural fluid were significantly greater in those samples from which a microbial pathogen was identified compared with those with negative cultures. This suggests that sTREM measured in pleural fluid may be of value in the diagnosis of complicated parapneumonic effusions or empyema. The assay technique allows rapid quantification of sTREM concentrations in pleural fluid before culture results are usually available, and this may guide clinicians during the evaluation of patients with pleural effusions. Concentrations of sTREM greater than 300 to 350 pg/mL in pleural fluid samples may predict the presence of a parapneumonic effusion and the need for early empiric antimicrobial therapy. Unnecessary therapy with antimicrobials could be avoided in those patients with lower pleural fluid sTREM concentrations. Interestingly, 14 patients had tuberculous effusions with a mean sTREM concentration of 776 ± 320 pg/mL. Although this represents a relatively small number of all patients diagnosed with parapneumonic effusions, it does suggest that sTREM concentrations are increased with tuberculosis infection of the pleural cavity. Furthermore, 2 patients with confirmed tuberculous peritonitis had peritoneal fluid sTREM concentrations of 146 and 4000 pg/mL. These findings contrast with previous studies that reported that sTREM is not up-regulated by intracellular pathogens such as M. tuberculosis,8 and that sTREM concentrations measured in tuberculous effusions, although elevated, were lower than those associated with parapneumonic effusions.13 However, up-regulation of TREM in patients with Ebola and Marburg virus infections14 underscores the need for further studies to determine the spectrum of pathogens that may trigger TREM expression in vivo. In keeping with previous reports, ADA concentrations in pleural fluid greater than 100 IU/L may be diagnostic of tuberculous pleuritis. Limitations of our study include the possibility that positive microbial cultures could represent contamination of fluid samples, or that a true infection may be present in the absence of identifiable microorganisms. However, as a criterion standard for diagnosing infection may not exist for every clinical scenario, these difficulties reflect a common clinical dilemma and underscore the need for diagnostic tests of greater specificity.
In conclusion, the current study has identified a potentially useful role for sTREM during the laboratory evaluation of patients with suspected meningitis, peritonitis, and pleuritis. Although a pilot study, the findings are novel and pave the way for future trials to clearly define the clinical use of sTREM as a marker of infection. The importance of differentiating infectious from noninfectious causes of localized and systemic inflammation highlights the value of a marker, such as sTREM, with the potential toreliably predict the presence of microbial pathogens in body fluids.
© 2008 Lippincott Williams & Wilkins, Inc.